UnstableLlama/Qwen3.6-35B-A3B-exl3-2.08bpw
6B • Updated • 20 • 1
Carlo Moro
cnmoro
AI & ML interests
None yet
Recent Activity
liked a model about 8 hours ago
UnstableLlama/Qwen3.6-35B-A3B-exl3-2.08bpw liked a model about 9 hours ago
spiritbuun/Qwen3.5-27B-DFlash-GGUF liked a model 4 days ago
yonigozlan/sam3-litetext-s0Organizations
liked a
model about 8 hours ago
liked a
model about 9 hours ago
liked a
model 4 days ago
reacted to sergiopaniego's post with 🔥 7 days ago
Post
1127
Earlier this month, Apple introduced Simple Self-Distillation: a fine-tuning method that improves models on coding tasks just by sampling from the model and training on its own outputs with plain cross-entropy
And… it's already supported in TRL, built by Kashif Rasul. you can really feel the pace of development in the team 🐎
Paper by Ruixiang ZHANG, He Bai, Huangjie Zheng, Navdeep Jaitly, Ronan Collobert, Yizhe Zhang at Apple 🍎
How it works: the model generates completions at a training-time temperature (T_train) with top_k/top_p truncation, then fine-tunes on them with plain cross-entropy. no labels or verifier needed
You can try it right away with this ready-to-run example (Qwen3-4B on rStar-Coder):
https://github.com/huggingface/trl/blob/main/trl/experimental/ssd/ssd.py
or benchmark a checkpoint with the eval script:
https://github.com/huggingface/trl/blob/main/trl/experimental/ssd/ssd_eval.py
One neat insight from the paper: T_train and T_eval compose into an effective T_eff = T_train × T_eval, so a broad band of configs works well. even very noisy samples still help
Want to dig deeper?
Paper: Embarrassingly Simple Self-Distillation Improves Code Generation (2604.01193)
Trainer docs: https://huggingface.co/docs/trl/main/en/ssd_trainer
And… it's already supported in TRL, built by Kashif Rasul. you can really feel the pace of development in the team 🐎
Paper by Ruixiang ZHANG, He Bai, Huangjie Zheng, Navdeep Jaitly, Ronan Collobert, Yizhe Zhang at Apple 🍎
How it works: the model generates completions at a training-time temperature (T_train) with top_k/top_p truncation, then fine-tunes on them with plain cross-entropy. no labels or verifier needed
You can try it right away with this ready-to-run example (Qwen3-4B on rStar-Coder):
https://github.com/huggingface/trl/blob/main/trl/experimental/ssd/ssd.py
or benchmark a checkpoint with the eval script:
https://github.com/huggingface/trl/blob/main/trl/experimental/ssd/ssd_eval.py
One neat insight from the paper: T_train and T_eval compose into an effective T_eff = T_train × T_eval, so a broad band of configs works well. even very noisy samples still help
Want to dig deeper?
Paper: Embarrassingly Simple Self-Distillation Improves Code Generation (2604.01193)
Trainer docs: https://huggingface.co/docs/trl/main/en/ssd_trainer
liked a
model 8 days ago
updated a
model 9 days ago
liked a
model 10 days ago
liked 2
models 11 days ago
updated 2
models 11 days ago
published 2
models 11 days ago
Any plans on releasing flashhead for qwen3.5 models?
reacted to JonnaMat's post with 🚀 13 days ago
Post
1230
⚡ FlashHead: Fast LM Head Inference - Now a Simple vLLM Plugin
flash-head replaces the dense LM head with a two-stage retrieval pipeline - up to 2x inference speedup, training-free. Previously required custom Docker images; now it's just:
✨ The plugin activates automatically via vLLM's
🧩 Supported models (full collection):
Qwen Qwen3, 
meta-llama Llama3, 
google Gemma3, 
nvidia Cosmos-Reason2 - BF16 and W4A16 variants.
https://huggingface.co/collections/embedl/flashhead
📊 embedl/Edge-Inference-Benchmarks
🔧 Benchmark it yourself:
FlashHead shines at low batch sizes; the typical real-time / on-device use case. 🚀
flash-head replaces the dense LM head with a two-stage retrieval pipeline - up to 2x inference speedup, training-free. Previously required custom Docker images; now it's just:
pip install flash-head
vllm serve embedl/Qwen3-1.7B-FlashHead-W4A16✨ The plugin activates automatically via vLLM's
vllm.general_plugins entry point. No source patches, no custom imports.🧩 Supported models (full collection):
https://huggingface.co/collections/embedl/flashhead
📊 embedl/Edge-Inference-Benchmarks
🔧 Benchmark it yourself:
vllm bench latency --model embedl/Qwen3-1.7B-FlashHead-W4A16 --batch-size 1
# Baseline comparison
FLASHHEAD_ENABLED=0 vllm bench latency --model embedl/Qwen3-1.7B-FlashHead-W4A16 --batch-size 1FlashHead shines at low batch sizes; the typical real-time / on-device use case. 🚀
- 2 replies
liked a
dataset 13 days ago
liked 3
models 14 days ago
reacted to fffiloni's post with 🚀 14 days ago
Post
3150
✨ PASD Magnify is back on Hugging Face Spaces
fffiloni/PASD
PASD isn’t recent, but still delivers strong results — worth restoring rather than replacing.
Getting it to run again wasn’t a simple dependency issue.
It relied on parts of diffusers that no longer exist, while moving to Gradio 6 forced a much newer HF stack — and I couldn’t modify the original source directly.
Recreating the old environment wasn’t practical.
So I patched the downloaded code at runtime before import and made it compatible with today’s stack.
That ended up being the only approach that held without forking or freezing everything to outdated versions.
If you’ve used it before (or are curious), feel free to give it another try.
fffiloni/PASD
PASD isn’t recent, but still delivers strong results — worth restoring rather than replacing.
Getting it to run again wasn’t a simple dependency issue.
It relied on parts of diffusers that no longer exist, while moving to Gradio 6 forced a much newer HF stack — and I couldn’t modify the original source directly.
Recreating the old environment wasn’t practical.
So I patched the downloaded code at runtime before import and made it compatible with today’s stack.
That ended up being the only approach that held without forking or freezing everything to outdated versions.
If you’ve used it before (or are curious), feel free to give it another try.