Hugging Face's logo

vqstudio
/
Bielik-1.5B-v3.0-Instruct-MLX-4bit

Text Generation
MLX
Safetensors
Polish
llama
finetuned
conversational
4-bit precision
Model card Files
xet
 Community
vqstudio commited on
Commit
77f20f0
·
verified ·
1 Parent(s): 71d9073

Add files using upload-large-folder tool

Browse files
Files changed (2) hide show
  1. README.md +1 -1
  2. config.json +4 -2
README.md CHANGED
@@ -23,7 +23,7 @@ pipeline_tag: text-generation
23
 
24
  This model [vqstudio/Bielik-1.5B-v3.0-Instruct-MLX-4bit](https://huggingface.co/vqstudio/Bielik-1.5B-v3.0-Instruct-MLX-4bit) was
25
  converted to MLX format from [speakleash/Bielik-1.5B-v3.0-Instruct](https://huggingface.co/speakleash/Bielik-1.5B-v3.0-Instruct)
26
- using mlx-lm version **0.26.3**.
27
 
28
  ## Use with mlx
29
 
 
23
 
24
  This model [vqstudio/Bielik-1.5B-v3.0-Instruct-MLX-4bit](https://huggingface.co/vqstudio/Bielik-1.5B-v3.0-Instruct-MLX-4bit) was
25
  converted to MLX format from [speakleash/Bielik-1.5B-v3.0-Instruct](https://huggingface.co/speakleash/Bielik-1.5B-v3.0-Instruct)
26
+ using mlx-lm version **0.28.3**.
27
 
28
  ## Use with mlx
29
 
config.json CHANGED
@@ -24,11 +24,13 @@
24
  "pretraining_tp": 1,
25
  "quantization": {
26
  "group_size": 64,
27
- "bits": 4
 
28
  },
29
  "quantization_config": {
30
  "group_size": 64,
31
- "bits": 4
 
32
  },
33
  "rms_norm_eps": 1e-06,
34
  "rope_scaling": null,
 
24
  "pretraining_tp": 1,
25
  "quantization": {
26
  "group_size": 64,
27
+ "bits": 4,
28
+ "mode": "affine"
29
  },
30
  "quantization_config": {
31
  "group_size": 64,
32
+ "bits": 4,
33
+ "mode": "affine"
34
  },
35
  "rms_norm_eps": 1e-06,
36
  "rope_scaling": null,