yachiyo8080 helcig commited on
Commit
980e00b
·
0 Parent(s):

Duplicate from ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF

Browse files

Co-authored-by: Michael Helcig <helcig@users.noreply.huggingface.co>

.gitattributes ADDED
@@ -0,0 +1,45 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ assets/banner.png filter=lfs diff=lfs merge=lfs -text
37
+ Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
38
+ Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
39
+ IQ2_XS/Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
40
+ IQ2_XS/Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00002-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
41
+ IQ3_XXS/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
42
+ IQ3_XXS/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00002-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
43
+ mmproj-Qwen3.8-Flash-Next-BF16.gguf filter=lfs diff=lfs merge=lfs -text
44
+ IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
45
+ IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00002-of-00002.gguf filter=lfs diff=lfs merge=lfs -text
IQ2_XS/Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92cee27ae5bbadcd732416a0f7a7f0acc092399dbbe8f5a5efa707c2ec0a49d7
3
+ size 39225954592
IQ2_XS/Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00002-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113
3
+ size 28800138432
IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3
3
+ size 54817524224
IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00002-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113
3
+ size 28800138432
IQ3_XXS/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:219ea929900dfa9ef091f3aa473fdba6874b65fcb36526d7d851ac9e95856d15
3
+ size 47039860096
IQ3_XXS/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00002-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113
3
+ size 28800138432
Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69820c02ec7d0b45ef2ebb19d6620299db749fe2aded7f39f93c6b88b199b720
3
+ size 37623740192
Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113
3
+ size 28800138432
README.md ADDED
@@ -0,0 +1,236 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen3.8-Flash-Next
3
+ base_model_relation: quantized
4
+ pipeline_tag: image-text-to-text
5
+ library_name: gguf
6
+ license: apache-2.0
7
+ tags:
8
+ - gguf
9
+ - gsq
10
+ - rco
11
+ - quantization
12
+ - mixed-precision
13
+ - ist-daslab
14
+ - moe
15
+ - multimodal
16
+ - vision
17
+ ---
18
+
19
+ <!--
20
+ GSQ-RCO GGUF release card for Qwen3.8-Flash-Next.
21
+ Mirrors the Qwen3.8-27B card.
22
+ NB: the YAML block must remain the very first bytes of the file (HF requirement).
23
+ -->
24
+
25
+ <div align="center">
26
+
27
+ <a href="https://github.com/IST-DASLab"><img src="assets/banner.png" alt="GGUF, GSQ-RCO dynamic non-uniform quantization" width="100%"/></a>
28
+
29
+ <br/>
30
+
31
+ # Qwen3.8-Flash-Next &middot; GSQ-RCO GGUFs
32
+
33
+ **Non-uniform GGUF quantizations** of a 512-expert MoE, produced with **GSQ** and **RCO**, with a vision projector for multimodal use.
34
+
35
+ [![arXiv: GSQ](https://img.shields.io/badge/arXiv-GSQ_2604.18556-b31b1b?logo=arxiv&logoColor=white)](https://arxiv.org/abs/2604.18556)
36
+ [![arXiv: RCO](https://img.shields.io/badge/arXiv-RCO_2605.00649-b31b1b?logo=arxiv&logoColor=white)](https://arxiv.org/abs/2605.00649)
37
+ [![GSQ code](https://img.shields.io/badge/code-GSQ-181717?logo=github&logoColor=white)](https://github.com/IST-DASLab/GSQ)
38
+ [![RCO code](https://img.shields.io/badge/code-RCO-181717?logo=github&logoColor=white)](https://github.com/IST-DASLab/RCO)
39
+ [![DASLab](https://img.shields.io/badge/DASLab-GitHub-101048?logo=github&logoColor=white)](https://github.com/IST-DASLab)
40
+ [![license](https://img.shields.io/badge/license-apache--2.0-19a34a)](#license)
41
+
42
+ </div>
43
+
44
+ ![Task average vs bit-width](assets/plots/Qwen3.8-Flash-Next-task_avg_vs_avg_bit_width.png)
45
+
46
+ ---
47
+
48
+ ## Overview
49
+
50
+ This repository provides GGUF quantizations of **Qwen3.8-Flash-Next** at four sizes, together with the model's vision projector (`mmproj`) for multimodal use. In contrast to uniform quantization, which applies a single quantization type to all weight tensors, each model here assigns a separate quantization type to every tensor. The assignment is obtained by a gradient-based search that allocates precision according to per-tensor sensitivity, subject to a total size budget. The resulting files are standard GGUF and run unmodified in `llama.cpp`, Ollama, and LM Studio.
51
+
52
+ A code-specialized variant is released separately as [Qwen3.8-Flash-Next-GSQ-RCO-Coder](https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF), which retains 256 of the 512 routed experts per layer at an effective 1.89 bpw over the original parameters, for a 58.4 GB total.
53
+
54
+ Flash-Next is a sparse mixture-of-experts model: **512 routed experts per layer across 48 layers**, of which 10 are active per token. The routed expert matrices dominate the parameter count, so they also dominate the size budget: 95% of the searchable weights live in the experts. The search therefore spends most of its freedom there, assigning a quantization type **per layer** (not per expert, which GGUF cannot express).
55
+
56
+ > **Method summary.** GSQ provides accurate low-bit scalar quantization of each tensor at a given quantization type; RCO assigns the per-tensor quantization types under a size budget. Together they yield a non-uniform GGUF at the requested size.
57
+
58
+ | Method | Description |
59
+ |---|---|
60
+ | **GSQ** (Gumbel-Softmax Quantization, [paper](https://arxiv.org/abs/2604.18556), [code](https://github.com/IST-DASLab/GSQ)) | Post-training scalar quantization that jointly learns the per-coordinate grid assignments and the per-group scales via a Gumbel-Softmax relaxation. GSQ closes most of the gap between scalar and vector quantization at 2 to 3 bits while remaining deployable in standard scalar formats such as GGUF. |
61
+ | **RCO** (Riemannian Constrained Optimization, [paper](https://arxiv.org/abs/2605.00649), [code](https://github.com/IST-DASLab/RCO)) | Assigns one of K quantization types to each of N tensors under a total size budget. The budget constraint is reformulated as a smooth Riemannian manifold in logit space, which permits gradient-based optimization directly on the task loss while enforcing the budget exactly, without constraint-specific hyperparameter tuning. |
62
+
63
+ Both methods were developed at the [Deep Algorithms and Systems Lab (DASLab)](https://github.com/IST-DASLab), Institute of Science and Technology Austria.
64
+
65
+ ---
66
+
67
+ ## Available files
68
+
69
+ Files follow the convention **`<model>-GSQ-RCO-<type>.gguf`**, where the suffix names the quantization class; the table lists each file's average bit-width over the transformer weights.
70
+
71
+ The `mmproj` file carries the vision encoder and projector at BF16; one copy serves all quantizations. Each quantized model ships as **two shards**. The second shard is identical across all four variants: it holds the per-layer n-gram embedding table (`per_layer_token_embd`, 51.2B parameters) at IQ4_NL. That table is a lookup, not a matmul weight, so it is held at a fixed 4.5 bpw in every build and excluded from the search. It is quantized in the evaluated models, not just the shipped ones.
72
+
73
+ | Directory | bpw | Shard 1 (weights) | Shard 2 (n-gram) | Total | Notes |
74
+ |---|---|---|---|---|---|
75
+ | `Q2_0/` | 2.40 | 37.6 GB | 28.8 GB | **66.4 GB** | Fastest; avoids lookup-table formats |
76
+ | `IQ2_XS/` | 2.50 | 39.2 GB | 28.8 GB | **68.0 GB** | Smallest at equal quality |
77
+ | `IQ3_XXS/` | 3.00 | 47.0 GB | 28.8 GB | **75.8 GB** | Matches the base model on AIME25 |
78
+ | `IQ3_S/` | 3.50 | 54.8 GB | 28.8 GB | **83.6 GB** | Recommended; matches or exceeds the base model on every task |
79
+ | `mmproj-Qwen3.8-Flash-Next-BF16.gguf` | 16 | 0.91 GB | n/a | **0.91 GB** | Vision encoder + projector, for multimodal use |
80
+
81
+ **Q2_0** is built for speed. It avoids the quantization formats that rely on large lookup tables: those formats pack more accuracy into a given bit-width, but decoding them costs real time, and on this model that cost dominates inference. Q2_0 delivers **3.4x the prompt throughput** and **1.9x lower end-to-end latency** than IQ2_XS at a slightly smaller file size, and its decode rate stays flat across workloads instead of varying with the content. The trade is a little quality: 89.07 task average against 89.16 for IQ2_XS, and 3.5 points below IQ3_XXS. Pick it when throughput matters most, and see [Performance](#performance) for the measurements.
82
+
83
+ The IQ3_S model is the strongest operating point: it matches the base model exactly on AIME25 (100.00), exceeds it on GPQA-Diamond (92.93 against 91.92) and is within 0.57 points on LiveCodeBench v6, at under a quarter of the BF16 size. IQ3_XXS remains the better choice when memory is tight: it gives up 1.52 points on GPQA-Diamond and 0.57 on LiveCodeBench v6 for 7.8 GB less.
84
+
85
+ ### Memory requirements
86
+
87
+ Only the first shard holds transformer weights and needs to be resident. The second shard is the n-gram lookup table: it is read sparsely, one row per token, so it can stay memory-mapped on disk and does not need to occupy VRAM. Run with `-lm mmap --lazy-mode on` to get this behaviour, plan for **shard 1 in VRAM**, and keep shard 2 on an SSD so the paging stays cheap.
88
+
89
+ | Model | Must be resident (shard 1) | Can stay on disk (shard 2) | Total download |
90
+ |---|---|---|---|
91
+ | `Q2_0` | 37.6 GB | 28.8 GB | 66.4 GB |
92
+ | `IQ2_XS` | 39.2 GB | 28.8 GB | 68.0 GB |
93
+ | `IQ3_XXS` | 47.0 GB | 28.8 GB | 75.8 GB |
94
+ | `IQ3_S` | 54.8 GB | 28.8 GB | 83.6 GB |
95
+
96
+ Add headroom for the KV cache and the vision projector (0.91 GB) if used. Keeping the n-gram table in RAM rather than on disk removes the paging cost entirely and is worth it where memory allows.
97
+
98
+ ---
99
+
100
+ ## Performance
101
+
102
+ Measured with `llama.cpp` over 55 prompts spanning eleven categories (coding, humanities, math, QA, RAG, reasoning, STEM, writing, multilingual, summarization, roleplay).
103
+
104
+ | Model | Prompt t/s | Decode t/s | Avg latency |
105
+ |---|---|---|---|
106
+ | `Q2_0` | **367.49** | **93.79** | **6.70 s** |
107
+ | `IQ2_XS` | 108.19 | 70.30 | 12.68 s |
108
+
109
+ ![Prefill throughput by category](assets/plots/Qwen3.8-Flash-Next-prefill_throughput.png)
110
+
111
+ The gap comes from the quantization formats rather than the size: the two files differ by only 1.6 GB, and Q2_0 is the smaller of the two. Q2_0's decode rate is also far more stable across categories, 92.5 to 94.5 t/s (a 2.0 t/s spread), where IQ2_XS swings from 49.4 to 94.5 t/s (a 45.1 t/s spread), because the lookup cost varies with how much of each layer a request touches.
112
+
113
+ The prefill advantage is uneven, and largest where prompts are long: 9.6x on RAG, 6.8x on writing and 6.2x on coding. On short-prompt, reasoning-heavy categories the two are level or Q2_0 is marginally behind (0.8x to 0.9x on stem, math and reasoning), where prefill is too brief for the format's decoding cost to matter.
114
+
115
+ ---
116
+
117
+ ## Results
118
+
119
+ All models are evaluated against the **BF16** base model. We report the average over five zero-shot tasks (arc_easy, arc_challenge, hellaswag, winogrande, piqa), recovery (zero-shot average relative to BF16), and three reasoning and generation benchmarks: **AIME25**, **GPQA-Diamond**, and **LiveCodeBench v6**. Task average is the mean of the three. Sizes are the full download, including the n-gram shard; the bit-width column is the average over the transformer weights, which is what the search controls.
120
+
121
+ | Variant | bpw | GB | ZS avg↑ | recovery | AIME25↑ | GPQA-D↑ | LCB v6↑ | Task avg↑ |
122
+ |---|---|---|---|---|---|---|---|---|
123
+ | BF16 | 16.00 | 354 | 76.94 | 100.0% | 100.00 | 91.92 | 87.43 | 93.12 |
124
+ | **GSQ-RCO Q2_0** | 2.40 | 66.4 | **78.00** | **101.4%** | 96.67 | 89.39 | 81.14 | 89.07 |
125
+ | **GSQ-RCO IQ2_XS** | 2.50 | 68.0 | 77.16 | 100.3% | 96.67 | 87.37 | 83.43 | 89.16 |
126
+ | **GSQ-RCO IQ3_XXS** | 3.00 | 75.8 | 77.23 | 100.4% | **100.00** | 91.41 | 86.29 | 92.57 |
127
+ | **GSQ-RCO IQ3_S** | 3.50 | 83.6 | n/a | n/a | **100.00** | **92.93** | **86.86** | **93.26** |
128
+
129
+ All three models exceed the BF16 base on the zero-shot average (100.3% to 101.4%), which is common for careful low-bit quantization on these five tasks: the differences are within roughly one standard error and should be read as parity rather than improvement. The same applies where IQ3_S exceeds the base on GPQA-Diamond and on the task average: recoveries slightly above 100% reflect benchmark variance, not a model that is better than the one it was quantized from. The reasoning and generation benchmarks separate the models far more clearly.
130
+
131
+ At 3.00 bpw, IQ3_XXS reaches **99.4%** of the base model's task average (92.57 against 93.12) while being **4.7x smaller** (75.8 GB against 354 GB). It matches the base exactly on AIME25 (100.00), and trails by 0.51 on GPQA-Diamond and 1.14 on LiveCodeBench v6. At 2.50 bpw, IQ2_XS retains 95.7% of the task average at 68.0 GB, and the smallest build, Q2_0, holds 96.67 on AIME25 at 66.4 GB, a 5.3x reduction.
132
+
133
+ The two sub-2.5 bpw builds land within 0.1 points of each other on task average despite different allocations: Q2_0 is stronger on GPQA-Diamond (89.39 against 87.37) while IQ2_XS is stronger on LiveCodeBench v6 (83.43 against 81.14).
134
+
135
+ ![AIME25 vs bit-width](assets/plots/Qwen3.8-Flash-Next-aime25_vs_avg_bit_width.png)
136
+
137
+ ![GPQA-Diamond vs bit-width](assets/plots/Qwen3.8-Flash-Next-gpqa_diamond_vs_avg_bit_width.png)
138
+
139
+ ![LiveCodeBench v6 vs bit-width](assets/plots/Qwen3.8-Flash-Next-lcb_vs_avg_bit_width.png)
140
+
141
+ ---
142
+
143
+ ## Usage
144
+
145
+ Both shards must be downloaded; `llama.cpp` loads the split automatically when given the first one.
146
+
147
+ ### llama.cpp
148
+ ```bash
149
+ # download (requires: pip install -U "huggingface_hub[cli]")
150
+ hf download <this-repo> --include "IQ3_XXS/*" --local-dir .
151
+
152
+ llama-cli -m IQ3_XXS/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf \
153
+ -lm mmap --lazy-mode on \
154
+ -p "Explain mixed-precision quantization." -ngl 99
155
+ ```
156
+ `-lm mmap --lazy-mode on` keeps the n-gram table memory-mapped on disk instead of loading it into memory, which removes 28.8 GB from the resident footprint. See [Memory requirements](#memory-requirements).
157
+
158
+ ### Strata Engine
159
+
160
+ Many users report strong results with [Strata](https://github.com/Niko1221/Strata), a third-party engine that runs these GGUFs on a single NVIDIA GPU from 8 GB of VRAM, with a one-click installer and an OpenAI-compatible endpoint on localhost.
161
+
162
+ ### Vision (multimodal)
163
+ ```bash
164
+ hf download <this-repo> mmproj-Qwen3.8-Flash-Next-BF16.gguf --local-dir .
165
+
166
+ llama-mtmd-cli -m IQ3_XXS/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf \
167
+ --mmproj mmproj-Qwen3.8-Flash-Next-BF16.gguf \
168
+ -lm mmap --lazy-mode on \
169
+ --image photo.jpg -p "Describe this image."
170
+ ```
171
+ The projector was converted directly from the base checkpoint and is shared by all four quantizations.
172
+
173
+ ### Ollama
174
+ ```bash
175
+ ollama run hf.co/<this-repo> # pick the directory matching your memory budget
176
+ ```
177
+
178
+ ### LM Studio
179
+ Search the repo name, then pick a `GSQ-RCO-*` build from the file list.
180
+
181
+ ---
182
+
183
+ ## Quantization procedure
184
+
185
+ 1. **Per-tensor database.** Each weight tensor is quantized at every candidate GGUF quantization type with GSQ, yielding a searchable database of quantized tensor variants.
186
+ 2. **RCO search.** The budget-constrained Riemannian search assigns one quantization type per tensor such that the whole-file average bit-width meets the target.
187
+ 3. **Assembly.** The selected per-tensor variants are stitched into a single standard GGUF file.
188
+
189
+ For this model the search covers **352 tensors**: the 304 dense tensors (attention, SSM projections, shared experts, embeddings, output head) and the 48 fused routed-expert matrices, one per layer. Two classes draw on separate quantization ladders, because the routed experts sit at a much lower bit-width than the dense path at any given budget. `ffn_down_exps` searches over a much smaller ladder than the other expert matrices. Its 640 rows are not divisible by 256, which rules out every block-256 K and I quant format and leaves only Q2_0 (block 64) and IQ4_NL (block 32). In the two smaller builds every layer lands on Q2_0; in IQ3_XXS the search lifts 18 of the 48 layers to IQ4_NL.
190
+
191
+ Reference implementations: **GSQ** at [IST-DASLab/GSQ](https://github.com/IST-DASLab/GSQ) and **RCO** at [IST-DASLab/RCO](https://github.com/IST-DASLab/RCO).
192
+
193
+ ### Reproducibility artifacts
194
+
195
+ Each released GGUF ships the files needed to audit how it was built:
196
+
197
+ | File | Contents |
198
+ |---|---|
199
+ | `tensor-allocation/<model>.rco-allocation.txt` | The quantization type assigned to every tensor in that file, with a quant-type histogram and the target bit-width. This is the RCO search result, so the allocation can be inspected without opening the model. |
200
+
201
+ ---
202
+
203
+ ## Citation
204
+
205
+ If you use these models or methods, please cite both papers:
206
+
207
+ ```bibtex
208
+ @article{gsq2026,
209
+ title = {GSQ: Highly-Accurate Low-Precision Scalar Quantization for LLMs via Gumbel-Softmax Sampling},
210
+ author = {Dadgarnia, Alireza and Tabesh, Soroush and Nikdan, Mahdi and Helcig, Michael and Kurtic, Eldar and Kleinegger, Maximilian and Alistarh, Dan},
211
+ journal= {arXiv preprint arXiv:2604.18556},
212
+ year = {2026}
213
+ }
214
+ @article{rco2026,
215
+ title = {Model Compression with Exact Budget Constraints via Riemannian Manifolds},
216
+ author = {Helcig, Michael and Alistarh, Dan},
217
+ journal= {arXiv preprint arXiv:2605.00649},
218
+ year = {2026}
219
+ }
220
+ ```
221
+
222
+ ---
223
+
224
+ ## Acknowledgements
225
+
226
+ We thank [Verda](https://verda.com/) and Scientific Computing at the Institute of Science and Technology Austria for providing the compute resources used to produce these models.
227
+
228
+ ---
229
+
230
+ ## License
231
+
232
+ These quantized weights inherit the license of the base model (**Qwen3.8-Flash-Next**). The GSQ-RCO tooling is released by the Deep Algorithms and Systems Lab under its repository license.
233
+
234
+ <div align="center">
235
+ <sub>Built with <b>GSQ</b> and <b>RCO</b> at the <a href="https://github.com/IST-DASLab">Deep Algorithms and Systems Lab</a> &middot; Institute of Science and Technology Austria</sub>
236
+ </div>
assets/banner.png ADDED

Git LFS Details

  • SHA256: 42ce97689e7497dcc46a2fe4701e70ee39ad856f9edffaae3b3220ea5628e076
  • Pointer size: 131 Bytes
  • Size of remote file: 310 kB
assets/plots/Qwen3.8-Flash-Next-aime25_vs_avg_bit_width.png ADDED
assets/plots/Qwen3.8-Flash-Next-gpqa_diamond_vs_avg_bit_width.png ADDED
assets/plots/Qwen3.8-Flash-Next-lcb_vs_avg_bit_width.png ADDED
assets/plots/Qwen3.8-Flash-Next-prefill_throughput.png ADDED
assets/plots/Qwen3.8-Flash-Next-task_avg_vs_avg_bit_width.png ADDED
mmproj-Qwen3.8-Flash-Next-BF16.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1a82259702816a5330d7bd7607cd9676b11780e79ff7348c21103ff3ce49bd0
3
+ size 907543008
tensor-allocation/Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.rco-allocation.txt ADDED
@@ -0,0 +1,1235 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # RCO per-tensor GGUF allocation
2
+ # Repository: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF
3
+ # GGUF: Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.gguf
4
+ # Source: https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF/resolve/main/Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.gguf
5
+ # Target whole-file bpw: 2.50
6
+ # GGUF version: 3
7
+ # GGUF metadata entries: 73
8
+ # Tensor count: 1223
9
+ # Quant-type counts: BF16=484, F16=1, F32=292, IQ1_M=6, IQ2_S=68, IQ2_XXS=22, IQ3_S=44, IQ4_NL=36, IQ4_XS=181, Q2_0=53, Q6_K=29, Q8_0=7
10
+ # Format: <GGUF tensor name>: <GGML quantization/storage type>
11
+ # Includes fixed non-RCO tensors (e.g. F32/BF16) so the released GGUF is fully reproducible.
12
+ #
13
+ output.weight: IQ4_XS
14
+ output_hc_down.weight: BF16
15
+ output_hc_norm.weight: F32
16
+ output_hc_up.weight: BF16
17
+ token_embd.weight: IQ4_XS
18
+ blk.0.attn_gate.weight: IQ3_S
19
+ blk.0.attn_qkv.weight: IQ4_XS
20
+ blk.0.ffn_down_exps.weight: Q2_0
21
+ blk.0.ffn_down_shexp.weight: IQ4_NL
22
+ blk.0.ffn_gate_exps.weight: IQ2_S
23
+ blk.0.ffn_gate_inp.weight: BF16
24
+ blk.0.ffn_gate_inp_shexp.weight: BF16
25
+ blk.0.ffn_gate_shexp.weight: IQ3_S
26
+ blk.0.ffn_up_exps.weight: IQ2_S
27
+ blk.0.ffn_up_shexp.weight: IQ4_XS
28
+ blk.0.hc_attn_down.weight: BF16
29
+ blk.0.hc_attn_inject.weight: BF16
30
+ blk.0.hc_attn_norm.weight: F32
31
+ blk.0.hc_attn_up.weight: BF16
32
+ blk.0.hc_ffn_down.weight: BF16
33
+ blk.0.hc_ffn_inject.weight: BF16
34
+ blk.0.hc_ffn_norm.weight: F32
35
+ blk.0.hc_ffn_up.weight: BF16
36
+ blk.0.ssm_a: F32
37
+ blk.0.ssm_alpha.weight: BF16
38
+ blk.0.ssm_beta.weight: BF16
39
+ blk.0.ssm_conv1d.weight: F32
40
+ blk.0.ssm_dt.bias: F32
41
+ blk.0.ssm_norm.weight: F32
42
+ blk.0.ssm_out.weight: IQ4_XS
43
+ blk.1.attn_gate.weight: IQ3_S
44
+ blk.1.attn_qkv.weight: IQ4_XS
45
+ blk.1.ffn_down_exps.weight: Q2_0
46
+ blk.1.ffn_down_shexp.weight: IQ4_NL
47
+ blk.1.ffn_gate_exps.weight: IQ2_XXS
48
+ blk.1.ffn_gate_inp.weight: BF16
49
+ blk.1.ffn_gate_inp_shexp.weight: BF16
50
+ blk.1.ffn_gate_shexp.weight: Q6_K
51
+ blk.1.ffn_up_exps.weight: IQ2_XXS
52
+ blk.1.ffn_up_shexp.weight: IQ4_XS
53
+ blk.1.hc_attn_down.weight: BF16
54
+ blk.1.hc_attn_inject.weight: BF16
55
+ blk.1.hc_attn_norm.weight: F32
56
+ blk.1.hc_attn_up.weight: BF16
57
+ blk.1.hc_ffn_down.weight: BF16
58
+ blk.1.hc_ffn_inject.weight: BF16
59
+ blk.1.hc_ffn_norm.weight: F32
60
+ blk.1.hc_ffn_up.weight: BF16
61
+ blk.1.ple_conv1d.weight: F16
62
+ blk.1.ple_key.weight: BF16
63
+ blk.1.ple_norm_conv.weight: F32
64
+ blk.1.ple_norm_key.weight: F32
65
+ blk.1.ple_norm_query.weight: F32
66
+ blk.1.ple_value.weight: BF16
67
+ blk.1.ssm_a: F32
68
+ blk.1.ssm_alpha.weight: BF16
69
+ blk.1.ssm_beta.weight: BF16
70
+ blk.1.ssm_conv1d.weight: F32
71
+ blk.1.ssm_dt.bias: F32
72
+ blk.1.ssm_norm.weight: F32
73
+ blk.1.ssm_out.weight: IQ4_XS
74
+ blk.2.attn_gate.weight: IQ4_XS
75
+ blk.2.attn_qkv.weight: IQ4_XS
76
+ blk.2.ffn_down_exps.weight: Q2_0
77
+ blk.2.ffn_down_shexp.weight: IQ4_NL
78
+ blk.2.ffn_gate_exps.weight: IQ2_S
79
+ blk.2.ffn_gate_inp.weight: BF16
80
+ blk.2.ffn_gate_inp_shexp.weight: BF16
81
+ blk.2.ffn_gate_shexp.weight: IQ4_XS
82
+ blk.2.ffn_up_exps.weight: IQ2_S
83
+ blk.2.ffn_up_shexp.weight: IQ4_XS
84
+ blk.2.hc_attn_down.weight: BF16
85
+ blk.2.hc_attn_inject.weight: BF16
86
+ blk.2.hc_attn_norm.weight: F32
87
+ blk.2.hc_attn_up.weight: BF16
88
+ blk.2.hc_ffn_down.weight: BF16
89
+ blk.2.hc_ffn_inject.weight: BF16
90
+ blk.2.hc_ffn_norm.weight: F32
91
+ blk.2.hc_ffn_up.weight: BF16
92
+ blk.2.ssm_a: F32
93
+ blk.2.ssm_alpha.weight: BF16
94
+ blk.2.ssm_beta.weight: BF16
95
+ blk.2.ssm_conv1d.weight: F32
96
+ blk.2.ssm_dt.bias: F32
97
+ blk.2.ssm_norm.weight: F32
98
+ blk.2.ssm_out.weight: IQ4_XS
99
+ blk.3.attn_k.weight: Q6_K
100
+ blk.3.attn_k_norm.weight: F32
101
+ blk.3.attn_output.weight: IQ3_S
102
+ blk.3.attn_q.weight: IQ4_XS
103
+ blk.3.attn_q_norm.weight: F32
104
+ blk.3.attn_v.weight: IQ4_XS
105
+ blk.3.ffn_down_exps.weight: Q2_0
106
+ blk.3.ffn_down_shexp.weight: Q8_0
107
+ blk.3.ffn_gate_exps.weight: IQ2_S
108
+ blk.3.ffn_gate_inp.weight: BF16
109
+ blk.3.ffn_gate_inp_shexp.weight: BF16
110
+ blk.3.ffn_gate_shexp.weight: IQ4_XS
111
+ blk.3.ffn_up_exps.weight: IQ2_S
112
+ blk.3.ffn_up_shexp.weight: IQ4_XS
113
+ blk.3.hc_attn_down.weight: BF16
114
+ blk.3.hc_attn_inject.weight: BF16
115
+ blk.3.hc_attn_norm.weight: F32
116
+ blk.3.hc_attn_up.weight: BF16
117
+ blk.3.hc_ffn_down.weight: BF16
118
+ blk.3.hc_ffn_inject.weight: BF16
119
+ blk.3.hc_ffn_norm.weight: F32
120
+ blk.3.hc_ffn_up.weight: BF16
121
+ blk.3.indexer.k_norm.weight: F32
122
+ blk.3.indexer.k_proj.weight: BF16
123
+ blk.3.indexer.q_norm.weight: F32
124
+ blk.3.indexer.q_proj.weight: BF16
125
+ blk.4.attn_gate.weight: IQ3_S
126
+ blk.4.attn_qkv.weight: IQ4_XS
127
+ blk.4.ffn_down_exps.weight: Q2_0
128
+ blk.4.ffn_down_shexp.weight: IQ4_NL
129
+ blk.4.ffn_gate_exps.weight: IQ2_XXS
130
+ blk.4.ffn_gate_inp.weight: BF16
131
+ blk.4.ffn_gate_inp_shexp.weight: BF16
132
+ blk.4.ffn_gate_shexp.weight: IQ4_XS
133
+ blk.4.ffn_up_exps.weight: IQ2_XXS
134
+ blk.4.ffn_up_shexp.weight: IQ3_S
135
+ blk.4.hc_attn_down.weight: BF16
136
+ blk.4.hc_attn_inject.weight: BF16
137
+ blk.4.hc_attn_norm.weight: F32
138
+ blk.4.hc_attn_up.weight: BF16
139
+ blk.4.hc_ffn_down.weight: BF16
140
+ blk.4.hc_ffn_inject.weight: BF16
141
+ blk.4.hc_ffn_norm.weight: F32
142
+ blk.4.hc_ffn_up.weight: BF16
143
+ blk.4.ssm_a: F32
144
+ blk.4.ssm_alpha.weight: BF16
145
+ blk.4.ssm_beta.weight: BF16
146
+ blk.4.ssm_conv1d.weight: F32
147
+ blk.4.ssm_dt.bias: F32
148
+ blk.4.ssm_norm.weight: F32
149
+ blk.4.ssm_out.weight: IQ4_XS
150
+ blk.5.attn_gate.weight: IQ3_S
151
+ blk.5.attn_qkv.weight: IQ4_XS
152
+ blk.5.ffn_down_exps.weight: Q2_0
153
+ blk.5.ffn_down_shexp.weight: IQ4_NL
154
+ blk.5.ffn_gate_exps.weight: IQ2_S
155
+ blk.5.ffn_gate_inp.weight: BF16
156
+ blk.5.ffn_gate_inp_shexp.weight: BF16
157
+ blk.5.ffn_gate_shexp.weight: IQ4_XS
158
+ blk.5.ffn_up_exps.weight: IQ2_S
159
+ blk.5.ffn_up_shexp.weight: Q6_K
160
+ blk.5.hc_attn_down.weight: BF16
161
+ blk.5.hc_attn_inject.weight: BF16
162
+ blk.5.hc_attn_norm.weight: F32
163
+ blk.5.hc_attn_up.weight: BF16
164
+ blk.5.hc_ffn_down.weight: BF16
165
+ blk.5.hc_ffn_inject.weight: BF16
166
+ blk.5.hc_ffn_norm.weight: F32
167
+ blk.5.hc_ffn_up.weight: BF16
168
+ blk.5.ssm_a: F32
169
+ blk.5.ssm_alpha.weight: BF16
170
+ blk.5.ssm_beta.weight: BF16
171
+ blk.5.ssm_conv1d.weight: F32
172
+ blk.5.ssm_dt.bias: F32
173
+ blk.5.ssm_norm.weight: F32
174
+ blk.5.ssm_out.weight: IQ4_XS
175
+ blk.6.attn_gate.weight: IQ4_XS
176
+ blk.6.attn_qkv.weight: IQ4_XS
177
+ blk.6.ffn_down_exps.weight: Q2_0
178
+ blk.6.ffn_down_shexp.weight: Q2_0
179
+ blk.6.ffn_gate_exps.weight: IQ2_S
180
+ blk.6.ffn_gate_inp.weight: BF16
181
+ blk.6.ffn_gate_inp_shexp.weight: BF16
182
+ blk.6.ffn_gate_shexp.weight: IQ4_XS
183
+ blk.6.ffn_up_exps.weight: IQ2_S
184
+ blk.6.ffn_up_shexp.weight: IQ4_XS
185
+ blk.6.hc_attn_down.weight: BF16
186
+ blk.6.hc_attn_inject.weight: BF16
187
+ blk.6.hc_attn_norm.weight: F32
188
+ blk.6.hc_attn_up.weight: BF16
189
+ blk.6.hc_ffn_down.weight: BF16
190
+ blk.6.hc_ffn_inject.weight: BF16
191
+ blk.6.hc_ffn_norm.weight: F32
192
+ blk.6.hc_ffn_up.weight: BF16
193
+ blk.6.ssm_a: F32
194
+ blk.6.ssm_alpha.weight: BF16
195
+ blk.6.ssm_beta.weight: BF16
196
+ blk.6.ssm_conv1d.weight: F32
197
+ blk.6.ssm_dt.bias: F32
198
+ blk.6.ssm_norm.weight: F32
199
+ blk.6.ssm_out.weight: IQ4_XS
200
+ blk.7.attn_k.weight: Q6_K
201
+ blk.7.attn_k_norm.weight: F32
202
+ blk.7.attn_output.weight: IQ4_XS
203
+ blk.7.attn_q.weight: IQ4_XS
204
+ blk.7.attn_q_norm.weight: F32
205
+ blk.7.attn_v.weight: Q6_K
206
+ blk.7.ffn_down_exps.weight: Q2_0
207
+ blk.7.ffn_down_shexp.weight: Q2_0
208
+ blk.7.ffn_gate_exps.weight: IQ2_S
209
+ blk.7.ffn_gate_inp.weight: BF16
210
+ blk.7.ffn_gate_inp_shexp.weight: BF16
211
+ blk.7.ffn_gate_shexp.weight: Q6_K
212
+ blk.7.ffn_up_exps.weight: IQ2_S
213
+ blk.7.ffn_up_shexp.weight: IQ4_XS
214
+ blk.7.hc_attn_down.weight: BF16
215
+ blk.7.hc_attn_inject.weight: BF16
216
+ blk.7.hc_attn_norm.weight: F32
217
+ blk.7.hc_attn_up.weight: BF16
218
+ blk.7.hc_ffn_down.weight: BF16
219
+ blk.7.hc_ffn_inject.weight: BF16
220
+ blk.7.hc_ffn_norm.weight: F32
221
+ blk.7.hc_ffn_up.weight: BF16
222
+ blk.7.indexer.k_norm.weight: F32
223
+ blk.7.indexer.k_proj.weight: BF16
224
+ blk.7.indexer.q_norm.weight: F32
225
+ blk.7.indexer.q_proj.weight: BF16
226
+ blk.8.attn_gate.weight: IQ4_XS
227
+ blk.8.attn_qkv.weight: IQ4_XS
228
+ blk.8.ffn_down_exps.weight: Q2_0
229
+ blk.8.ffn_down_shexp.weight: IQ4_NL
230
+ blk.8.ffn_gate_exps.weight: IQ1_M
231
+ blk.8.ffn_gate_inp.weight: BF16
232
+ blk.8.ffn_gate_inp_shexp.weight: BF16
233
+ blk.8.ffn_gate_shexp.weight: IQ4_XS
234
+ blk.8.ffn_up_exps.weight: IQ1_M
235
+ blk.8.ffn_up_shexp.weight: IQ4_XS
236
+ blk.8.hc_attn_down.weight: BF16
237
+ blk.8.hc_attn_inject.weight: BF16
238
+ blk.8.hc_attn_norm.weight: F32
239
+ blk.8.hc_attn_up.weight: BF16
240
+ blk.8.hc_ffn_down.weight: BF16
241
+ blk.8.hc_ffn_inject.weight: BF16
242
+ blk.8.hc_ffn_norm.weight: F32
243
+ blk.8.hc_ffn_up.weight: BF16
244
+ blk.8.ssm_a: F32
245
+ blk.8.ssm_alpha.weight: BF16
246
+ blk.8.ssm_beta.weight: BF16
247
+ blk.8.ssm_conv1d.weight: F32
248
+ blk.8.ssm_dt.bias: F32
249
+ blk.8.ssm_norm.weight: F32
250
+ blk.8.ssm_out.weight: IQ3_S
251
+ blk.9.attn_gate.weight: IQ4_XS
252
+ blk.9.attn_qkv.weight: IQ4_XS
253
+ blk.9.ffn_down_exps.weight: Q2_0
254
+ blk.9.ffn_down_shexp.weight: Q8_0
255
+ blk.9.ffn_gate_exps.weight: IQ2_XXS
256
+ blk.9.ffn_gate_inp.weight: BF16
257
+ blk.9.ffn_gate_inp_shexp.weight: BF16
258
+ blk.9.ffn_gate_shexp.weight: IQ3_S
259
+ blk.9.ffn_up_exps.weight: IQ2_XXS
260
+ blk.9.ffn_up_shexp.weight: IQ3_S
261
+ blk.9.hc_attn_down.weight: BF16
262
+ blk.9.hc_attn_inject.weight: BF16
263
+ blk.9.hc_attn_norm.weight: F32
264
+ blk.9.hc_attn_up.weight: BF16
265
+ blk.9.hc_ffn_down.weight: BF16
266
+ blk.9.hc_ffn_inject.weight: BF16
267
+ blk.9.hc_ffn_norm.weight: F32
268
+ blk.9.hc_ffn_up.weight: BF16
269
+ blk.9.ssm_a: F32
270
+ blk.9.ssm_alpha.weight: BF16
271
+ blk.9.ssm_beta.weight: BF16
272
+ blk.9.ssm_conv1d.weight: F32
273
+ blk.9.ssm_dt.bias: F32
274
+ blk.9.ssm_norm.weight: F32
275
+ blk.9.ssm_out.weight: IQ4_XS
276
+ blk.10.attn_gate.weight: IQ4_XS
277
+ blk.10.attn_qkv.weight: IQ4_XS
278
+ blk.10.ffn_down_exps.weight: Q2_0
279
+ blk.10.ffn_down_shexp.weight: IQ4_NL
280
+ blk.10.ffn_gate_exps.weight: IQ2_XXS
281
+ blk.10.ffn_gate_inp.weight: BF16
282
+ blk.10.ffn_gate_inp_shexp.weight: BF16
283
+ blk.10.ffn_gate_shexp.weight: IQ4_XS
284
+ blk.10.ffn_up_exps.weight: IQ2_XXS
285
+ blk.10.ffn_up_shexp.weight: IQ3_S
286
+ blk.10.hc_attn_down.weight: BF16
287
+ blk.10.hc_attn_inject.weight: BF16
288
+ blk.10.hc_attn_norm.weight: F32
289
+ blk.10.hc_attn_up.weight: BF16
290
+ blk.10.hc_ffn_down.weight: BF16
291
+ blk.10.hc_ffn_inject.weight: BF16
292
+ blk.10.hc_ffn_norm.weight: F32
293
+ blk.10.hc_ffn_up.weight: BF16
294
+ blk.10.ssm_a: F32
295
+ blk.10.ssm_alpha.weight: BF16
296
+ blk.10.ssm_beta.weight: BF16
297
+ blk.10.ssm_conv1d.weight: F32
298
+ blk.10.ssm_dt.bias: F32
299
+ blk.10.ssm_norm.weight: F32
300
+ blk.10.ssm_out.weight: IQ4_XS
301
+ blk.11.attn_k.weight: IQ4_XS
302
+ blk.11.attn_k_norm.weight: F32
303
+ blk.11.attn_output.weight: IQ4_XS
304
+ blk.11.attn_q.weight: IQ4_XS
305
+ blk.11.attn_q_norm.weight: F32
306
+ blk.11.attn_v.weight: IQ4_XS
307
+ blk.11.ffn_down_exps.weight: Q2_0
308
+ blk.11.ffn_down_shexp.weight: Q2_0
309
+ blk.11.ffn_gate_exps.weight: IQ2_XXS
310
+ blk.11.ffn_gate_inp.weight: BF16
311
+ blk.11.ffn_gate_inp_shexp.weight: BF16
312
+ blk.11.ffn_gate_shexp.weight: IQ4_XS
313
+ blk.11.ffn_up_exps.weight: IQ2_XXS
314
+ blk.11.ffn_up_shexp.weight: Q6_K
315
+ blk.11.hc_attn_down.weight: BF16
316
+ blk.11.hc_attn_inject.weight: BF16
317
+ blk.11.hc_attn_norm.weight: F32
318
+ blk.11.hc_attn_up.weight: BF16
319
+ blk.11.hc_ffn_down.weight: BF16
320
+ blk.11.hc_ffn_inject.weight: BF16
321
+ blk.11.hc_ffn_norm.weight: F32
322
+ blk.11.hc_ffn_up.weight: BF16
323
+ blk.11.indexer.k_norm.weight: F32
324
+ blk.11.indexer.k_proj.weight: BF16
325
+ blk.11.indexer.q_norm.weight: F32
326
+ blk.11.indexer.q_proj.weight: BF16
327
+ blk.12.attn_gate.weight: IQ4_XS
328
+ blk.12.attn_qkv.weight: IQ4_XS
329
+ blk.12.ffn_down_exps.weight: Q2_0
330
+ blk.12.ffn_down_shexp.weight: IQ4_NL
331
+ blk.12.ffn_gate_exps.weight: IQ2_S
332
+ blk.12.ffn_gate_inp.weight: BF16
333
+ blk.12.ffn_gate_inp_shexp.weight: BF16
334
+ blk.12.ffn_gate_shexp.weight: Q6_K
335
+ blk.12.ffn_up_exps.weight: IQ2_S
336
+ blk.12.ffn_up_shexp.weight: IQ3_S
337
+ blk.12.hc_attn_down.weight: BF16
338
+ blk.12.hc_attn_inject.weight: BF16
339
+ blk.12.hc_attn_norm.weight: F32
340
+ blk.12.hc_attn_up.weight: BF16
341
+ blk.12.hc_ffn_down.weight: BF16
342
+ blk.12.hc_ffn_inject.weight: BF16
343
+ blk.12.hc_ffn_norm.weight: F32
344
+ blk.12.hc_ffn_up.weight: BF16
345
+ blk.12.ssm_a: F32
346
+ blk.12.ssm_alpha.weight: BF16
347
+ blk.12.ssm_beta.weight: BF16
348
+ blk.12.ssm_conv1d.weight: F32
349
+ blk.12.ssm_dt.bias: F32
350
+ blk.12.ssm_norm.weight: F32
351
+ blk.12.ssm_out.weight: IQ4_XS
352
+ blk.13.attn_gate.weight: IQ4_XS
353
+ blk.13.attn_qkv.weight: IQ4_XS
354
+ blk.13.ffn_down_exps.weight: Q2_0
355
+ blk.13.ffn_down_shexp.weight: Q8_0
356
+ blk.13.ffn_gate_exps.weight: IQ1_M
357
+ blk.13.ffn_gate_inp.weight: BF16
358
+ blk.13.ffn_gate_inp_shexp.weight: BF16
359
+ blk.13.ffn_gate_shexp.weight: IQ3_S
360
+ blk.13.ffn_up_exps.weight: IQ1_M
361
+ blk.13.ffn_up_shexp.weight: IQ3_S
362
+ blk.13.hc_attn_down.weight: BF16
363
+ blk.13.hc_attn_inject.weight: BF16
364
+ blk.13.hc_attn_norm.weight: F32
365
+ blk.13.hc_attn_up.weight: BF16
366
+ blk.13.hc_ffn_down.weight: BF16
367
+ blk.13.hc_ffn_inject.weight: BF16
368
+ blk.13.hc_ffn_norm.weight: F32
369
+ blk.13.hc_ffn_up.weight: BF16
370
+ blk.13.ssm_a: F32
371
+ blk.13.ssm_alpha.weight: BF16
372
+ blk.13.ssm_beta.weight: BF16
373
+ blk.13.ssm_conv1d.weight: F32
374
+ blk.13.ssm_dt.bias: F32
375
+ blk.13.ssm_norm.weight: F32
376
+ blk.13.ssm_out.weight: IQ4_XS
377
+ blk.14.attn_gate.weight: IQ3_S
378
+ blk.14.attn_qkv.weight: IQ4_XS
379
+ blk.14.ffn_down_exps.weight: Q2_0
380
+ blk.14.ffn_down_shexp.weight: Q8_0
381
+ blk.14.ffn_gate_exps.weight: IQ2_XXS
382
+ blk.14.ffn_gate_inp.weight: BF16
383
+ blk.14.ffn_gate_inp_shexp.weight: BF16
384
+ blk.14.ffn_gate_shexp.weight: IQ4_XS
385
+ blk.14.ffn_up_exps.weight: IQ2_XXS
386
+ blk.14.ffn_up_shexp.weight: IQ4_XS
387
+ blk.14.hc_attn_down.weight: BF16
388
+ blk.14.hc_attn_inject.weight: BF16
389
+ blk.14.hc_attn_norm.weight: F32
390
+ blk.14.hc_attn_up.weight: BF16
391
+ blk.14.hc_ffn_down.weight: BF16
392
+ blk.14.hc_ffn_inject.weight: BF16
393
+ blk.14.hc_ffn_norm.weight: F32
394
+ blk.14.hc_ffn_up.weight: BF16
395
+ blk.14.ssm_a: F32
396
+ blk.14.ssm_alpha.weight: BF16
397
+ blk.14.ssm_beta.weight: BF16
398
+ blk.14.ssm_conv1d.weight: F32
399
+ blk.14.ssm_dt.bias: F32
400
+ blk.14.ssm_norm.weight: F32
401
+ blk.14.ssm_out.weight: IQ4_XS
402
+ blk.15.attn_k.weight: IQ3_S
403
+ blk.15.attn_k_norm.weight: F32
404
+ blk.15.attn_output.weight: Q6_K
405
+ blk.15.attn_q.weight: IQ4_XS
406
+ blk.15.attn_q_norm.weight: F32
407
+ blk.15.attn_v.weight: IQ4_XS
408
+ blk.15.ffn_down_exps.weight: Q2_0
409
+ blk.15.ffn_down_shexp.weight: IQ4_NL
410
+ blk.15.ffn_gate_exps.weight: IQ2_S
411
+ blk.15.ffn_gate_inp.weight: BF16
412
+ blk.15.ffn_gate_inp_shexp.weight: BF16
413
+ blk.15.ffn_gate_shexp.weight: IQ3_S
414
+ blk.15.ffn_up_exps.weight: IQ2_S
415
+ blk.15.ffn_up_shexp.weight: Q6_K
416
+ blk.15.hc_attn_down.weight: BF16
417
+ blk.15.hc_attn_inject.weight: BF16
418
+ blk.15.hc_attn_norm.weight: F32
419
+ blk.15.hc_attn_up.weight: BF16
420
+ blk.15.hc_ffn_down.weight: BF16
421
+ blk.15.hc_ffn_inject.weight: BF16
422
+ blk.15.hc_ffn_norm.weight: F32
423
+ blk.15.hc_ffn_up.weight: BF16
424
+ blk.15.indexer.k_norm.weight: F32
425
+ blk.15.indexer.k_proj.weight: BF16
426
+ blk.15.indexer.q_norm.weight: F32
427
+ blk.15.indexer.q_proj.weight: BF16
428
+ blk.16.attn_gate.weight: IQ4_XS
429
+ blk.16.attn_qkv.weight: IQ4_XS
430
+ blk.16.ffn_down_exps.weight: Q2_0
431
+ blk.16.ffn_down_shexp.weight: Q8_0
432
+ blk.16.ffn_gate_exps.weight: IQ2_S
433
+ blk.16.ffn_gate_inp.weight: BF16
434
+ blk.16.ffn_gate_inp_shexp.weight: BF16
435
+ blk.16.ffn_gate_shexp.weight: Q6_K
436
+ blk.16.ffn_up_exps.weight: IQ2_S
437
+ blk.16.ffn_up_shexp.weight: Q6_K
438
+ blk.16.hc_attn_down.weight: BF16
439
+ blk.16.hc_attn_inject.weight: BF16
440
+ blk.16.hc_attn_norm.weight: F32
441
+ blk.16.hc_attn_up.weight: BF16
442
+ blk.16.hc_ffn_down.weight: BF16
443
+ blk.16.hc_ffn_inject.weight: BF16
444
+ blk.16.hc_ffn_norm.weight: F32
445
+ blk.16.hc_ffn_up.weight: BF16
446
+ blk.16.ssm_a: F32
447
+ blk.16.ssm_alpha.weight: BF16
448
+ blk.16.ssm_beta.weight: BF16
449
+ blk.16.ssm_conv1d.weight: F32
450
+ blk.16.ssm_dt.bias: F32
451
+ blk.16.ssm_norm.weight: F32
452
+ blk.16.ssm_out.weight: IQ4_XS
453
+ blk.17.attn_gate.weight: IQ3_S
454
+ blk.17.attn_qkv.weight: IQ4_XS
455
+ blk.17.ffn_down_exps.weight: Q2_0
456
+ blk.17.ffn_down_shexp.weight: IQ4_NL
457
+ blk.17.ffn_gate_exps.weight: IQ2_S
458
+ blk.17.ffn_gate_inp.weight: BF16
459
+ blk.17.ffn_gate_inp_shexp.weight: BF16
460
+ blk.17.ffn_gate_shexp.weight: IQ4_XS
461
+ blk.17.ffn_up_exps.weight: IQ2_S
462
+ blk.17.ffn_up_shexp.weight: IQ4_XS
463
+ blk.17.hc_attn_down.weight: BF16
464
+ blk.17.hc_attn_inject.weight: BF16
465
+ blk.17.hc_attn_norm.weight: F32
466
+ blk.17.hc_attn_up.weight: BF16
467
+ blk.17.hc_ffn_down.weight: BF16
468
+ blk.17.hc_ffn_inject.weight: BF16
469
+ blk.17.hc_ffn_norm.weight: F32
470
+ blk.17.hc_ffn_up.weight: BF16
471
+ blk.17.ssm_a: F32
472
+ blk.17.ssm_alpha.weight: BF16
473
+ blk.17.ssm_beta.weight: BF16
474
+ blk.17.ssm_conv1d.weight: F32
475
+ blk.17.ssm_dt.bias: F32
476
+ blk.17.ssm_norm.weight: F32
477
+ blk.17.ssm_out.weight: IQ4_XS
478
+ blk.18.attn_gate.weight: IQ4_XS
479
+ blk.18.attn_qkv.weight: IQ4_XS
480
+ blk.18.ffn_down_exps.weight: Q2_0
481
+ blk.18.ffn_down_shexp.weight: IQ4_NL
482
+ blk.18.ffn_gate_exps.weight: IQ2_XXS
483
+ blk.18.ffn_gate_inp.weight: BF16
484
+ blk.18.ffn_gate_inp_shexp.weight: BF16
485
+ blk.18.ffn_gate_shexp.weight: IQ3_S
486
+ blk.18.ffn_up_exps.weight: IQ2_XXS
487
+ blk.18.ffn_up_shexp.weight: IQ3_S
488
+ blk.18.hc_attn_down.weight: BF16
489
+ blk.18.hc_attn_inject.weight: BF16
490
+ blk.18.hc_attn_norm.weight: F32
491
+ blk.18.hc_attn_up.weight: BF16
492
+ blk.18.hc_ffn_down.weight: BF16
493
+ blk.18.hc_ffn_inject.weight: BF16
494
+ blk.18.hc_ffn_norm.weight: F32
495
+ blk.18.hc_ffn_up.weight: BF16
496
+ blk.18.ssm_a: F32
497
+ blk.18.ssm_alpha.weight: BF16
498
+ blk.18.ssm_beta.weight: BF16
499
+ blk.18.ssm_conv1d.weight: F32
500
+ blk.18.ssm_dt.bias: F32
501
+ blk.18.ssm_norm.weight: F32
502
+ blk.18.ssm_out.weight: IQ4_XS
503
+ blk.19.attn_k.weight: IQ4_XS
504
+ blk.19.attn_k_norm.weight: F32
505
+ blk.19.attn_output.weight: IQ4_XS
506
+ blk.19.attn_q.weight: IQ4_XS
507
+ blk.19.attn_q_norm.weight: F32
508
+ blk.19.attn_v.weight: IQ4_XS
509
+ blk.19.ffn_down_exps.weight: Q2_0
510
+ blk.19.ffn_down_shexp.weight: IQ4_NL
511
+ blk.19.ffn_gate_exps.weight: IQ2_XXS
512
+ blk.19.ffn_gate_inp.weight: BF16
513
+ blk.19.ffn_gate_inp_shexp.weight: BF16
514
+ blk.19.ffn_gate_shexp.weight: IQ4_XS
515
+ blk.19.ffn_up_exps.weight: IQ2_XXS
516
+ blk.19.ffn_up_shexp.weight: IQ3_S
517
+ blk.19.hc_attn_down.weight: BF16
518
+ blk.19.hc_attn_inject.weight: BF16
519
+ blk.19.hc_attn_norm.weight: F32
520
+ blk.19.hc_attn_up.weight: BF16
521
+ blk.19.hc_ffn_down.weight: BF16
522
+ blk.19.hc_ffn_inject.weight: BF16
523
+ blk.19.hc_ffn_norm.weight: F32
524
+ blk.19.hc_ffn_up.weight: BF16
525
+ blk.19.indexer.k_norm.weight: F32
526
+ blk.19.indexer.k_proj.weight: BF16
527
+ blk.19.indexer.q_norm.weight: F32
528
+ blk.19.indexer.q_proj.weight: BF16
529
+ blk.20.attn_gate.weight: IQ4_XS
530
+ blk.20.attn_qkv.weight: IQ4_XS
531
+ blk.20.ffn_down_exps.weight: Q2_0
532
+ blk.20.ffn_down_shexp.weight: Q8_0
533
+ blk.20.ffn_gate_exps.weight: IQ2_S
534
+ blk.20.ffn_gate_inp.weight: BF16
535
+ blk.20.ffn_gate_inp_shexp.weight: BF16
536
+ blk.20.ffn_gate_shexp.weight: IQ4_XS
537
+ blk.20.ffn_up_exps.weight: IQ2_S
538
+ blk.20.ffn_up_shexp.weight: IQ4_XS
539
+ blk.20.hc_attn_down.weight: BF16
540
+ blk.20.hc_attn_inject.weight: BF16
541
+ blk.20.hc_attn_norm.weight: F32
542
+ blk.20.hc_attn_up.weight: BF16
543
+ blk.20.hc_ffn_down.weight: BF16
544
+ blk.20.hc_ffn_inject.weight: BF16
545
+ blk.20.hc_ffn_norm.weight: F32
546
+ blk.20.hc_ffn_up.weight: BF16
547
+ blk.20.ssm_a: F32
548
+ blk.20.ssm_alpha.weight: BF16
549
+ blk.20.ssm_beta.weight: BF16
550
+ blk.20.ssm_conv1d.weight: F32
551
+ blk.20.ssm_dt.bias: F32
552
+ blk.20.ssm_norm.weight: F32
553
+ blk.20.ssm_out.weight: IQ4_XS
554
+ blk.21.attn_gate.weight: IQ3_S
555
+ blk.21.attn_qkv.weight: IQ4_XS
556
+ blk.21.ffn_down_exps.weight: Q2_0
557
+ blk.21.ffn_down_shexp.weight: Q8_0
558
+ blk.21.ffn_gate_exps.weight: IQ2_S
559
+ blk.21.ffn_gate_inp.weight: BF16
560
+ blk.21.ffn_gate_inp_shexp.weight: BF16
561
+ blk.21.ffn_gate_shexp.weight: IQ4_XS
562
+ blk.21.ffn_up_exps.weight: IQ2_S
563
+ blk.21.ffn_up_shexp.weight: IQ3_S
564
+ blk.21.hc_attn_down.weight: BF16
565
+ blk.21.hc_attn_inject.weight: BF16
566
+ blk.21.hc_attn_norm.weight: F32
567
+ blk.21.hc_attn_up.weight: BF16
568
+ blk.21.hc_ffn_down.weight: BF16
569
+ blk.21.hc_ffn_inject.weight: BF16
570
+ blk.21.hc_ffn_norm.weight: F32
571
+ blk.21.hc_ffn_up.weight: BF16
572
+ blk.21.ssm_a: F32
573
+ blk.21.ssm_alpha.weight: BF16
574
+ blk.21.ssm_beta.weight: BF16
575
+ blk.21.ssm_conv1d.weight: F32
576
+ blk.21.ssm_dt.bias: F32
577
+ blk.21.ssm_norm.weight: F32
578
+ blk.21.ssm_out.weight: IQ3_S
579
+ blk.22.attn_gate.weight: IQ3_S
580
+ blk.22.attn_qkv.weight: IQ4_XS
581
+ blk.22.ffn_down_exps.weight: Q2_0
582
+ blk.22.ffn_down_shexp.weight: Q2_0
583
+ blk.22.ffn_gate_exps.weight: IQ2_S
584
+ blk.22.ffn_gate_inp.weight: BF16
585
+ blk.22.ffn_gate_inp_shexp.weight: BF16
586
+ blk.22.ffn_gate_shexp.weight: IQ4_XS
587
+ blk.22.ffn_up_exps.weight: IQ2_S
588
+ blk.22.ffn_up_shexp.weight: IQ4_XS
589
+ blk.22.hc_attn_down.weight: BF16
590
+ blk.22.hc_attn_inject.weight: BF16
591
+ blk.22.hc_attn_norm.weight: F32
592
+ blk.22.hc_attn_up.weight: BF16
593
+ blk.22.hc_ffn_down.weight: BF16
594
+ blk.22.hc_ffn_inject.weight: BF16
595
+ blk.22.hc_ffn_norm.weight: F32
596
+ blk.22.hc_ffn_up.weight: BF16
597
+ blk.22.ssm_a: F32
598
+ blk.22.ssm_alpha.weight: BF16
599
+ blk.22.ssm_beta.weight: BF16
600
+ blk.22.ssm_conv1d.weight: F32
601
+ blk.22.ssm_dt.bias: F32
602
+ blk.22.ssm_norm.weight: F32
603
+ blk.22.ssm_out.weight: IQ4_XS
604
+ blk.23.attn_k.weight: IQ4_XS
605
+ blk.23.attn_k_norm.weight: F32
606
+ blk.23.attn_output.weight: IQ4_XS
607
+ blk.23.attn_q.weight: IQ4_XS
608
+ blk.23.attn_q_norm.weight: F32
609
+ blk.23.attn_v.weight: Q6_K
610
+ blk.23.ffn_down_exps.weight: Q2_0
611
+ blk.23.ffn_down_shexp.weight: IQ4_NL
612
+ blk.23.ffn_gate_exps.weight: IQ2_S
613
+ blk.23.ffn_gate_inp.weight: BF16
614
+ blk.23.ffn_gate_inp_shexp.weight: BF16
615
+ blk.23.ffn_gate_shexp.weight: IQ3_S
616
+ blk.23.ffn_up_exps.weight: IQ2_S
617
+ blk.23.ffn_up_shexp.weight: IQ4_XS
618
+ blk.23.hc_attn_down.weight: BF16
619
+ blk.23.hc_attn_inject.weight: BF16
620
+ blk.23.hc_attn_norm.weight: F32
621
+ blk.23.hc_attn_up.weight: BF16
622
+ blk.23.hc_ffn_down.weight: BF16
623
+ blk.23.hc_ffn_inject.weight: BF16
624
+ blk.23.hc_ffn_norm.weight: F32
625
+ blk.23.hc_ffn_up.weight: BF16
626
+ blk.23.indexer.k_norm.weight: F32
627
+ blk.23.indexer.k_proj.weight: BF16
628
+ blk.23.indexer.q_norm.weight: F32
629
+ blk.23.indexer.q_proj.weight: BF16
630
+ blk.24.attn_gate.weight: Q6_K
631
+ blk.24.attn_qkv.weight: IQ4_XS
632
+ blk.24.ffn_down_exps.weight: Q2_0
633
+ blk.24.ffn_down_shexp.weight: IQ4_NL
634
+ blk.24.ffn_gate_exps.weight: IQ2_S
635
+ blk.24.ffn_gate_inp.weight: BF16
636
+ blk.24.ffn_gate_inp_shexp.weight: BF16
637
+ blk.24.ffn_gate_shexp.weight: IQ4_XS
638
+ blk.24.ffn_up_exps.weight: IQ2_S
639
+ blk.24.ffn_up_shexp.weight: IQ4_XS
640
+ blk.24.hc_attn_down.weight: BF16
641
+ blk.24.hc_attn_inject.weight: BF16
642
+ blk.24.hc_attn_norm.weight: F32
643
+ blk.24.hc_attn_up.weight: BF16
644
+ blk.24.hc_ffn_down.weight: BF16
645
+ blk.24.hc_ffn_inject.weight: BF16
646
+ blk.24.hc_ffn_norm.weight: F32
647
+ blk.24.hc_ffn_up.weight: BF16
648
+ blk.24.ssm_a: F32
649
+ blk.24.ssm_alpha.weight: BF16
650
+ blk.24.ssm_beta.weight: BF16
651
+ blk.24.ssm_conv1d.weight: F32
652
+ blk.24.ssm_dt.bias: F32
653
+ blk.24.ssm_norm.weight: F32
654
+ blk.24.ssm_out.weight: IQ3_S
655
+ blk.25.attn_gate.weight: Q6_K
656
+ blk.25.attn_qkv.weight: IQ4_XS
657
+ blk.25.ffn_down_exps.weight: Q2_0
658
+ blk.25.ffn_down_shexp.weight: IQ4_NL
659
+ blk.25.ffn_gate_exps.weight: IQ2_S
660
+ blk.25.ffn_gate_inp.weight: BF16
661
+ blk.25.ffn_gate_inp_shexp.weight: BF16
662
+ blk.25.ffn_gate_shexp.weight: IQ3_S
663
+ blk.25.ffn_up_exps.weight: IQ2_S
664
+ blk.25.ffn_up_shexp.weight: IQ4_XS
665
+ blk.25.hc_attn_down.weight: BF16
666
+ blk.25.hc_attn_inject.weight: BF16
667
+ blk.25.hc_attn_norm.weight: F32
668
+ blk.25.hc_attn_up.weight: BF16
669
+ blk.25.hc_ffn_down.weight: BF16
670
+ blk.25.hc_ffn_inject.weight: BF16
671
+ blk.25.hc_ffn_norm.weight: F32
672
+ blk.25.hc_ffn_up.weight: BF16
673
+ blk.25.ssm_a: F32
674
+ blk.25.ssm_alpha.weight: BF16
675
+ blk.25.ssm_beta.weight: BF16
676
+ blk.25.ssm_conv1d.weight: F32
677
+ blk.25.ssm_dt.bias: F32
678
+ blk.25.ssm_norm.weight: F32
679
+ blk.25.ssm_out.weight: IQ4_XS
680
+ blk.26.attn_gate.weight: IQ3_S
681
+ blk.26.attn_qkv.weight: IQ4_XS
682
+ blk.26.ffn_down_exps.weight: Q2_0
683
+ blk.26.ffn_down_shexp.weight: IQ4_NL
684
+ blk.26.ffn_gate_exps.weight: IQ2_S
685
+ blk.26.ffn_gate_inp.weight: BF16
686
+ blk.26.ffn_gate_inp_shexp.weight: BF16
687
+ blk.26.ffn_gate_shexp.weight: IQ4_XS
688
+ blk.26.ffn_up_exps.weight: IQ2_S
689
+ blk.26.ffn_up_shexp.weight: IQ3_S
690
+ blk.26.hc_attn_down.weight: BF16
691
+ blk.26.hc_attn_inject.weight: BF16
692
+ blk.26.hc_attn_norm.weight: F32
693
+ blk.26.hc_attn_up.weight: BF16
694
+ blk.26.hc_ffn_down.weight: BF16
695
+ blk.26.hc_ffn_inject.weight: BF16
696
+ blk.26.hc_ffn_norm.weight: F32
697
+ blk.26.hc_ffn_up.weight: BF16
698
+ blk.26.ssm_a: F32
699
+ blk.26.ssm_alpha.weight: BF16
700
+ blk.26.ssm_beta.weight: BF16
701
+ blk.26.ssm_conv1d.weight: F32
702
+ blk.26.ssm_dt.bias: F32
703
+ blk.26.ssm_norm.weight: F32
704
+ blk.26.ssm_out.weight: IQ4_XS
705
+ blk.27.attn_k.weight: IQ4_XS
706
+ blk.27.attn_k_norm.weight: F32
707
+ blk.27.attn_output.weight: IQ4_XS
708
+ blk.27.attn_q.weight: IQ4_XS
709
+ blk.27.attn_q_norm.weight: F32
710
+ blk.27.attn_v.weight: IQ4_XS
711
+ blk.27.ffn_down_exps.weight: Q2_0
712
+ blk.27.ffn_down_shexp.weight: IQ4_NL
713
+ blk.27.ffn_gate_exps.weight: IQ2_XXS
714
+ blk.27.ffn_gate_inp.weight: BF16
715
+ blk.27.ffn_gate_inp_shexp.weight: BF16
716
+ blk.27.ffn_gate_shexp.weight: Q6_K
717
+ blk.27.ffn_up_exps.weight: IQ2_XXS
718
+ blk.27.ffn_up_shexp.weight: IQ3_S
719
+ blk.27.hc_attn_down.weight: BF16
720
+ blk.27.hc_attn_inject.weight: BF16
721
+ blk.27.hc_attn_norm.weight: F32
722
+ blk.27.hc_attn_up.weight: BF16
723
+ blk.27.hc_ffn_down.weight: BF16
724
+ blk.27.hc_ffn_inject.weight: BF16
725
+ blk.27.hc_ffn_norm.weight: F32
726
+ blk.27.hc_ffn_up.weight: BF16
727
+ blk.27.indexer.k_norm.weight: F32
728
+ blk.27.indexer.k_proj.weight: BF16
729
+ blk.27.indexer.q_norm.weight: F32
730
+ blk.27.indexer.q_proj.weight: BF16
731
+ blk.28.attn_gate.weight: IQ4_XS
732
+ blk.28.attn_qkv.weight: IQ4_XS
733
+ blk.28.ffn_down_exps.weight: Q2_0
734
+ blk.28.ffn_down_shexp.weight: IQ4_NL
735
+ blk.28.ffn_gate_exps.weight: IQ2_S
736
+ blk.28.ffn_gate_inp.weight: BF16
737
+ blk.28.ffn_gate_inp_shexp.weight: BF16
738
+ blk.28.ffn_gate_shexp.weight: IQ3_S
739
+ blk.28.ffn_up_exps.weight: IQ2_S
740
+ blk.28.ffn_up_shexp.weight: Q6_K
741
+ blk.28.hc_attn_down.weight: BF16
742
+ blk.28.hc_attn_inject.weight: BF16
743
+ blk.28.hc_attn_norm.weight: F32
744
+ blk.28.hc_attn_up.weight: BF16
745
+ blk.28.hc_ffn_down.weight: BF16
746
+ blk.28.hc_ffn_inject.weight: BF16
747
+ blk.28.hc_ffn_norm.weight: F32
748
+ blk.28.hc_ffn_up.weight: BF16
749
+ blk.28.ssm_a: F32
750
+ blk.28.ssm_alpha.weight: BF16
751
+ blk.28.ssm_beta.weight: BF16
752
+ blk.28.ssm_conv1d.weight: F32
753
+ blk.28.ssm_dt.bias: F32
754
+ blk.28.ssm_norm.weight: F32
755
+ blk.28.ssm_out.weight: IQ4_XS
756
+ blk.29.attn_gate.weight: IQ4_XS
757
+ blk.29.attn_qkv.weight: IQ4_XS
758
+ blk.29.ffn_down_exps.weight: Q2_0
759
+ blk.29.ffn_down_shexp.weight: IQ4_NL
760
+ blk.29.ffn_gate_exps.weight: IQ2_XXS
761
+ blk.29.ffn_gate_inp.weight: BF16
762
+ blk.29.ffn_gate_inp_shexp.weight: BF16
763
+ blk.29.ffn_gate_shexp.weight: IQ4_XS
764
+ blk.29.ffn_up_exps.weight: IQ2_XXS
765
+ blk.29.ffn_up_shexp.weight: IQ4_XS
766
+ blk.29.hc_attn_down.weight: BF16
767
+ blk.29.hc_attn_inject.weight: BF16
768
+ blk.29.hc_attn_norm.weight: F32
769
+ blk.29.hc_attn_up.weight: BF16
770
+ blk.29.hc_ffn_down.weight: BF16
771
+ blk.29.hc_ffn_inject.weight: BF16
772
+ blk.29.hc_ffn_norm.weight: F32
773
+ blk.29.hc_ffn_up.weight: BF16
774
+ blk.29.ssm_a: F32
775
+ blk.29.ssm_alpha.weight: BF16
776
+ blk.29.ssm_beta.weight: BF16
777
+ blk.29.ssm_conv1d.weight: F32
778
+ blk.29.ssm_dt.bias: F32
779
+ blk.29.ssm_norm.weight: F32
780
+ blk.29.ssm_out.weight: IQ3_S
781
+ blk.30.attn_gate.weight: IQ4_XS
782
+ blk.30.attn_qkv.weight: IQ4_XS
783
+ blk.30.ffn_down_exps.weight: Q2_0
784
+ blk.30.ffn_down_shexp.weight: Q2_0
785
+ blk.30.ffn_gate_exps.weight: IQ2_XXS
786
+ blk.30.ffn_gate_inp.weight: BF16
787
+ blk.30.ffn_gate_inp_shexp.weight: BF16
788
+ blk.30.ffn_gate_shexp.weight: IQ4_XS
789
+ blk.30.ffn_up_exps.weight: IQ2_XXS
790
+ blk.30.ffn_up_shexp.weight: IQ4_XS
791
+ blk.30.hc_attn_down.weight: BF16
792
+ blk.30.hc_attn_inject.weight: BF16
793
+ blk.30.hc_attn_norm.weight: F32
794
+ blk.30.hc_attn_up.weight: BF16
795
+ blk.30.hc_ffn_down.weight: BF16
796
+ blk.30.hc_ffn_inject.weight: BF16
797
+ blk.30.hc_ffn_norm.weight: F32
798
+ blk.30.hc_ffn_up.weight: BF16
799
+ blk.30.ssm_a: F32
800
+ blk.30.ssm_alpha.weight: BF16
801
+ blk.30.ssm_beta.weight: BF16
802
+ blk.30.ssm_conv1d.weight: F32
803
+ blk.30.ssm_dt.bias: F32
804
+ blk.30.ssm_norm.weight: F32
805
+ blk.30.ssm_out.weight: IQ4_XS
806
+ blk.31.attn_k.weight: IQ4_XS
807
+ blk.31.attn_k_norm.weight: F32
808
+ blk.31.attn_output.weight: IQ4_XS
809
+ blk.31.attn_q.weight: IQ4_XS
810
+ blk.31.attn_q_norm.weight: F32
811
+ blk.31.attn_v.weight: Q6_K
812
+ blk.31.ffn_down_exps.weight: Q2_0
813
+ blk.31.ffn_down_shexp.weight: IQ4_NL
814
+ blk.31.ffn_gate_exps.weight: IQ2_S
815
+ blk.31.ffn_gate_inp.weight: BF16
816
+ blk.31.ffn_gate_inp_shexp.weight: BF16
817
+ blk.31.ffn_gate_shexp.weight: IQ4_XS
818
+ blk.31.ffn_up_exps.weight: IQ2_S
819
+ blk.31.ffn_up_shexp.weight: IQ3_S
820
+ blk.31.hc_attn_down.weight: BF16
821
+ blk.31.hc_attn_inject.weight: BF16
822
+ blk.31.hc_attn_norm.weight: F32
823
+ blk.31.hc_attn_up.weight: BF16
824
+ blk.31.hc_ffn_down.weight: BF16
825
+ blk.31.hc_ffn_inject.weight: BF16
826
+ blk.31.hc_ffn_norm.weight: F32
827
+ blk.31.hc_ffn_up.weight: BF16
828
+ blk.31.indexer.k_norm.weight: F32
829
+ blk.31.indexer.k_proj.weight: BF16
830
+ blk.31.indexer.q_norm.weight: F32
831
+ blk.31.indexer.q_proj.weight: BF16
832
+ blk.32.attn_gate.weight: IQ4_XS
833
+ blk.32.attn_qkv.weight: IQ4_XS
834
+ blk.32.ffn_down_exps.weight: Q2_0
835
+ blk.32.ffn_down_shexp.weight: IQ4_NL
836
+ blk.32.ffn_gate_exps.weight: IQ2_S
837
+ blk.32.ffn_gate_inp.weight: BF16
838
+ blk.32.ffn_gate_inp_shexp.weight: BF16
839
+ blk.32.ffn_gate_shexp.weight: IQ4_XS
840
+ blk.32.ffn_up_exps.weight: IQ2_S
841
+ blk.32.ffn_up_shexp.weight: Q6_K
842
+ blk.32.hc_attn_down.weight: BF16
843
+ blk.32.hc_attn_inject.weight: BF16
844
+ blk.32.hc_attn_norm.weight: F32
845
+ blk.32.hc_attn_up.weight: BF16
846
+ blk.32.hc_ffn_down.weight: BF16
847
+ blk.32.hc_ffn_inject.weight: BF16
848
+ blk.32.hc_ffn_norm.weight: F32
849
+ blk.32.hc_ffn_up.weight: BF16
850
+ blk.32.ssm_a: F32
851
+ blk.32.ssm_alpha.weight: BF16
852
+ blk.32.ssm_beta.weight: BF16
853
+ blk.32.ssm_conv1d.weight: F32
854
+ blk.32.ssm_dt.bias: F32
855
+ blk.32.ssm_norm.weight: F32
856
+ blk.32.ssm_out.weight: IQ4_XS
857
+ blk.33.attn_gate.weight: IQ3_S
858
+ blk.33.attn_qkv.weight: IQ4_XS
859
+ blk.33.ffn_down_exps.weight: Q2_0
860
+ blk.33.ffn_down_shexp.weight: IQ4_NL
861
+ blk.33.ffn_gate_exps.weight: IQ2_S
862
+ blk.33.ffn_gate_inp.weight: BF16
863
+ blk.33.ffn_gate_inp_shexp.weight: BF16
864
+ blk.33.ffn_gate_shexp.weight: IQ4_XS
865
+ blk.33.ffn_up_exps.weight: IQ2_S
866
+ blk.33.ffn_up_shexp.weight: IQ4_XS
867
+ blk.33.hc_attn_down.weight: BF16
868
+ blk.33.hc_attn_inject.weight: BF16
869
+ blk.33.hc_attn_norm.weight: F32
870
+ blk.33.hc_attn_up.weight: BF16
871
+ blk.33.hc_ffn_down.weight: BF16
872
+ blk.33.hc_ffn_inject.weight: BF16
873
+ blk.33.hc_ffn_norm.weight: F32
874
+ blk.33.hc_ffn_up.weight: BF16
875
+ blk.33.ssm_a: F32
876
+ blk.33.ssm_alpha.weight: BF16
877
+ blk.33.ssm_beta.weight: BF16
878
+ blk.33.ssm_conv1d.weight: F32
879
+ blk.33.ssm_dt.bias: F32
880
+ blk.33.ssm_norm.weight: F32
881
+ blk.33.ssm_out.weight: Q6_K
882
+ blk.34.attn_gate.weight: IQ4_XS
883
+ blk.34.attn_qkv.weight: IQ4_XS
884
+ blk.34.ffn_down_exps.weight: Q2_0
885
+ blk.34.ffn_down_shexp.weight: IQ4_NL
886
+ blk.34.ffn_gate_exps.weight: IQ2_S
887
+ blk.34.ffn_gate_inp.weight: BF16
888
+ blk.34.ffn_gate_inp_shexp.weight: BF16
889
+ blk.34.ffn_gate_shexp.weight: IQ4_XS
890
+ blk.34.ffn_up_exps.weight: IQ2_S
891
+ blk.34.ffn_up_shexp.weight: IQ4_XS
892
+ blk.34.hc_attn_down.weight: BF16
893
+ blk.34.hc_attn_inject.weight: BF16
894
+ blk.34.hc_attn_norm.weight: F32
895
+ blk.34.hc_attn_up.weight: BF16
896
+ blk.34.hc_ffn_down.weight: BF16
897
+ blk.34.hc_ffn_inject.weight: BF16
898
+ blk.34.hc_ffn_norm.weight: F32
899
+ blk.34.hc_ffn_up.weight: BF16
900
+ blk.34.ssm_a: F32
901
+ blk.34.ssm_alpha.weight: BF16
902
+ blk.34.ssm_beta.weight: BF16
903
+ blk.34.ssm_conv1d.weight: F32
904
+ blk.34.ssm_dt.bias: F32
905
+ blk.34.ssm_norm.weight: F32
906
+ blk.34.ssm_out.weight: IQ4_XS
907
+ blk.35.attn_k.weight: IQ4_XS
908
+ blk.35.attn_k_norm.weight: F32
909
+ blk.35.attn_output.weight: IQ4_XS
910
+ blk.35.attn_q.weight: IQ3_S
911
+ blk.35.attn_q_norm.weight: F32
912
+ blk.35.attn_v.weight: IQ4_XS
913
+ blk.35.ffn_down_exps.weight: Q2_0
914
+ blk.35.ffn_down_shexp.weight: IQ4_NL
915
+ blk.35.ffn_gate_exps.weight: IQ2_S
916
+ blk.35.ffn_gate_inp.weight: BF16
917
+ blk.35.ffn_gate_inp_shexp.weight: BF16
918
+ blk.35.ffn_gate_shexp.weight: IQ4_XS
919
+ blk.35.ffn_up_exps.weight: IQ2_S
920
+ blk.35.ffn_up_shexp.weight: IQ4_XS
921
+ blk.35.hc_attn_down.weight: BF16
922
+ blk.35.hc_attn_inject.weight: BF16
923
+ blk.35.hc_attn_norm.weight: F32
924
+ blk.35.hc_attn_up.weight: BF16
925
+ blk.35.hc_ffn_down.weight: BF16
926
+ blk.35.hc_ffn_inject.weight: BF16
927
+ blk.35.hc_ffn_norm.weight: F32
928
+ blk.35.hc_ffn_up.weight: BF16
929
+ blk.35.indexer.k_norm.weight: F32
930
+ blk.35.indexer.k_proj.weight: BF16
931
+ blk.35.indexer.q_norm.weight: F32
932
+ blk.35.indexer.q_proj.weight: BF16
933
+ blk.36.attn_gate.weight: IQ4_XS
934
+ blk.36.attn_qkv.weight: IQ4_XS
935
+ blk.36.ffn_down_exps.weight: Q2_0
936
+ blk.36.ffn_down_shexp.weight: IQ4_NL
937
+ blk.36.ffn_gate_exps.weight: IQ2_S
938
+ blk.36.ffn_gate_inp.weight: BF16
939
+ blk.36.ffn_gate_inp_shexp.weight: BF16
940
+ blk.36.ffn_gate_shexp.weight: Q6_K
941
+ blk.36.ffn_up_exps.weight: IQ2_S
942
+ blk.36.ffn_up_shexp.weight: Q6_K
943
+ blk.36.hc_attn_down.weight: BF16
944
+ blk.36.hc_attn_inject.weight: BF16
945
+ blk.36.hc_attn_norm.weight: F32
946
+ blk.36.hc_attn_up.weight: BF16
947
+ blk.36.hc_ffn_down.weight: BF16
948
+ blk.36.hc_ffn_inject.weight: BF16
949
+ blk.36.hc_ffn_norm.weight: F32
950
+ blk.36.hc_ffn_up.weight: BF16
951
+ blk.36.ssm_a: F32
952
+ blk.36.ssm_alpha.weight: BF16
953
+ blk.36.ssm_beta.weight: BF16
954
+ blk.36.ssm_conv1d.weight: F32
955
+ blk.36.ssm_dt.bias: F32
956
+ blk.36.ssm_norm.weight: F32
957
+ blk.36.ssm_out.weight: IQ4_XS
958
+ blk.37.attn_gate.weight: IQ3_S
959
+ blk.37.attn_qkv.weight: IQ4_XS
960
+ blk.37.ffn_down_exps.weight: Q2_0
961
+ blk.37.ffn_down_shexp.weight: IQ4_NL
962
+ blk.37.ffn_gate_exps.weight: IQ1_M
963
+ blk.37.ffn_gate_inp.weight: BF16
964
+ blk.37.ffn_gate_inp_shexp.weight: BF16
965
+ blk.37.ffn_gate_shexp.weight: IQ4_XS
966
+ blk.37.ffn_up_exps.weight: IQ1_M
967
+ blk.37.ffn_up_shexp.weight: IQ3_S
968
+ blk.37.hc_attn_down.weight: BF16
969
+ blk.37.hc_attn_inject.weight: BF16
970
+ blk.37.hc_attn_norm.weight: F32
971
+ blk.37.hc_attn_up.weight: BF16
972
+ blk.37.hc_ffn_down.weight: BF16
973
+ blk.37.hc_ffn_inject.weight: BF16
974
+ blk.37.hc_ffn_norm.weight: F32
975
+ blk.37.hc_ffn_up.weight: BF16
976
+ blk.37.ssm_a: F32
977
+ blk.37.ssm_alpha.weight: BF16
978
+ blk.37.ssm_beta.weight: BF16
979
+ blk.37.ssm_conv1d.weight: F32
980
+ blk.37.ssm_dt.bias: F32
981
+ blk.37.ssm_norm.weight: F32
982
+ blk.37.ssm_out.weight: IQ4_XS
983
+ blk.38.attn_gate.weight: IQ4_XS
984
+ blk.38.attn_qkv.weight: IQ4_XS
985
+ blk.38.ffn_down_exps.weight: Q2_0
986
+ blk.38.ffn_down_shexp.weight: IQ4_NL
987
+ blk.38.ffn_gate_exps.weight: IQ2_S
988
+ blk.38.ffn_gate_inp.weight: BF16
989
+ blk.38.ffn_gate_inp_shexp.weight: BF16
990
+ blk.38.ffn_gate_shexp.weight: IQ4_XS
991
+ blk.38.ffn_up_exps.weight: IQ2_S
992
+ blk.38.ffn_up_shexp.weight: IQ4_XS
993
+ blk.38.hc_attn_down.weight: BF16
994
+ blk.38.hc_attn_inject.weight: BF16
995
+ blk.38.hc_attn_norm.weight: F32
996
+ blk.38.hc_attn_up.weight: BF16
997
+ blk.38.hc_ffn_down.weight: BF16
998
+ blk.38.hc_ffn_inject.weight: BF16
999
+ blk.38.hc_ffn_norm.weight: F32
1000
+ blk.38.hc_ffn_up.weight: BF16
1001
+ blk.38.ssm_a: F32
1002
+ blk.38.ssm_alpha.weight: BF16
1003
+ blk.38.ssm_beta.weight: BF16
1004
+ blk.38.ssm_conv1d.weight: F32
1005
+ blk.38.ssm_dt.bias: F32
1006
+ blk.38.ssm_norm.weight: F32
1007
+ blk.38.ssm_out.weight: IQ4_XS
1008
+ blk.39.attn_k.weight: IQ4_XS
1009
+ blk.39.attn_k_norm.weight: F32
1010
+ blk.39.attn_output.weight: IQ4_XS
1011
+ blk.39.attn_q.weight: IQ4_XS
1012
+ blk.39.attn_q_norm.weight: F32
1013
+ blk.39.attn_v.weight: Q6_K
1014
+ blk.39.ffn_down_exps.weight: Q2_0
1015
+ blk.39.ffn_down_shexp.weight: IQ4_NL
1016
+ blk.39.ffn_gate_exps.weight: IQ2_S
1017
+ blk.39.ffn_gate_inp.weight: BF16
1018
+ blk.39.ffn_gate_inp_shexp.weight: BF16
1019
+ blk.39.ffn_gate_shexp.weight: IQ4_XS
1020
+ blk.39.ffn_up_exps.weight: IQ2_S
1021
+ blk.39.ffn_up_shexp.weight: IQ4_XS
1022
+ blk.39.hc_attn_down.weight: BF16
1023
+ blk.39.hc_attn_inject.weight: BF16
1024
+ blk.39.hc_attn_norm.weight: F32
1025
+ blk.39.hc_attn_up.weight: BF16
1026
+ blk.39.hc_ffn_down.weight: BF16
1027
+ blk.39.hc_ffn_inject.weight: BF16
1028
+ blk.39.hc_ffn_norm.weight: F32
1029
+ blk.39.hc_ffn_up.weight: BF16
1030
+ blk.39.indexer.k_norm.weight: F32
1031
+ blk.39.indexer.k_proj.weight: BF16
1032
+ blk.39.indexer.q_norm.weight: F32
1033
+ blk.39.indexer.q_proj.weight: BF16
1034
+ blk.40.attn_gate.weight: IQ4_XS
1035
+ blk.40.attn_qkv.weight: IQ3_S
1036
+ blk.40.ffn_down_exps.weight: Q2_0
1037
+ blk.40.ffn_down_shexp.weight: IQ4_NL
1038
+ blk.40.ffn_gate_exps.weight: IQ2_S
1039
+ blk.40.ffn_gate_inp.weight: BF16
1040
+ blk.40.ffn_gate_inp_shexp.weight: BF16
1041
+ blk.40.ffn_gate_shexp.weight: IQ4_XS
1042
+ blk.40.ffn_up_exps.weight: IQ2_S
1043
+ blk.40.ffn_up_shexp.weight: IQ3_S
1044
+ blk.40.hc_attn_down.weight: BF16
1045
+ blk.40.hc_attn_inject.weight: BF16
1046
+ blk.40.hc_attn_norm.weight: F32
1047
+ blk.40.hc_attn_up.weight: BF16
1048
+ blk.40.hc_ffn_down.weight: BF16
1049
+ blk.40.hc_ffn_inject.weight: BF16
1050
+ blk.40.hc_ffn_norm.weight: F32
1051
+ blk.40.hc_ffn_up.weight: BF16
1052
+ blk.40.ssm_a: F32
1053
+ blk.40.ssm_alpha.weight: BF16
1054
+ blk.40.ssm_beta.weight: BF16
1055
+ blk.40.ssm_conv1d.weight: F32
1056
+ blk.40.ssm_dt.bias: F32
1057
+ blk.40.ssm_norm.weight: F32
1058
+ blk.40.ssm_out.weight: IQ3_S
1059
+ blk.41.attn_gate.weight: IQ4_XS
1060
+ blk.41.attn_qkv.weight: IQ4_XS
1061
+ blk.41.ffn_down_exps.weight: Q2_0
1062
+ blk.41.ffn_down_shexp.weight: IQ4_NL
1063
+ blk.41.ffn_gate_exps.weight: IQ2_S
1064
+ blk.41.ffn_gate_inp.weight: BF16
1065
+ blk.41.ffn_gate_inp_shexp.weight: BF16
1066
+ blk.41.ffn_gate_shexp.weight: IQ4_XS
1067
+ blk.41.ffn_up_exps.weight: IQ2_S
1068
+ blk.41.ffn_up_shexp.weight: Q6_K
1069
+ blk.41.hc_attn_down.weight: BF16
1070
+ blk.41.hc_attn_inject.weight: BF16
1071
+ blk.41.hc_attn_norm.weight: F32
1072
+ blk.41.hc_attn_up.weight: BF16
1073
+ blk.41.hc_ffn_down.weight: BF16
1074
+ blk.41.hc_ffn_inject.weight: BF16
1075
+ blk.41.hc_ffn_norm.weight: F32
1076
+ blk.41.hc_ffn_up.weight: BF16
1077
+ blk.41.ssm_a: F32
1078
+ blk.41.ssm_alpha.weight: BF16
1079
+ blk.41.ssm_beta.weight: BF16
1080
+ blk.41.ssm_conv1d.weight: F32
1081
+ blk.41.ssm_dt.bias: F32
1082
+ blk.41.ssm_norm.weight: F32
1083
+ blk.41.ssm_out.weight: IQ3_S
1084
+ blk.42.attn_gate.weight: IQ4_XS
1085
+ blk.42.attn_qkv.weight: IQ4_XS
1086
+ blk.42.ffn_down_exps.weight: Q2_0
1087
+ blk.42.ffn_down_shexp.weight: IQ4_NL
1088
+ blk.42.ffn_gate_exps.weight: IQ2_S
1089
+ blk.42.ffn_gate_inp.weight: BF16
1090
+ blk.42.ffn_gate_inp_shexp.weight: BF16
1091
+ blk.42.ffn_gate_shexp.weight: Q6_K
1092
+ blk.42.ffn_up_exps.weight: IQ2_S
1093
+ blk.42.ffn_up_shexp.weight: IQ4_XS
1094
+ blk.42.hc_attn_down.weight: BF16
1095
+ blk.42.hc_attn_inject.weight: BF16
1096
+ blk.42.hc_attn_norm.weight: F32
1097
+ blk.42.hc_attn_up.weight: BF16
1098
+ blk.42.hc_ffn_down.weight: BF16
1099
+ blk.42.hc_ffn_inject.weight: BF16
1100
+ blk.42.hc_ffn_norm.weight: F32
1101
+ blk.42.hc_ffn_up.weight: BF16
1102
+ blk.42.ssm_a: F32
1103
+ blk.42.ssm_alpha.weight: BF16
1104
+ blk.42.ssm_beta.weight: BF16
1105
+ blk.42.ssm_conv1d.weight: F32
1106
+ blk.42.ssm_dt.bias: F32
1107
+ blk.42.ssm_norm.weight: F32
1108
+ blk.42.ssm_out.weight: IQ4_XS
1109
+ blk.43.attn_k.weight: IQ4_XS
1110
+ blk.43.attn_k_norm.weight: F32
1111
+ blk.43.attn_output.weight: IQ4_XS
1112
+ blk.43.attn_q.weight: IQ4_XS
1113
+ blk.43.attn_q_norm.weight: F32
1114
+ blk.43.attn_v.weight: IQ4_XS
1115
+ blk.43.ffn_down_exps.weight: Q2_0
1116
+ blk.43.ffn_down_shexp.weight: IQ4_NL
1117
+ blk.43.ffn_gate_exps.weight: IQ2_S
1118
+ blk.43.ffn_gate_inp.weight: BF16
1119
+ blk.43.ffn_gate_inp_shexp.weight: BF16
1120
+ blk.43.ffn_gate_shexp.weight: IQ3_S
1121
+ blk.43.ffn_up_exps.weight: IQ2_S
1122
+ blk.43.ffn_up_shexp.weight: IQ4_XS
1123
+ blk.43.hc_attn_down.weight: BF16
1124
+ blk.43.hc_attn_inject.weight: BF16
1125
+ blk.43.hc_attn_norm.weight: F32
1126
+ blk.43.hc_attn_up.weight: BF16
1127
+ blk.43.hc_ffn_down.weight: BF16
1128
+ blk.43.hc_ffn_inject.weight: BF16
1129
+ blk.43.hc_ffn_norm.weight: F32
1130
+ blk.43.hc_ffn_up.weight: BF16
1131
+ blk.43.indexer.k_norm.weight: F32
1132
+ blk.43.indexer.k_proj.weight: BF16
1133
+ blk.43.indexer.q_norm.weight: F32
1134
+ blk.43.indexer.q_proj.weight: BF16
1135
+ blk.44.attn_gate.weight: IQ4_XS
1136
+ blk.44.attn_qkv.weight: IQ4_XS
1137
+ blk.44.ffn_down_exps.weight: Q2_0
1138
+ blk.44.ffn_down_shexp.weight: IQ4_NL
1139
+ blk.44.ffn_gate_exps.weight: IQ2_S
1140
+ blk.44.ffn_gate_inp.weight: BF16
1141
+ blk.44.ffn_gate_inp_shexp.weight: BF16
1142
+ blk.44.ffn_gate_shexp.weight: Q6_K
1143
+ blk.44.ffn_up_exps.weight: IQ2_S
1144
+ blk.44.ffn_up_shexp.weight: IQ4_XS
1145
+ blk.44.hc_attn_down.weight: BF16
1146
+ blk.44.hc_attn_inject.weight: BF16
1147
+ blk.44.hc_attn_norm.weight: F32
1148
+ blk.44.hc_attn_up.weight: BF16
1149
+ blk.44.hc_ffn_down.weight: BF16
1150
+ blk.44.hc_ffn_inject.weight: BF16
1151
+ blk.44.hc_ffn_norm.weight: F32
1152
+ blk.44.hc_ffn_up.weight: BF16
1153
+ blk.44.ssm_a: F32
1154
+ blk.44.ssm_alpha.weight: BF16
1155
+ blk.44.ssm_beta.weight: BF16
1156
+ blk.44.ssm_conv1d.weight: F32
1157
+ blk.44.ssm_dt.bias: F32
1158
+ blk.44.ssm_norm.weight: F32
1159
+ blk.44.ssm_out.weight: IQ4_XS
1160
+ blk.45.attn_gate.weight: Q6_K
1161
+ blk.45.attn_qkv.weight: IQ4_XS
1162
+ blk.45.ffn_down_exps.weight: Q2_0
1163
+ blk.45.ffn_down_shexp.weight: IQ4_NL
1164
+ blk.45.ffn_gate_exps.weight: IQ2_S
1165
+ blk.45.ffn_gate_inp.weight: BF16
1166
+ blk.45.ffn_gate_inp_shexp.weight: BF16
1167
+ blk.45.ffn_gate_shexp.weight: IQ4_XS
1168
+ blk.45.ffn_up_exps.weight: IQ2_S
1169
+ blk.45.ffn_up_shexp.weight: IQ3_S
1170
+ blk.45.hc_attn_down.weight: BF16
1171
+ blk.45.hc_attn_inject.weight: BF16
1172
+ blk.45.hc_attn_norm.weight: F32
1173
+ blk.45.hc_attn_up.weight: BF16
1174
+ blk.45.hc_ffn_down.weight: BF16
1175
+ blk.45.hc_ffn_inject.weight: BF16
1176
+ blk.45.hc_ffn_norm.weight: F32
1177
+ blk.45.hc_ffn_up.weight: BF16
1178
+ blk.45.ssm_a: F32
1179
+ blk.45.ssm_alpha.weight: BF16
1180
+ blk.45.ssm_beta.weight: BF16
1181
+ blk.45.ssm_conv1d.weight: F32
1182
+ blk.45.ssm_dt.bias: F32
1183
+ blk.45.ssm_norm.weight: F32
1184
+ blk.45.ssm_out.weight: IQ4_XS
1185
+ blk.46.attn_gate.weight: IQ4_XS
1186
+ blk.46.attn_qkv.weight: IQ4_XS
1187
+ blk.46.ffn_down_exps.weight: Q2_0
1188
+ blk.46.ffn_down_shexp.weight: IQ4_NL
1189
+ blk.46.ffn_gate_exps.weight: IQ2_S
1190
+ blk.46.ffn_gate_inp.weight: BF16
1191
+ blk.46.ffn_gate_inp_shexp.weight: BF16
1192
+ blk.46.ffn_gate_shexp.weight: IQ4_XS
1193
+ blk.46.ffn_up_exps.weight: IQ2_S
1194
+ blk.46.ffn_up_shexp.weight: IQ4_XS
1195
+ blk.46.hc_attn_down.weight: BF16
1196
+ blk.46.hc_attn_inject.weight: BF16
1197
+ blk.46.hc_attn_norm.weight: F32
1198
+ blk.46.hc_attn_up.weight: BF16
1199
+ blk.46.hc_ffn_down.weight: BF16
1200
+ blk.46.hc_ffn_inject.weight: BF16
1201
+ blk.46.hc_ffn_norm.weight: F32
1202
+ blk.46.hc_ffn_up.weight: BF16
1203
+ blk.46.ssm_a: F32
1204
+ blk.46.ssm_alpha.weight: BF16
1205
+ blk.46.ssm_beta.weight: BF16
1206
+ blk.46.ssm_conv1d.weight: F32
1207
+ blk.46.ssm_dt.bias: F32
1208
+ blk.46.ssm_norm.weight: F32
1209
+ blk.46.ssm_out.weight: IQ4_XS
1210
+ blk.47.attn_k.weight: Q6_K
1211
+ blk.47.attn_k_norm.weight: F32
1212
+ blk.47.attn_output.weight: IQ4_XS
1213
+ blk.47.attn_q.weight: IQ4_XS
1214
+ blk.47.attn_q_norm.weight: F32
1215
+ blk.47.attn_v.weight: Q6_K
1216
+ blk.47.ffn_down_exps.weight: Q2_0
1217
+ blk.47.ffn_down_shexp.weight: IQ4_NL
1218
+ blk.47.ffn_gate_exps.weight: IQ2_S
1219
+ blk.47.ffn_gate_inp.weight: BF16
1220
+ blk.47.ffn_gate_inp_shexp.weight: BF16
1221
+ blk.47.ffn_gate_shexp.weight: IQ4_XS
1222
+ blk.47.ffn_up_exps.weight: IQ2_S
1223
+ blk.47.ffn_up_shexp.weight: IQ4_XS
1224
+ blk.47.hc_attn_down.weight: BF16
1225
+ blk.47.hc_attn_inject.weight: BF16
1226
+ blk.47.hc_attn_norm.weight: F32
1227
+ blk.47.hc_attn_up.weight: BF16
1228
+ blk.47.hc_ffn_down.weight: BF16
1229
+ blk.47.hc_ffn_inject.weight: BF16
1230
+ blk.47.hc_ffn_norm.weight: F32
1231
+ blk.47.hc_ffn_up.weight: BF16
1232
+ blk.47.indexer.k_norm.weight: F32
1233
+ blk.47.indexer.k_proj.weight: BF16
1234
+ blk.47.indexer.q_norm.weight: F32
1235
+ blk.47.indexer.q_proj.weight: BF16
tensor-allocation/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.rco-allocation.txt ADDED
@@ -0,0 +1,1235 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # RCO per-tensor GGUF allocation
2
+ # Repository: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF
3
+ # GGUF: Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf
4
+ # Source: https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF/resolve/main/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf
5
+ # Target whole-file bpw: 3.50
6
+ # GGUF version: 3
7
+ # GGUF metadata entries: 67
8
+ # Tensor count: 1223
9
+ # Quant-type counts: BF16=484, F16=1, F32=292, IQ2_S=40, IQ3_S=20, IQ3_XXS=34, IQ4_NL=86, IQ4_XS=45, Q2_0=9, Q4_K=47, Q5_K=35, Q6_K=129, Q8_0=1
10
+ # Format: <GGUF tensor name>: <GGML quantization/storage type>
11
+ # Includes fixed non-RCO tensors (e.g. F32/BF16) so the released GGUF is fully reproducible.
12
+ #
13
+ output.weight: Q6_K
14
+ output_hc_down.weight: BF16
15
+ output_hc_norm.weight: F32
16
+ output_hc_up.weight: BF16
17
+ token_embd.weight: IQ4_XS
18
+ blk.0.attn_gate.weight: Q4_K
19
+ blk.0.attn_qkv.weight: Q6_K
20
+ blk.0.ffn_down_exps.weight: IQ4_NL
21
+ blk.0.ffn_down_shexp.weight: IQ4_NL
22
+ blk.0.ffn_gate_exps.weight: IQ3_XXS
23
+ blk.0.ffn_gate_inp.weight: BF16
24
+ blk.0.ffn_gate_inp_shexp.weight: BF16
25
+ blk.0.ffn_gate_shexp.weight: Q4_K
26
+ blk.0.ffn_up_exps.weight: IQ3_XXS
27
+ blk.0.ffn_up_shexp.weight: IQ4_XS
28
+ blk.0.hc_attn_down.weight: BF16
29
+ blk.0.hc_attn_inject.weight: BF16
30
+ blk.0.hc_attn_norm.weight: F32
31
+ blk.0.hc_attn_up.weight: BF16
32
+ blk.0.hc_ffn_down.weight: BF16
33
+ blk.0.hc_ffn_inject.weight: BF16
34
+ blk.0.hc_ffn_norm.weight: F32
35
+ blk.0.hc_ffn_up.weight: BF16
36
+ blk.0.ssm_a: F32
37
+ blk.0.ssm_alpha.weight: BF16
38
+ blk.0.ssm_beta.weight: BF16
39
+ blk.0.ssm_conv1d.weight: F32
40
+ blk.0.ssm_dt.bias: F32
41
+ blk.0.ssm_norm.weight: F32
42
+ blk.0.ssm_out.weight: Q6_K
43
+ blk.1.attn_gate.weight: Q6_K
44
+ blk.1.attn_qkv.weight: Q5_K
45
+ blk.1.ffn_down_exps.weight: Q2_0
46
+ blk.1.ffn_down_shexp.weight: IQ4_NL
47
+ blk.1.ffn_gate_exps.weight: IQ2_S
48
+ blk.1.ffn_gate_inp.weight: BF16
49
+ blk.1.ffn_gate_inp_shexp.weight: BF16
50
+ blk.1.ffn_gate_shexp.weight: IQ4_XS
51
+ blk.1.ffn_up_exps.weight: IQ2_S
52
+ blk.1.ffn_up_shexp.weight: Q4_K
53
+ blk.1.hc_attn_down.weight: BF16
54
+ blk.1.hc_attn_inject.weight: BF16
55
+ blk.1.hc_attn_norm.weight: F32
56
+ blk.1.hc_attn_up.weight: BF16
57
+ blk.1.hc_ffn_down.weight: BF16
58
+ blk.1.hc_ffn_inject.weight: BF16
59
+ blk.1.hc_ffn_norm.weight: F32
60
+ blk.1.hc_ffn_up.weight: BF16
61
+ blk.1.ple_conv1d.weight: F16
62
+ blk.1.ple_key.weight: BF16
63
+ blk.1.ple_norm_conv.weight: F32
64
+ blk.1.ple_norm_key.weight: F32
65
+ blk.1.ple_norm_query.weight: F32
66
+ blk.1.ple_value.weight: BF16
67
+ blk.1.ssm_a: F32
68
+ blk.1.ssm_alpha.weight: BF16
69
+ blk.1.ssm_beta.weight: BF16
70
+ blk.1.ssm_conv1d.weight: F32
71
+ blk.1.ssm_dt.bias: F32
72
+ blk.1.ssm_norm.weight: F32
73
+ blk.1.ssm_out.weight: Q6_K
74
+ blk.2.attn_gate.weight: IQ4_XS
75
+ blk.2.attn_qkv.weight: Q6_K
76
+ blk.2.ffn_down_exps.weight: IQ4_NL
77
+ blk.2.ffn_down_shexp.weight: IQ4_NL
78
+ blk.2.ffn_gate_exps.weight: IQ2_S
79
+ blk.2.ffn_gate_inp.weight: BF16
80
+ blk.2.ffn_gate_inp_shexp.weight: BF16
81
+ blk.2.ffn_gate_shexp.weight: Q4_K
82
+ blk.2.ffn_up_exps.weight: IQ2_S
83
+ blk.2.ffn_up_shexp.weight: Q6_K
84
+ blk.2.hc_attn_down.weight: BF16
85
+ blk.2.hc_attn_inject.weight: BF16
86
+ blk.2.hc_attn_norm.weight: F32
87
+ blk.2.hc_attn_up.weight: BF16
88
+ blk.2.hc_ffn_down.weight: BF16
89
+ blk.2.hc_ffn_inject.weight: BF16
90
+ blk.2.hc_ffn_norm.weight: F32
91
+ blk.2.hc_ffn_up.weight: BF16
92
+ blk.2.ssm_a: F32
93
+ blk.2.ssm_alpha.weight: BF16
94
+ blk.2.ssm_beta.weight: BF16
95
+ blk.2.ssm_conv1d.weight: F32
96
+ blk.2.ssm_dt.bias: F32
97
+ blk.2.ssm_norm.weight: F32
98
+ blk.2.ssm_out.weight: Q6_K
99
+ blk.3.attn_k.weight: Q6_K
100
+ blk.3.attn_k_norm.weight: F32
101
+ blk.3.attn_output.weight: Q4_K
102
+ blk.3.attn_q.weight: IQ4_XS
103
+ blk.3.attn_q_norm.weight: F32
104
+ blk.3.attn_v.weight: Q6_K
105
+ blk.3.ffn_down_exps.weight: IQ4_NL
106
+ blk.3.ffn_down_shexp.weight: IQ4_NL
107
+ blk.3.ffn_gate_exps.weight: IQ2_S
108
+ blk.3.ffn_gate_inp.weight: BF16
109
+ blk.3.ffn_gate_inp_shexp.weight: BF16
110
+ blk.3.ffn_gate_shexp.weight: IQ4_XS
111
+ blk.3.ffn_up_exps.weight: IQ2_S
112
+ blk.3.ffn_up_shexp.weight: Q4_K
113
+ blk.3.hc_attn_down.weight: BF16
114
+ blk.3.hc_attn_inject.weight: BF16
115
+ blk.3.hc_attn_norm.weight: F32
116
+ blk.3.hc_attn_up.weight: BF16
117
+ blk.3.hc_ffn_down.weight: BF16
118
+ blk.3.hc_ffn_inject.weight: BF16
119
+ blk.3.hc_ffn_norm.weight: F32
120
+ blk.3.hc_ffn_up.weight: BF16
121
+ blk.3.indexer.k_norm.weight: F32
122
+ blk.3.indexer.k_proj.weight: BF16
123
+ blk.3.indexer.q_norm.weight: F32
124
+ blk.3.indexer.q_proj.weight: BF16
125
+ blk.4.attn_gate.weight: Q6_K
126
+ blk.4.attn_qkv.weight: Q6_K
127
+ blk.4.ffn_down_exps.weight: IQ4_NL
128
+ blk.4.ffn_down_shexp.weight: IQ4_NL
129
+ blk.4.ffn_gate_exps.weight: IQ2_S
130
+ blk.4.ffn_gate_inp.weight: BF16
131
+ blk.4.ffn_gate_inp_shexp.weight: BF16
132
+ blk.4.ffn_gate_shexp.weight: Q6_K
133
+ blk.4.ffn_up_exps.weight: IQ2_S
134
+ blk.4.ffn_up_shexp.weight: Q5_K
135
+ blk.4.hc_attn_down.weight: BF16
136
+ blk.4.hc_attn_inject.weight: BF16
137
+ blk.4.hc_attn_norm.weight: F32
138
+ blk.4.hc_attn_up.weight: BF16
139
+ blk.4.hc_ffn_down.weight: BF16
140
+ blk.4.hc_ffn_inject.weight: BF16
141
+ blk.4.hc_ffn_norm.weight: F32
142
+ blk.4.hc_ffn_up.weight: BF16
143
+ blk.4.ssm_a: F32
144
+ blk.4.ssm_alpha.weight: BF16
145
+ blk.4.ssm_beta.weight: BF16
146
+ blk.4.ssm_conv1d.weight: F32
147
+ blk.4.ssm_dt.bias: F32
148
+ blk.4.ssm_norm.weight: F32
149
+ blk.4.ssm_out.weight: Q6_K
150
+ blk.5.attn_gate.weight: Q4_K
151
+ blk.5.attn_qkv.weight: Q6_K
152
+ blk.5.ffn_down_exps.weight: IQ4_NL
153
+ blk.5.ffn_down_shexp.weight: IQ4_NL
154
+ blk.5.ffn_gate_exps.weight: IQ2_S
155
+ blk.5.ffn_gate_inp.weight: BF16
156
+ blk.5.ffn_gate_inp_shexp.weight: BF16
157
+ blk.5.ffn_gate_shexp.weight: IQ4_XS
158
+ blk.5.ffn_up_exps.weight: IQ2_S
159
+ blk.5.ffn_up_shexp.weight: Q6_K
160
+ blk.5.hc_attn_down.weight: BF16
161
+ blk.5.hc_attn_inject.weight: BF16
162
+ blk.5.hc_attn_norm.weight: F32
163
+ blk.5.hc_attn_up.weight: BF16
164
+ blk.5.hc_ffn_down.weight: BF16
165
+ blk.5.hc_ffn_inject.weight: BF16
166
+ blk.5.hc_ffn_norm.weight: F32
167
+ blk.5.hc_ffn_up.weight: BF16
168
+ blk.5.ssm_a: F32
169
+ blk.5.ssm_alpha.weight: BF16
170
+ blk.5.ssm_beta.weight: BF16
171
+ blk.5.ssm_conv1d.weight: F32
172
+ blk.5.ssm_dt.bias: F32
173
+ blk.5.ssm_norm.weight: F32
174
+ blk.5.ssm_out.weight: Q4_K
175
+ blk.6.attn_gate.weight: IQ4_XS
176
+ blk.6.attn_qkv.weight: Q6_K
177
+ blk.6.ffn_down_exps.weight: IQ4_NL
178
+ blk.6.ffn_down_shexp.weight: IQ4_NL
179
+ blk.6.ffn_gate_exps.weight: IQ2_S
180
+ blk.6.ffn_gate_inp.weight: BF16
181
+ blk.6.ffn_gate_inp_shexp.weight: BF16
182
+ blk.6.ffn_gate_shexp.weight: Q6_K
183
+ blk.6.ffn_up_exps.weight: IQ2_S
184
+ blk.6.ffn_up_shexp.weight: Q4_K
185
+ blk.6.hc_attn_down.weight: BF16
186
+ blk.6.hc_attn_inject.weight: BF16
187
+ blk.6.hc_attn_norm.weight: F32
188
+ blk.6.hc_attn_up.weight: BF16
189
+ blk.6.hc_ffn_down.weight: BF16
190
+ blk.6.hc_ffn_inject.weight: BF16
191
+ blk.6.hc_ffn_norm.weight: F32
192
+ blk.6.hc_ffn_up.weight: BF16
193
+ blk.6.ssm_a: F32
194
+ blk.6.ssm_alpha.weight: BF16
195
+ blk.6.ssm_beta.weight: BF16
196
+ blk.6.ssm_conv1d.weight: F32
197
+ blk.6.ssm_dt.bias: F32
198
+ blk.6.ssm_norm.weight: F32
199
+ blk.6.ssm_out.weight: Q6_K
200
+ blk.7.attn_k.weight: IQ4_XS
201
+ blk.7.attn_k_norm.weight: F32
202
+ blk.7.attn_output.weight: Q6_K
203
+ blk.7.attn_q.weight: IQ4_XS
204
+ blk.7.attn_q_norm.weight: F32
205
+ blk.7.attn_v.weight: Q6_K
206
+ blk.7.ffn_down_exps.weight: IQ4_NL
207
+ blk.7.ffn_down_shexp.weight: IQ4_NL
208
+ blk.7.ffn_gate_exps.weight: IQ3_XXS
209
+ blk.7.ffn_gate_inp.weight: BF16
210
+ blk.7.ffn_gate_inp_shexp.weight: BF16
211
+ blk.7.ffn_gate_shexp.weight: IQ4_XS
212
+ blk.7.ffn_up_exps.weight: IQ3_XXS
213
+ blk.7.ffn_up_shexp.weight: IQ4_XS
214
+ blk.7.hc_attn_down.weight: BF16
215
+ blk.7.hc_attn_inject.weight: BF16
216
+ blk.7.hc_attn_norm.weight: F32
217
+ blk.7.hc_attn_up.weight: BF16
218
+ blk.7.hc_ffn_down.weight: BF16
219
+ blk.7.hc_ffn_inject.weight: BF16
220
+ blk.7.hc_ffn_norm.weight: F32
221
+ blk.7.hc_ffn_up.weight: BF16
222
+ blk.7.indexer.k_norm.weight: F32
223
+ blk.7.indexer.k_proj.weight: BF16
224
+ blk.7.indexer.q_norm.weight: F32
225
+ blk.7.indexer.q_proj.weight: BF16
226
+ blk.8.attn_gate.weight: IQ4_XS
227
+ blk.8.attn_qkv.weight: Q6_K
228
+ blk.8.ffn_down_exps.weight: Q2_0
229
+ blk.8.ffn_down_shexp.weight: IQ4_NL
230
+ blk.8.ffn_gate_exps.weight: IQ2_S
231
+ blk.8.ffn_gate_inp.weight: BF16
232
+ blk.8.ffn_gate_inp_shexp.weight: BF16
233
+ blk.8.ffn_gate_shexp.weight: IQ4_XS
234
+ blk.8.ffn_up_exps.weight: IQ2_S
235
+ blk.8.ffn_up_shexp.weight: Q4_K
236
+ blk.8.hc_attn_down.weight: BF16
237
+ blk.8.hc_attn_inject.weight: BF16
238
+ blk.8.hc_attn_norm.weight: F32
239
+ blk.8.hc_attn_up.weight: BF16
240
+ blk.8.hc_ffn_down.weight: BF16
241
+ blk.8.hc_ffn_inject.weight: BF16
242
+ blk.8.hc_ffn_norm.weight: F32
243
+ blk.8.hc_ffn_up.weight: BF16
244
+ blk.8.ssm_a: F32
245
+ blk.8.ssm_alpha.weight: BF16
246
+ blk.8.ssm_beta.weight: BF16
247
+ blk.8.ssm_conv1d.weight: F32
248
+ blk.8.ssm_dt.bias: F32
249
+ blk.8.ssm_norm.weight: F32
250
+ blk.8.ssm_out.weight: Q4_K
251
+ blk.9.attn_gate.weight: Q4_K
252
+ blk.9.attn_qkv.weight: Q6_K
253
+ blk.9.ffn_down_exps.weight: IQ4_NL
254
+ blk.9.ffn_down_shexp.weight: IQ4_NL
255
+ blk.9.ffn_gate_exps.weight: IQ2_S
256
+ blk.9.ffn_gate_inp.weight: BF16
257
+ blk.9.ffn_gate_inp_shexp.weight: BF16
258
+ blk.9.ffn_gate_shexp.weight: Q4_K
259
+ blk.9.ffn_up_exps.weight: IQ2_S
260
+ blk.9.ffn_up_shexp.weight: IQ4_XS
261
+ blk.9.hc_attn_down.weight: BF16
262
+ blk.9.hc_attn_inject.weight: BF16
263
+ blk.9.hc_attn_norm.weight: F32
264
+ blk.9.hc_attn_up.weight: BF16
265
+ blk.9.hc_ffn_down.weight: BF16
266
+ blk.9.hc_ffn_inject.weight: BF16
267
+ blk.9.hc_ffn_norm.weight: F32
268
+ blk.9.hc_ffn_up.weight: BF16
269
+ blk.9.ssm_a: F32
270
+ blk.9.ssm_alpha.weight: BF16
271
+ blk.9.ssm_beta.weight: BF16
272
+ blk.9.ssm_conv1d.weight: F32
273
+ blk.9.ssm_dt.bias: F32
274
+ blk.9.ssm_norm.weight: F32
275
+ blk.9.ssm_out.weight: Q6_K
276
+ blk.10.attn_gate.weight: Q6_K
277
+ blk.10.attn_qkv.weight: Q5_K
278
+ blk.10.ffn_down_exps.weight: Q2_0
279
+ blk.10.ffn_down_shexp.weight: IQ4_NL
280
+ blk.10.ffn_gate_exps.weight: IQ2_S
281
+ blk.10.ffn_gate_inp.weight: BF16
282
+ blk.10.ffn_gate_inp_shexp.weight: BF16
283
+ blk.10.ffn_gate_shexp.weight: Q4_K
284
+ blk.10.ffn_up_exps.weight: IQ2_S
285
+ blk.10.ffn_up_shexp.weight: Q4_K
286
+ blk.10.hc_attn_down.weight: BF16
287
+ blk.10.hc_attn_inject.weight: BF16
288
+ blk.10.hc_attn_norm.weight: F32
289
+ blk.10.hc_attn_up.weight: BF16
290
+ blk.10.hc_ffn_down.weight: BF16
291
+ blk.10.hc_ffn_inject.weight: BF16
292
+ blk.10.hc_ffn_norm.weight: F32
293
+ blk.10.hc_ffn_up.weight: BF16
294
+ blk.10.ssm_a: F32
295
+ blk.10.ssm_alpha.weight: BF16
296
+ blk.10.ssm_beta.weight: BF16
297
+ blk.10.ssm_conv1d.weight: F32
298
+ blk.10.ssm_dt.bias: F32
299
+ blk.10.ssm_norm.weight: F32
300
+ blk.10.ssm_out.weight: Q6_K
301
+ blk.11.attn_k.weight: Q6_K
302
+ blk.11.attn_k_norm.weight: F32
303
+ blk.11.attn_output.weight: Q6_K
304
+ blk.11.attn_q.weight: Q4_K
305
+ blk.11.attn_q_norm.weight: F32
306
+ blk.11.attn_v.weight: Q5_K
307
+ blk.11.ffn_down_exps.weight: IQ4_NL
308
+ blk.11.ffn_down_shexp.weight: IQ4_NL
309
+ blk.11.ffn_gate_exps.weight: IQ2_S
310
+ blk.11.ffn_gate_inp.weight: BF16
311
+ blk.11.ffn_gate_inp_shexp.weight: BF16
312
+ blk.11.ffn_gate_shexp.weight: Q4_K
313
+ blk.11.ffn_up_exps.weight: IQ2_S
314
+ blk.11.ffn_up_shexp.weight: Q6_K
315
+ blk.11.hc_attn_down.weight: BF16
316
+ blk.11.hc_attn_inject.weight: BF16
317
+ blk.11.hc_attn_norm.weight: F32
318
+ blk.11.hc_attn_up.weight: BF16
319
+ blk.11.hc_ffn_down.weight: BF16
320
+ blk.11.hc_ffn_inject.weight: BF16
321
+ blk.11.hc_ffn_norm.weight: F32
322
+ blk.11.hc_ffn_up.weight: BF16
323
+ blk.11.indexer.k_norm.weight: F32
324
+ blk.11.indexer.k_proj.weight: BF16
325
+ blk.11.indexer.q_norm.weight: F32
326
+ blk.11.indexer.q_proj.weight: BF16
327
+ blk.12.attn_gate.weight: IQ4_XS
328
+ blk.12.attn_qkv.weight: Q5_K
329
+ blk.12.ffn_down_exps.weight: Q2_0
330
+ blk.12.ffn_down_shexp.weight: IQ4_NL
331
+ blk.12.ffn_gate_exps.weight: IQ3_XXS
332
+ blk.12.ffn_gate_inp.weight: BF16
333
+ blk.12.ffn_gate_inp_shexp.weight: BF16
334
+ blk.12.ffn_gate_shexp.weight: Q6_K
335
+ blk.12.ffn_up_exps.weight: IQ3_XXS
336
+ blk.12.ffn_up_shexp.weight: Q6_K
337
+ blk.12.hc_attn_down.weight: BF16
338
+ blk.12.hc_attn_inject.weight: BF16
339
+ blk.12.hc_attn_norm.weight: F32
340
+ blk.12.hc_attn_up.weight: BF16
341
+ blk.12.hc_ffn_down.weight: BF16
342
+ blk.12.hc_ffn_inject.weight: BF16
343
+ blk.12.hc_ffn_norm.weight: F32
344
+ blk.12.hc_ffn_up.weight: BF16
345
+ blk.12.ssm_a: F32
346
+ blk.12.ssm_alpha.weight: BF16
347
+ blk.12.ssm_beta.weight: BF16
348
+ blk.12.ssm_conv1d.weight: F32
349
+ blk.12.ssm_dt.bias: F32
350
+ blk.12.ssm_norm.weight: F32
351
+ blk.12.ssm_out.weight: Q6_K
352
+ blk.13.attn_gate.weight: Q4_K
353
+ blk.13.attn_qkv.weight: Q4_K
354
+ blk.13.ffn_down_exps.weight: Q2_0
355
+ blk.13.ffn_down_shexp.weight: IQ4_NL
356
+ blk.13.ffn_gate_exps.weight: IQ2_S
357
+ blk.13.ffn_gate_inp.weight: BF16
358
+ blk.13.ffn_gate_inp_shexp.weight: BF16
359
+ blk.13.ffn_gate_shexp.weight: Q4_K
360
+ blk.13.ffn_up_exps.weight: IQ2_S
361
+ blk.13.ffn_up_shexp.weight: IQ4_XS
362
+ blk.13.hc_attn_down.weight: BF16
363
+ blk.13.hc_attn_inject.weight: BF16
364
+ blk.13.hc_attn_norm.weight: F32
365
+ blk.13.hc_attn_up.weight: BF16
366
+ blk.13.hc_ffn_down.weight: BF16
367
+ blk.13.hc_ffn_inject.weight: BF16
368
+ blk.13.hc_ffn_norm.weight: F32
369
+ blk.13.hc_ffn_up.weight: BF16
370
+ blk.13.ssm_a: F32
371
+ blk.13.ssm_alpha.weight: BF16
372
+ blk.13.ssm_beta.weight: BF16
373
+ blk.13.ssm_conv1d.weight: F32
374
+ blk.13.ssm_dt.bias: F32
375
+ blk.13.ssm_norm.weight: F32
376
+ blk.13.ssm_out.weight: Q6_K
377
+ blk.14.attn_gate.weight: Q5_K
378
+ blk.14.attn_qkv.weight: Q6_K
379
+ blk.14.ffn_down_exps.weight: IQ4_NL
380
+ blk.14.ffn_down_shexp.weight: IQ4_NL
381
+ blk.14.ffn_gate_exps.weight: IQ3_XXS
382
+ blk.14.ffn_gate_inp.weight: BF16
383
+ blk.14.ffn_gate_inp_shexp.weight: BF16
384
+ blk.14.ffn_gate_shexp.weight: IQ4_XS
385
+ blk.14.ffn_up_exps.weight: IQ3_XXS
386
+ blk.14.ffn_up_shexp.weight: IQ4_XS
387
+ blk.14.hc_attn_down.weight: BF16
388
+ blk.14.hc_attn_inject.weight: BF16
389
+ blk.14.hc_attn_norm.weight: F32
390
+ blk.14.hc_attn_up.weight: BF16
391
+ blk.14.hc_ffn_down.weight: BF16
392
+ blk.14.hc_ffn_inject.weight: BF16
393
+ blk.14.hc_ffn_norm.weight: F32
394
+ blk.14.hc_ffn_up.weight: BF16
395
+ blk.14.ssm_a: F32
396
+ blk.14.ssm_alpha.weight: BF16
397
+ blk.14.ssm_beta.weight: BF16
398
+ blk.14.ssm_conv1d.weight: F32
399
+ blk.14.ssm_dt.bias: F32
400
+ blk.14.ssm_norm.weight: F32
401
+ blk.14.ssm_out.weight: Q4_K
402
+ blk.15.attn_k.weight: Q6_K
403
+ blk.15.attn_k_norm.weight: F32
404
+ blk.15.attn_output.weight: Q6_K
405
+ blk.15.attn_q.weight: IQ4_XS
406
+ blk.15.attn_q_norm.weight: F32
407
+ blk.15.attn_v.weight: Q6_K
408
+ blk.15.ffn_down_exps.weight: IQ4_NL
409
+ blk.15.ffn_down_shexp.weight: IQ4_NL
410
+ blk.15.ffn_gate_exps.weight: IQ2_S
411
+ blk.15.ffn_gate_inp.weight: BF16
412
+ blk.15.ffn_gate_inp_shexp.weight: BF16
413
+ blk.15.ffn_gate_shexp.weight: Q6_K
414
+ blk.15.ffn_up_exps.weight: IQ2_S
415
+ blk.15.ffn_up_shexp.weight: IQ4_XS
416
+ blk.15.hc_attn_down.weight: BF16
417
+ blk.15.hc_attn_inject.weight: BF16
418
+ blk.15.hc_attn_norm.weight: F32
419
+ blk.15.hc_attn_up.weight: BF16
420
+ blk.15.hc_ffn_down.weight: BF16
421
+ blk.15.hc_ffn_inject.weight: BF16
422
+ blk.15.hc_ffn_norm.weight: F32
423
+ blk.15.hc_ffn_up.weight: BF16
424
+ blk.15.indexer.k_norm.weight: F32
425
+ blk.15.indexer.k_proj.weight: BF16
426
+ blk.15.indexer.q_norm.weight: F32
427
+ blk.15.indexer.q_proj.weight: BF16
428
+ blk.16.attn_gate.weight: IQ4_XS
429
+ blk.16.attn_qkv.weight: Q6_K
430
+ blk.16.ffn_down_exps.weight: IQ4_NL
431
+ blk.16.ffn_down_shexp.weight: IQ4_NL
432
+ blk.16.ffn_gate_exps.weight: IQ3_XXS
433
+ blk.16.ffn_gate_inp.weight: BF16
434
+ blk.16.ffn_gate_inp_shexp.weight: BF16
435
+ blk.16.ffn_gate_shexp.weight: Q4_K
436
+ blk.16.ffn_up_exps.weight: IQ3_XXS
437
+ blk.16.ffn_up_shexp.weight: Q4_K
438
+ blk.16.hc_attn_down.weight: BF16
439
+ blk.16.hc_attn_inject.weight: BF16
440
+ blk.16.hc_attn_norm.weight: F32
441
+ blk.16.hc_attn_up.weight: BF16
442
+ blk.16.hc_ffn_down.weight: BF16
443
+ blk.16.hc_ffn_inject.weight: BF16
444
+ blk.16.hc_ffn_norm.weight: F32
445
+ blk.16.hc_ffn_up.weight: BF16
446
+ blk.16.ssm_a: F32
447
+ blk.16.ssm_alpha.weight: BF16
448
+ blk.16.ssm_beta.weight: BF16
449
+ blk.16.ssm_conv1d.weight: F32
450
+ blk.16.ssm_dt.bias: F32
451
+ blk.16.ssm_norm.weight: F32
452
+ blk.16.ssm_out.weight: Q6_K
453
+ blk.17.attn_gate.weight: Q6_K
454
+ blk.17.attn_qkv.weight: Q6_K
455
+ blk.17.ffn_down_exps.weight: IQ4_NL
456
+ blk.17.ffn_down_shexp.weight: IQ4_NL
457
+ blk.17.ffn_gate_exps.weight: IQ3_S
458
+ blk.17.ffn_gate_inp.weight: BF16
459
+ blk.17.ffn_gate_inp_shexp.weight: BF16
460
+ blk.17.ffn_gate_shexp.weight: IQ4_XS
461
+ blk.17.ffn_up_exps.weight: IQ3_S
462
+ blk.17.ffn_up_shexp.weight: Q4_K
463
+ blk.17.hc_attn_down.weight: BF16
464
+ blk.17.hc_attn_inject.weight: BF16
465
+ blk.17.hc_attn_norm.weight: F32
466
+ blk.17.hc_attn_up.weight: BF16
467
+ blk.17.hc_ffn_down.weight: BF16
468
+ blk.17.hc_ffn_inject.weight: BF16
469
+ blk.17.hc_ffn_norm.weight: F32
470
+ blk.17.hc_ffn_up.weight: BF16
471
+ blk.17.ssm_a: F32
472
+ blk.17.ssm_alpha.weight: BF16
473
+ blk.17.ssm_beta.weight: BF16
474
+ blk.17.ssm_conv1d.weight: F32
475
+ blk.17.ssm_dt.bias: F32
476
+ blk.17.ssm_norm.weight: F32
477
+ blk.17.ssm_out.weight: Q6_K
478
+ blk.18.attn_gate.weight: IQ4_XS
479
+ blk.18.attn_qkv.weight: Q6_K
480
+ blk.18.ffn_down_exps.weight: IQ4_NL
481
+ blk.18.ffn_down_shexp.weight: IQ4_NL
482
+ blk.18.ffn_gate_exps.weight: IQ3_XXS
483
+ blk.18.ffn_gate_inp.weight: BF16
484
+ blk.18.ffn_gate_inp_shexp.weight: BF16
485
+ blk.18.ffn_gate_shexp.weight: Q5_K
486
+ blk.18.ffn_up_exps.weight: IQ3_XXS
487
+ blk.18.ffn_up_shexp.weight: Q5_K
488
+ blk.18.hc_attn_down.weight: BF16
489
+ blk.18.hc_attn_inject.weight: BF16
490
+ blk.18.hc_attn_norm.weight: F32
491
+ blk.18.hc_attn_up.weight: BF16
492
+ blk.18.hc_ffn_down.weight: BF16
493
+ blk.18.hc_ffn_inject.weight: BF16
494
+ blk.18.hc_ffn_norm.weight: F32
495
+ blk.18.hc_ffn_up.weight: BF16
496
+ blk.18.ssm_a: F32
497
+ blk.18.ssm_alpha.weight: BF16
498
+ blk.18.ssm_beta.weight: BF16
499
+ blk.18.ssm_conv1d.weight: F32
500
+ blk.18.ssm_dt.bias: F32
501
+ blk.18.ssm_norm.weight: F32
502
+ blk.18.ssm_out.weight: Q5_K
503
+ blk.19.attn_k.weight: Q6_K
504
+ blk.19.attn_k_norm.weight: F32
505
+ blk.19.attn_output.weight: Q6_K
506
+ blk.19.attn_q.weight: Q5_K
507
+ blk.19.attn_q_norm.weight: F32
508
+ blk.19.attn_v.weight: Q5_K
509
+ blk.19.ffn_down_exps.weight: IQ4_NL
510
+ blk.19.ffn_down_shexp.weight: IQ4_NL
511
+ blk.19.ffn_gate_exps.weight: IQ2_S
512
+ blk.19.ffn_gate_inp.weight: BF16
513
+ blk.19.ffn_gate_inp_shexp.weight: BF16
514
+ blk.19.ffn_gate_shexp.weight: Q6_K
515
+ blk.19.ffn_up_exps.weight: IQ2_S
516
+ blk.19.ffn_up_shexp.weight: Q6_K
517
+ blk.19.hc_attn_down.weight: BF16
518
+ blk.19.hc_attn_inject.weight: BF16
519
+ blk.19.hc_attn_norm.weight: F32
520
+ blk.19.hc_attn_up.weight: BF16
521
+ blk.19.hc_ffn_down.weight: BF16
522
+ blk.19.hc_ffn_inject.weight: BF16
523
+ blk.19.hc_ffn_norm.weight: F32
524
+ blk.19.hc_ffn_up.weight: BF16
525
+ blk.19.indexer.k_norm.weight: F32
526
+ blk.19.indexer.k_proj.weight: BF16
527
+ blk.19.indexer.q_norm.weight: F32
528
+ blk.19.indexer.q_proj.weight: BF16
529
+ blk.20.attn_gate.weight: Q6_K
530
+ blk.20.attn_qkv.weight: Q6_K
531
+ blk.20.ffn_down_exps.weight: Q2_0
532
+ blk.20.ffn_down_shexp.weight: IQ4_NL
533
+ blk.20.ffn_gate_exps.weight: IQ3_XXS
534
+ blk.20.ffn_gate_inp.weight: BF16
535
+ blk.20.ffn_gate_inp_shexp.weight: BF16
536
+ blk.20.ffn_gate_shexp.weight: IQ4_XS
537
+ blk.20.ffn_up_exps.weight: IQ3_XXS
538
+ blk.20.ffn_up_shexp.weight: Q4_K
539
+ blk.20.hc_attn_down.weight: BF16
540
+ blk.20.hc_attn_inject.weight: BF16
541
+ blk.20.hc_attn_norm.weight: F32
542
+ blk.20.hc_attn_up.weight: BF16
543
+ blk.20.hc_ffn_down.weight: BF16
544
+ blk.20.hc_ffn_inject.weight: BF16
545
+ blk.20.hc_ffn_norm.weight: F32
546
+ blk.20.hc_ffn_up.weight: BF16
547
+ blk.20.ssm_a: F32
548
+ blk.20.ssm_alpha.weight: BF16
549
+ blk.20.ssm_beta.weight: BF16
550
+ blk.20.ssm_conv1d.weight: F32
551
+ blk.20.ssm_dt.bias: F32
552
+ blk.20.ssm_norm.weight: F32
553
+ blk.20.ssm_out.weight: Q6_K
554
+ blk.21.attn_gate.weight: Q6_K
555
+ blk.21.attn_qkv.weight: Q6_K
556
+ blk.21.ffn_down_exps.weight: Q2_0
557
+ blk.21.ffn_down_shexp.weight: IQ4_NL
558
+ blk.21.ffn_gate_exps.weight: IQ3_S
559
+ blk.21.ffn_gate_inp.weight: BF16
560
+ blk.21.ffn_gate_inp_shexp.weight: BF16
561
+ blk.21.ffn_gate_shexp.weight: Q4_K
562
+ blk.21.ffn_up_exps.weight: IQ3_S
563
+ blk.21.ffn_up_shexp.weight: IQ4_XS
564
+ blk.21.hc_attn_down.weight: BF16
565
+ blk.21.hc_attn_inject.weight: BF16
566
+ blk.21.hc_attn_norm.weight: F32
567
+ blk.21.hc_attn_up.weight: BF16
568
+ blk.21.hc_ffn_down.weight: BF16
569
+ blk.21.hc_ffn_inject.weight: BF16
570
+ blk.21.hc_ffn_norm.weight: F32
571
+ blk.21.hc_ffn_up.weight: BF16
572
+ blk.21.ssm_a: F32
573
+ blk.21.ssm_alpha.weight: BF16
574
+ blk.21.ssm_beta.weight: BF16
575
+ blk.21.ssm_conv1d.weight: F32
576
+ blk.21.ssm_dt.bias: F32
577
+ blk.21.ssm_norm.weight: F32
578
+ blk.21.ssm_out.weight: Q5_K
579
+ blk.22.attn_gate.weight: Q4_K
580
+ blk.22.attn_qkv.weight: Q6_K
581
+ blk.22.ffn_down_exps.weight: IQ4_NL
582
+ blk.22.ffn_down_shexp.weight: IQ4_NL
583
+ blk.22.ffn_gate_exps.weight: IQ3_XXS
584
+ blk.22.ffn_gate_inp.weight: BF16
585
+ blk.22.ffn_gate_inp_shexp.weight: BF16
586
+ blk.22.ffn_gate_shexp.weight: Q6_K
587
+ blk.22.ffn_up_exps.weight: IQ3_XXS
588
+ blk.22.ffn_up_shexp.weight: IQ4_XS
589
+ blk.22.hc_attn_down.weight: BF16
590
+ blk.22.hc_attn_inject.weight: BF16
591
+ blk.22.hc_attn_norm.weight: F32
592
+ blk.22.hc_attn_up.weight: BF16
593
+ blk.22.hc_ffn_down.weight: BF16
594
+ blk.22.hc_ffn_inject.weight: BF16
595
+ blk.22.hc_ffn_norm.weight: F32
596
+ blk.22.hc_ffn_up.weight: BF16
597
+ blk.22.ssm_a: F32
598
+ blk.22.ssm_alpha.weight: BF16
599
+ blk.22.ssm_beta.weight: BF16
600
+ blk.22.ssm_conv1d.weight: F32
601
+ blk.22.ssm_dt.bias: F32
602
+ blk.22.ssm_norm.weight: F32
603
+ blk.22.ssm_out.weight: Q6_K
604
+ blk.23.attn_k.weight: Q6_K
605
+ blk.23.attn_k_norm.weight: F32
606
+ blk.23.attn_output.weight: Q6_K
607
+ blk.23.attn_q.weight: Q5_K
608
+ blk.23.attn_q_norm.weight: F32
609
+ blk.23.attn_v.weight: Q6_K
610
+ blk.23.ffn_down_exps.weight: IQ4_NL
611
+ blk.23.ffn_down_shexp.weight: IQ4_NL
612
+ blk.23.ffn_gate_exps.weight: IQ2_S
613
+ blk.23.ffn_gate_inp.weight: BF16
614
+ blk.23.ffn_gate_inp_shexp.weight: BF16
615
+ blk.23.ffn_gate_shexp.weight: Q5_K
616
+ blk.23.ffn_up_exps.weight: IQ2_S
617
+ blk.23.ffn_up_shexp.weight: Q5_K
618
+ blk.23.hc_attn_down.weight: BF16
619
+ blk.23.hc_attn_inject.weight: BF16
620
+ blk.23.hc_attn_norm.weight: F32
621
+ blk.23.hc_attn_up.weight: BF16
622
+ blk.23.hc_ffn_down.weight: BF16
623
+ blk.23.hc_ffn_inject.weight: BF16
624
+ blk.23.hc_ffn_norm.weight: F32
625
+ blk.23.hc_ffn_up.weight: BF16
626
+ blk.23.indexer.k_norm.weight: F32
627
+ blk.23.indexer.k_proj.weight: BF16
628
+ blk.23.indexer.q_norm.weight: F32
629
+ blk.23.indexer.q_proj.weight: BF16
630
+ blk.24.attn_gate.weight: Q6_K
631
+ blk.24.attn_qkv.weight: Q6_K
632
+ blk.24.ffn_down_exps.weight: IQ4_NL
633
+ blk.24.ffn_down_shexp.weight: IQ4_NL
634
+ blk.24.ffn_gate_exps.weight: IQ2_S
635
+ blk.24.ffn_gate_inp.weight: BF16
636
+ blk.24.ffn_gate_inp_shexp.weight: BF16
637
+ blk.24.ffn_gate_shexp.weight: IQ4_XS
638
+ blk.24.ffn_up_exps.weight: IQ2_S
639
+ blk.24.ffn_up_shexp.weight: Q5_K
640
+ blk.24.hc_attn_down.weight: BF16
641
+ blk.24.hc_attn_inject.weight: BF16
642
+ blk.24.hc_attn_norm.weight: F32
643
+ blk.24.hc_attn_up.weight: BF16
644
+ blk.24.hc_ffn_down.weight: BF16
645
+ blk.24.hc_ffn_inject.weight: BF16
646
+ blk.24.hc_ffn_norm.weight: F32
647
+ blk.24.hc_ffn_up.weight: BF16
648
+ blk.24.ssm_a: F32
649
+ blk.24.ssm_alpha.weight: BF16
650
+ blk.24.ssm_beta.weight: BF16
651
+ blk.24.ssm_conv1d.weight: F32
652
+ blk.24.ssm_dt.bias: F32
653
+ blk.24.ssm_norm.weight: F32
654
+ blk.24.ssm_out.weight: Q6_K
655
+ blk.25.attn_gate.weight: Q6_K
656
+ blk.25.attn_qkv.weight: Q6_K
657
+ blk.25.ffn_down_exps.weight: IQ4_NL
658
+ blk.25.ffn_down_shexp.weight: IQ4_NL
659
+ blk.25.ffn_gate_exps.weight: IQ3_S
660
+ blk.25.ffn_gate_inp.weight: BF16
661
+ blk.25.ffn_gate_inp_shexp.weight: BF16
662
+ blk.25.ffn_gate_shexp.weight: IQ4_XS
663
+ blk.25.ffn_up_exps.weight: IQ3_S
664
+ blk.25.ffn_up_shexp.weight: Q6_K
665
+ blk.25.hc_attn_down.weight: BF16
666
+ blk.25.hc_attn_inject.weight: BF16
667
+ blk.25.hc_attn_norm.weight: F32
668
+ blk.25.hc_attn_up.weight: BF16
669
+ blk.25.hc_ffn_down.weight: BF16
670
+ blk.25.hc_ffn_inject.weight: BF16
671
+ blk.25.hc_ffn_norm.weight: F32
672
+ blk.25.hc_ffn_up.weight: BF16
673
+ blk.25.ssm_a: F32
674
+ blk.25.ssm_alpha.weight: BF16
675
+ blk.25.ssm_beta.weight: BF16
676
+ blk.25.ssm_conv1d.weight: F32
677
+ blk.25.ssm_dt.bias: F32
678
+ blk.25.ssm_norm.weight: F32
679
+ blk.25.ssm_out.weight: Q6_K
680
+ blk.26.attn_gate.weight: Q6_K
681
+ blk.26.attn_qkv.weight: Q5_K
682
+ blk.26.ffn_down_exps.weight: IQ4_NL
683
+ blk.26.ffn_down_shexp.weight: IQ4_NL
684
+ blk.26.ffn_gate_exps.weight: IQ2_S
685
+ blk.26.ffn_gate_inp.weight: BF16
686
+ blk.26.ffn_gate_inp_shexp.weight: BF16
687
+ blk.26.ffn_gate_shexp.weight: Q6_K
688
+ blk.26.ffn_up_exps.weight: IQ2_S
689
+ blk.26.ffn_up_shexp.weight: Q6_K
690
+ blk.26.hc_attn_down.weight: BF16
691
+ blk.26.hc_attn_inject.weight: BF16
692
+ blk.26.hc_attn_norm.weight: F32
693
+ blk.26.hc_attn_up.weight: BF16
694
+ blk.26.hc_ffn_down.weight: BF16
695
+ blk.26.hc_ffn_inject.weight: BF16
696
+ blk.26.hc_ffn_norm.weight: F32
697
+ blk.26.hc_ffn_up.weight: BF16
698
+ blk.26.ssm_a: F32
699
+ blk.26.ssm_alpha.weight: BF16
700
+ blk.26.ssm_beta.weight: BF16
701
+ blk.26.ssm_conv1d.weight: F32
702
+ blk.26.ssm_dt.bias: F32
703
+ blk.26.ssm_norm.weight: F32
704
+ blk.26.ssm_out.weight: Q5_K
705
+ blk.27.attn_k.weight: Q6_K
706
+ blk.27.attn_k_norm.weight: F32
707
+ blk.27.attn_output.weight: Q6_K
708
+ blk.27.attn_q.weight: Q5_K
709
+ blk.27.attn_q_norm.weight: F32
710
+ blk.27.attn_v.weight: Q6_K
711
+ blk.27.ffn_down_exps.weight: IQ4_NL
712
+ blk.27.ffn_down_shexp.weight: IQ4_NL
713
+ blk.27.ffn_gate_exps.weight: IQ3_S
714
+ blk.27.ffn_gate_inp.weight: BF16
715
+ blk.27.ffn_gate_inp_shexp.weight: BF16
716
+ blk.27.ffn_gate_shexp.weight: IQ4_XS
717
+ blk.27.ffn_up_exps.weight: IQ3_S
718
+ blk.27.ffn_up_shexp.weight: IQ4_XS
719
+ blk.27.hc_attn_down.weight: BF16
720
+ blk.27.hc_attn_inject.weight: BF16
721
+ blk.27.hc_attn_norm.weight: F32
722
+ blk.27.hc_attn_up.weight: BF16
723
+ blk.27.hc_ffn_down.weight: BF16
724
+ blk.27.hc_ffn_inject.weight: BF16
725
+ blk.27.hc_ffn_norm.weight: F32
726
+ blk.27.hc_ffn_up.weight: BF16
727
+ blk.27.indexer.k_norm.weight: F32
728
+ blk.27.indexer.k_proj.weight: BF16
729
+ blk.27.indexer.q_norm.weight: F32
730
+ blk.27.indexer.q_proj.weight: BF16
731
+ blk.28.attn_gate.weight: Q6_K
732
+ blk.28.attn_qkv.weight: Q6_K
733
+ blk.28.ffn_down_exps.weight: IQ4_NL
734
+ blk.28.ffn_down_shexp.weight: IQ4_NL
735
+ blk.28.ffn_gate_exps.weight: IQ3_S
736
+ blk.28.ffn_gate_inp.weight: BF16
737
+ blk.28.ffn_gate_inp_shexp.weight: BF16
738
+ blk.28.ffn_gate_shexp.weight: Q6_K
739
+ blk.28.ffn_up_exps.weight: IQ3_S
740
+ blk.28.ffn_up_shexp.weight: Q6_K
741
+ blk.28.hc_attn_down.weight: BF16
742
+ blk.28.hc_attn_inject.weight: BF16
743
+ blk.28.hc_attn_norm.weight: F32
744
+ blk.28.hc_attn_up.weight: BF16
745
+ blk.28.hc_ffn_down.weight: BF16
746
+ blk.28.hc_ffn_inject.weight: BF16
747
+ blk.28.hc_ffn_norm.weight: F32
748
+ blk.28.hc_ffn_up.weight: BF16
749
+ blk.28.ssm_a: F32
750
+ blk.28.ssm_alpha.weight: BF16
751
+ blk.28.ssm_beta.weight: BF16
752
+ blk.28.ssm_conv1d.weight: F32
753
+ blk.28.ssm_dt.bias: F32
754
+ blk.28.ssm_norm.weight: F32
755
+ blk.28.ssm_out.weight: Q6_K
756
+ blk.29.attn_gate.weight: Q6_K
757
+ blk.29.attn_qkv.weight: Q5_K
758
+ blk.29.ffn_down_exps.weight: IQ4_NL
759
+ blk.29.ffn_down_shexp.weight: IQ4_NL
760
+ blk.29.ffn_gate_exps.weight: IQ3_S
761
+ blk.29.ffn_gate_inp.weight: BF16
762
+ blk.29.ffn_gate_inp_shexp.weight: BF16
763
+ blk.29.ffn_gate_shexp.weight: Q5_K
764
+ blk.29.ffn_up_exps.weight: IQ3_S
765
+ blk.29.ffn_up_shexp.weight: IQ4_XS
766
+ blk.29.hc_attn_down.weight: BF16
767
+ blk.29.hc_attn_inject.weight: BF16
768
+ blk.29.hc_attn_norm.weight: F32
769
+ blk.29.hc_attn_up.weight: BF16
770
+ blk.29.hc_ffn_down.weight: BF16
771
+ blk.29.hc_ffn_inject.weight: BF16
772
+ blk.29.hc_ffn_norm.weight: F32
773
+ blk.29.hc_ffn_up.weight: BF16
774
+ blk.29.ssm_a: F32
775
+ blk.29.ssm_alpha.weight: BF16
776
+ blk.29.ssm_beta.weight: BF16
777
+ blk.29.ssm_conv1d.weight: F32
778
+ blk.29.ssm_dt.bias: F32
779
+ blk.29.ssm_norm.weight: F32
780
+ blk.29.ssm_out.weight: Q6_K
781
+ blk.30.attn_gate.weight: Q4_K
782
+ blk.30.attn_qkv.weight: Q6_K
783
+ blk.30.ffn_down_exps.weight: Q2_0
784
+ blk.30.ffn_down_shexp.weight: IQ4_NL
785
+ blk.30.ffn_gate_exps.weight: IQ3_XXS
786
+ blk.30.ffn_gate_inp.weight: BF16
787
+ blk.30.ffn_gate_inp_shexp.weight: BF16
788
+ blk.30.ffn_gate_shexp.weight: IQ4_XS
789
+ blk.30.ffn_up_exps.weight: IQ3_XXS
790
+ blk.30.ffn_up_shexp.weight: Q4_K
791
+ blk.30.hc_attn_down.weight: BF16
792
+ blk.30.hc_attn_inject.weight: BF16
793
+ blk.30.hc_attn_norm.weight: F32
794
+ blk.30.hc_attn_up.weight: BF16
795
+ blk.30.hc_ffn_down.weight: BF16
796
+ blk.30.hc_ffn_inject.weight: BF16
797
+ blk.30.hc_ffn_norm.weight: F32
798
+ blk.30.hc_ffn_up.weight: BF16
799
+ blk.30.ssm_a: F32
800
+ blk.30.ssm_alpha.weight: BF16
801
+ blk.30.ssm_beta.weight: BF16
802
+ blk.30.ssm_conv1d.weight: F32
803
+ blk.30.ssm_dt.bias: F32
804
+ blk.30.ssm_norm.weight: F32
805
+ blk.30.ssm_out.weight: Q6_K
806
+ blk.31.attn_k.weight: Q6_K
807
+ blk.31.attn_k_norm.weight: F32
808
+ blk.31.attn_output.weight: Q6_K
809
+ blk.31.attn_q.weight: Q6_K
810
+ blk.31.attn_q_norm.weight: F32
811
+ blk.31.attn_v.weight: Q6_K
812
+ blk.31.ffn_down_exps.weight: IQ4_NL
813
+ blk.31.ffn_down_shexp.weight: IQ4_NL
814
+ blk.31.ffn_gate_exps.weight: IQ3_XXS
815
+ blk.31.ffn_gate_inp.weight: BF16
816
+ blk.31.ffn_gate_inp_shexp.weight: BF16
817
+ blk.31.ffn_gate_shexp.weight: IQ4_XS
818
+ blk.31.ffn_up_exps.weight: IQ3_XXS
819
+ blk.31.ffn_up_shexp.weight: Q5_K
820
+ blk.31.hc_attn_down.weight: BF16
821
+ blk.31.hc_attn_inject.weight: BF16
822
+ blk.31.hc_attn_norm.weight: F32
823
+ blk.31.hc_attn_up.weight: BF16
824
+ blk.31.hc_ffn_down.weight: BF16
825
+ blk.31.hc_ffn_inject.weight: BF16
826
+ blk.31.hc_ffn_norm.weight: F32
827
+ blk.31.hc_ffn_up.weight: BF16
828
+ blk.31.indexer.k_norm.weight: F32
829
+ blk.31.indexer.k_proj.weight: BF16
830
+ blk.31.indexer.q_norm.weight: F32
831
+ blk.31.indexer.q_proj.weight: BF16
832
+ blk.32.attn_gate.weight: Q5_K
833
+ blk.32.attn_qkv.weight: Q6_K
834
+ blk.32.ffn_down_exps.weight: IQ4_NL
835
+ blk.32.ffn_down_shexp.weight: IQ4_NL
836
+ blk.32.ffn_gate_exps.weight: IQ2_S
837
+ blk.32.ffn_gate_inp.weight: BF16
838
+ blk.32.ffn_gate_inp_shexp.weight: BF16
839
+ blk.32.ffn_gate_shexp.weight: Q6_K
840
+ blk.32.ffn_up_exps.weight: IQ2_S
841
+ blk.32.ffn_up_shexp.weight: Q4_K
842
+ blk.32.hc_attn_down.weight: BF16
843
+ blk.32.hc_attn_inject.weight: BF16
844
+ blk.32.hc_attn_norm.weight: F32
845
+ blk.32.hc_attn_up.weight: BF16
846
+ blk.32.hc_ffn_down.weight: BF16
847
+ blk.32.hc_ffn_inject.weight: BF16
848
+ blk.32.hc_ffn_norm.weight: F32
849
+ blk.32.hc_ffn_up.weight: BF16
850
+ blk.32.ssm_a: F32
851
+ blk.32.ssm_alpha.weight: BF16
852
+ blk.32.ssm_beta.weight: BF16
853
+ blk.32.ssm_conv1d.weight: F32
854
+ blk.32.ssm_dt.bias: F32
855
+ blk.32.ssm_norm.weight: F32
856
+ blk.32.ssm_out.weight: Q6_K
857
+ blk.33.attn_gate.weight: Q6_K
858
+ blk.33.attn_qkv.weight: IQ4_XS
859
+ blk.33.ffn_down_exps.weight: IQ4_NL
860
+ blk.33.ffn_down_shexp.weight: IQ4_NL
861
+ blk.33.ffn_gate_exps.weight: IQ3_XXS
862
+ blk.33.ffn_gate_inp.weight: BF16
863
+ blk.33.ffn_gate_inp_shexp.weight: BF16
864
+ blk.33.ffn_gate_shexp.weight: IQ4_XS
865
+ blk.33.ffn_up_exps.weight: IQ3_XXS
866
+ blk.33.ffn_up_shexp.weight: Q4_K
867
+ blk.33.hc_attn_down.weight: BF16
868
+ blk.33.hc_attn_inject.weight: BF16
869
+ blk.33.hc_attn_norm.weight: F32
870
+ blk.33.hc_attn_up.weight: BF16
871
+ blk.33.hc_ffn_down.weight: BF16
872
+ blk.33.hc_ffn_inject.weight: BF16
873
+ blk.33.hc_ffn_norm.weight: F32
874
+ blk.33.hc_ffn_up.weight: BF16
875
+ blk.33.ssm_a: F32
876
+ blk.33.ssm_alpha.weight: BF16
877
+ blk.33.ssm_beta.weight: BF16
878
+ blk.33.ssm_conv1d.weight: F32
879
+ blk.33.ssm_dt.bias: F32
880
+ blk.33.ssm_norm.weight: F32
881
+ blk.33.ssm_out.weight: Q6_K
882
+ blk.34.attn_gate.weight: IQ4_XS
883
+ blk.34.attn_qkv.weight: Q5_K
884
+ blk.34.ffn_down_exps.weight: IQ4_NL
885
+ blk.34.ffn_down_shexp.weight: IQ4_NL
886
+ blk.34.ffn_gate_exps.weight: IQ2_S
887
+ blk.34.ffn_gate_inp.weight: BF16
888
+ blk.34.ffn_gate_inp_shexp.weight: BF16
889
+ blk.34.ffn_gate_shexp.weight: Q4_K
890
+ blk.34.ffn_up_exps.weight: IQ2_S
891
+ blk.34.ffn_up_shexp.weight: Q5_K
892
+ blk.34.hc_attn_down.weight: BF16
893
+ blk.34.hc_attn_inject.weight: BF16
894
+ blk.34.hc_attn_norm.weight: F32
895
+ blk.34.hc_attn_up.weight: BF16
896
+ blk.34.hc_ffn_down.weight: BF16
897
+ blk.34.hc_ffn_inject.weight: BF16
898
+ blk.34.hc_ffn_norm.weight: F32
899
+ blk.34.hc_ffn_up.weight: BF16
900
+ blk.34.ssm_a: F32
901
+ blk.34.ssm_alpha.weight: BF16
902
+ blk.34.ssm_beta.weight: BF16
903
+ blk.34.ssm_conv1d.weight: F32
904
+ blk.34.ssm_dt.bias: F32
905
+ blk.34.ssm_norm.weight: F32
906
+ blk.34.ssm_out.weight: Q6_K
907
+ blk.35.attn_k.weight: Q6_K
908
+ blk.35.attn_k_norm.weight: F32
909
+ blk.35.attn_output.weight: Q6_K
910
+ blk.35.attn_q.weight: Q6_K
911
+ blk.35.attn_q_norm.weight: F32
912
+ blk.35.attn_v.weight: Q6_K
913
+ blk.35.ffn_down_exps.weight: IQ4_NL
914
+ blk.35.ffn_down_shexp.weight: IQ4_NL
915
+ blk.35.ffn_gate_exps.weight: IQ3_S
916
+ blk.35.ffn_gate_inp.weight: BF16
917
+ blk.35.ffn_gate_inp_shexp.weight: BF16
918
+ blk.35.ffn_gate_shexp.weight: IQ4_XS
919
+ blk.35.ffn_up_exps.weight: IQ3_S
920
+ blk.35.ffn_up_shexp.weight: Q6_K
921
+ blk.35.hc_attn_down.weight: BF16
922
+ blk.35.hc_attn_inject.weight: BF16
923
+ blk.35.hc_attn_norm.weight: F32
924
+ blk.35.hc_attn_up.weight: BF16
925
+ blk.35.hc_ffn_down.weight: BF16
926
+ blk.35.hc_ffn_inject.weight: BF16
927
+ blk.35.hc_ffn_norm.weight: F32
928
+ blk.35.hc_ffn_up.weight: BF16
929
+ blk.35.indexer.k_norm.weight: F32
930
+ blk.35.indexer.k_proj.weight: BF16
931
+ blk.35.indexer.q_norm.weight: F32
932
+ blk.35.indexer.q_proj.weight: BF16
933
+ blk.36.attn_gate.weight: Q4_K
934
+ blk.36.attn_qkv.weight: Q6_K
935
+ blk.36.ffn_down_exps.weight: Q2_0
936
+ blk.36.ffn_down_shexp.weight: IQ4_NL
937
+ blk.36.ffn_gate_exps.weight: IQ2_S
938
+ blk.36.ffn_gate_inp.weight: BF16
939
+ blk.36.ffn_gate_inp_shexp.weight: BF16
940
+ blk.36.ffn_gate_shexp.weight: Q6_K
941
+ blk.36.ffn_up_exps.weight: IQ2_S
942
+ blk.36.ffn_up_shexp.weight: Q6_K
943
+ blk.36.hc_attn_down.weight: BF16
944
+ blk.36.hc_attn_inject.weight: BF16
945
+ blk.36.hc_attn_norm.weight: F32
946
+ blk.36.hc_attn_up.weight: BF16
947
+ blk.36.hc_ffn_down.weight: BF16
948
+ blk.36.hc_ffn_inject.weight: BF16
949
+ blk.36.hc_ffn_norm.weight: F32
950
+ blk.36.hc_ffn_up.weight: BF16
951
+ blk.36.ssm_a: F32
952
+ blk.36.ssm_alpha.weight: BF16
953
+ blk.36.ssm_beta.weight: BF16
954
+ blk.36.ssm_conv1d.weight: F32
955
+ blk.36.ssm_dt.bias: F32
956
+ blk.36.ssm_norm.weight: F32
957
+ blk.36.ssm_out.weight: Q6_K
958
+ blk.37.attn_gate.weight: Q4_K
959
+ blk.37.attn_qkv.weight: Q6_K
960
+ blk.37.ffn_down_exps.weight: IQ4_NL
961
+ blk.37.ffn_down_shexp.weight: IQ4_NL
962
+ blk.37.ffn_gate_exps.weight: IQ3_XXS
963
+ blk.37.ffn_gate_inp.weight: BF16
964
+ blk.37.ffn_gate_inp_shexp.weight: BF16
965
+ blk.37.ffn_gate_shexp.weight: Q6_K
966
+ blk.37.ffn_up_exps.weight: IQ3_XXS
967
+ blk.37.ffn_up_shexp.weight: Q5_K
968
+ blk.37.hc_attn_down.weight: BF16
969
+ blk.37.hc_attn_inject.weight: BF16
970
+ blk.37.hc_attn_norm.weight: F32
971
+ blk.37.hc_attn_up.weight: BF16
972
+ blk.37.hc_ffn_down.weight: BF16
973
+ blk.37.hc_ffn_inject.weight: BF16
974
+ blk.37.hc_ffn_norm.weight: F32
975
+ blk.37.hc_ffn_up.weight: BF16
976
+ blk.37.ssm_a: F32
977
+ blk.37.ssm_alpha.weight: BF16
978
+ blk.37.ssm_beta.weight: BF16
979
+ blk.37.ssm_conv1d.weight: F32
980
+ blk.37.ssm_dt.bias: F32
981
+ blk.37.ssm_norm.weight: F32
982
+ blk.37.ssm_out.weight: Q6_K
983
+ blk.38.attn_gate.weight: Q5_K
984
+ blk.38.attn_qkv.weight: Q5_K
985
+ blk.38.ffn_down_exps.weight: IQ4_NL
986
+ blk.38.ffn_down_shexp.weight: IQ4_NL
987
+ blk.38.ffn_gate_exps.weight: IQ3_XXS
988
+ blk.38.ffn_gate_inp.weight: BF16
989
+ blk.38.ffn_gate_inp_shexp.weight: BF16
990
+ blk.38.ffn_gate_shexp.weight: Q4_K
991
+ blk.38.ffn_up_exps.weight: IQ3_XXS
992
+ blk.38.ffn_up_shexp.weight: IQ4_XS
993
+ blk.38.hc_attn_down.weight: BF16
994
+ blk.38.hc_attn_inject.weight: BF16
995
+ blk.38.hc_attn_norm.weight: F32
996
+ blk.38.hc_attn_up.weight: BF16
997
+ blk.38.hc_ffn_down.weight: BF16
998
+ blk.38.hc_ffn_inject.weight: BF16
999
+ blk.38.hc_ffn_norm.weight: F32
1000
+ blk.38.hc_ffn_up.weight: BF16
1001
+ blk.38.ssm_a: F32
1002
+ blk.38.ssm_alpha.weight: BF16
1003
+ blk.38.ssm_beta.weight: BF16
1004
+ blk.38.ssm_conv1d.weight: F32
1005
+ blk.38.ssm_dt.bias: F32
1006
+ blk.38.ssm_norm.weight: F32
1007
+ blk.38.ssm_out.weight: Q6_K
1008
+ blk.39.attn_k.weight: Q6_K
1009
+ blk.39.attn_k_norm.weight: F32
1010
+ blk.39.attn_output.weight: Q6_K
1011
+ blk.39.attn_q.weight: Q6_K
1012
+ blk.39.attn_q_norm.weight: F32
1013
+ blk.39.attn_v.weight: Q6_K
1014
+ blk.39.ffn_down_exps.weight: IQ4_NL
1015
+ blk.39.ffn_down_shexp.weight: IQ4_NL
1016
+ blk.39.ffn_gate_exps.weight: IQ2_S
1017
+ blk.39.ffn_gate_inp.weight: BF16
1018
+ blk.39.ffn_gate_inp_shexp.weight: BF16
1019
+ blk.39.ffn_gate_shexp.weight: Q4_K
1020
+ blk.39.ffn_up_exps.weight: IQ2_S
1021
+ blk.39.ffn_up_shexp.weight: Q4_K
1022
+ blk.39.hc_attn_down.weight: BF16
1023
+ blk.39.hc_attn_inject.weight: BF16
1024
+ blk.39.hc_attn_norm.weight: F32
1025
+ blk.39.hc_attn_up.weight: BF16
1026
+ blk.39.hc_ffn_down.weight: BF16
1027
+ blk.39.hc_ffn_inject.weight: BF16
1028
+ blk.39.hc_ffn_norm.weight: F32
1029
+ blk.39.hc_ffn_up.weight: BF16
1030
+ blk.39.indexer.k_norm.weight: F32
1031
+ blk.39.indexer.k_proj.weight: BF16
1032
+ blk.39.indexer.q_norm.weight: F32
1033
+ blk.39.indexer.q_proj.weight: BF16
1034
+ blk.40.attn_gate.weight: Q6_K
1035
+ blk.40.attn_qkv.weight: Q6_K
1036
+ blk.40.ffn_down_exps.weight: IQ4_NL
1037
+ blk.40.ffn_down_shexp.weight: IQ4_NL
1038
+ blk.40.ffn_gate_exps.weight: IQ3_S
1039
+ blk.40.ffn_gate_inp.weight: BF16
1040
+ blk.40.ffn_gate_inp_shexp.weight: BF16
1041
+ blk.40.ffn_gate_shexp.weight: Q5_K
1042
+ blk.40.ffn_up_exps.weight: IQ3_S
1043
+ blk.40.ffn_up_shexp.weight: IQ4_XS
1044
+ blk.40.hc_attn_down.weight: BF16
1045
+ blk.40.hc_attn_inject.weight: BF16
1046
+ blk.40.hc_attn_norm.weight: F32
1047
+ blk.40.hc_attn_up.weight: BF16
1048
+ blk.40.hc_ffn_down.weight: BF16
1049
+ blk.40.hc_ffn_inject.weight: BF16
1050
+ blk.40.hc_ffn_norm.weight: F32
1051
+ blk.40.hc_ffn_up.weight: BF16
1052
+ blk.40.ssm_a: F32
1053
+ blk.40.ssm_alpha.weight: BF16
1054
+ blk.40.ssm_beta.weight: BF16
1055
+ blk.40.ssm_conv1d.weight: F32
1056
+ blk.40.ssm_dt.bias: F32
1057
+ blk.40.ssm_norm.weight: F32
1058
+ blk.40.ssm_out.weight: Q6_K
1059
+ blk.41.attn_gate.weight: Q4_K
1060
+ blk.41.attn_qkv.weight: Q4_K
1061
+ blk.41.ffn_down_exps.weight: IQ4_NL
1062
+ blk.41.ffn_down_shexp.weight: IQ4_NL
1063
+ blk.41.ffn_gate_exps.weight: IQ3_XXS
1064
+ blk.41.ffn_gate_inp.weight: BF16
1065
+ blk.41.ffn_gate_inp_shexp.weight: BF16
1066
+ blk.41.ffn_gate_shexp.weight: Q4_K
1067
+ blk.41.ffn_up_exps.weight: IQ3_XXS
1068
+ blk.41.ffn_up_shexp.weight: Q4_K
1069
+ blk.41.hc_attn_down.weight: BF16
1070
+ blk.41.hc_attn_inject.weight: BF16
1071
+ blk.41.hc_attn_norm.weight: F32
1072
+ blk.41.hc_attn_up.weight: BF16
1073
+ blk.41.hc_ffn_down.weight: BF16
1074
+ blk.41.hc_ffn_inject.weight: BF16
1075
+ blk.41.hc_ffn_norm.weight: F32
1076
+ blk.41.hc_ffn_up.weight: BF16
1077
+ blk.41.ssm_a: F32
1078
+ blk.41.ssm_alpha.weight: BF16
1079
+ blk.41.ssm_beta.weight: BF16
1080
+ blk.41.ssm_conv1d.weight: F32
1081
+ blk.41.ssm_dt.bias: F32
1082
+ blk.41.ssm_norm.weight: F32
1083
+ blk.41.ssm_out.weight: Q6_K
1084
+ blk.42.attn_gate.weight: IQ4_XS
1085
+ blk.42.attn_qkv.weight: Q5_K
1086
+ blk.42.ffn_down_exps.weight: IQ4_NL
1087
+ blk.42.ffn_down_shexp.weight: IQ4_NL
1088
+ blk.42.ffn_gate_exps.weight: IQ3_XXS
1089
+ blk.42.ffn_gate_inp.weight: BF16
1090
+ blk.42.ffn_gate_inp_shexp.weight: BF16
1091
+ blk.42.ffn_gate_shexp.weight: Q6_K
1092
+ blk.42.ffn_up_exps.weight: IQ3_XXS
1093
+ blk.42.ffn_up_shexp.weight: IQ4_XS
1094
+ blk.42.hc_attn_down.weight: BF16
1095
+ blk.42.hc_attn_inject.weight: BF16
1096
+ blk.42.hc_attn_norm.weight: F32
1097
+ blk.42.hc_attn_up.weight: BF16
1098
+ blk.42.hc_ffn_down.weight: BF16
1099
+ blk.42.hc_ffn_inject.weight: BF16
1100
+ blk.42.hc_ffn_norm.weight: F32
1101
+ blk.42.hc_ffn_up.weight: BF16
1102
+ blk.42.ssm_a: F32
1103
+ blk.42.ssm_alpha.weight: BF16
1104
+ blk.42.ssm_beta.weight: BF16
1105
+ blk.42.ssm_conv1d.weight: F32
1106
+ blk.42.ssm_dt.bias: F32
1107
+ blk.42.ssm_norm.weight: F32
1108
+ blk.42.ssm_out.weight: Q6_K
1109
+ blk.43.attn_k.weight: Q6_K
1110
+ blk.43.attn_k_norm.weight: F32
1111
+ blk.43.attn_output.weight: Q6_K
1112
+ blk.43.attn_q.weight: Q6_K
1113
+ blk.43.attn_q_norm.weight: F32
1114
+ blk.43.attn_v.weight: Q6_K
1115
+ blk.43.ffn_down_exps.weight: IQ4_NL
1116
+ blk.43.ffn_down_shexp.weight: IQ4_NL
1117
+ blk.43.ffn_gate_exps.weight: IQ3_XXS
1118
+ blk.43.ffn_gate_inp.weight: BF16
1119
+ blk.43.ffn_gate_inp_shexp.weight: BF16
1120
+ blk.43.ffn_gate_shexp.weight: Q6_K
1121
+ blk.43.ffn_up_exps.weight: IQ3_XXS
1122
+ blk.43.ffn_up_shexp.weight: Q6_K
1123
+ blk.43.hc_attn_down.weight: BF16
1124
+ blk.43.hc_attn_inject.weight: BF16
1125
+ blk.43.hc_attn_norm.weight: F32
1126
+ blk.43.hc_attn_up.weight: BF16
1127
+ blk.43.hc_ffn_down.weight: BF16
1128
+ blk.43.hc_ffn_inject.weight: BF16
1129
+ blk.43.hc_ffn_norm.weight: F32
1130
+ blk.43.hc_ffn_up.weight: BF16
1131
+ blk.43.indexer.k_norm.weight: F32
1132
+ blk.43.indexer.k_proj.weight: BF16
1133
+ blk.43.indexer.q_norm.weight: F32
1134
+ blk.43.indexer.q_proj.weight: BF16
1135
+ blk.44.attn_gate.weight: Q4_K
1136
+ blk.44.attn_qkv.weight: Q5_K
1137
+ blk.44.ffn_down_exps.weight: IQ4_NL
1138
+ blk.44.ffn_down_shexp.weight: IQ4_NL
1139
+ blk.44.ffn_gate_exps.weight: IQ3_S
1140
+ blk.44.ffn_gate_inp.weight: BF16
1141
+ blk.44.ffn_gate_inp_shexp.weight: BF16
1142
+ blk.44.ffn_gate_shexp.weight: Q4_K
1143
+ blk.44.ffn_up_exps.weight: IQ3_S
1144
+ blk.44.ffn_up_shexp.weight: Q4_K
1145
+ blk.44.hc_attn_down.weight: BF16
1146
+ blk.44.hc_attn_inject.weight: BF16
1147
+ blk.44.hc_attn_norm.weight: F32
1148
+ blk.44.hc_attn_up.weight: BF16
1149
+ blk.44.hc_ffn_down.weight: BF16
1150
+ blk.44.hc_ffn_inject.weight: BF16
1151
+ blk.44.hc_ffn_norm.weight: F32
1152
+ blk.44.hc_ffn_up.weight: BF16
1153
+ blk.44.ssm_a: F32
1154
+ blk.44.ssm_alpha.weight: BF16
1155
+ blk.44.ssm_beta.weight: BF16
1156
+ blk.44.ssm_conv1d.weight: F32
1157
+ blk.44.ssm_dt.bias: F32
1158
+ blk.44.ssm_norm.weight: F32
1159
+ blk.44.ssm_out.weight: Q6_K
1160
+ blk.45.attn_gate.weight: Q6_K
1161
+ blk.45.attn_qkv.weight: IQ4_XS
1162
+ blk.45.ffn_down_exps.weight: IQ4_NL
1163
+ blk.45.ffn_down_shexp.weight: IQ4_NL
1164
+ blk.45.ffn_gate_exps.weight: IQ3_S
1165
+ blk.45.ffn_gate_inp.weight: BF16
1166
+ blk.45.ffn_gate_inp_shexp.weight: BF16
1167
+ blk.45.ffn_gate_shexp.weight: Q4_K
1168
+ blk.45.ffn_up_exps.weight: IQ3_S
1169
+ blk.45.ffn_up_shexp.weight: Q6_K
1170
+ blk.45.hc_attn_down.weight: BF16
1171
+ blk.45.hc_attn_inject.weight: BF16
1172
+ blk.45.hc_attn_norm.weight: F32
1173
+ blk.45.hc_attn_up.weight: BF16
1174
+ blk.45.hc_ffn_down.weight: BF16
1175
+ blk.45.hc_ffn_inject.weight: BF16
1176
+ blk.45.hc_ffn_norm.weight: F32
1177
+ blk.45.hc_ffn_up.weight: BF16
1178
+ blk.45.ssm_a: F32
1179
+ blk.45.ssm_alpha.weight: BF16
1180
+ blk.45.ssm_beta.weight: BF16
1181
+ blk.45.ssm_conv1d.weight: F32
1182
+ blk.45.ssm_dt.bias: F32
1183
+ blk.45.ssm_norm.weight: F32
1184
+ blk.45.ssm_out.weight: Q6_K
1185
+ blk.46.attn_gate.weight: Q5_K
1186
+ blk.46.attn_qkv.weight: Q5_K
1187
+ blk.46.ffn_down_exps.weight: IQ4_NL
1188
+ blk.46.ffn_down_shexp.weight: IQ4_NL
1189
+ blk.46.ffn_gate_exps.weight: IQ3_XXS
1190
+ blk.46.ffn_gate_inp.weight: BF16
1191
+ blk.46.ffn_gate_inp_shexp.weight: BF16
1192
+ blk.46.ffn_gate_shexp.weight: Q4_K
1193
+ blk.46.ffn_up_exps.weight: IQ3_XXS
1194
+ blk.46.ffn_up_shexp.weight: Q6_K
1195
+ blk.46.hc_attn_down.weight: BF16
1196
+ blk.46.hc_attn_inject.weight: BF16
1197
+ blk.46.hc_attn_norm.weight: F32
1198
+ blk.46.hc_attn_up.weight: BF16
1199
+ blk.46.hc_ffn_down.weight: BF16
1200
+ blk.46.hc_ffn_inject.weight: BF16
1201
+ blk.46.hc_ffn_norm.weight: F32
1202
+ blk.46.hc_ffn_up.weight: BF16
1203
+ blk.46.ssm_a: F32
1204
+ blk.46.ssm_alpha.weight: BF16
1205
+ blk.46.ssm_beta.weight: BF16
1206
+ blk.46.ssm_conv1d.weight: F32
1207
+ blk.46.ssm_dt.bias: F32
1208
+ blk.46.ssm_norm.weight: F32
1209
+ blk.46.ssm_out.weight: Q5_K
1210
+ blk.47.attn_k.weight: Q6_K
1211
+ blk.47.attn_k_norm.weight: F32
1212
+ blk.47.attn_output.weight: Q6_K
1213
+ blk.47.attn_q.weight: Q6_K
1214
+ blk.47.attn_q_norm.weight: F32
1215
+ blk.47.attn_v.weight: Q6_K
1216
+ blk.47.ffn_down_exps.weight: IQ4_NL
1217
+ blk.47.ffn_down_shexp.weight: Q8_0
1218
+ blk.47.ffn_gate_exps.weight: IQ4_XS
1219
+ blk.47.ffn_gate_inp.weight: BF16
1220
+ blk.47.ffn_gate_inp_shexp.weight: BF16
1221
+ blk.47.ffn_gate_shexp.weight: Q4_K
1222
+ blk.47.ffn_up_exps.weight: IQ4_XS
1223
+ blk.47.ffn_up_shexp.weight: Q5_K
1224
+ blk.47.hc_attn_down.weight: BF16
1225
+ blk.47.hc_attn_inject.weight: BF16
1226
+ blk.47.hc_attn_norm.weight: F32
1227
+ blk.47.hc_attn_up.weight: BF16
1228
+ blk.47.hc_ffn_down.weight: BF16
1229
+ blk.47.hc_ffn_inject.weight: BF16
1230
+ blk.47.hc_ffn_norm.weight: F32
1231
+ blk.47.hc_ffn_up.weight: BF16
1232
+ blk.47.indexer.k_norm.weight: F32
1233
+ blk.47.indexer.k_proj.weight: BF16
1234
+ blk.47.indexer.q_norm.weight: F32
1235
+ blk.47.indexer.q_proj.weight: BF16
tensor-allocation/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.rco-allocation.txt ADDED
@@ -0,0 +1,1235 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # RCO per-tensor GGUF allocation
2
+ # Repository: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF
3
+ # GGUF: Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf
4
+ # Source: https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF/resolve/main/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf
5
+ # Target whole-file bpw: 3.00
6
+ # GGUF version: 3
7
+ # GGUF metadata entries: 67
8
+ # Tensor count: 1223
9
+ # Quant-type counts: BF16=484, F16=1, F32=292, IQ2_S=20, IQ2_XS=20, IQ2_XXS=18, IQ3_S=82, IQ3_XXS=12, IQ4_NL=58, IQ4_XS=67, Q2_0=38, Q4_K=45, Q5_K=15, Q6_K=71
10
+ # Format: <GGUF tensor name>: <GGML quantization/storage type>
11
+ # Includes fixed non-RCO tensors (e.g. F32/BF16) so the released GGUF is fully reproducible.
12
+ #
13
+ output.weight: Q5_K
14
+ output_hc_down.weight: BF16
15
+ output_hc_norm.weight: F32
16
+ output_hc_up.weight: BF16
17
+ token_embd.weight: IQ3_S
18
+ blk.0.attn_gate.weight: IQ3_S
19
+ blk.0.attn_qkv.weight: IQ4_XS
20
+ blk.0.ffn_down_exps.weight: IQ4_NL
21
+ blk.0.ffn_down_shexp.weight: IQ4_NL
22
+ blk.0.ffn_gate_exps.weight: IQ2_XS
23
+ blk.0.ffn_gate_inp.weight: BF16
24
+ blk.0.ffn_gate_inp_shexp.weight: BF16
25
+ blk.0.ffn_gate_shexp.weight: IQ4_XS
26
+ blk.0.ffn_up_exps.weight: IQ2_XS
27
+ blk.0.ffn_up_shexp.weight: IQ3_S
28
+ blk.0.hc_attn_down.weight: BF16
29
+ blk.0.hc_attn_inject.weight: BF16
30
+ blk.0.hc_attn_norm.weight: F32
31
+ blk.0.hc_attn_up.weight: BF16
32
+ blk.0.hc_ffn_down.weight: BF16
33
+ blk.0.hc_ffn_inject.weight: BF16
34
+ blk.0.hc_ffn_norm.weight: F32
35
+ blk.0.hc_ffn_up.weight: BF16
36
+ blk.0.ssm_a: F32
37
+ blk.0.ssm_alpha.weight: BF16
38
+ blk.0.ssm_beta.weight: BF16
39
+ blk.0.ssm_conv1d.weight: F32
40
+ blk.0.ssm_dt.bias: F32
41
+ blk.0.ssm_norm.weight: F32
42
+ blk.0.ssm_out.weight: IQ4_XS
43
+ blk.1.attn_gate.weight: Q6_K
44
+ blk.1.attn_qkv.weight: IQ4_XS
45
+ blk.1.ffn_down_exps.weight: Q2_0
46
+ blk.1.ffn_down_shexp.weight: IQ4_NL
47
+ blk.1.ffn_gate_exps.weight: IQ2_XXS
48
+ blk.1.ffn_gate_inp.weight: BF16
49
+ blk.1.ffn_gate_inp_shexp.weight: BF16
50
+ blk.1.ffn_gate_shexp.weight: IQ4_XS
51
+ blk.1.ffn_up_exps.weight: IQ2_XXS
52
+ blk.1.ffn_up_shexp.weight: IQ3_S
53
+ blk.1.hc_attn_down.weight: BF16
54
+ blk.1.hc_attn_inject.weight: BF16
55
+ blk.1.hc_attn_norm.weight: F32
56
+ blk.1.hc_attn_up.weight: BF16
57
+ blk.1.hc_ffn_down.weight: BF16
58
+ blk.1.hc_ffn_inject.weight: BF16
59
+ blk.1.hc_ffn_norm.weight: F32
60
+ blk.1.hc_ffn_up.weight: BF16
61
+ blk.1.ple_conv1d.weight: F16
62
+ blk.1.ple_key.weight: BF16
63
+ blk.1.ple_norm_conv.weight: F32
64
+ blk.1.ple_norm_key.weight: F32
65
+ blk.1.ple_norm_query.weight: F32
66
+ blk.1.ple_value.weight: BF16
67
+ blk.1.ssm_a: F32
68
+ blk.1.ssm_alpha.weight: BF16
69
+ blk.1.ssm_beta.weight: BF16
70
+ blk.1.ssm_conv1d.weight: F32
71
+ blk.1.ssm_dt.bias: F32
72
+ blk.1.ssm_norm.weight: F32
73
+ blk.1.ssm_out.weight: IQ3_S
74
+ blk.2.attn_gate.weight: IQ3_S
75
+ blk.2.attn_qkv.weight: IQ4_XS
76
+ blk.2.ffn_down_exps.weight: IQ4_NL
77
+ blk.2.ffn_down_shexp.weight: IQ4_NL
78
+ blk.2.ffn_gate_exps.weight: IQ2_XS
79
+ blk.2.ffn_gate_inp.weight: BF16
80
+ blk.2.ffn_gate_inp_shexp.weight: BF16
81
+ blk.2.ffn_gate_shexp.weight: Q4_K
82
+ blk.2.ffn_up_exps.weight: IQ2_XS
83
+ blk.2.ffn_up_shexp.weight: Q5_K
84
+ blk.2.hc_attn_down.weight: BF16
85
+ blk.2.hc_attn_inject.weight: BF16
86
+ blk.2.hc_attn_norm.weight: F32
87
+ blk.2.hc_attn_up.weight: BF16
88
+ blk.2.hc_ffn_down.weight: BF16
89
+ blk.2.hc_ffn_inject.weight: BF16
90
+ blk.2.hc_ffn_norm.weight: F32
91
+ blk.2.hc_ffn_up.weight: BF16
92
+ blk.2.ssm_a: F32
93
+ blk.2.ssm_alpha.weight: BF16
94
+ blk.2.ssm_beta.weight: BF16
95
+ blk.2.ssm_conv1d.weight: F32
96
+ blk.2.ssm_dt.bias: F32
97
+ blk.2.ssm_norm.weight: F32
98
+ blk.2.ssm_out.weight: Q4_K
99
+ blk.3.attn_k.weight: IQ4_XS
100
+ blk.3.attn_k_norm.weight: F32
101
+ blk.3.attn_output.weight: Q6_K
102
+ blk.3.attn_q.weight: IQ4_XS
103
+ blk.3.attn_q_norm.weight: F32
104
+ blk.3.attn_v.weight: Q6_K
105
+ blk.3.ffn_down_exps.weight: Q2_0
106
+ blk.3.ffn_down_shexp.weight: Q2_0
107
+ blk.3.ffn_gate_exps.weight: IQ2_XXS
108
+ blk.3.ffn_gate_inp.weight: BF16
109
+ blk.3.ffn_gate_inp_shexp.weight: BF16
110
+ blk.3.ffn_gate_shexp.weight: Q4_K
111
+ blk.3.ffn_up_exps.weight: IQ2_XXS
112
+ blk.3.ffn_up_shexp.weight: Q4_K
113
+ blk.3.hc_attn_down.weight: BF16
114
+ blk.3.hc_attn_inject.weight: BF16
115
+ blk.3.hc_attn_norm.weight: F32
116
+ blk.3.hc_attn_up.weight: BF16
117
+ blk.3.hc_ffn_down.weight: BF16
118
+ blk.3.hc_ffn_inject.weight: BF16
119
+ blk.3.hc_ffn_norm.weight: F32
120
+ blk.3.hc_ffn_up.weight: BF16
121
+ blk.3.indexer.k_norm.weight: F32
122
+ blk.3.indexer.k_proj.weight: BF16
123
+ blk.3.indexer.q_norm.weight: F32
124
+ blk.3.indexer.q_proj.weight: BF16
125
+ blk.4.attn_gate.weight: IQ3_S
126
+ blk.4.attn_qkv.weight: Q6_K
127
+ blk.4.ffn_down_exps.weight: IQ4_NL
128
+ blk.4.ffn_down_shexp.weight: Q2_0
129
+ blk.4.ffn_gate_exps.weight: IQ2_XS
130
+ blk.4.ffn_gate_inp.weight: BF16
131
+ blk.4.ffn_gate_inp_shexp.weight: BF16
132
+ blk.4.ffn_gate_shexp.weight: IQ4_XS
133
+ blk.4.ffn_up_exps.weight: IQ2_XS
134
+ blk.4.ffn_up_shexp.weight: IQ3_S
135
+ blk.4.hc_attn_down.weight: BF16
136
+ blk.4.hc_attn_inject.weight: BF16
137
+ blk.4.hc_attn_norm.weight: F32
138
+ blk.4.hc_attn_up.weight: BF16
139
+ blk.4.hc_ffn_down.weight: BF16
140
+ blk.4.hc_ffn_inject.weight: BF16
141
+ blk.4.hc_ffn_norm.weight: F32
142
+ blk.4.hc_ffn_up.weight: BF16
143
+ blk.4.ssm_a: F32
144
+ blk.4.ssm_alpha.weight: BF16
145
+ blk.4.ssm_beta.weight: BF16
146
+ blk.4.ssm_conv1d.weight: F32
147
+ blk.4.ssm_dt.bias: F32
148
+ blk.4.ssm_norm.weight: F32
149
+ blk.4.ssm_out.weight: IQ4_XS
150
+ blk.5.attn_gate.weight: IQ3_S
151
+ blk.5.attn_qkv.weight: Q6_K
152
+ blk.5.ffn_down_exps.weight: Q2_0
153
+ blk.5.ffn_down_shexp.weight: IQ4_NL
154
+ blk.5.ffn_gate_exps.weight: IQ2_XS
155
+ blk.5.ffn_gate_inp.weight: BF16
156
+ blk.5.ffn_gate_inp_shexp.weight: BF16
157
+ blk.5.ffn_gate_shexp.weight: IQ4_XS
158
+ blk.5.ffn_up_exps.weight: IQ2_XS
159
+ blk.5.ffn_up_shexp.weight: IQ3_S
160
+ blk.5.hc_attn_down.weight: BF16
161
+ blk.5.hc_attn_inject.weight: BF16
162
+ blk.5.hc_attn_norm.weight: F32
163
+ blk.5.hc_attn_up.weight: BF16
164
+ blk.5.hc_ffn_down.weight: BF16
165
+ blk.5.hc_ffn_inject.weight: BF16
166
+ blk.5.hc_ffn_norm.weight: F32
167
+ blk.5.hc_ffn_up.weight: BF16
168
+ blk.5.ssm_a: F32
169
+ blk.5.ssm_alpha.weight: BF16
170
+ blk.5.ssm_beta.weight: BF16
171
+ blk.5.ssm_conv1d.weight: F32
172
+ blk.5.ssm_dt.bias: F32
173
+ blk.5.ssm_norm.weight: F32
174
+ blk.5.ssm_out.weight: IQ4_XS
175
+ blk.6.attn_gate.weight: IQ4_XS
176
+ blk.6.attn_qkv.weight: IQ4_XS
177
+ blk.6.ffn_down_exps.weight: Q2_0
178
+ blk.6.ffn_down_shexp.weight: Q2_0
179
+ blk.6.ffn_gate_exps.weight: IQ2_XXS
180
+ blk.6.ffn_gate_inp.weight: BF16
181
+ blk.6.ffn_gate_inp_shexp.weight: BF16
182
+ blk.6.ffn_gate_shexp.weight: IQ4_XS
183
+ blk.6.ffn_up_exps.weight: IQ2_XXS
184
+ blk.6.ffn_up_shexp.weight: Q6_K
185
+ blk.6.hc_attn_down.weight: BF16
186
+ blk.6.hc_attn_inject.weight: BF16
187
+ blk.6.hc_attn_norm.weight: F32
188
+ blk.6.hc_attn_up.weight: BF16
189
+ blk.6.hc_ffn_down.weight: BF16
190
+ blk.6.hc_ffn_inject.weight: BF16
191
+ blk.6.hc_ffn_norm.weight: F32
192
+ blk.6.hc_ffn_up.weight: BF16
193
+ blk.6.ssm_a: F32
194
+ blk.6.ssm_alpha.weight: BF16
195
+ blk.6.ssm_beta.weight: BF16
196
+ blk.6.ssm_conv1d.weight: F32
197
+ blk.6.ssm_dt.bias: F32
198
+ blk.6.ssm_norm.weight: F32
199
+ blk.6.ssm_out.weight: Q5_K
200
+ blk.7.attn_k.weight: Q4_K
201
+ blk.7.attn_k_norm.weight: F32
202
+ blk.7.attn_output.weight: Q5_K
203
+ blk.7.attn_q.weight: IQ4_XS
204
+ blk.7.attn_q_norm.weight: F32
205
+ blk.7.attn_v.weight: Q6_K
206
+ blk.7.ffn_down_exps.weight: Q2_0
207
+ blk.7.ffn_down_shexp.weight: IQ4_NL
208
+ blk.7.ffn_gate_exps.weight: IQ2_XXS
209
+ blk.7.ffn_gate_inp.weight: BF16
210
+ blk.7.ffn_gate_inp_shexp.weight: BF16
211
+ blk.7.ffn_gate_shexp.weight: IQ3_S
212
+ blk.7.ffn_up_exps.weight: IQ2_XXS
213
+ blk.7.ffn_up_shexp.weight: IQ4_XS
214
+ blk.7.hc_attn_down.weight: BF16
215
+ blk.7.hc_attn_inject.weight: BF16
216
+ blk.7.hc_attn_norm.weight: F32
217
+ blk.7.hc_attn_up.weight: BF16
218
+ blk.7.hc_ffn_down.weight: BF16
219
+ blk.7.hc_ffn_inject.weight: BF16
220
+ blk.7.hc_ffn_norm.weight: F32
221
+ blk.7.hc_ffn_up.weight: BF16
222
+ blk.7.indexer.k_norm.weight: F32
223
+ blk.7.indexer.k_proj.weight: BF16
224
+ blk.7.indexer.q_norm.weight: F32
225
+ blk.7.indexer.q_proj.weight: BF16
226
+ blk.8.attn_gate.weight: IQ4_XS
227
+ blk.8.attn_qkv.weight: IQ4_XS
228
+ blk.8.ffn_down_exps.weight: Q2_0
229
+ blk.8.ffn_down_shexp.weight: IQ4_NL
230
+ blk.8.ffn_gate_exps.weight: IQ3_S
231
+ blk.8.ffn_gate_inp.weight: BF16
232
+ blk.8.ffn_gate_inp_shexp.weight: BF16
233
+ blk.8.ffn_gate_shexp.weight: IQ3_S
234
+ blk.8.ffn_up_exps.weight: IQ3_S
235
+ blk.8.ffn_up_shexp.weight: IQ3_S
236
+ blk.8.hc_attn_down.weight: BF16
237
+ blk.8.hc_attn_inject.weight: BF16
238
+ blk.8.hc_attn_norm.weight: F32
239
+ blk.8.hc_attn_up.weight: BF16
240
+ blk.8.hc_ffn_down.weight: BF16
241
+ blk.8.hc_ffn_inject.weight: BF16
242
+ blk.8.hc_ffn_norm.weight: F32
243
+ blk.8.hc_ffn_up.weight: BF16
244
+ blk.8.ssm_a: F32
245
+ blk.8.ssm_alpha.weight: BF16
246
+ blk.8.ssm_beta.weight: BF16
247
+ blk.8.ssm_conv1d.weight: F32
248
+ blk.8.ssm_dt.bias: F32
249
+ blk.8.ssm_norm.weight: F32
250
+ blk.8.ssm_out.weight: IQ4_XS
251
+ blk.9.attn_gate.weight: IQ4_XS
252
+ blk.9.attn_qkv.weight: Q6_K
253
+ blk.9.ffn_down_exps.weight: Q2_0
254
+ blk.9.ffn_down_shexp.weight: IQ4_NL
255
+ blk.9.ffn_gate_exps.weight: IQ2_XS
256
+ blk.9.ffn_gate_inp.weight: BF16
257
+ blk.9.ffn_gate_inp_shexp.weight: BF16
258
+ blk.9.ffn_gate_shexp.weight: Q4_K
259
+ blk.9.ffn_up_exps.weight: IQ2_XS
260
+ blk.9.ffn_up_shexp.weight: IQ3_S
261
+ blk.9.hc_attn_down.weight: BF16
262
+ blk.9.hc_attn_inject.weight: BF16
263
+ blk.9.hc_attn_norm.weight: F32
264
+ blk.9.hc_attn_up.weight: BF16
265
+ blk.9.hc_ffn_down.weight: BF16
266
+ blk.9.hc_ffn_inject.weight: BF16
267
+ blk.9.hc_ffn_norm.weight: F32
268
+ blk.9.hc_ffn_up.weight: BF16
269
+ blk.9.ssm_a: F32
270
+ blk.9.ssm_alpha.weight: BF16
271
+ blk.9.ssm_beta.weight: BF16
272
+ blk.9.ssm_conv1d.weight: F32
273
+ blk.9.ssm_dt.bias: F32
274
+ blk.9.ssm_norm.weight: F32
275
+ blk.9.ssm_out.weight: IQ4_XS
276
+ blk.10.attn_gate.weight: Q6_K
277
+ blk.10.attn_qkv.weight: Q6_K
278
+ blk.10.ffn_down_exps.weight: Q2_0
279
+ blk.10.ffn_down_shexp.weight: IQ4_NL
280
+ blk.10.ffn_gate_exps.weight: IQ2_XS
281
+ blk.10.ffn_gate_inp.weight: BF16
282
+ blk.10.ffn_gate_inp_shexp.weight: BF16
283
+ blk.10.ffn_gate_shexp.weight: IQ3_S
284
+ blk.10.ffn_up_exps.weight: IQ2_XS
285
+ blk.10.ffn_up_shexp.weight: IQ4_XS
286
+ blk.10.hc_attn_down.weight: BF16
287
+ blk.10.hc_attn_inject.weight: BF16
288
+ blk.10.hc_attn_norm.weight: F32
289
+ blk.10.hc_attn_up.weight: BF16
290
+ blk.10.hc_ffn_down.weight: BF16
291
+ blk.10.hc_ffn_inject.weight: BF16
292
+ blk.10.hc_ffn_norm.weight: F32
293
+ blk.10.hc_ffn_up.weight: BF16
294
+ blk.10.ssm_a: F32
295
+ blk.10.ssm_alpha.weight: BF16
296
+ blk.10.ssm_beta.weight: BF16
297
+ blk.10.ssm_conv1d.weight: F32
298
+ blk.10.ssm_dt.bias: F32
299
+ blk.10.ssm_norm.weight: F32
300
+ blk.10.ssm_out.weight: Q5_K
301
+ blk.11.attn_k.weight: Q4_K
302
+ blk.11.attn_k_norm.weight: F32
303
+ blk.11.attn_output.weight: Q6_K
304
+ blk.11.attn_q.weight: Q4_K
305
+ blk.11.attn_q_norm.weight: F32
306
+ blk.11.attn_v.weight: Q4_K
307
+ blk.11.ffn_down_exps.weight: Q2_0
308
+ blk.11.ffn_down_shexp.weight: IQ4_NL
309
+ blk.11.ffn_gate_exps.weight: IQ2_XXS
310
+ blk.11.ffn_gate_inp.weight: BF16
311
+ blk.11.ffn_gate_inp_shexp.weight: BF16
312
+ blk.11.ffn_gate_shexp.weight: Q4_K
313
+ blk.11.ffn_up_exps.weight: IQ2_XXS
314
+ blk.11.ffn_up_shexp.weight: Q4_K
315
+ blk.11.hc_attn_down.weight: BF16
316
+ blk.11.hc_attn_inject.weight: BF16
317
+ blk.11.hc_attn_norm.weight: F32
318
+ blk.11.hc_attn_up.weight: BF16
319
+ blk.11.hc_ffn_down.weight: BF16
320
+ blk.11.hc_ffn_inject.weight: BF16
321
+ blk.11.hc_ffn_norm.weight: F32
322
+ blk.11.hc_ffn_up.weight: BF16
323
+ blk.11.indexer.k_norm.weight: F32
324
+ blk.11.indexer.k_proj.weight: BF16
325
+ blk.11.indexer.q_norm.weight: F32
326
+ blk.11.indexer.q_proj.weight: BF16
327
+ blk.12.attn_gate.weight: IQ3_S
328
+ blk.12.attn_qkv.weight: Q4_K
329
+ blk.12.ffn_down_exps.weight: Q2_0
330
+ blk.12.ffn_down_shexp.weight: Q2_0
331
+ blk.12.ffn_gate_exps.weight: IQ2_S
332
+ blk.12.ffn_gate_inp.weight: BF16
333
+ blk.12.ffn_gate_inp_shexp.weight: BF16
334
+ blk.12.ffn_gate_shexp.weight: IQ3_S
335
+ blk.12.ffn_up_exps.weight: IQ2_S
336
+ blk.12.ffn_up_shexp.weight: IQ3_S
337
+ blk.12.hc_attn_down.weight: BF16
338
+ blk.12.hc_attn_inject.weight: BF16
339
+ blk.12.hc_attn_norm.weight: F32
340
+ blk.12.hc_attn_up.weight: BF16
341
+ blk.12.hc_ffn_down.weight: BF16
342
+ blk.12.hc_ffn_inject.weight: BF16
343
+ blk.12.hc_ffn_norm.weight: F32
344
+ blk.12.hc_ffn_up.weight: BF16
345
+ blk.12.ssm_a: F32
346
+ blk.12.ssm_alpha.weight: BF16
347
+ blk.12.ssm_beta.weight: BF16
348
+ blk.12.ssm_conv1d.weight: F32
349
+ blk.12.ssm_dt.bias: F32
350
+ blk.12.ssm_norm.weight: F32
351
+ blk.12.ssm_out.weight: Q6_K
352
+ blk.13.attn_gate.weight: IQ4_XS
353
+ blk.13.attn_qkv.weight: Q4_K
354
+ blk.13.ffn_down_exps.weight: Q2_0
355
+ blk.13.ffn_down_shexp.weight: Q2_0
356
+ blk.13.ffn_gate_exps.weight: IQ2_XXS
357
+ blk.13.ffn_gate_inp.weight: BF16
358
+ blk.13.ffn_gate_inp_shexp.weight: BF16
359
+ blk.13.ffn_gate_shexp.weight: IQ3_S
360
+ blk.13.ffn_up_exps.weight: IQ2_XXS
361
+ blk.13.ffn_up_shexp.weight: IQ3_S
362
+ blk.13.hc_attn_down.weight: BF16
363
+ blk.13.hc_attn_inject.weight: BF16
364
+ blk.13.hc_attn_norm.weight: F32
365
+ blk.13.hc_attn_up.weight: BF16
366
+ blk.13.hc_ffn_down.weight: BF16
367
+ blk.13.hc_ffn_inject.weight: BF16
368
+ blk.13.hc_ffn_norm.weight: F32
369
+ blk.13.hc_ffn_up.weight: BF16
370
+ blk.13.ssm_a: F32
371
+ blk.13.ssm_alpha.weight: BF16
372
+ blk.13.ssm_beta.weight: BF16
373
+ blk.13.ssm_conv1d.weight: F32
374
+ blk.13.ssm_dt.bias: F32
375
+ blk.13.ssm_norm.weight: F32
376
+ blk.13.ssm_out.weight: Q6_K
377
+ blk.14.attn_gate.weight: Q4_K
378
+ blk.14.attn_qkv.weight: Q4_K
379
+ blk.14.ffn_down_exps.weight: Q2_0
380
+ blk.14.ffn_down_shexp.weight: Q2_0
381
+ blk.14.ffn_gate_exps.weight: IQ2_XS
382
+ blk.14.ffn_gate_inp.weight: BF16
383
+ blk.14.ffn_gate_inp_shexp.weight: BF16
384
+ blk.14.ffn_gate_shexp.weight: Q5_K
385
+ blk.14.ffn_up_exps.weight: IQ2_XS
386
+ blk.14.ffn_up_shexp.weight: Q4_K
387
+ blk.14.hc_attn_down.weight: BF16
388
+ blk.14.hc_attn_inject.weight: BF16
389
+ blk.14.hc_attn_norm.weight: F32
390
+ blk.14.hc_attn_up.weight: BF16
391
+ blk.14.hc_ffn_down.weight: BF16
392
+ blk.14.hc_ffn_inject.weight: BF16
393
+ blk.14.hc_ffn_norm.weight: F32
394
+ blk.14.hc_ffn_up.weight: BF16
395
+ blk.14.ssm_a: F32
396
+ blk.14.ssm_alpha.weight: BF16
397
+ blk.14.ssm_beta.weight: BF16
398
+ blk.14.ssm_conv1d.weight: F32
399
+ blk.14.ssm_dt.bias: F32
400
+ blk.14.ssm_norm.weight: F32
401
+ blk.14.ssm_out.weight: Q5_K
402
+ blk.15.attn_k.weight: Q6_K
403
+ blk.15.attn_k_norm.weight: F32
404
+ blk.15.attn_output.weight: Q6_K
405
+ blk.15.attn_q.weight: IQ4_XS
406
+ blk.15.attn_q_norm.weight: F32
407
+ blk.15.attn_v.weight: IQ4_XS
408
+ blk.15.ffn_down_exps.weight: Q2_0
409
+ blk.15.ffn_down_shexp.weight: IQ4_NL
410
+ blk.15.ffn_gate_exps.weight: IQ2_S
411
+ blk.15.ffn_gate_inp.weight: BF16
412
+ blk.15.ffn_gate_inp_shexp.weight: BF16
413
+ blk.15.ffn_gate_shexp.weight: Q4_K
414
+ blk.15.ffn_up_exps.weight: IQ2_S
415
+ blk.15.ffn_up_shexp.weight: IQ3_S
416
+ blk.15.hc_attn_down.weight: BF16
417
+ blk.15.hc_attn_inject.weight: BF16
418
+ blk.15.hc_attn_norm.weight: F32
419
+ blk.15.hc_attn_up.weight: BF16
420
+ blk.15.hc_ffn_down.weight: BF16
421
+ blk.15.hc_ffn_inject.weight: BF16
422
+ blk.15.hc_ffn_norm.weight: F32
423
+ blk.15.hc_ffn_up.weight: BF16
424
+ blk.15.indexer.k_norm.weight: F32
425
+ blk.15.indexer.k_proj.weight: BF16
426
+ blk.15.indexer.q_norm.weight: F32
427
+ blk.15.indexer.q_proj.weight: BF16
428
+ blk.16.attn_gate.weight: IQ4_XS
429
+ blk.16.attn_qkv.weight: Q5_K
430
+ blk.16.ffn_down_exps.weight: Q2_0
431
+ blk.16.ffn_down_shexp.weight: IQ4_NL
432
+ blk.16.ffn_gate_exps.weight: IQ2_XXS
433
+ blk.16.ffn_gate_inp.weight: BF16
434
+ blk.16.ffn_gate_inp_shexp.weight: BF16
435
+ blk.16.ffn_gate_shexp.weight: Q4_K
436
+ blk.16.ffn_up_exps.weight: IQ2_XXS
437
+ blk.16.ffn_up_shexp.weight: IQ3_S
438
+ blk.16.hc_attn_down.weight: BF16
439
+ blk.16.hc_attn_inject.weight: BF16
440
+ blk.16.hc_attn_norm.weight: F32
441
+ blk.16.hc_attn_up.weight: BF16
442
+ blk.16.hc_ffn_down.weight: BF16
443
+ blk.16.hc_ffn_inject.weight: BF16
444
+ blk.16.hc_ffn_norm.weight: F32
445
+ blk.16.hc_ffn_up.weight: BF16
446
+ blk.16.ssm_a: F32
447
+ blk.16.ssm_alpha.weight: BF16
448
+ blk.16.ssm_beta.weight: BF16
449
+ blk.16.ssm_conv1d.weight: F32
450
+ blk.16.ssm_dt.bias: F32
451
+ blk.16.ssm_norm.weight: F32
452
+ blk.16.ssm_out.weight: Q6_K
453
+ blk.17.attn_gate.weight: Q4_K
454
+ blk.17.attn_qkv.weight: IQ4_XS
455
+ blk.17.ffn_down_exps.weight: Q2_0
456
+ blk.17.ffn_down_shexp.weight: IQ4_NL
457
+ blk.17.ffn_gate_exps.weight: IQ2_S
458
+ blk.17.ffn_gate_inp.weight: BF16
459
+ blk.17.ffn_gate_inp_shexp.weight: BF16
460
+ blk.17.ffn_gate_shexp.weight: IQ4_XS
461
+ blk.17.ffn_up_exps.weight: IQ2_S
462
+ blk.17.ffn_up_shexp.weight: IQ4_XS
463
+ blk.17.hc_attn_down.weight: BF16
464
+ blk.17.hc_attn_inject.weight: BF16
465
+ blk.17.hc_attn_norm.weight: F32
466
+ blk.17.hc_attn_up.weight: BF16
467
+ blk.17.hc_ffn_down.weight: BF16
468
+ blk.17.hc_ffn_inject.weight: BF16
469
+ blk.17.hc_ffn_norm.weight: F32
470
+ blk.17.hc_ffn_up.weight: BF16
471
+ blk.17.ssm_a: F32
472
+ blk.17.ssm_alpha.weight: BF16
473
+ blk.17.ssm_beta.weight: BF16
474
+ blk.17.ssm_conv1d.weight: F32
475
+ blk.17.ssm_dt.bias: F32
476
+ blk.17.ssm_norm.weight: F32
477
+ blk.17.ssm_out.weight: Q6_K
478
+ blk.18.attn_gate.weight: IQ4_XS
479
+ blk.18.attn_qkv.weight: Q4_K
480
+ blk.18.ffn_down_exps.weight: Q2_0
481
+ blk.18.ffn_down_shexp.weight: IQ4_NL
482
+ blk.18.ffn_gate_exps.weight: IQ3_S
483
+ blk.18.ffn_gate_inp.weight: BF16
484
+ blk.18.ffn_gate_inp_shexp.weight: BF16
485
+ blk.18.ffn_gate_shexp.weight: IQ3_S
486
+ blk.18.ffn_up_exps.weight: IQ3_S
487
+ blk.18.ffn_up_shexp.weight: IQ3_S
488
+ blk.18.hc_attn_down.weight: BF16
489
+ blk.18.hc_attn_inject.weight: BF16
490
+ blk.18.hc_attn_norm.weight: F32
491
+ blk.18.hc_attn_up.weight: BF16
492
+ blk.18.hc_ffn_down.weight: BF16
493
+ blk.18.hc_ffn_inject.weight: BF16
494
+ blk.18.hc_ffn_norm.weight: F32
495
+ blk.18.hc_ffn_up.weight: BF16
496
+ blk.18.ssm_a: F32
497
+ blk.18.ssm_alpha.weight: BF16
498
+ blk.18.ssm_beta.weight: BF16
499
+ blk.18.ssm_conv1d.weight: F32
500
+ blk.18.ssm_dt.bias: F32
501
+ blk.18.ssm_norm.weight: F32
502
+ blk.18.ssm_out.weight: Q6_K
503
+ blk.19.attn_k.weight: IQ4_XS
504
+ blk.19.attn_k_norm.weight: F32
505
+ blk.19.attn_output.weight: Q6_K
506
+ blk.19.attn_q.weight: Q4_K
507
+ blk.19.attn_q_norm.weight: F32
508
+ blk.19.attn_v.weight: IQ4_XS
509
+ blk.19.ffn_down_exps.weight: Q2_0
510
+ blk.19.ffn_down_shexp.weight: Q2_0
511
+ blk.19.ffn_gate_exps.weight: IQ2_XS
512
+ blk.19.ffn_gate_inp.weight: BF16
513
+ blk.19.ffn_gate_inp_shexp.weight: BF16
514
+ blk.19.ffn_gate_shexp.weight: IQ4_XS
515
+ blk.19.ffn_up_exps.weight: IQ2_XS
516
+ blk.19.ffn_up_shexp.weight: Q4_K
517
+ blk.19.hc_attn_down.weight: BF16
518
+ blk.19.hc_attn_inject.weight: BF16
519
+ blk.19.hc_attn_norm.weight: F32
520
+ blk.19.hc_attn_up.weight: BF16
521
+ blk.19.hc_ffn_down.weight: BF16
522
+ blk.19.hc_ffn_inject.weight: BF16
523
+ blk.19.hc_ffn_norm.weight: F32
524
+ blk.19.hc_ffn_up.weight: BF16
525
+ blk.19.indexer.k_norm.weight: F32
526
+ blk.19.indexer.k_proj.weight: BF16
527
+ blk.19.indexer.q_norm.weight: F32
528
+ blk.19.indexer.q_proj.weight: BF16
529
+ blk.20.attn_gate.weight: IQ4_XS
530
+ blk.20.attn_qkv.weight: IQ4_XS
531
+ blk.20.ffn_down_exps.weight: Q2_0
532
+ blk.20.ffn_down_shexp.weight: Q2_0
533
+ blk.20.ffn_gate_exps.weight: IQ3_S
534
+ blk.20.ffn_gate_inp.weight: BF16
535
+ blk.20.ffn_gate_inp_shexp.weight: BF16
536
+ blk.20.ffn_gate_shexp.weight: Q4_K
537
+ blk.20.ffn_up_exps.weight: IQ3_S
538
+ blk.20.ffn_up_shexp.weight: IQ3_S
539
+ blk.20.hc_attn_down.weight: BF16
540
+ blk.20.hc_attn_inject.weight: BF16
541
+ blk.20.hc_attn_norm.weight: F32
542
+ blk.20.hc_attn_up.weight: BF16
543
+ blk.20.hc_ffn_down.weight: BF16
544
+ blk.20.hc_ffn_inject.weight: BF16
545
+ blk.20.hc_ffn_norm.weight: F32
546
+ blk.20.hc_ffn_up.weight: BF16
547
+ blk.20.ssm_a: F32
548
+ blk.20.ssm_alpha.weight: BF16
549
+ blk.20.ssm_beta.weight: BF16
550
+ blk.20.ssm_conv1d.weight: F32
551
+ blk.20.ssm_dt.bias: F32
552
+ blk.20.ssm_norm.weight: F32
553
+ blk.20.ssm_out.weight: Q6_K
554
+ blk.21.attn_gate.weight: IQ3_S
555
+ blk.21.attn_qkv.weight: Q4_K
556
+ blk.21.ffn_down_exps.weight: Q2_0
557
+ blk.21.ffn_down_shexp.weight: IQ4_NL
558
+ blk.21.ffn_gate_exps.weight: IQ2_S
559
+ blk.21.ffn_gate_inp.weight: BF16
560
+ blk.21.ffn_gate_inp_shexp.weight: BF16
561
+ blk.21.ffn_gate_shexp.weight: IQ4_XS
562
+ blk.21.ffn_up_exps.weight: IQ2_S
563
+ blk.21.ffn_up_shexp.weight: IQ4_XS
564
+ blk.21.hc_attn_down.weight: BF16
565
+ blk.21.hc_attn_inject.weight: BF16
566
+ blk.21.hc_attn_norm.weight: F32
567
+ blk.21.hc_attn_up.weight: BF16
568
+ blk.21.hc_ffn_down.weight: BF16
569
+ blk.21.hc_ffn_inject.weight: BF16
570
+ blk.21.hc_ffn_norm.weight: F32
571
+ blk.21.hc_ffn_up.weight: BF16
572
+ blk.21.ssm_a: F32
573
+ blk.21.ssm_alpha.weight: BF16
574
+ blk.21.ssm_beta.weight: BF16
575
+ blk.21.ssm_conv1d.weight: F32
576
+ blk.21.ssm_dt.bias: F32
577
+ blk.21.ssm_norm.weight: F32
578
+ blk.21.ssm_out.weight: IQ3_S
579
+ blk.22.attn_gate.weight: IQ3_S
580
+ blk.22.attn_qkv.weight: Q6_K
581
+ blk.22.ffn_down_exps.weight: Q2_0
582
+ blk.22.ffn_down_shexp.weight: IQ4_NL
583
+ blk.22.ffn_gate_exps.weight: IQ3_S
584
+ blk.22.ffn_gate_inp.weight: BF16
585
+ blk.22.ffn_gate_inp_shexp.weight: BF16
586
+ blk.22.ffn_gate_shexp.weight: IQ4_XS
587
+ blk.22.ffn_up_exps.weight: IQ3_S
588
+ blk.22.ffn_up_shexp.weight: Q6_K
589
+ blk.22.hc_attn_down.weight: BF16
590
+ blk.22.hc_attn_inject.weight: BF16
591
+ blk.22.hc_attn_norm.weight: F32
592
+ blk.22.hc_attn_up.weight: BF16
593
+ blk.22.hc_ffn_down.weight: BF16
594
+ blk.22.hc_ffn_inject.weight: BF16
595
+ blk.22.hc_ffn_norm.weight: F32
596
+ blk.22.hc_ffn_up.weight: BF16
597
+ blk.22.ssm_a: F32
598
+ blk.22.ssm_alpha.weight: BF16
599
+ blk.22.ssm_beta.weight: BF16
600
+ blk.22.ssm_conv1d.weight: F32
601
+ blk.22.ssm_dt.bias: F32
602
+ blk.22.ssm_norm.weight: F32
603
+ blk.22.ssm_out.weight: Q4_K
604
+ blk.23.attn_k.weight: Q6_K
605
+ blk.23.attn_k_norm.weight: F32
606
+ blk.23.attn_output.weight: Q6_K
607
+ blk.23.attn_q.weight: IQ4_XS
608
+ blk.23.attn_q_norm.weight: F32
609
+ blk.23.attn_v.weight: Q6_K
610
+ blk.23.ffn_down_exps.weight: Q2_0
611
+ blk.23.ffn_down_shexp.weight: IQ4_NL
612
+ blk.23.ffn_gate_exps.weight: IQ3_S
613
+ blk.23.ffn_gate_inp.weight: BF16
614
+ blk.23.ffn_gate_inp_shexp.weight: BF16
615
+ blk.23.ffn_gate_shexp.weight: IQ4_XS
616
+ blk.23.ffn_up_exps.weight: IQ3_S
617
+ blk.23.ffn_up_shexp.weight: IQ4_XS
618
+ blk.23.hc_attn_down.weight: BF16
619
+ blk.23.hc_attn_inject.weight: BF16
620
+ blk.23.hc_attn_norm.weight: F32
621
+ blk.23.hc_attn_up.weight: BF16
622
+ blk.23.hc_ffn_down.weight: BF16
623
+ blk.23.hc_ffn_inject.weight: BF16
624
+ blk.23.hc_ffn_norm.weight: F32
625
+ blk.23.hc_ffn_up.weight: BF16
626
+ blk.23.indexer.k_norm.weight: F32
627
+ blk.23.indexer.k_proj.weight: BF16
628
+ blk.23.indexer.q_norm.weight: F32
629
+ blk.23.indexer.q_proj.weight: BF16
630
+ blk.24.attn_gate.weight: Q4_K
631
+ blk.24.attn_qkv.weight: Q6_K
632
+ blk.24.ffn_down_exps.weight: Q2_0
633
+ blk.24.ffn_down_shexp.weight: IQ4_NL
634
+ blk.24.ffn_gate_exps.weight: IQ3_S
635
+ blk.24.ffn_gate_inp.weight: BF16
636
+ blk.24.ffn_gate_inp_shexp.weight: BF16
637
+ blk.24.ffn_gate_shexp.weight: IQ3_S
638
+ blk.24.ffn_up_exps.weight: IQ3_S
639
+ blk.24.ffn_up_shexp.weight: IQ3_S
640
+ blk.24.hc_attn_down.weight: BF16
641
+ blk.24.hc_attn_inject.weight: BF16
642
+ blk.24.hc_attn_norm.weight: F32
643
+ blk.24.hc_attn_up.weight: BF16
644
+ blk.24.hc_ffn_down.weight: BF16
645
+ blk.24.hc_ffn_inject.weight: BF16
646
+ blk.24.hc_ffn_norm.weight: F32
647
+ blk.24.hc_ffn_up.weight: BF16
648
+ blk.24.ssm_a: F32
649
+ blk.24.ssm_alpha.weight: BF16
650
+ blk.24.ssm_beta.weight: BF16
651
+ blk.24.ssm_conv1d.weight: F32
652
+ blk.24.ssm_dt.bias: F32
653
+ blk.24.ssm_norm.weight: F32
654
+ blk.24.ssm_out.weight: IQ3_S
655
+ blk.25.attn_gate.weight: Q6_K
656
+ blk.25.attn_qkv.weight: Q6_K
657
+ blk.25.ffn_down_exps.weight: Q2_0
658
+ blk.25.ffn_down_shexp.weight: IQ4_NL
659
+ blk.25.ffn_gate_exps.weight: IQ2_XS
660
+ blk.25.ffn_gate_inp.weight: BF16
661
+ blk.25.ffn_gate_inp_shexp.weight: BF16
662
+ blk.25.ffn_gate_shexp.weight: IQ4_XS
663
+ blk.25.ffn_up_exps.weight: IQ2_XS
664
+ blk.25.ffn_up_shexp.weight: IQ4_XS
665
+ blk.25.hc_attn_down.weight: BF16
666
+ blk.25.hc_attn_inject.weight: BF16
667
+ blk.25.hc_attn_norm.weight: F32
668
+ blk.25.hc_attn_up.weight: BF16
669
+ blk.25.hc_ffn_down.weight: BF16
670
+ blk.25.hc_ffn_inject.weight: BF16
671
+ blk.25.hc_ffn_norm.weight: F32
672
+ blk.25.hc_ffn_up.weight: BF16
673
+ blk.25.ssm_a: F32
674
+ blk.25.ssm_alpha.weight: BF16
675
+ blk.25.ssm_beta.weight: BF16
676
+ blk.25.ssm_conv1d.weight: F32
677
+ blk.25.ssm_dt.bias: F32
678
+ blk.25.ssm_norm.weight: F32
679
+ blk.25.ssm_out.weight: Q6_K
680
+ blk.26.attn_gate.weight: Q4_K
681
+ blk.26.attn_qkv.weight: Q4_K
682
+ blk.26.ffn_down_exps.weight: Q2_0
683
+ blk.26.ffn_down_shexp.weight: IQ4_NL
684
+ blk.26.ffn_gate_exps.weight: IQ2_S
685
+ blk.26.ffn_gate_inp.weight: BF16
686
+ blk.26.ffn_gate_inp_shexp.weight: BF16
687
+ blk.26.ffn_gate_shexp.weight: Q4_K
688
+ blk.26.ffn_up_exps.weight: IQ2_S
689
+ blk.26.ffn_up_shexp.weight: IQ3_S
690
+ blk.26.hc_attn_down.weight: BF16
691
+ blk.26.hc_attn_inject.weight: BF16
692
+ blk.26.hc_attn_norm.weight: F32
693
+ blk.26.hc_attn_up.weight: BF16
694
+ blk.26.hc_ffn_down.weight: BF16
695
+ blk.26.hc_ffn_inject.weight: BF16
696
+ blk.26.hc_ffn_norm.weight: F32
697
+ blk.26.hc_ffn_up.weight: BF16
698
+ blk.26.ssm_a: F32
699
+ blk.26.ssm_alpha.weight: BF16
700
+ blk.26.ssm_beta.weight: BF16
701
+ blk.26.ssm_conv1d.weight: F32
702
+ blk.26.ssm_dt.bias: F32
703
+ blk.26.ssm_norm.weight: F32
704
+ blk.26.ssm_out.weight: Q4_K
705
+ blk.27.attn_k.weight: Q6_K
706
+ blk.27.attn_k_norm.weight: F32
707
+ blk.27.attn_output.weight: Q6_K
708
+ blk.27.attn_q.weight: Q6_K
709
+ blk.27.attn_q_norm.weight: F32
710
+ blk.27.attn_v.weight: Q6_K
711
+ blk.27.ffn_down_exps.weight: Q2_0
712
+ blk.27.ffn_down_shexp.weight: IQ4_NL
713
+ blk.27.ffn_gate_exps.weight: IQ2_XXS
714
+ blk.27.ffn_gate_inp.weight: BF16
715
+ blk.27.ffn_gate_inp_shexp.weight: BF16
716
+ blk.27.ffn_gate_shexp.weight: IQ4_XS
717
+ blk.27.ffn_up_exps.weight: IQ2_XXS
718
+ blk.27.ffn_up_shexp.weight: IQ3_S
719
+ blk.27.hc_attn_down.weight: BF16
720
+ blk.27.hc_attn_inject.weight: BF16
721
+ blk.27.hc_attn_norm.weight: F32
722
+ blk.27.hc_attn_up.weight: BF16
723
+ blk.27.hc_ffn_down.weight: BF16
724
+ blk.27.hc_ffn_inject.weight: BF16
725
+ blk.27.hc_ffn_norm.weight: F32
726
+ blk.27.hc_ffn_up.weight: BF16
727
+ blk.27.indexer.k_norm.weight: F32
728
+ blk.27.indexer.k_proj.weight: BF16
729
+ blk.27.indexer.q_norm.weight: F32
730
+ blk.27.indexer.q_proj.weight: BF16
731
+ blk.28.attn_gate.weight: Q6_K
732
+ blk.28.attn_qkv.weight: IQ4_XS
733
+ blk.28.ffn_down_exps.weight: IQ4_NL
734
+ blk.28.ffn_down_shexp.weight: IQ4_NL
735
+ blk.28.ffn_gate_exps.weight: IQ3_S
736
+ blk.28.ffn_gate_inp.weight: BF16
737
+ blk.28.ffn_gate_inp_shexp.weight: BF16
738
+ blk.28.ffn_gate_shexp.weight: IQ3_S
739
+ blk.28.ffn_up_exps.weight: IQ3_S
740
+ blk.28.ffn_up_shexp.weight: IQ3_S
741
+ blk.28.hc_attn_down.weight: BF16
742
+ blk.28.hc_attn_inject.weight: BF16
743
+ blk.28.hc_attn_norm.weight: F32
744
+ blk.28.hc_attn_up.weight: BF16
745
+ blk.28.hc_ffn_down.weight: BF16
746
+ blk.28.hc_ffn_inject.weight: BF16
747
+ blk.28.hc_ffn_norm.weight: F32
748
+ blk.28.hc_ffn_up.weight: BF16
749
+ blk.28.ssm_a: F32
750
+ blk.28.ssm_alpha.weight: BF16
751
+ blk.28.ssm_beta.weight: BF16
752
+ blk.28.ssm_conv1d.weight: F32
753
+ blk.28.ssm_dt.bias: F32
754
+ blk.28.ssm_norm.weight: F32
755
+ blk.28.ssm_out.weight: Q6_K
756
+ blk.29.attn_gate.weight: Q4_K
757
+ blk.29.attn_qkv.weight: Q5_K
758
+ blk.29.ffn_down_exps.weight: IQ4_NL
759
+ blk.29.ffn_down_shexp.weight: IQ4_NL
760
+ blk.29.ffn_gate_exps.weight: IQ2_S
761
+ blk.29.ffn_gate_inp.weight: BF16
762
+ blk.29.ffn_gate_inp_shexp.weight: BF16
763
+ blk.29.ffn_gate_shexp.weight: IQ4_XS
764
+ blk.29.ffn_up_exps.weight: IQ2_S
765
+ blk.29.ffn_up_shexp.weight: IQ3_S
766
+ blk.29.hc_attn_down.weight: BF16
767
+ blk.29.hc_attn_inject.weight: BF16
768
+ blk.29.hc_attn_norm.weight: F32
769
+ blk.29.hc_attn_up.weight: BF16
770
+ blk.29.hc_ffn_down.weight: BF16
771
+ blk.29.hc_ffn_inject.weight: BF16
772
+ blk.29.hc_ffn_norm.weight: F32
773
+ blk.29.hc_ffn_up.weight: BF16
774
+ blk.29.ssm_a: F32
775
+ blk.29.ssm_alpha.weight: BF16
776
+ blk.29.ssm_beta.weight: BF16
777
+ blk.29.ssm_conv1d.weight: F32
778
+ blk.29.ssm_dt.bias: F32
779
+ blk.29.ssm_norm.weight: F32
780
+ blk.29.ssm_out.weight: Q6_K
781
+ blk.30.attn_gate.weight: Q4_K
782
+ blk.30.attn_qkv.weight: Q6_K
783
+ blk.30.ffn_down_exps.weight: IQ4_NL
784
+ blk.30.ffn_down_shexp.weight: IQ4_NL
785
+ blk.30.ffn_gate_exps.weight: IQ3_S
786
+ blk.30.ffn_gate_inp.weight: BF16
787
+ blk.30.ffn_gate_inp_shexp.weight: BF16
788
+ blk.30.ffn_gate_shexp.weight: IQ4_XS
789
+ blk.30.ffn_up_exps.weight: IQ3_S
790
+ blk.30.ffn_up_shexp.weight: Q4_K
791
+ blk.30.hc_attn_down.weight: BF16
792
+ blk.30.hc_attn_inject.weight: BF16
793
+ blk.30.hc_attn_norm.weight: F32
794
+ blk.30.hc_attn_up.weight: BF16
795
+ blk.30.hc_ffn_down.weight: BF16
796
+ blk.30.hc_ffn_inject.weight: BF16
797
+ blk.30.hc_ffn_norm.weight: F32
798
+ blk.30.hc_ffn_up.weight: BF16
799
+ blk.30.ssm_a: F32
800
+ blk.30.ssm_alpha.weight: BF16
801
+ blk.30.ssm_beta.weight: BF16
802
+ blk.30.ssm_conv1d.weight: F32
803
+ blk.30.ssm_dt.bias: F32
804
+ blk.30.ssm_norm.weight: F32
805
+ blk.30.ssm_out.weight: Q6_K
806
+ blk.31.attn_k.weight: Q6_K
807
+ blk.31.attn_k_norm.weight: F32
808
+ blk.31.attn_output.weight: Q6_K
809
+ blk.31.attn_q.weight: Q6_K
810
+ blk.31.attn_q_norm.weight: F32
811
+ blk.31.attn_v.weight: Q6_K
812
+ blk.31.ffn_down_exps.weight: IQ4_NL
813
+ blk.31.ffn_down_shexp.weight: IQ4_NL
814
+ blk.31.ffn_gate_exps.weight: IQ2_XXS
815
+ blk.31.ffn_gate_inp.weight: BF16
816
+ blk.31.ffn_gate_inp_shexp.weight: BF16
817
+ blk.31.ffn_gate_shexp.weight: IQ4_XS
818
+ blk.31.ffn_up_exps.weight: IQ2_XXS
819
+ blk.31.ffn_up_shexp.weight: IQ3_S
820
+ blk.31.hc_attn_down.weight: BF16
821
+ blk.31.hc_attn_inject.weight: BF16
822
+ blk.31.hc_attn_norm.weight: F32
823
+ blk.31.hc_attn_up.weight: BF16
824
+ blk.31.hc_ffn_down.weight: BF16
825
+ blk.31.hc_ffn_inject.weight: BF16
826
+ blk.31.hc_ffn_norm.weight: F32
827
+ blk.31.hc_ffn_up.weight: BF16
828
+ blk.31.indexer.k_norm.weight: F32
829
+ blk.31.indexer.k_proj.weight: BF16
830
+ blk.31.indexer.q_norm.weight: F32
831
+ blk.31.indexer.q_proj.weight: BF16
832
+ blk.32.attn_gate.weight: Q4_K
833
+ blk.32.attn_qkv.weight: Q4_K
834
+ blk.32.ffn_down_exps.weight: IQ4_NL
835
+ blk.32.ffn_down_shexp.weight: IQ4_NL
836
+ blk.32.ffn_gate_exps.weight: IQ3_S
837
+ blk.32.ffn_gate_inp.weight: BF16
838
+ blk.32.ffn_gate_inp_shexp.weight: BF16
839
+ blk.32.ffn_gate_shexp.weight: Q4_K
840
+ blk.32.ffn_up_exps.weight: IQ3_S
841
+ blk.32.ffn_up_shexp.weight: IQ3_S
842
+ blk.32.hc_attn_down.weight: BF16
843
+ blk.32.hc_attn_inject.weight: BF16
844
+ blk.32.hc_attn_norm.weight: F32
845
+ blk.32.hc_attn_up.weight: BF16
846
+ blk.32.hc_ffn_down.weight: BF16
847
+ blk.32.hc_ffn_inject.weight: BF16
848
+ blk.32.hc_ffn_norm.weight: F32
849
+ blk.32.hc_ffn_up.weight: BF16
850
+ blk.32.ssm_a: F32
851
+ blk.32.ssm_alpha.weight: BF16
852
+ blk.32.ssm_beta.weight: BF16
853
+ blk.32.ssm_conv1d.weight: F32
854
+ blk.32.ssm_dt.bias: F32
855
+ blk.32.ssm_norm.weight: F32
856
+ blk.32.ssm_out.weight: Q6_K
857
+ blk.33.attn_gate.weight: IQ3_S
858
+ blk.33.attn_qkv.weight: IQ4_XS
859
+ blk.33.ffn_down_exps.weight: Q2_0
860
+ blk.33.ffn_down_shexp.weight: IQ4_NL
861
+ blk.33.ffn_gate_exps.weight: IQ3_S
862
+ blk.33.ffn_gate_inp.weight: BF16
863
+ blk.33.ffn_gate_inp_shexp.weight: BF16
864
+ blk.33.ffn_gate_shexp.weight: IQ4_XS
865
+ blk.33.ffn_up_exps.weight: IQ3_S
866
+ blk.33.ffn_up_shexp.weight: Q4_K
867
+ blk.33.hc_attn_down.weight: BF16
868
+ blk.33.hc_attn_inject.weight: BF16
869
+ blk.33.hc_attn_norm.weight: F32
870
+ blk.33.hc_attn_up.weight: BF16
871
+ blk.33.hc_ffn_down.weight: BF16
872
+ blk.33.hc_ffn_inject.weight: BF16
873
+ blk.33.hc_ffn_norm.weight: F32
874
+ blk.33.hc_ffn_up.weight: BF16
875
+ blk.33.ssm_a: F32
876
+ blk.33.ssm_alpha.weight: BF16
877
+ blk.33.ssm_beta.weight: BF16
878
+ blk.33.ssm_conv1d.weight: F32
879
+ blk.33.ssm_dt.bias: F32
880
+ blk.33.ssm_norm.weight: F32
881
+ blk.33.ssm_out.weight: Q6_K
882
+ blk.34.attn_gate.weight: Q5_K
883
+ blk.34.attn_qkv.weight: Q6_K
884
+ blk.34.ffn_down_exps.weight: IQ4_NL
885
+ blk.34.ffn_down_shexp.weight: IQ4_NL
886
+ blk.34.ffn_gate_exps.weight: IQ3_S
887
+ blk.34.ffn_gate_inp.weight: BF16
888
+ blk.34.ffn_gate_inp_shexp.weight: BF16
889
+ blk.34.ffn_gate_shexp.weight: Q4_K
890
+ blk.34.ffn_up_exps.weight: IQ3_S
891
+ blk.34.ffn_up_shexp.weight: IQ4_XS
892
+ blk.34.hc_attn_down.weight: BF16
893
+ blk.34.hc_attn_inject.weight: BF16
894
+ blk.34.hc_attn_norm.weight: F32
895
+ blk.34.hc_attn_up.weight: BF16
896
+ blk.34.hc_ffn_down.weight: BF16
897
+ blk.34.hc_ffn_inject.weight: BF16
898
+ blk.34.hc_ffn_norm.weight: F32
899
+ blk.34.hc_ffn_up.weight: BF16
900
+ blk.34.ssm_a: F32
901
+ blk.34.ssm_alpha.weight: BF16
902
+ blk.34.ssm_beta.weight: BF16
903
+ blk.34.ssm_conv1d.weight: F32
904
+ blk.34.ssm_dt.bias: F32
905
+ blk.34.ssm_norm.weight: F32
906
+ blk.34.ssm_out.weight: Q6_K
907
+ blk.35.attn_k.weight: Q6_K
908
+ blk.35.attn_k_norm.weight: F32
909
+ blk.35.attn_output.weight: Q6_K
910
+ blk.35.attn_q.weight: Q4_K
911
+ blk.35.attn_q_norm.weight: F32
912
+ blk.35.attn_v.weight: Q6_K
913
+ blk.35.ffn_down_exps.weight: IQ4_NL
914
+ blk.35.ffn_down_shexp.weight: IQ4_NL
915
+ blk.35.ffn_gate_exps.weight: IQ3_XXS
916
+ blk.35.ffn_gate_inp.weight: BF16
917
+ blk.35.ffn_gate_inp_shexp.weight: BF16
918
+ blk.35.ffn_gate_shexp.weight: IQ4_XS
919
+ blk.35.ffn_up_exps.weight: IQ3_XXS
920
+ blk.35.ffn_up_shexp.weight: Q6_K
921
+ blk.35.hc_attn_down.weight: BF16
922
+ blk.35.hc_attn_inject.weight: BF16
923
+ blk.35.hc_attn_norm.weight: F32
924
+ blk.35.hc_attn_up.weight: BF16
925
+ blk.35.hc_ffn_down.weight: BF16
926
+ blk.35.hc_ffn_inject.weight: BF16
927
+ blk.35.hc_ffn_norm.weight: F32
928
+ blk.35.hc_ffn_up.weight: BF16
929
+ blk.35.indexer.k_norm.weight: F32
930
+ blk.35.indexer.k_proj.weight: BF16
931
+ blk.35.indexer.q_norm.weight: F32
932
+ blk.35.indexer.q_proj.weight: BF16
933
+ blk.36.attn_gate.weight: IQ4_XS
934
+ blk.36.attn_qkv.weight: Q6_K
935
+ blk.36.ffn_down_exps.weight: Q2_0
936
+ blk.36.ffn_down_shexp.weight: IQ4_NL
937
+ blk.36.ffn_gate_exps.weight: IQ3_XXS
938
+ blk.36.ffn_gate_inp.weight: BF16
939
+ blk.36.ffn_gate_inp_shexp.weight: BF16
940
+ blk.36.ffn_gate_shexp.weight: IQ3_S
941
+ blk.36.ffn_up_exps.weight: IQ3_XXS
942
+ blk.36.ffn_up_shexp.weight: IQ3_S
943
+ blk.36.hc_attn_down.weight: BF16
944
+ blk.36.hc_attn_inject.weight: BF16
945
+ blk.36.hc_attn_norm.weight: F32
946
+ blk.36.hc_attn_up.weight: BF16
947
+ blk.36.hc_ffn_down.weight: BF16
948
+ blk.36.hc_ffn_inject.weight: BF16
949
+ blk.36.hc_ffn_norm.weight: F32
950
+ blk.36.hc_ffn_up.weight: BF16
951
+ blk.36.ssm_a: F32
952
+ blk.36.ssm_alpha.weight: BF16
953
+ blk.36.ssm_beta.weight: BF16
954
+ blk.36.ssm_conv1d.weight: F32
955
+ blk.36.ssm_dt.bias: F32
956
+ blk.36.ssm_norm.weight: F32
957
+ blk.36.ssm_out.weight: Q6_K
958
+ blk.37.attn_gate.weight: IQ4_XS
959
+ blk.37.attn_qkv.weight: Q4_K
960
+ blk.37.ffn_down_exps.weight: IQ4_NL
961
+ blk.37.ffn_down_shexp.weight: IQ4_NL
962
+ blk.37.ffn_gate_exps.weight: IQ3_S
963
+ blk.37.ffn_gate_inp.weight: BF16
964
+ blk.37.ffn_gate_inp_shexp.weight: BF16
965
+ blk.37.ffn_gate_shexp.weight: Q4_K
966
+ blk.37.ffn_up_exps.weight: IQ3_S
967
+ blk.37.ffn_up_shexp.weight: IQ3_S
968
+ blk.37.hc_attn_down.weight: BF16
969
+ blk.37.hc_attn_inject.weight: BF16
970
+ blk.37.hc_attn_norm.weight: F32
971
+ blk.37.hc_attn_up.weight: BF16
972
+ blk.37.hc_ffn_down.weight: BF16
973
+ blk.37.hc_ffn_inject.weight: BF16
974
+ blk.37.hc_ffn_norm.weight: F32
975
+ blk.37.hc_ffn_up.weight: BF16
976
+ blk.37.ssm_a: F32
977
+ blk.37.ssm_alpha.weight: BF16
978
+ blk.37.ssm_beta.weight: BF16
979
+ blk.37.ssm_conv1d.weight: F32
980
+ blk.37.ssm_dt.bias: F32
981
+ blk.37.ssm_norm.weight: F32
982
+ blk.37.ssm_out.weight: IQ4_XS
983
+ blk.38.attn_gate.weight: Q5_K
984
+ blk.38.attn_qkv.weight: Q6_K
985
+ blk.38.ffn_down_exps.weight: IQ4_NL
986
+ blk.38.ffn_down_shexp.weight: IQ4_NL
987
+ blk.38.ffn_gate_exps.weight: IQ2_S
988
+ blk.38.ffn_gate_inp.weight: BF16
989
+ blk.38.ffn_gate_inp_shexp.weight: BF16
990
+ blk.38.ffn_gate_shexp.weight: IQ3_S
991
+ blk.38.ffn_up_exps.weight: IQ2_S
992
+ blk.38.ffn_up_shexp.weight: IQ3_S
993
+ blk.38.hc_attn_down.weight: BF16
994
+ blk.38.hc_attn_inject.weight: BF16
995
+ blk.38.hc_attn_norm.weight: F32
996
+ blk.38.hc_attn_up.weight: BF16
997
+ blk.38.hc_ffn_down.weight: BF16
998
+ blk.38.hc_ffn_inject.weight: BF16
999
+ blk.38.hc_ffn_norm.weight: F32
1000
+ blk.38.hc_ffn_up.weight: BF16
1001
+ blk.38.ssm_a: F32
1002
+ blk.38.ssm_alpha.weight: BF16
1003
+ blk.38.ssm_beta.weight: BF16
1004
+ blk.38.ssm_conv1d.weight: F32
1005
+ blk.38.ssm_dt.bias: F32
1006
+ blk.38.ssm_norm.weight: F32
1007
+ blk.38.ssm_out.weight: Q6_K
1008
+ blk.39.attn_k.weight: Q6_K
1009
+ blk.39.attn_k_norm.weight: F32
1010
+ blk.39.attn_output.weight: Q6_K
1011
+ blk.39.attn_q.weight: Q6_K
1012
+ blk.39.attn_q_norm.weight: F32
1013
+ blk.39.attn_v.weight: Q6_K
1014
+ blk.39.ffn_down_exps.weight: IQ4_NL
1015
+ blk.39.ffn_down_shexp.weight: IQ4_NL
1016
+ blk.39.ffn_gate_exps.weight: IQ2_XS
1017
+ blk.39.ffn_gate_inp.weight: BF16
1018
+ blk.39.ffn_gate_inp_shexp.weight: BF16
1019
+ blk.39.ffn_gate_shexp.weight: IQ4_XS
1020
+ blk.39.ffn_up_exps.weight: IQ2_XS
1021
+ blk.39.ffn_up_shexp.weight: IQ4_XS
1022
+ blk.39.hc_attn_down.weight: BF16
1023
+ blk.39.hc_attn_inject.weight: BF16
1024
+ blk.39.hc_attn_norm.weight: F32
1025
+ blk.39.hc_attn_up.weight: BF16
1026
+ blk.39.hc_ffn_down.weight: BF16
1027
+ blk.39.hc_ffn_inject.weight: BF16
1028
+ blk.39.hc_ffn_norm.weight: F32
1029
+ blk.39.hc_ffn_up.weight: BF16
1030
+ blk.39.indexer.k_norm.weight: F32
1031
+ blk.39.indexer.k_proj.weight: BF16
1032
+ blk.39.indexer.q_norm.weight: F32
1033
+ blk.39.indexer.q_proj.weight: BF16
1034
+ blk.40.attn_gate.weight: IQ3_S
1035
+ blk.40.attn_qkv.weight: Q6_K
1036
+ blk.40.ffn_down_exps.weight: Q2_0
1037
+ blk.40.ffn_down_shexp.weight: IQ4_NL
1038
+ blk.40.ffn_gate_exps.weight: IQ2_S
1039
+ blk.40.ffn_gate_inp.weight: BF16
1040
+ blk.40.ffn_gate_inp_shexp.weight: BF16
1041
+ blk.40.ffn_gate_shexp.weight: IQ4_XS
1042
+ blk.40.ffn_up_exps.weight: IQ2_S
1043
+ blk.40.ffn_up_shexp.weight: IQ3_S
1044
+ blk.40.hc_attn_down.weight: BF16
1045
+ blk.40.hc_attn_inject.weight: BF16
1046
+ blk.40.hc_attn_norm.weight: F32
1047
+ blk.40.hc_attn_up.weight: BF16
1048
+ blk.40.hc_ffn_down.weight: BF16
1049
+ blk.40.hc_ffn_inject.weight: BF16
1050
+ blk.40.hc_ffn_norm.weight: F32
1051
+ blk.40.hc_ffn_up.weight: BF16
1052
+ blk.40.ssm_a: F32
1053
+ blk.40.ssm_alpha.weight: BF16
1054
+ blk.40.ssm_beta.weight: BF16
1055
+ blk.40.ssm_conv1d.weight: F32
1056
+ blk.40.ssm_dt.bias: F32
1057
+ blk.40.ssm_norm.weight: F32
1058
+ blk.40.ssm_out.weight: Q6_K
1059
+ blk.41.attn_gate.weight: IQ3_S
1060
+ blk.41.attn_qkv.weight: IQ4_XS
1061
+ blk.41.ffn_down_exps.weight: Q2_0
1062
+ blk.41.ffn_down_shexp.weight: IQ4_NL
1063
+ blk.41.ffn_gate_exps.weight: IQ3_XXS
1064
+ blk.41.ffn_gate_inp.weight: BF16
1065
+ blk.41.ffn_gate_inp_shexp.weight: BF16
1066
+ blk.41.ffn_gate_shexp.weight: Q4_K
1067
+ blk.41.ffn_up_exps.weight: IQ3_XXS
1068
+ blk.41.ffn_up_shexp.weight: IQ3_S
1069
+ blk.41.hc_attn_down.weight: BF16
1070
+ blk.41.hc_attn_inject.weight: BF16
1071
+ blk.41.hc_attn_norm.weight: F32
1072
+ blk.41.hc_attn_up.weight: BF16
1073
+ blk.41.hc_ffn_down.weight: BF16
1074
+ blk.41.hc_ffn_inject.weight: BF16
1075
+ blk.41.hc_ffn_norm.weight: F32
1076
+ blk.41.hc_ffn_up.weight: BF16
1077
+ blk.41.ssm_a: F32
1078
+ blk.41.ssm_alpha.weight: BF16
1079
+ blk.41.ssm_beta.weight: BF16
1080
+ blk.41.ssm_conv1d.weight: F32
1081
+ blk.41.ssm_dt.bias: F32
1082
+ blk.41.ssm_norm.weight: F32
1083
+ blk.41.ssm_out.weight: Q4_K
1084
+ blk.42.attn_gate.weight: IQ4_XS
1085
+ blk.42.attn_qkv.weight: Q6_K
1086
+ blk.42.ffn_down_exps.weight: IQ4_NL
1087
+ blk.42.ffn_down_shexp.weight: IQ4_NL
1088
+ blk.42.ffn_gate_exps.weight: IQ2_S
1089
+ blk.42.ffn_gate_inp.weight: BF16
1090
+ blk.42.ffn_gate_inp_shexp.weight: BF16
1091
+ blk.42.ffn_gate_shexp.weight: IQ3_S
1092
+ blk.42.ffn_up_exps.weight: IQ2_S
1093
+ blk.42.ffn_up_shexp.weight: IQ4_XS
1094
+ blk.42.hc_attn_down.weight: BF16
1095
+ blk.42.hc_attn_inject.weight: BF16
1096
+ blk.42.hc_attn_norm.weight: F32
1097
+ blk.42.hc_attn_up.weight: BF16
1098
+ blk.42.hc_ffn_down.weight: BF16
1099
+ blk.42.hc_ffn_inject.weight: BF16
1100
+ blk.42.hc_ffn_norm.weight: F32
1101
+ blk.42.hc_ffn_up.weight: BF16
1102
+ blk.42.ssm_a: F32
1103
+ blk.42.ssm_alpha.weight: BF16
1104
+ blk.42.ssm_beta.weight: BF16
1105
+ blk.42.ssm_conv1d.weight: F32
1106
+ blk.42.ssm_dt.bias: F32
1107
+ blk.42.ssm_norm.weight: F32
1108
+ blk.42.ssm_out.weight: Q6_K
1109
+ blk.43.attn_k.weight: Q5_K
1110
+ blk.43.attn_k_norm.weight: F32
1111
+ blk.43.attn_output.weight: Q6_K
1112
+ blk.43.attn_q.weight: Q6_K
1113
+ blk.43.attn_q_norm.weight: F32
1114
+ blk.43.attn_v.weight: Q6_K
1115
+ blk.43.ffn_down_exps.weight: IQ4_NL
1116
+ blk.43.ffn_down_shexp.weight: IQ4_NL
1117
+ blk.43.ffn_gate_exps.weight: IQ3_S
1118
+ blk.43.ffn_gate_inp.weight: BF16
1119
+ blk.43.ffn_gate_inp_shexp.weight: BF16
1120
+ blk.43.ffn_gate_shexp.weight: IQ3_S
1121
+ blk.43.ffn_up_exps.weight: IQ3_S
1122
+ blk.43.ffn_up_shexp.weight: IQ3_S
1123
+ blk.43.hc_attn_down.weight: BF16
1124
+ blk.43.hc_attn_inject.weight: BF16
1125
+ blk.43.hc_attn_norm.weight: F32
1126
+ blk.43.hc_attn_up.weight: BF16
1127
+ blk.43.hc_ffn_down.weight: BF16
1128
+ blk.43.hc_ffn_inject.weight: BF16
1129
+ blk.43.hc_ffn_norm.weight: F32
1130
+ blk.43.hc_ffn_up.weight: BF16
1131
+ blk.43.indexer.k_norm.weight: F32
1132
+ blk.43.indexer.k_proj.weight: BF16
1133
+ blk.43.indexer.q_norm.weight: F32
1134
+ blk.43.indexer.q_proj.weight: BF16
1135
+ blk.44.attn_gate.weight: IQ4_XS
1136
+ blk.44.attn_qkv.weight: Q6_K
1137
+ blk.44.ffn_down_exps.weight: IQ4_NL
1138
+ blk.44.ffn_down_shexp.weight: IQ4_NL
1139
+ blk.44.ffn_gate_exps.weight: IQ3_XXS
1140
+ blk.44.ffn_gate_inp.weight: BF16
1141
+ blk.44.ffn_gate_inp_shexp.weight: BF16
1142
+ blk.44.ffn_gate_shexp.weight: IQ3_S
1143
+ blk.44.ffn_up_exps.weight: IQ3_XXS
1144
+ blk.44.ffn_up_shexp.weight: Q4_K
1145
+ blk.44.hc_attn_down.weight: BF16
1146
+ blk.44.hc_attn_inject.weight: BF16
1147
+ blk.44.hc_attn_norm.weight: F32
1148
+ blk.44.hc_attn_up.weight: BF16
1149
+ blk.44.hc_ffn_down.weight: BF16
1150
+ blk.44.hc_ffn_inject.weight: BF16
1151
+ blk.44.hc_ffn_norm.weight: F32
1152
+ blk.44.hc_ffn_up.weight: BF16
1153
+ blk.44.ssm_a: F32
1154
+ blk.44.ssm_alpha.weight: BF16
1155
+ blk.44.ssm_beta.weight: BF16
1156
+ blk.44.ssm_conv1d.weight: F32
1157
+ blk.44.ssm_dt.bias: F32
1158
+ blk.44.ssm_norm.weight: F32
1159
+ blk.44.ssm_out.weight: Q6_K
1160
+ blk.45.attn_gate.weight: Q6_K
1161
+ blk.45.attn_qkv.weight: IQ4_XS
1162
+ blk.45.ffn_down_exps.weight: Q2_0
1163
+ blk.45.ffn_down_shexp.weight: IQ4_NL
1164
+ blk.45.ffn_gate_exps.weight: IQ3_XXS
1165
+ blk.45.ffn_gate_inp.weight: BF16
1166
+ blk.45.ffn_gate_inp_shexp.weight: BF16
1167
+ blk.45.ffn_gate_shexp.weight: Q6_K
1168
+ blk.45.ffn_up_exps.weight: IQ3_XXS
1169
+ blk.45.ffn_up_shexp.weight: IQ3_S
1170
+ blk.45.hc_attn_down.weight: BF16
1171
+ blk.45.hc_attn_inject.weight: BF16
1172
+ blk.45.hc_attn_norm.weight: F32
1173
+ blk.45.hc_attn_up.weight: BF16
1174
+ blk.45.hc_ffn_down.weight: BF16
1175
+ blk.45.hc_ffn_inject.weight: BF16
1176
+ blk.45.hc_ffn_norm.weight: F32
1177
+ blk.45.hc_ffn_up.weight: BF16
1178
+ blk.45.ssm_a: F32
1179
+ blk.45.ssm_alpha.weight: BF16
1180
+ blk.45.ssm_beta.weight: BF16
1181
+ blk.45.ssm_conv1d.weight: F32
1182
+ blk.45.ssm_dt.bias: F32
1183
+ blk.45.ssm_norm.weight: F32
1184
+ blk.45.ssm_out.weight: Q5_K
1185
+ blk.46.attn_gate.weight: Q5_K
1186
+ blk.46.attn_qkv.weight: Q5_K
1187
+ blk.46.ffn_down_exps.weight: IQ4_NL
1188
+ blk.46.ffn_down_shexp.weight: IQ4_NL
1189
+ blk.46.ffn_gate_exps.weight: IQ2_S
1190
+ blk.46.ffn_gate_inp.weight: BF16
1191
+ blk.46.ffn_gate_inp_shexp.weight: BF16
1192
+ blk.46.ffn_gate_shexp.weight: IQ4_XS
1193
+ blk.46.ffn_up_exps.weight: IQ2_S
1194
+ blk.46.ffn_up_shexp.weight: IQ3_S
1195
+ blk.46.hc_attn_down.weight: BF16
1196
+ blk.46.hc_attn_inject.weight: BF16
1197
+ blk.46.hc_attn_norm.weight: F32
1198
+ blk.46.hc_attn_up.weight: BF16
1199
+ blk.46.hc_ffn_down.weight: BF16
1200
+ blk.46.hc_ffn_inject.weight: BF16
1201
+ blk.46.hc_ffn_norm.weight: F32
1202
+ blk.46.hc_ffn_up.weight: BF16
1203
+ blk.46.ssm_a: F32
1204
+ blk.46.ssm_alpha.weight: BF16
1205
+ blk.46.ssm_beta.weight: BF16
1206
+ blk.46.ssm_conv1d.weight: F32
1207
+ blk.46.ssm_dt.bias: F32
1208
+ blk.46.ssm_norm.weight: F32
1209
+ blk.46.ssm_out.weight: Q6_K
1210
+ blk.47.attn_k.weight: IQ3_S
1211
+ blk.47.attn_k_norm.weight: F32
1212
+ blk.47.attn_output.weight: Q4_K
1213
+ blk.47.attn_q.weight: IQ3_S
1214
+ blk.47.attn_q_norm.weight: F32
1215
+ blk.47.attn_v.weight: Q6_K
1216
+ blk.47.ffn_down_exps.weight: IQ4_NL
1217
+ blk.47.ffn_down_shexp.weight: IQ4_NL
1218
+ blk.47.ffn_gate_exps.weight: IQ3_XXS
1219
+ blk.47.ffn_gate_inp.weight: BF16
1220
+ blk.47.ffn_gate_inp_shexp.weight: BF16
1221
+ blk.47.ffn_gate_shexp.weight: IQ4_XS
1222
+ blk.47.ffn_up_exps.weight: IQ3_XXS
1223
+ blk.47.ffn_up_shexp.weight: IQ4_XS
1224
+ blk.47.hc_attn_down.weight: BF16
1225
+ blk.47.hc_attn_inject.weight: BF16
1226
+ blk.47.hc_attn_norm.weight: F32
1227
+ blk.47.hc_attn_up.weight: BF16
1228
+ blk.47.hc_ffn_down.weight: BF16
1229
+ blk.47.hc_ffn_inject.weight: BF16
1230
+ blk.47.hc_ffn_norm.weight: F32
1231
+ blk.47.hc_ffn_up.weight: BF16
1232
+ blk.47.indexer.k_norm.weight: F32
1233
+ blk.47.indexer.k_proj.weight: BF16
1234
+ blk.47.indexer.q_norm.weight: F32
1235
+ blk.47.indexer.q_proj.weight: BF16
tensor-allocation/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.rco-allocation.txt ADDED
@@ -0,0 +1,1235 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # RCO per-tensor GGUF allocation
2
+ # Repository: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF
3
+ # GGUF: Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf
4
+ # Source: https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF/resolve/main/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf
5
+ # Target whole-file bpw: 2.40
6
+ # GGUF version: 3
7
+ # GGUF metadata entries: 73
8
+ # Tensor count: 1223
9
+ # Quant-type counts: BF16=483, F16=1, F32=292, IQ4_NL=7, IQ4_XS=56, Q2_0=202, Q3_K=91, Q4_0=16, Q4_K=38, Q5_0=7, Q5_K=16, Q6_K=12, Q8_0=2
10
+ # Format: <GGUF tensor name>: <GGML quantization/storage type>
11
+ # Includes fixed non-RCO tensors (e.g. F32/BF16) so the released GGUF is fully reproducible.
12
+ #
13
+ output.weight: Q5_K
14
+ output_hc_down.weight: BF16
15
+ output_hc_norm.weight: F32
16
+ output_hc_up.weight: BF16
17
+ token_embd.weight: Q3_K
18
+ blk.0.attn_gate.weight: Q4_K
19
+ blk.0.attn_qkv.weight: IQ4_XS
20
+ blk.0.ffn_down_exps.weight: Q2_0
21
+ blk.0.ffn_down_shexp.weight: Q5_0
22
+ blk.0.ffn_gate_exps.weight: Q2_0
23
+ blk.0.ffn_gate_inp.weight: BF16
24
+ blk.0.ffn_gate_inp_shexp.weight: BF16
25
+ blk.0.ffn_gate_shexp.weight: IQ4_XS
26
+ blk.0.ffn_up_exps.weight: Q2_0
27
+ blk.0.ffn_up_shexp.weight: Q3_K
28
+ blk.0.hc_attn_down.weight: BF16
29
+ blk.0.hc_attn_inject.weight: BF16
30
+ blk.0.hc_attn_norm.weight: F32
31
+ blk.0.hc_attn_up.weight: BF16
32
+ blk.0.hc_ffn_down.weight: BF16
33
+ blk.0.hc_ffn_inject.weight: BF16
34
+ blk.0.hc_ffn_norm.weight: F32
35
+ blk.0.hc_ffn_up.weight: BF16
36
+ blk.0.ssm_a: F32
37
+ blk.0.ssm_alpha.weight: BF16
38
+ blk.0.ssm_beta.weight: BF16
39
+ blk.0.ssm_conv1d.weight: F32
40
+ blk.0.ssm_dt.bias: F32
41
+ blk.0.ssm_norm.weight: F32
42
+ blk.0.ssm_out.weight: Q5_K
43
+ blk.1.attn_gate.weight: IQ4_XS
44
+ blk.1.attn_qkv.weight: IQ4_XS
45
+ blk.1.ffn_down_exps.weight: Q2_0
46
+ blk.1.ffn_down_shexp.weight: Q2_0
47
+ blk.1.ffn_gate_exps.weight: Q2_0
48
+ blk.1.ffn_gate_inp.weight: BF16
49
+ blk.1.ffn_gate_inp_shexp.weight: BF16
50
+ blk.1.ffn_gate_shexp.weight: Q2_0
51
+ blk.1.ffn_up_exps.weight: Q2_0
52
+ blk.1.ffn_up_shexp.weight: Q2_0
53
+ blk.1.hc_attn_down.weight: BF16
54
+ blk.1.hc_attn_inject.weight: BF16
55
+ blk.1.hc_attn_norm.weight: F32
56
+ blk.1.hc_attn_up.weight: BF16
57
+ blk.1.hc_ffn_down.weight: BF16
58
+ blk.1.hc_ffn_inject.weight: BF16
59
+ blk.1.hc_ffn_norm.weight: F32
60
+ blk.1.hc_ffn_up.weight: BF16
61
+ blk.1.ple_conv1d.weight: F16
62
+ blk.1.ple_key.weight: Q2_0
63
+ blk.1.ple_norm_conv.weight: F32
64
+ blk.1.ple_norm_key.weight: F32
65
+ blk.1.ple_norm_query.weight: F32
66
+ blk.1.ple_value.weight: BF16
67
+ blk.1.ssm_a: F32
68
+ blk.1.ssm_alpha.weight: BF16
69
+ blk.1.ssm_beta.weight: BF16
70
+ blk.1.ssm_conv1d.weight: F32
71
+ blk.1.ssm_dt.bias: F32
72
+ blk.1.ssm_norm.weight: F32
73
+ blk.1.ssm_out.weight: Q4_K
74
+ blk.2.attn_gate.weight: IQ4_XS
75
+ blk.2.attn_qkv.weight: IQ4_XS
76
+ blk.2.ffn_down_exps.weight: Q2_0
77
+ blk.2.ffn_down_shexp.weight: Q4_0
78
+ blk.2.ffn_gate_exps.weight: Q2_0
79
+ blk.2.ffn_gate_inp.weight: BF16
80
+ blk.2.ffn_gate_inp_shexp.weight: BF16
81
+ blk.2.ffn_gate_shexp.weight: IQ4_XS
82
+ blk.2.ffn_up_exps.weight: Q2_0
83
+ blk.2.ffn_up_shexp.weight: Q2_0
84
+ blk.2.hc_attn_down.weight: BF16
85
+ blk.2.hc_attn_inject.weight: BF16
86
+ blk.2.hc_attn_norm.weight: F32
87
+ blk.2.hc_attn_up.weight: BF16
88
+ blk.2.hc_ffn_down.weight: BF16
89
+ blk.2.hc_ffn_inject.weight: BF16
90
+ blk.2.hc_ffn_norm.weight: F32
91
+ blk.2.hc_ffn_up.weight: BF16
92
+ blk.2.ssm_a: F32
93
+ blk.2.ssm_alpha.weight: BF16
94
+ blk.2.ssm_beta.weight: BF16
95
+ blk.2.ssm_conv1d.weight: F32
96
+ blk.2.ssm_dt.bias: F32
97
+ blk.2.ssm_norm.weight: F32
98
+ blk.2.ssm_out.weight: IQ4_XS
99
+ blk.3.attn_k.weight: Q3_K
100
+ blk.3.attn_k_norm.weight: F32
101
+ blk.3.attn_output.weight: Q4_K
102
+ blk.3.attn_q.weight: Q2_0
103
+ blk.3.attn_q_norm.weight: F32
104
+ blk.3.attn_v.weight: Q5_K
105
+ blk.3.ffn_down_exps.weight: Q2_0
106
+ blk.3.ffn_down_shexp.weight: Q4_0
107
+ blk.3.ffn_gate_exps.weight: Q2_0
108
+ blk.3.ffn_gate_inp.weight: BF16
109
+ blk.3.ffn_gate_inp_shexp.weight: BF16
110
+ blk.3.ffn_gate_shexp.weight: Q2_0
111
+ blk.3.ffn_up_exps.weight: Q2_0
112
+ blk.3.ffn_up_shexp.weight: Q2_0
113
+ blk.3.hc_attn_down.weight: BF16
114
+ blk.3.hc_attn_inject.weight: BF16
115
+ blk.3.hc_attn_norm.weight: F32
116
+ blk.3.hc_attn_up.weight: BF16
117
+ blk.3.hc_ffn_down.weight: BF16
118
+ blk.3.hc_ffn_inject.weight: BF16
119
+ blk.3.hc_ffn_norm.weight: F32
120
+ blk.3.hc_ffn_up.weight: BF16
121
+ blk.3.indexer.k_norm.weight: F32
122
+ blk.3.indexer.k_proj.weight: BF16
123
+ blk.3.indexer.q_norm.weight: F32
124
+ blk.3.indexer.q_proj.weight: BF16
125
+ blk.4.attn_gate.weight: Q3_K
126
+ blk.4.attn_qkv.weight: IQ4_XS
127
+ blk.4.ffn_down_exps.weight: Q2_0
128
+ blk.4.ffn_down_shexp.weight: Q2_0
129
+ blk.4.ffn_gate_exps.weight: Q2_0
130
+ blk.4.ffn_gate_inp.weight: BF16
131
+ blk.4.ffn_gate_inp_shexp.weight: BF16
132
+ blk.4.ffn_gate_shexp.weight: Q2_0
133
+ blk.4.ffn_up_exps.weight: Q2_0
134
+ blk.4.ffn_up_shexp.weight: Q2_0
135
+ blk.4.hc_attn_down.weight: BF16
136
+ blk.4.hc_attn_inject.weight: BF16
137
+ blk.4.hc_attn_norm.weight: F32
138
+ blk.4.hc_attn_up.weight: BF16
139
+ blk.4.hc_ffn_down.weight: BF16
140
+ blk.4.hc_ffn_inject.weight: BF16
141
+ blk.4.hc_ffn_norm.weight: F32
142
+ blk.4.hc_ffn_up.weight: BF16
143
+ blk.4.ssm_a: F32
144
+ blk.4.ssm_alpha.weight: BF16
145
+ blk.4.ssm_beta.weight: BF16
146
+ blk.4.ssm_conv1d.weight: F32
147
+ blk.4.ssm_dt.bias: F32
148
+ blk.4.ssm_norm.weight: F32
149
+ blk.4.ssm_out.weight: IQ4_XS
150
+ blk.5.attn_gate.weight: Q3_K
151
+ blk.5.attn_qkv.weight: Q3_K
152
+ blk.5.ffn_down_exps.weight: Q2_0
153
+ blk.5.ffn_down_shexp.weight: Q4_0
154
+ blk.5.ffn_gate_exps.weight: Q2_0
155
+ blk.5.ffn_gate_inp.weight: BF16
156
+ blk.5.ffn_gate_inp_shexp.weight: BF16
157
+ blk.5.ffn_gate_shexp.weight: Q2_0
158
+ blk.5.ffn_up_exps.weight: Q2_0
159
+ blk.5.ffn_up_shexp.weight: Q6_K
160
+ blk.5.hc_attn_down.weight: BF16
161
+ blk.5.hc_attn_inject.weight: BF16
162
+ blk.5.hc_attn_norm.weight: F32
163
+ blk.5.hc_attn_up.weight: BF16
164
+ blk.5.hc_ffn_down.weight: BF16
165
+ blk.5.hc_ffn_inject.weight: BF16
166
+ blk.5.hc_ffn_norm.weight: F32
167
+ blk.5.hc_ffn_up.weight: BF16
168
+ blk.5.ssm_a: F32
169
+ blk.5.ssm_alpha.weight: BF16
170
+ blk.5.ssm_beta.weight: BF16
171
+ blk.5.ssm_conv1d.weight: F32
172
+ blk.5.ssm_dt.bias: F32
173
+ blk.5.ssm_norm.weight: F32
174
+ blk.5.ssm_out.weight: IQ4_XS
175
+ blk.6.attn_gate.weight: Q3_K
176
+ blk.6.attn_qkv.weight: Q3_K
177
+ blk.6.ffn_down_exps.weight: Q2_0
178
+ blk.6.ffn_down_shexp.weight: Q2_0
179
+ blk.6.ffn_gate_exps.weight: Q2_0
180
+ blk.6.ffn_gate_inp.weight: BF16
181
+ blk.6.ffn_gate_inp_shexp.weight: BF16
182
+ blk.6.ffn_gate_shexp.weight: Q2_0
183
+ blk.6.ffn_up_exps.weight: Q2_0
184
+ blk.6.ffn_up_shexp.weight: Q4_K
185
+ blk.6.hc_attn_down.weight: BF16
186
+ blk.6.hc_attn_inject.weight: BF16
187
+ blk.6.hc_attn_norm.weight: F32
188
+ blk.6.hc_attn_up.weight: BF16
189
+ blk.6.hc_ffn_down.weight: BF16
190
+ blk.6.hc_ffn_inject.weight: BF16
191
+ blk.6.hc_ffn_norm.weight: F32
192
+ blk.6.hc_ffn_up.weight: BF16
193
+ blk.6.ssm_a: F32
194
+ blk.6.ssm_alpha.weight: BF16
195
+ blk.6.ssm_beta.weight: BF16
196
+ blk.6.ssm_conv1d.weight: F32
197
+ blk.6.ssm_dt.bias: F32
198
+ blk.6.ssm_norm.weight: F32
199
+ blk.6.ssm_out.weight: Q3_K
200
+ blk.7.attn_k.weight: Q4_K
201
+ blk.7.attn_k_norm.weight: F32
202
+ blk.7.attn_output.weight: Q4_K
203
+ blk.7.attn_q.weight: Q2_0
204
+ blk.7.attn_q_norm.weight: F32
205
+ blk.7.attn_v.weight: Q3_K
206
+ blk.7.ffn_down_exps.weight: Q2_0
207
+ blk.7.ffn_down_shexp.weight: Q2_0
208
+ blk.7.ffn_gate_exps.weight: Q2_0
209
+ blk.7.ffn_gate_inp.weight: BF16
210
+ blk.7.ffn_gate_inp_shexp.weight: BF16
211
+ blk.7.ffn_gate_shexp.weight: Q2_0
212
+ blk.7.ffn_up_exps.weight: Q2_0
213
+ blk.7.ffn_up_shexp.weight: Q2_0
214
+ blk.7.hc_attn_down.weight: BF16
215
+ blk.7.hc_attn_inject.weight: BF16
216
+ blk.7.hc_attn_norm.weight: F32
217
+ blk.7.hc_attn_up.weight: BF16
218
+ blk.7.hc_ffn_down.weight: BF16
219
+ blk.7.hc_ffn_inject.weight: BF16
220
+ blk.7.hc_ffn_norm.weight: F32
221
+ blk.7.hc_ffn_up.weight: BF16
222
+ blk.7.indexer.k_norm.weight: F32
223
+ blk.7.indexer.k_proj.weight: BF16
224
+ blk.7.indexer.q_norm.weight: F32
225
+ blk.7.indexer.q_proj.weight: BF16
226
+ blk.8.attn_gate.weight: Q3_K
227
+ blk.8.attn_qkv.weight: Q3_K
228
+ blk.8.ffn_down_exps.weight: Q2_0
229
+ blk.8.ffn_down_shexp.weight: Q5_0
230
+ blk.8.ffn_gate_exps.weight: Q2_0
231
+ blk.8.ffn_gate_inp.weight: BF16
232
+ blk.8.ffn_gate_inp_shexp.weight: BF16
233
+ blk.8.ffn_gate_shexp.weight: Q2_0
234
+ blk.8.ffn_up_exps.weight: Q2_0
235
+ blk.8.ffn_up_shexp.weight: Q2_0
236
+ blk.8.hc_attn_down.weight: BF16
237
+ blk.8.hc_attn_inject.weight: BF16
238
+ blk.8.hc_attn_norm.weight: F32
239
+ blk.8.hc_attn_up.weight: BF16
240
+ blk.8.hc_ffn_down.weight: BF16
241
+ blk.8.hc_ffn_inject.weight: BF16
242
+ blk.8.hc_ffn_norm.weight: F32
243
+ blk.8.hc_ffn_up.weight: BF16
244
+ blk.8.ssm_a: F32
245
+ blk.8.ssm_alpha.weight: BF16
246
+ blk.8.ssm_beta.weight: BF16
247
+ blk.8.ssm_conv1d.weight: F32
248
+ blk.8.ssm_dt.bias: F32
249
+ blk.8.ssm_norm.weight: F32
250
+ blk.8.ssm_out.weight: IQ4_XS
251
+ blk.9.attn_gate.weight: Q2_0
252
+ blk.9.attn_qkv.weight: Q2_0
253
+ blk.9.ffn_down_exps.weight: Q2_0
254
+ blk.9.ffn_down_shexp.weight: Q2_0
255
+ blk.9.ffn_gate_exps.weight: Q2_0
256
+ blk.9.ffn_gate_inp.weight: BF16
257
+ blk.9.ffn_gate_inp_shexp.weight: BF16
258
+ blk.9.ffn_gate_shexp.weight: IQ4_XS
259
+ blk.9.ffn_up_exps.weight: Q2_0
260
+ blk.9.ffn_up_shexp.weight: Q2_0
261
+ blk.9.hc_attn_down.weight: BF16
262
+ blk.9.hc_attn_inject.weight: BF16
263
+ blk.9.hc_attn_norm.weight: F32
264
+ blk.9.hc_attn_up.weight: BF16
265
+ blk.9.hc_ffn_down.weight: BF16
266
+ blk.9.hc_ffn_inject.weight: BF16
267
+ blk.9.hc_ffn_norm.weight: F32
268
+ blk.9.hc_ffn_up.weight: BF16
269
+ blk.9.ssm_a: F32
270
+ blk.9.ssm_alpha.weight: BF16
271
+ blk.9.ssm_beta.weight: BF16
272
+ blk.9.ssm_conv1d.weight: F32
273
+ blk.9.ssm_dt.bias: F32
274
+ blk.9.ssm_norm.weight: F32
275
+ blk.9.ssm_out.weight: Q3_K
276
+ blk.10.attn_gate.weight: Q2_0
277
+ blk.10.attn_qkv.weight: Q3_K
278
+ blk.10.ffn_down_exps.weight: Q2_0
279
+ blk.10.ffn_down_shexp.weight: IQ4_NL
280
+ blk.10.ffn_gate_exps.weight: Q2_0
281
+ blk.10.ffn_gate_inp.weight: BF16
282
+ blk.10.ffn_gate_inp_shexp.weight: BF16
283
+ blk.10.ffn_gate_shexp.weight: Q2_0
284
+ blk.10.ffn_up_exps.weight: Q2_0
285
+ blk.10.ffn_up_shexp.weight: Q3_K
286
+ blk.10.hc_attn_down.weight: BF16
287
+ blk.10.hc_attn_inject.weight: BF16
288
+ blk.10.hc_attn_norm.weight: F32
289
+ blk.10.hc_attn_up.weight: BF16
290
+ blk.10.hc_ffn_down.weight: BF16
291
+ blk.10.hc_ffn_inject.weight: BF16
292
+ blk.10.hc_ffn_norm.weight: F32
293
+ blk.10.hc_ffn_up.weight: BF16
294
+ blk.10.ssm_a: F32
295
+ blk.10.ssm_alpha.weight: BF16
296
+ blk.10.ssm_beta.weight: BF16
297
+ blk.10.ssm_conv1d.weight: F32
298
+ blk.10.ssm_dt.bias: F32
299
+ blk.10.ssm_norm.weight: F32
300
+ blk.10.ssm_out.weight: IQ4_XS
301
+ blk.11.attn_k.weight: IQ4_XS
302
+ blk.11.attn_k_norm.weight: F32
303
+ blk.11.attn_output.weight: Q6_K
304
+ blk.11.attn_q.weight: Q3_K
305
+ blk.11.attn_q_norm.weight: F32
306
+ blk.11.attn_v.weight: IQ4_XS
307
+ blk.11.ffn_down_exps.weight: Q2_0
308
+ blk.11.ffn_down_shexp.weight: Q2_0
309
+ blk.11.ffn_gate_exps.weight: Q2_0
310
+ blk.11.ffn_gate_inp.weight: BF16
311
+ blk.11.ffn_gate_inp_shexp.weight: BF16
312
+ blk.11.ffn_gate_shexp.weight: IQ4_XS
313
+ blk.11.ffn_up_exps.weight: Q2_0
314
+ blk.11.ffn_up_shexp.weight: Q3_K
315
+ blk.11.hc_attn_down.weight: BF16
316
+ blk.11.hc_attn_inject.weight: BF16
317
+ blk.11.hc_attn_norm.weight: F32
318
+ blk.11.hc_attn_up.weight: BF16
319
+ blk.11.hc_ffn_down.weight: BF16
320
+ blk.11.hc_ffn_inject.weight: BF16
321
+ blk.11.hc_ffn_norm.weight: F32
322
+ blk.11.hc_ffn_up.weight: BF16
323
+ blk.11.indexer.k_norm.weight: F32
324
+ blk.11.indexer.k_proj.weight: BF16
325
+ blk.11.indexer.q_norm.weight: F32
326
+ blk.11.indexer.q_proj.weight: BF16
327
+ blk.12.attn_gate.weight: Q2_0
328
+ blk.12.attn_qkv.weight: Q3_K
329
+ blk.12.ffn_down_exps.weight: Q2_0
330
+ blk.12.ffn_down_shexp.weight: Q2_0
331
+ blk.12.ffn_gate_exps.weight: Q2_0
332
+ blk.12.ffn_gate_inp.weight: BF16
333
+ blk.12.ffn_gate_inp_shexp.weight: BF16
334
+ blk.12.ffn_gate_shexp.weight: Q3_K
335
+ blk.12.ffn_up_exps.weight: Q2_0
336
+ blk.12.ffn_up_shexp.weight: Q2_0
337
+ blk.12.hc_attn_down.weight: BF16
338
+ blk.12.hc_attn_inject.weight: BF16
339
+ blk.12.hc_attn_norm.weight: F32
340
+ blk.12.hc_attn_up.weight: BF16
341
+ blk.12.hc_ffn_down.weight: BF16
342
+ blk.12.hc_ffn_inject.weight: BF16
343
+ blk.12.hc_ffn_norm.weight: F32
344
+ blk.12.hc_ffn_up.weight: BF16
345
+ blk.12.ssm_a: F32
346
+ blk.12.ssm_alpha.weight: BF16
347
+ blk.12.ssm_beta.weight: BF16
348
+ blk.12.ssm_conv1d.weight: F32
349
+ blk.12.ssm_dt.bias: F32
350
+ blk.12.ssm_norm.weight: F32
351
+ blk.12.ssm_out.weight: IQ4_XS
352
+ blk.13.attn_gate.weight: Q3_K
353
+ blk.13.attn_qkv.weight: Q3_K
354
+ blk.13.ffn_down_exps.weight: Q2_0
355
+ blk.13.ffn_down_shexp.weight: Q2_0
356
+ blk.13.ffn_gate_exps.weight: Q2_0
357
+ blk.13.ffn_gate_inp.weight: BF16
358
+ blk.13.ffn_gate_inp_shexp.weight: BF16
359
+ blk.13.ffn_gate_shexp.weight: Q4_K
360
+ blk.13.ffn_up_exps.weight: Q2_0
361
+ blk.13.ffn_up_shexp.weight: Q3_K
362
+ blk.13.hc_attn_down.weight: BF16
363
+ blk.13.hc_attn_inject.weight: BF16
364
+ blk.13.hc_attn_norm.weight: F32
365
+ blk.13.hc_attn_up.weight: BF16
366
+ blk.13.hc_ffn_down.weight: BF16
367
+ blk.13.hc_ffn_inject.weight: BF16
368
+ blk.13.hc_ffn_norm.weight: F32
369
+ blk.13.hc_ffn_up.weight: BF16
370
+ blk.13.ssm_a: F32
371
+ blk.13.ssm_alpha.weight: BF16
372
+ blk.13.ssm_beta.weight: BF16
373
+ blk.13.ssm_conv1d.weight: F32
374
+ blk.13.ssm_dt.bias: F32
375
+ blk.13.ssm_norm.weight: F32
376
+ blk.13.ssm_out.weight: Q4_K
377
+ blk.14.attn_gate.weight: Q3_K
378
+ blk.14.attn_qkv.weight: Q3_K
379
+ blk.14.ffn_down_exps.weight: Q2_0
380
+ blk.14.ffn_down_shexp.weight: Q8_0
381
+ blk.14.ffn_gate_exps.weight: Q2_0
382
+ blk.14.ffn_gate_inp.weight: BF16
383
+ blk.14.ffn_gate_inp_shexp.weight: BF16
384
+ blk.14.ffn_gate_shexp.weight: Q3_K
385
+ blk.14.ffn_up_exps.weight: Q2_0
386
+ blk.14.ffn_up_shexp.weight: Q3_K
387
+ blk.14.hc_attn_down.weight: BF16
388
+ blk.14.hc_attn_inject.weight: BF16
389
+ blk.14.hc_attn_norm.weight: F32
390
+ blk.14.hc_attn_up.weight: BF16
391
+ blk.14.hc_ffn_down.weight: BF16
392
+ blk.14.hc_ffn_inject.weight: BF16
393
+ blk.14.hc_ffn_norm.weight: F32
394
+ blk.14.hc_ffn_up.weight: BF16
395
+ blk.14.ssm_a: F32
396
+ blk.14.ssm_alpha.weight: BF16
397
+ blk.14.ssm_beta.weight: BF16
398
+ blk.14.ssm_conv1d.weight: F32
399
+ blk.14.ssm_dt.bias: F32
400
+ blk.14.ssm_norm.weight: F32
401
+ blk.14.ssm_out.weight: Q4_K
402
+ blk.15.attn_k.weight: Q5_K
403
+ blk.15.attn_k_norm.weight: F32
404
+ blk.15.attn_output.weight: Q5_K
405
+ blk.15.attn_q.weight: IQ4_XS
406
+ blk.15.attn_q_norm.weight: F32
407
+ blk.15.attn_v.weight: Q5_K
408
+ blk.15.ffn_down_exps.weight: Q2_0
409
+ blk.15.ffn_down_shexp.weight: Q5_0
410
+ blk.15.ffn_gate_exps.weight: Q2_0
411
+ blk.15.ffn_gate_inp.weight: BF16
412
+ blk.15.ffn_gate_inp_shexp.weight: BF16
413
+ blk.15.ffn_gate_shexp.weight: Q2_0
414
+ blk.15.ffn_up_exps.weight: Q2_0
415
+ blk.15.ffn_up_shexp.weight: Q3_K
416
+ blk.15.hc_attn_down.weight: BF16
417
+ blk.15.hc_attn_inject.weight: BF16
418
+ blk.15.hc_attn_norm.weight: F32
419
+ blk.15.hc_attn_up.weight: BF16
420
+ blk.15.hc_ffn_down.weight: BF16
421
+ blk.15.hc_ffn_inject.weight: BF16
422
+ blk.15.hc_ffn_norm.weight: F32
423
+ blk.15.hc_ffn_up.weight: BF16
424
+ blk.15.indexer.k_norm.weight: F32
425
+ blk.15.indexer.k_proj.weight: BF16
426
+ blk.15.indexer.q_norm.weight: F32
427
+ blk.15.indexer.q_proj.weight: BF16
428
+ blk.16.attn_gate.weight: Q2_0
429
+ blk.16.attn_qkv.weight: Q3_K
430
+ blk.16.ffn_down_exps.weight: Q2_0
431
+ blk.16.ffn_down_shexp.weight: Q4_0
432
+ blk.16.ffn_gate_exps.weight: Q2_0
433
+ blk.16.ffn_gate_inp.weight: BF16
434
+ blk.16.ffn_gate_inp_shexp.weight: BF16
435
+ blk.16.ffn_gate_shexp.weight: Q3_K
436
+ blk.16.ffn_up_exps.weight: Q2_0
437
+ blk.16.ffn_up_shexp.weight: Q3_K
438
+ blk.16.hc_attn_down.weight: BF16
439
+ blk.16.hc_attn_inject.weight: BF16
440
+ blk.16.hc_attn_norm.weight: F32
441
+ blk.16.hc_attn_up.weight: BF16
442
+ blk.16.hc_ffn_down.weight: BF16
443
+ blk.16.hc_ffn_inject.weight: BF16
444
+ blk.16.hc_ffn_norm.weight: F32
445
+ blk.16.hc_ffn_up.weight: BF16
446
+ blk.16.ssm_a: F32
447
+ blk.16.ssm_alpha.weight: BF16
448
+ blk.16.ssm_beta.weight: BF16
449
+ blk.16.ssm_conv1d.weight: F32
450
+ blk.16.ssm_dt.bias: F32
451
+ blk.16.ssm_norm.weight: F32
452
+ blk.16.ssm_out.weight: Q4_K
453
+ blk.17.attn_gate.weight: Q3_K
454
+ blk.17.attn_qkv.weight: Q3_K
455
+ blk.17.ffn_down_exps.weight: Q2_0
456
+ blk.17.ffn_down_shexp.weight: Q4_0
457
+ blk.17.ffn_gate_exps.weight: Q2_0
458
+ blk.17.ffn_gate_inp.weight: BF16
459
+ blk.17.ffn_gate_inp_shexp.weight: BF16
460
+ blk.17.ffn_gate_shexp.weight: Q2_0
461
+ blk.17.ffn_up_exps.weight: Q2_0
462
+ blk.17.ffn_up_shexp.weight: Q2_0
463
+ blk.17.hc_attn_down.weight: BF16
464
+ blk.17.hc_attn_inject.weight: BF16
465
+ blk.17.hc_attn_norm.weight: F32
466
+ blk.17.hc_attn_up.weight: BF16
467
+ blk.17.hc_ffn_down.weight: BF16
468
+ blk.17.hc_ffn_inject.weight: BF16
469
+ blk.17.hc_ffn_norm.weight: F32
470
+ blk.17.hc_ffn_up.weight: BF16
471
+ blk.17.ssm_a: F32
472
+ blk.17.ssm_alpha.weight: BF16
473
+ blk.17.ssm_beta.weight: BF16
474
+ blk.17.ssm_conv1d.weight: F32
475
+ blk.17.ssm_dt.bias: F32
476
+ blk.17.ssm_norm.weight: F32
477
+ blk.17.ssm_out.weight: Q3_K
478
+ blk.18.attn_gate.weight: IQ4_XS
479
+ blk.18.attn_qkv.weight: Q3_K
480
+ blk.18.ffn_down_exps.weight: Q2_0
481
+ blk.18.ffn_down_shexp.weight: Q4_0
482
+ blk.18.ffn_gate_exps.weight: Q2_0
483
+ blk.18.ffn_gate_inp.weight: BF16
484
+ blk.18.ffn_gate_inp_shexp.weight: BF16
485
+ blk.18.ffn_gate_shexp.weight: Q2_0
486
+ blk.18.ffn_up_exps.weight: Q2_0
487
+ blk.18.ffn_up_shexp.weight: IQ4_XS
488
+ blk.18.hc_attn_down.weight: BF16
489
+ blk.18.hc_attn_inject.weight: BF16
490
+ blk.18.hc_attn_norm.weight: F32
491
+ blk.18.hc_attn_up.weight: BF16
492
+ blk.18.hc_ffn_down.weight: BF16
493
+ blk.18.hc_ffn_inject.weight: BF16
494
+ blk.18.hc_ffn_norm.weight: F32
495
+ blk.18.hc_ffn_up.weight: BF16
496
+ blk.18.ssm_a: F32
497
+ blk.18.ssm_alpha.weight: BF16
498
+ blk.18.ssm_beta.weight: BF16
499
+ blk.18.ssm_conv1d.weight: F32
500
+ blk.18.ssm_dt.bias: F32
501
+ blk.18.ssm_norm.weight: F32
502
+ blk.18.ssm_out.weight: Q4_K
503
+ blk.19.attn_k.weight: Q3_K
504
+ blk.19.attn_k_norm.weight: F32
505
+ blk.19.attn_output.weight: Q4_K
506
+ blk.19.attn_q.weight: Q3_K
507
+ blk.19.attn_q_norm.weight: F32
508
+ blk.19.attn_v.weight: Q4_K
509
+ blk.19.ffn_down_exps.weight: Q2_0
510
+ blk.19.ffn_down_shexp.weight: Q4_0
511
+ blk.19.ffn_gate_exps.weight: Q2_0
512
+ blk.19.ffn_gate_inp.weight: BF16
513
+ blk.19.ffn_gate_inp_shexp.weight: BF16
514
+ blk.19.ffn_gate_shexp.weight: Q2_0
515
+ blk.19.ffn_up_exps.weight: Q2_0
516
+ blk.19.ffn_up_shexp.weight: Q3_K
517
+ blk.19.hc_attn_down.weight: BF16
518
+ blk.19.hc_attn_inject.weight: BF16
519
+ blk.19.hc_attn_norm.weight: F32
520
+ blk.19.hc_attn_up.weight: BF16
521
+ blk.19.hc_ffn_down.weight: BF16
522
+ blk.19.hc_ffn_inject.weight: BF16
523
+ blk.19.hc_ffn_norm.weight: F32
524
+ blk.19.hc_ffn_up.weight: BF16
525
+ blk.19.indexer.k_norm.weight: F32
526
+ blk.19.indexer.k_proj.weight: BF16
527
+ blk.19.indexer.q_norm.weight: F32
528
+ blk.19.indexer.q_proj.weight: BF16
529
+ blk.20.attn_gate.weight: Q3_K
530
+ blk.20.attn_qkv.weight: Q3_K
531
+ blk.20.ffn_down_exps.weight: Q2_0
532
+ blk.20.ffn_down_shexp.weight: Q2_0
533
+ blk.20.ffn_gate_exps.weight: Q2_0
534
+ blk.20.ffn_gate_inp.weight: BF16
535
+ blk.20.ffn_gate_inp_shexp.weight: BF16
536
+ blk.20.ffn_gate_shexp.weight: Q3_K
537
+ blk.20.ffn_up_exps.weight: Q2_0
538
+ blk.20.ffn_up_shexp.weight: Q2_0
539
+ blk.20.hc_attn_down.weight: BF16
540
+ blk.20.hc_attn_inject.weight: BF16
541
+ blk.20.hc_attn_norm.weight: F32
542
+ blk.20.hc_attn_up.weight: BF16
543
+ blk.20.hc_ffn_down.weight: BF16
544
+ blk.20.hc_ffn_inject.weight: BF16
545
+ blk.20.hc_ffn_norm.weight: F32
546
+ blk.20.hc_ffn_up.weight: BF16
547
+ blk.20.ssm_a: F32
548
+ blk.20.ssm_alpha.weight: BF16
549
+ blk.20.ssm_beta.weight: BF16
550
+ blk.20.ssm_conv1d.weight: F32
551
+ blk.20.ssm_dt.bias: F32
552
+ blk.20.ssm_norm.weight: F32
553
+ blk.20.ssm_out.weight: IQ4_XS
554
+ blk.21.attn_gate.weight: Q3_K
555
+ blk.21.attn_qkv.weight: Q3_K
556
+ blk.21.ffn_down_exps.weight: Q2_0
557
+ blk.21.ffn_down_shexp.weight: Q4_0
558
+ blk.21.ffn_gate_exps.weight: Q2_0
559
+ blk.21.ffn_gate_inp.weight: BF16
560
+ blk.21.ffn_gate_inp_shexp.weight: BF16
561
+ blk.21.ffn_gate_shexp.weight: Q2_0
562
+ blk.21.ffn_up_exps.weight: Q2_0
563
+ blk.21.ffn_up_shexp.weight: Q4_K
564
+ blk.21.hc_attn_down.weight: BF16
565
+ blk.21.hc_attn_inject.weight: BF16
566
+ blk.21.hc_attn_norm.weight: F32
567
+ blk.21.hc_attn_up.weight: BF16
568
+ blk.21.hc_ffn_down.weight: BF16
569
+ blk.21.hc_ffn_inject.weight: BF16
570
+ blk.21.hc_ffn_norm.weight: F32
571
+ blk.21.hc_ffn_up.weight: BF16
572
+ blk.21.ssm_a: F32
573
+ blk.21.ssm_alpha.weight: BF16
574
+ blk.21.ssm_beta.weight: BF16
575
+ blk.21.ssm_conv1d.weight: F32
576
+ blk.21.ssm_dt.bias: F32
577
+ blk.21.ssm_norm.weight: F32
578
+ blk.21.ssm_out.weight: Q4_K
579
+ blk.22.attn_gate.weight: Q3_K
580
+ blk.22.attn_qkv.weight: Q3_K
581
+ blk.22.ffn_down_exps.weight: Q2_0
582
+ blk.22.ffn_down_shexp.weight: Q2_0
583
+ blk.22.ffn_gate_exps.weight: Q2_0
584
+ blk.22.ffn_gate_inp.weight: BF16
585
+ blk.22.ffn_gate_inp_shexp.weight: BF16
586
+ blk.22.ffn_gate_shexp.weight: Q2_0
587
+ blk.22.ffn_up_exps.weight: Q2_0
588
+ blk.22.ffn_up_shexp.weight: IQ4_XS
589
+ blk.22.hc_attn_down.weight: BF16
590
+ blk.22.hc_attn_inject.weight: BF16
591
+ blk.22.hc_attn_norm.weight: F32
592
+ blk.22.hc_attn_up.weight: BF16
593
+ blk.22.hc_ffn_down.weight: BF16
594
+ blk.22.hc_ffn_inject.weight: BF16
595
+ blk.22.hc_ffn_norm.weight: F32
596
+ blk.22.hc_ffn_up.weight: BF16
597
+ blk.22.ssm_a: F32
598
+ blk.22.ssm_alpha.weight: BF16
599
+ blk.22.ssm_beta.weight: BF16
600
+ blk.22.ssm_conv1d.weight: F32
601
+ blk.22.ssm_dt.bias: F32
602
+ blk.22.ssm_norm.weight: F32
603
+ blk.22.ssm_out.weight: Q4_K
604
+ blk.23.attn_k.weight: IQ4_XS
605
+ blk.23.attn_k_norm.weight: F32
606
+ blk.23.attn_output.weight: Q4_K
607
+ blk.23.attn_q.weight: Q3_K
608
+ blk.23.attn_q_norm.weight: F32
609
+ blk.23.attn_v.weight: Q4_K
610
+ blk.23.ffn_down_exps.weight: Q2_0
611
+ blk.23.ffn_down_shexp.weight: IQ4_NL
612
+ blk.23.ffn_gate_exps.weight: Q2_0
613
+ blk.23.ffn_gate_inp.weight: BF16
614
+ blk.23.ffn_gate_inp_shexp.weight: BF16
615
+ blk.23.ffn_gate_shexp.weight: Q4_K
616
+ blk.23.ffn_up_exps.weight: Q2_0
617
+ blk.23.ffn_up_shexp.weight: Q4_K
618
+ blk.23.hc_attn_down.weight: BF16
619
+ blk.23.hc_attn_inject.weight: BF16
620
+ blk.23.hc_attn_norm.weight: F32
621
+ blk.23.hc_attn_up.weight: BF16
622
+ blk.23.hc_ffn_down.weight: BF16
623
+ blk.23.hc_ffn_inject.weight: BF16
624
+ blk.23.hc_ffn_norm.weight: F32
625
+ blk.23.hc_ffn_up.weight: BF16
626
+ blk.23.indexer.k_norm.weight: F32
627
+ blk.23.indexer.k_proj.weight: BF16
628
+ blk.23.indexer.q_norm.weight: F32
629
+ blk.23.indexer.q_proj.weight: BF16
630
+ blk.24.attn_gate.weight: Q3_K
631
+ blk.24.attn_qkv.weight: IQ4_XS
632
+ blk.24.ffn_down_exps.weight: Q2_0
633
+ blk.24.ffn_down_shexp.weight: Q5_0
634
+ blk.24.ffn_gate_exps.weight: Q2_0
635
+ blk.24.ffn_gate_inp.weight: BF16
636
+ blk.24.ffn_gate_inp_shexp.weight: BF16
637
+ blk.24.ffn_gate_shexp.weight: Q3_K
638
+ blk.24.ffn_up_exps.weight: Q2_0
639
+ blk.24.ffn_up_shexp.weight: Q5_K
640
+ blk.24.hc_attn_down.weight: BF16
641
+ blk.24.hc_attn_inject.weight: BF16
642
+ blk.24.hc_attn_norm.weight: F32
643
+ blk.24.hc_attn_up.weight: BF16
644
+ blk.24.hc_ffn_down.weight: BF16
645
+ blk.24.hc_ffn_inject.weight: BF16
646
+ blk.24.hc_ffn_norm.weight: F32
647
+ blk.24.hc_ffn_up.weight: BF16
648
+ blk.24.ssm_a: F32
649
+ blk.24.ssm_alpha.weight: BF16
650
+ blk.24.ssm_beta.weight: BF16
651
+ blk.24.ssm_conv1d.weight: F32
652
+ blk.24.ssm_dt.bias: F32
653
+ blk.24.ssm_norm.weight: F32
654
+ blk.24.ssm_out.weight: IQ4_XS
655
+ blk.25.attn_gate.weight: Q3_K
656
+ blk.25.attn_qkv.weight: Q3_K
657
+ blk.25.ffn_down_exps.weight: Q2_0
658
+ blk.25.ffn_down_shexp.weight: Q2_0
659
+ blk.25.ffn_gate_exps.weight: Q2_0
660
+ blk.25.ffn_gate_inp.weight: BF16
661
+ blk.25.ffn_gate_inp_shexp.weight: BF16
662
+ blk.25.ffn_gate_shexp.weight: Q2_0
663
+ blk.25.ffn_up_exps.weight: Q2_0
664
+ blk.25.ffn_up_shexp.weight: Q3_K
665
+ blk.25.hc_attn_down.weight: BF16
666
+ blk.25.hc_attn_inject.weight: BF16
667
+ blk.25.hc_attn_norm.weight: F32
668
+ blk.25.hc_attn_up.weight: BF16
669
+ blk.25.hc_ffn_down.weight: BF16
670
+ blk.25.hc_ffn_inject.weight: BF16
671
+ blk.25.hc_ffn_norm.weight: F32
672
+ blk.25.hc_ffn_up.weight: BF16
673
+ blk.25.ssm_a: F32
674
+ blk.25.ssm_alpha.weight: BF16
675
+ blk.25.ssm_beta.weight: BF16
676
+ blk.25.ssm_conv1d.weight: F32
677
+ blk.25.ssm_dt.bias: F32
678
+ blk.25.ssm_norm.weight: F32
679
+ blk.25.ssm_out.weight: IQ4_XS
680
+ blk.26.attn_gate.weight: Q3_K
681
+ blk.26.attn_qkv.weight: Q4_K
682
+ blk.26.ffn_down_exps.weight: Q2_0
683
+ blk.26.ffn_down_shexp.weight: IQ4_NL
684
+ blk.26.ffn_gate_exps.weight: Q2_0
685
+ blk.26.ffn_gate_inp.weight: BF16
686
+ blk.26.ffn_gate_inp_shexp.weight: BF16
687
+ blk.26.ffn_gate_shexp.weight: Q2_0
688
+ blk.26.ffn_up_exps.weight: Q2_0
689
+ blk.26.ffn_up_shexp.weight: Q3_K
690
+ blk.26.hc_attn_down.weight: BF16
691
+ blk.26.hc_attn_inject.weight: BF16
692
+ blk.26.hc_attn_norm.weight: F32
693
+ blk.26.hc_attn_up.weight: BF16
694
+ blk.26.hc_ffn_down.weight: BF16
695
+ blk.26.hc_ffn_inject.weight: BF16
696
+ blk.26.hc_ffn_norm.weight: F32
697
+ blk.26.hc_ffn_up.weight: BF16
698
+ blk.26.ssm_a: F32
699
+ blk.26.ssm_alpha.weight: BF16
700
+ blk.26.ssm_beta.weight: BF16
701
+ blk.26.ssm_conv1d.weight: F32
702
+ blk.26.ssm_dt.bias: F32
703
+ blk.26.ssm_norm.weight: F32
704
+ blk.26.ssm_out.weight: Q5_K
705
+ blk.27.attn_k.weight: IQ4_XS
706
+ blk.27.attn_k_norm.weight: F32
707
+ blk.27.attn_output.weight: Q6_K
708
+ blk.27.attn_q.weight: Q3_K
709
+ blk.27.attn_q_norm.weight: F32
710
+ blk.27.attn_v.weight: IQ4_XS
711
+ blk.27.ffn_down_exps.weight: Q2_0
712
+ blk.27.ffn_down_shexp.weight: Q2_0
713
+ blk.27.ffn_gate_exps.weight: Q2_0
714
+ blk.27.ffn_gate_inp.weight: BF16
715
+ blk.27.ffn_gate_inp_shexp.weight: BF16
716
+ blk.27.ffn_gate_shexp.weight: Q3_K
717
+ blk.27.ffn_up_exps.weight: Q2_0
718
+ blk.27.ffn_up_shexp.weight: Q2_0
719
+ blk.27.hc_attn_down.weight: BF16
720
+ blk.27.hc_attn_inject.weight: BF16
721
+ blk.27.hc_attn_norm.weight: F32
722
+ blk.27.hc_attn_up.weight: BF16
723
+ blk.27.hc_ffn_down.weight: BF16
724
+ blk.27.hc_ffn_inject.weight: BF16
725
+ blk.27.hc_ffn_norm.weight: F32
726
+ blk.27.hc_ffn_up.weight: BF16
727
+ blk.27.indexer.k_norm.weight: F32
728
+ blk.27.indexer.k_proj.weight: BF16
729
+ blk.27.indexer.q_norm.weight: F32
730
+ blk.27.indexer.q_proj.weight: BF16
731
+ blk.28.attn_gate.weight: Q3_K
732
+ blk.28.attn_qkv.weight: Q3_K
733
+ blk.28.ffn_down_exps.weight: Q2_0
734
+ blk.28.ffn_down_shexp.weight: Q4_0
735
+ blk.28.ffn_gate_exps.weight: Q2_0
736
+ blk.28.ffn_gate_inp.weight: BF16
737
+ blk.28.ffn_gate_inp_shexp.weight: BF16
738
+ blk.28.ffn_gate_shexp.weight: Q6_K
739
+ blk.28.ffn_up_exps.weight: Q2_0
740
+ blk.28.ffn_up_shexp.weight: Q3_K
741
+ blk.28.hc_attn_down.weight: BF16
742
+ blk.28.hc_attn_inject.weight: BF16
743
+ blk.28.hc_attn_norm.weight: F32
744
+ blk.28.hc_attn_up.weight: BF16
745
+ blk.28.hc_ffn_down.weight: BF16
746
+ blk.28.hc_ffn_inject.weight: BF16
747
+ blk.28.hc_ffn_norm.weight: F32
748
+ blk.28.hc_ffn_up.weight: BF16
749
+ blk.28.ssm_a: F32
750
+ blk.28.ssm_alpha.weight: BF16
751
+ blk.28.ssm_beta.weight: BF16
752
+ blk.28.ssm_conv1d.weight: F32
753
+ blk.28.ssm_dt.bias: F32
754
+ blk.28.ssm_norm.weight: F32
755
+ blk.28.ssm_out.weight: IQ4_XS
756
+ blk.29.attn_gate.weight: Q3_K
757
+ blk.29.attn_qkv.weight: Q3_K
758
+ blk.29.ffn_down_exps.weight: Q2_0
759
+ blk.29.ffn_down_shexp.weight: Q5_0
760
+ blk.29.ffn_gate_exps.weight: Q2_0
761
+ blk.29.ffn_gate_inp.weight: BF16
762
+ blk.29.ffn_gate_inp_shexp.weight: BF16
763
+ blk.29.ffn_gate_shexp.weight: Q2_0
764
+ blk.29.ffn_up_exps.weight: Q2_0
765
+ blk.29.ffn_up_shexp.weight: Q3_K
766
+ blk.29.hc_attn_down.weight: BF16
767
+ blk.29.hc_attn_inject.weight: BF16
768
+ blk.29.hc_attn_norm.weight: F32
769
+ blk.29.hc_attn_up.weight: BF16
770
+ blk.29.hc_ffn_down.weight: BF16
771
+ blk.29.hc_ffn_inject.weight: BF16
772
+ blk.29.hc_ffn_norm.weight: F32
773
+ blk.29.hc_ffn_up.weight: BF16
774
+ blk.29.ssm_a: F32
775
+ blk.29.ssm_alpha.weight: BF16
776
+ blk.29.ssm_beta.weight: BF16
777
+ blk.29.ssm_conv1d.weight: F32
778
+ blk.29.ssm_dt.bias: F32
779
+ blk.29.ssm_norm.weight: F32
780
+ blk.29.ssm_out.weight: IQ4_XS
781
+ blk.30.attn_gate.weight: Q3_K
782
+ blk.30.attn_qkv.weight: IQ4_XS
783
+ blk.30.ffn_down_exps.weight: Q2_0
784
+ blk.30.ffn_down_shexp.weight: Q2_0
785
+ blk.30.ffn_gate_exps.weight: Q2_0
786
+ blk.30.ffn_gate_inp.weight: BF16
787
+ blk.30.ffn_gate_inp_shexp.weight: BF16
788
+ blk.30.ffn_gate_shexp.weight: Q3_K
789
+ blk.30.ffn_up_exps.weight: Q2_0
790
+ blk.30.ffn_up_shexp.weight: IQ4_XS
791
+ blk.30.hc_attn_down.weight: BF16
792
+ blk.30.hc_attn_inject.weight: BF16
793
+ blk.30.hc_attn_norm.weight: F32
794
+ blk.30.hc_attn_up.weight: BF16
795
+ blk.30.hc_ffn_down.weight: BF16
796
+ blk.30.hc_ffn_inject.weight: BF16
797
+ blk.30.hc_ffn_norm.weight: F32
798
+ blk.30.hc_ffn_up.weight: BF16
799
+ blk.30.ssm_a: F32
800
+ blk.30.ssm_alpha.weight: BF16
801
+ blk.30.ssm_beta.weight: BF16
802
+ blk.30.ssm_conv1d.weight: F32
803
+ blk.30.ssm_dt.bias: F32
804
+ blk.30.ssm_norm.weight: F32
805
+ blk.30.ssm_out.weight: IQ4_XS
806
+ blk.31.attn_k.weight: Q4_K
807
+ blk.31.attn_k_norm.weight: F32
808
+ blk.31.attn_output.weight: Q6_K
809
+ blk.31.attn_q.weight: IQ4_XS
810
+ blk.31.attn_q_norm.weight: F32
811
+ blk.31.attn_v.weight: Q6_K
812
+ blk.31.ffn_down_exps.weight: Q2_0
813
+ blk.31.ffn_down_shexp.weight: Q4_0
814
+ blk.31.ffn_gate_exps.weight: Q2_0
815
+ blk.31.ffn_gate_inp.weight: BF16
816
+ blk.31.ffn_gate_inp_shexp.weight: BF16
817
+ blk.31.ffn_gate_shexp.weight: Q4_K
818
+ blk.31.ffn_up_exps.weight: Q2_0
819
+ blk.31.ffn_up_shexp.weight: Q2_0
820
+ blk.31.hc_attn_down.weight: BF16
821
+ blk.31.hc_attn_inject.weight: BF16
822
+ blk.31.hc_attn_norm.weight: F32
823
+ blk.31.hc_attn_up.weight: BF16
824
+ blk.31.hc_ffn_down.weight: BF16
825
+ blk.31.hc_ffn_inject.weight: BF16
826
+ blk.31.hc_ffn_norm.weight: F32
827
+ blk.31.hc_ffn_up.weight: BF16
828
+ blk.31.indexer.k_norm.weight: F32
829
+ blk.31.indexer.k_proj.weight: BF16
830
+ blk.31.indexer.q_norm.weight: F32
831
+ blk.31.indexer.q_proj.weight: BF16
832
+ blk.32.attn_gate.weight: Q3_K
833
+ blk.32.attn_qkv.weight: IQ4_XS
834
+ blk.32.ffn_down_exps.weight: Q2_0
835
+ blk.32.ffn_down_shexp.weight: Q4_0
836
+ blk.32.ffn_gate_exps.weight: Q2_0
837
+ blk.32.ffn_gate_inp.weight: BF16
838
+ blk.32.ffn_gate_inp_shexp.weight: BF16
839
+ blk.32.ffn_gate_shexp.weight: Q2_0
840
+ blk.32.ffn_up_exps.weight: Q2_0
841
+ blk.32.ffn_up_shexp.weight: Q4_K
842
+ blk.32.hc_attn_down.weight: BF16
843
+ blk.32.hc_attn_inject.weight: BF16
844
+ blk.32.hc_attn_norm.weight: F32
845
+ blk.32.hc_attn_up.weight: BF16
846
+ blk.32.hc_ffn_down.weight: BF16
847
+ blk.32.hc_ffn_inject.weight: BF16
848
+ blk.32.hc_ffn_norm.weight: F32
849
+ blk.32.hc_ffn_up.weight: BF16
850
+ blk.32.ssm_a: F32
851
+ blk.32.ssm_alpha.weight: BF16
852
+ blk.32.ssm_beta.weight: BF16
853
+ blk.32.ssm_conv1d.weight: F32
854
+ blk.32.ssm_dt.bias: F32
855
+ blk.32.ssm_norm.weight: F32
856
+ blk.32.ssm_out.weight: IQ4_XS
857
+ blk.33.attn_gate.weight: Q3_K
858
+ blk.33.attn_qkv.weight: Q3_K
859
+ blk.33.ffn_down_exps.weight: Q2_0
860
+ blk.33.ffn_down_shexp.weight: Q2_0
861
+ blk.33.ffn_gate_exps.weight: Q2_0
862
+ blk.33.ffn_gate_inp.weight: BF16
863
+ blk.33.ffn_gate_inp_shexp.weight: BF16
864
+ blk.33.ffn_gate_shexp.weight: Q3_K
865
+ blk.33.ffn_up_exps.weight: Q2_0
866
+ blk.33.ffn_up_shexp.weight: Q3_K
867
+ blk.33.hc_attn_down.weight: BF16
868
+ blk.33.hc_attn_inject.weight: BF16
869
+ blk.33.hc_attn_norm.weight: F32
870
+ blk.33.hc_attn_up.weight: BF16
871
+ blk.33.hc_ffn_down.weight: BF16
872
+ blk.33.hc_ffn_inject.weight: BF16
873
+ blk.33.hc_ffn_norm.weight: F32
874
+ blk.33.hc_ffn_up.weight: BF16
875
+ blk.33.ssm_a: F32
876
+ blk.33.ssm_alpha.weight: BF16
877
+ blk.33.ssm_beta.weight: BF16
878
+ blk.33.ssm_conv1d.weight: F32
879
+ blk.33.ssm_dt.bias: F32
880
+ blk.33.ssm_norm.weight: F32
881
+ blk.33.ssm_out.weight: IQ4_XS
882
+ blk.34.attn_gate.weight: Q3_K
883
+ blk.34.attn_qkv.weight: IQ4_XS
884
+ blk.34.ffn_down_exps.weight: Q2_0
885
+ blk.34.ffn_down_shexp.weight: Q4_0
886
+ blk.34.ffn_gate_exps.weight: Q2_0
887
+ blk.34.ffn_gate_inp.weight: BF16
888
+ blk.34.ffn_gate_inp_shexp.weight: BF16
889
+ blk.34.ffn_gate_shexp.weight: Q4_K
890
+ blk.34.ffn_up_exps.weight: Q2_0
891
+ blk.34.ffn_up_shexp.weight: Q2_0
892
+ blk.34.hc_attn_down.weight: BF16
893
+ blk.34.hc_attn_inject.weight: BF16
894
+ blk.34.hc_attn_norm.weight: F32
895
+ blk.34.hc_attn_up.weight: BF16
896
+ blk.34.hc_ffn_down.weight: BF16
897
+ blk.34.hc_ffn_inject.weight: BF16
898
+ blk.34.hc_ffn_norm.weight: F32
899
+ blk.34.hc_ffn_up.weight: BF16
900
+ blk.34.ssm_a: F32
901
+ blk.34.ssm_alpha.weight: BF16
902
+ blk.34.ssm_beta.weight: BF16
903
+ blk.34.ssm_conv1d.weight: F32
904
+ blk.34.ssm_dt.bias: F32
905
+ blk.34.ssm_norm.weight: F32
906
+ blk.34.ssm_out.weight: Q5_K
907
+ blk.35.attn_k.weight: Q3_K
908
+ blk.35.attn_k_norm.weight: F32
909
+ blk.35.attn_output.weight: Q5_K
910
+ blk.35.attn_q.weight: Q3_K
911
+ blk.35.attn_q_norm.weight: F32
912
+ blk.35.attn_v.weight: IQ4_XS
913
+ blk.35.ffn_down_exps.weight: Q2_0
914
+ blk.35.ffn_down_shexp.weight: Q2_0
915
+ blk.35.ffn_gate_exps.weight: Q2_0
916
+ blk.35.ffn_gate_inp.weight: BF16
917
+ blk.35.ffn_gate_inp_shexp.weight: BF16
918
+ blk.35.ffn_gate_shexp.weight: Q2_0
919
+ blk.35.ffn_up_exps.weight: Q2_0
920
+ blk.35.ffn_up_shexp.weight: Q3_K
921
+ blk.35.hc_attn_down.weight: BF16
922
+ blk.35.hc_attn_inject.weight: BF16
923
+ blk.35.hc_attn_norm.weight: F32
924
+ blk.35.hc_attn_up.weight: BF16
925
+ blk.35.hc_ffn_down.weight: BF16
926
+ blk.35.hc_ffn_inject.weight: BF16
927
+ blk.35.hc_ffn_norm.weight: F32
928
+ blk.35.hc_ffn_up.weight: BF16
929
+ blk.35.indexer.k_norm.weight: F32
930
+ blk.35.indexer.k_proj.weight: BF16
931
+ blk.35.indexer.q_norm.weight: F32
932
+ blk.35.indexer.q_proj.weight: BF16
933
+ blk.36.attn_gate.weight: Q3_K
934
+ blk.36.attn_qkv.weight: IQ4_XS
935
+ blk.36.ffn_down_exps.weight: Q2_0
936
+ blk.36.ffn_down_shexp.weight: Q2_0
937
+ blk.36.ffn_gate_exps.weight: Q2_0
938
+ blk.36.ffn_gate_inp.weight: BF16
939
+ blk.36.ffn_gate_inp_shexp.weight: BF16
940
+ blk.36.ffn_gate_shexp.weight: Q4_K
941
+ blk.36.ffn_up_exps.weight: Q2_0
942
+ blk.36.ffn_up_shexp.weight: Q4_K
943
+ blk.36.hc_attn_down.weight: BF16
944
+ blk.36.hc_attn_inject.weight: BF16
945
+ blk.36.hc_attn_norm.weight: F32
946
+ blk.36.hc_attn_up.weight: BF16
947
+ blk.36.hc_ffn_down.weight: BF16
948
+ blk.36.hc_ffn_inject.weight: BF16
949
+ blk.36.hc_ffn_norm.weight: F32
950
+ blk.36.hc_ffn_up.weight: BF16
951
+ blk.36.ssm_a: F32
952
+ blk.36.ssm_alpha.weight: BF16
953
+ blk.36.ssm_beta.weight: BF16
954
+ blk.36.ssm_conv1d.weight: F32
955
+ blk.36.ssm_dt.bias: F32
956
+ blk.36.ssm_norm.weight: F32
957
+ blk.36.ssm_out.weight: Q4_K
958
+ blk.37.attn_gate.weight: Q3_K
959
+ blk.37.attn_qkv.weight: Q4_K
960
+ blk.37.ffn_down_exps.weight: Q2_0
961
+ blk.37.ffn_down_shexp.weight: Q5_0
962
+ blk.37.ffn_gate_exps.weight: Q2_0
963
+ blk.37.ffn_gate_inp.weight: BF16
964
+ blk.37.ffn_gate_inp_shexp.weight: BF16
965
+ blk.37.ffn_gate_shexp.weight: IQ4_XS
966
+ blk.37.ffn_up_exps.weight: Q2_0
967
+ blk.37.ffn_up_shexp.weight: Q4_K
968
+ blk.37.hc_attn_down.weight: BF16
969
+ blk.37.hc_attn_inject.weight: BF16
970
+ blk.37.hc_attn_norm.weight: F32
971
+ blk.37.hc_attn_up.weight: BF16
972
+ blk.37.hc_ffn_down.weight: BF16
973
+ blk.37.hc_ffn_inject.weight: BF16
974
+ blk.37.hc_ffn_norm.weight: F32
975
+ blk.37.hc_ffn_up.weight: BF16
976
+ blk.37.ssm_a: F32
977
+ blk.37.ssm_alpha.weight: BF16
978
+ blk.37.ssm_beta.weight: BF16
979
+ blk.37.ssm_conv1d.weight: F32
980
+ blk.37.ssm_dt.bias: F32
981
+ blk.37.ssm_norm.weight: F32
982
+ blk.37.ssm_out.weight: IQ4_XS
983
+ blk.38.attn_gate.weight: Q3_K
984
+ blk.38.attn_qkv.weight: IQ4_XS
985
+ blk.38.ffn_down_exps.weight: Q2_0
986
+ blk.38.ffn_down_shexp.weight: Q4_0
987
+ blk.38.ffn_gate_exps.weight: Q2_0
988
+ blk.38.ffn_gate_inp.weight: BF16
989
+ blk.38.ffn_gate_inp_shexp.weight: BF16
990
+ blk.38.ffn_gate_shexp.weight: IQ4_XS
991
+ blk.38.ffn_up_exps.weight: Q2_0
992
+ blk.38.ffn_up_shexp.weight: Q3_K
993
+ blk.38.hc_attn_down.weight: BF16
994
+ blk.38.hc_attn_inject.weight: BF16
995
+ blk.38.hc_attn_norm.weight: F32
996
+ blk.38.hc_attn_up.weight: BF16
997
+ blk.38.hc_ffn_down.weight: BF16
998
+ blk.38.hc_ffn_inject.weight: BF16
999
+ blk.38.hc_ffn_norm.weight: F32
1000
+ blk.38.hc_ffn_up.weight: BF16
1001
+ blk.38.ssm_a: F32
1002
+ blk.38.ssm_alpha.weight: BF16
1003
+ blk.38.ssm_beta.weight: BF16
1004
+ blk.38.ssm_conv1d.weight: F32
1005
+ blk.38.ssm_dt.bias: F32
1006
+ blk.38.ssm_norm.weight: F32
1007
+ blk.38.ssm_out.weight: Q6_K
1008
+ blk.39.attn_k.weight: Q5_K
1009
+ blk.39.attn_k_norm.weight: F32
1010
+ blk.39.attn_output.weight: Q5_K
1011
+ blk.39.attn_q.weight: Q3_K
1012
+ blk.39.attn_q_norm.weight: F32
1013
+ blk.39.attn_v.weight: Q6_K
1014
+ blk.39.ffn_down_exps.weight: Q2_0
1015
+ blk.39.ffn_down_shexp.weight: IQ4_NL
1016
+ blk.39.ffn_gate_exps.weight: Q2_0
1017
+ blk.39.ffn_gate_inp.weight: BF16
1018
+ blk.39.ffn_gate_inp_shexp.weight: BF16
1019
+ blk.39.ffn_gate_shexp.weight: Q3_K
1020
+ blk.39.ffn_up_exps.weight: Q2_0
1021
+ blk.39.ffn_up_shexp.weight: IQ4_XS
1022
+ blk.39.hc_attn_down.weight: BF16
1023
+ blk.39.hc_attn_inject.weight: BF16
1024
+ blk.39.hc_attn_norm.weight: F32
1025
+ blk.39.hc_attn_up.weight: BF16
1026
+ blk.39.hc_ffn_down.weight: BF16
1027
+ blk.39.hc_ffn_inject.weight: BF16
1028
+ blk.39.hc_ffn_norm.weight: F32
1029
+ blk.39.hc_ffn_up.weight: BF16
1030
+ blk.39.indexer.k_norm.weight: F32
1031
+ blk.39.indexer.k_proj.weight: BF16
1032
+ blk.39.indexer.q_norm.weight: F32
1033
+ blk.39.indexer.q_proj.weight: BF16
1034
+ blk.40.attn_gate.weight: IQ4_XS
1035
+ blk.40.attn_qkv.weight: Q3_K
1036
+ blk.40.ffn_down_exps.weight: Q2_0
1037
+ blk.40.ffn_down_shexp.weight: IQ4_NL
1038
+ blk.40.ffn_gate_exps.weight: Q2_0
1039
+ blk.40.ffn_gate_inp.weight: BF16
1040
+ blk.40.ffn_gate_inp_shexp.weight: BF16
1041
+ blk.40.ffn_gate_shexp.weight: Q4_K
1042
+ blk.40.ffn_up_exps.weight: Q2_0
1043
+ blk.40.ffn_up_shexp.weight: Q3_K
1044
+ blk.40.hc_attn_down.weight: BF16
1045
+ blk.40.hc_attn_inject.weight: BF16
1046
+ blk.40.hc_attn_norm.weight: F32
1047
+ blk.40.hc_attn_up.weight: BF16
1048
+ blk.40.hc_ffn_down.weight: BF16
1049
+ blk.40.hc_ffn_inject.weight: BF16
1050
+ blk.40.hc_ffn_norm.weight: F32
1051
+ blk.40.hc_ffn_up.weight: BF16
1052
+ blk.40.ssm_a: F32
1053
+ blk.40.ssm_alpha.weight: BF16
1054
+ blk.40.ssm_beta.weight: BF16
1055
+ blk.40.ssm_conv1d.weight: F32
1056
+ blk.40.ssm_dt.bias: F32
1057
+ blk.40.ssm_norm.weight: F32
1058
+ blk.40.ssm_out.weight: Q4_K
1059
+ blk.41.attn_gate.weight: Q3_K
1060
+ blk.41.attn_qkv.weight: Q4_K
1061
+ blk.41.ffn_down_exps.weight: Q2_0
1062
+ blk.41.ffn_down_shexp.weight: Q4_0
1063
+ blk.41.ffn_gate_exps.weight: Q2_0
1064
+ blk.41.ffn_gate_inp.weight: BF16
1065
+ blk.41.ffn_gate_inp_shexp.weight: BF16
1066
+ blk.41.ffn_gate_shexp.weight: Q3_K
1067
+ blk.41.ffn_up_exps.weight: Q2_0
1068
+ blk.41.ffn_up_shexp.weight: Q3_K
1069
+ blk.41.hc_attn_down.weight: BF16
1070
+ blk.41.hc_attn_inject.weight: BF16
1071
+ blk.41.hc_attn_norm.weight: F32
1072
+ blk.41.hc_attn_up.weight: BF16
1073
+ blk.41.hc_ffn_down.weight: BF16
1074
+ blk.41.hc_ffn_inject.weight: BF16
1075
+ blk.41.hc_ffn_norm.weight: F32
1076
+ blk.41.hc_ffn_up.weight: BF16
1077
+ blk.41.ssm_a: F32
1078
+ blk.41.ssm_alpha.weight: BF16
1079
+ blk.41.ssm_beta.weight: BF16
1080
+ blk.41.ssm_conv1d.weight: F32
1081
+ blk.41.ssm_dt.bias: F32
1082
+ blk.41.ssm_norm.weight: F32
1083
+ blk.41.ssm_out.weight: Q4_K
1084
+ blk.42.attn_gate.weight: Q3_K
1085
+ blk.42.attn_qkv.weight: Q4_K
1086
+ blk.42.ffn_down_exps.weight: Q2_0
1087
+ blk.42.ffn_down_shexp.weight: Q4_0
1088
+ blk.42.ffn_gate_exps.weight: Q2_0
1089
+ blk.42.ffn_gate_inp.weight: BF16
1090
+ blk.42.ffn_gate_inp_shexp.weight: BF16
1091
+ blk.42.ffn_gate_shexp.weight: Q2_0
1092
+ blk.42.ffn_up_exps.weight: Q2_0
1093
+ blk.42.ffn_up_shexp.weight: Q3_K
1094
+ blk.42.hc_attn_down.weight: BF16
1095
+ blk.42.hc_attn_inject.weight: BF16
1096
+ blk.42.hc_attn_norm.weight: F32
1097
+ blk.42.hc_attn_up.weight: BF16
1098
+ blk.42.hc_ffn_down.weight: BF16
1099
+ blk.42.hc_ffn_inject.weight: BF16
1100
+ blk.42.hc_ffn_norm.weight: F32
1101
+ blk.42.hc_ffn_up.weight: BF16
1102
+ blk.42.ssm_a: F32
1103
+ blk.42.ssm_alpha.weight: BF16
1104
+ blk.42.ssm_beta.weight: BF16
1105
+ blk.42.ssm_conv1d.weight: F32
1106
+ blk.42.ssm_dt.bias: F32
1107
+ blk.42.ssm_norm.weight: F32
1108
+ blk.42.ssm_out.weight: IQ4_XS
1109
+ blk.43.attn_k.weight: IQ4_XS
1110
+ blk.43.attn_k_norm.weight: F32
1111
+ blk.43.attn_output.weight: Q6_K
1112
+ blk.43.attn_q.weight: IQ4_XS
1113
+ blk.43.attn_q_norm.weight: F32
1114
+ blk.43.attn_v.weight: Q6_K
1115
+ blk.43.ffn_down_exps.weight: Q2_0
1116
+ blk.43.ffn_down_shexp.weight: IQ4_NL
1117
+ blk.43.ffn_gate_exps.weight: Q2_0
1118
+ blk.43.ffn_gate_inp.weight: BF16
1119
+ blk.43.ffn_gate_inp_shexp.weight: BF16
1120
+ blk.43.ffn_gate_shexp.weight: Q2_0
1121
+ blk.43.ffn_up_exps.weight: Q2_0
1122
+ blk.43.ffn_up_shexp.weight: IQ4_XS
1123
+ blk.43.hc_attn_down.weight: BF16
1124
+ blk.43.hc_attn_inject.weight: BF16
1125
+ blk.43.hc_attn_norm.weight: F32
1126
+ blk.43.hc_attn_up.weight: BF16
1127
+ blk.43.hc_ffn_down.weight: BF16
1128
+ blk.43.hc_ffn_inject.weight: BF16
1129
+ blk.43.hc_ffn_norm.weight: F32
1130
+ blk.43.hc_ffn_up.weight: BF16
1131
+ blk.43.indexer.k_norm.weight: F32
1132
+ blk.43.indexer.k_proj.weight: BF16
1133
+ blk.43.indexer.q_norm.weight: F32
1134
+ blk.43.indexer.q_proj.weight: BF16
1135
+ blk.44.attn_gate.weight: Q3_K
1136
+ blk.44.attn_qkv.weight: IQ4_XS
1137
+ blk.44.ffn_down_exps.weight: Q2_0
1138
+ blk.44.ffn_down_shexp.weight: IQ4_NL
1139
+ blk.44.ffn_gate_exps.weight: Q2_0
1140
+ blk.44.ffn_gate_inp.weight: BF16
1141
+ blk.44.ffn_gate_inp_shexp.weight: BF16
1142
+ blk.44.ffn_gate_shexp.weight: Q3_K
1143
+ blk.44.ffn_up_exps.weight: Q2_0
1144
+ blk.44.ffn_up_shexp.weight: Q3_K
1145
+ blk.44.hc_attn_down.weight: BF16
1146
+ blk.44.hc_attn_inject.weight: BF16
1147
+ blk.44.hc_attn_norm.weight: F32
1148
+ blk.44.hc_attn_up.weight: BF16
1149
+ blk.44.hc_ffn_down.weight: BF16
1150
+ blk.44.hc_ffn_inject.weight: BF16
1151
+ blk.44.hc_ffn_norm.weight: F32
1152
+ blk.44.hc_ffn_up.weight: BF16
1153
+ blk.44.ssm_a: F32
1154
+ blk.44.ssm_alpha.weight: BF16
1155
+ blk.44.ssm_beta.weight: BF16
1156
+ blk.44.ssm_conv1d.weight: F32
1157
+ blk.44.ssm_dt.bias: F32
1158
+ blk.44.ssm_norm.weight: F32
1159
+ blk.44.ssm_out.weight: Q6_K
1160
+ blk.45.attn_gate.weight: Q4_K
1161
+ blk.45.attn_qkv.weight: IQ4_XS
1162
+ blk.45.ffn_down_exps.weight: Q2_0
1163
+ blk.45.ffn_down_shexp.weight: Q5_0
1164
+ blk.45.ffn_gate_exps.weight: Q2_0
1165
+ blk.45.ffn_gate_inp.weight: BF16
1166
+ blk.45.ffn_gate_inp_shexp.weight: BF16
1167
+ blk.45.ffn_gate_shexp.weight: Q3_K
1168
+ blk.45.ffn_up_exps.weight: Q2_0
1169
+ blk.45.ffn_up_shexp.weight: Q3_K
1170
+ blk.45.hc_attn_down.weight: BF16
1171
+ blk.45.hc_attn_inject.weight: BF16
1172
+ blk.45.hc_attn_norm.weight: F32
1173
+ blk.45.hc_attn_up.weight: BF16
1174
+ blk.45.hc_ffn_down.weight: BF16
1175
+ blk.45.hc_ffn_inject.weight: BF16
1176
+ blk.45.hc_ffn_norm.weight: F32
1177
+ blk.45.hc_ffn_up.weight: BF16
1178
+ blk.45.ssm_a: F32
1179
+ blk.45.ssm_alpha.weight: BF16
1180
+ blk.45.ssm_beta.weight: BF16
1181
+ blk.45.ssm_conv1d.weight: F32
1182
+ blk.45.ssm_dt.bias: F32
1183
+ blk.45.ssm_norm.weight: F32
1184
+ blk.45.ssm_out.weight: IQ4_XS
1185
+ blk.46.attn_gate.weight: Q4_K
1186
+ blk.46.attn_qkv.weight: IQ4_XS
1187
+ blk.46.ffn_down_exps.weight: Q2_0
1188
+ blk.46.ffn_down_shexp.weight: Q4_0
1189
+ blk.46.ffn_gate_exps.weight: Q2_0
1190
+ blk.46.ffn_gate_inp.weight: BF16
1191
+ blk.46.ffn_gate_inp_shexp.weight: BF16
1192
+ blk.46.ffn_gate_shexp.weight: Q5_K
1193
+ blk.46.ffn_up_exps.weight: Q2_0
1194
+ blk.46.ffn_up_shexp.weight: Q3_K
1195
+ blk.46.hc_attn_down.weight: BF16
1196
+ blk.46.hc_attn_inject.weight: BF16
1197
+ blk.46.hc_attn_norm.weight: F32
1198
+ blk.46.hc_attn_up.weight: BF16
1199
+ blk.46.hc_ffn_down.weight: BF16
1200
+ blk.46.hc_ffn_inject.weight: BF16
1201
+ blk.46.hc_ffn_norm.weight: F32
1202
+ blk.46.hc_ffn_up.weight: BF16
1203
+ blk.46.ssm_a: F32
1204
+ blk.46.ssm_alpha.weight: BF16
1205
+ blk.46.ssm_beta.weight: BF16
1206
+ blk.46.ssm_conv1d.weight: F32
1207
+ blk.46.ssm_dt.bias: F32
1208
+ blk.46.ssm_norm.weight: F32
1209
+ blk.46.ssm_out.weight: Q4_K
1210
+ blk.47.attn_k.weight: Q6_K
1211
+ blk.47.attn_k_norm.weight: F32
1212
+ blk.47.attn_output.weight: Q5_K
1213
+ blk.47.attn_q.weight: IQ4_XS
1214
+ blk.47.attn_q_norm.weight: F32
1215
+ blk.47.attn_v.weight: Q5_K
1216
+ blk.47.ffn_down_exps.weight: Q2_0
1217
+ blk.47.ffn_down_shexp.weight: Q8_0
1218
+ blk.47.ffn_gate_exps.weight: Q2_0
1219
+ blk.47.ffn_gate_inp.weight: BF16
1220
+ blk.47.ffn_gate_inp_shexp.weight: BF16
1221
+ blk.47.ffn_gate_shexp.weight: Q3_K
1222
+ blk.47.ffn_up_exps.weight: Q2_0
1223
+ blk.47.ffn_up_shexp.weight: Q5_K
1224
+ blk.47.hc_attn_down.weight: BF16
1225
+ blk.47.hc_attn_inject.weight: BF16
1226
+ blk.47.hc_attn_norm.weight: F32
1227
+ blk.47.hc_attn_up.weight: BF16
1228
+ blk.47.hc_ffn_down.weight: BF16
1229
+ blk.47.hc_ffn_inject.weight: BF16
1230
+ blk.47.hc_ffn_norm.weight: F32
1231
+ blk.47.hc_ffn_up.weight: BF16
1232
+ blk.47.indexer.k_norm.weight: F32
1233
+ blk.47.indexer.k_proj.weight: BF16
1234
+ blk.47.indexer.q_norm.weight: F32
1235
+ blk.47.indexer.q_proj.weight: BF16