Hugging Face's logo
Massimo Roberto Scamarcia's picture
32 10 94

Massimo Roberto Scamarcia PRO

mrs83
BaCreative's profile picture maxxafits00's profile picture Ephrem0001's profile picture
·

AI & ML interests

Natural Language Processing, Text Generation, Question Answering, Data Augmentation, Knowledge Transfer, Chain-of-Thought, ResearchOps, MLOps

Recent Activity

reacted to qgallouedec's post with 🚀 about 22 hours ago
TRL v1.3 ships day-one training support for Qwen 3.6 🚀 The new Qwen 3.6 family (`Qwen/Qwen3.6-27B`, `Qwen/Qwen3.6-35B-A3B`) reuses the Qwen3.5-MoE architecture but ships a slightly different chat template, so we updated the stack end-to-end: new training template with `{% generation %}` markers, tool-call response schema routing, tiny test models for the VLM matrix. SFT with assistant-only loss works out of the box: ```python from trl import SFTConfig, SFTTrainer trainer = SFTTrainer( model="Qwen/Qwen3.6-27B", args=SFTConfig(assistant_only_loss=True), train_dataset=dataset, ) trainer.train() ``` So does GRPO tool-calling — just hand `tools=[...]` to `GRPOTrainer`. v1.3 also brings a new experimental TPO trainer (Triple Preference Optimization), speculative decoding in `trl vllm-serve` (Qwen3 MTP / Eagle3 drafts), 12 more KTO ↔ DPO alignment PRs (KTO promotion to stable is now in reach), three more `{% generation %}` chat templates (Gemma/Gemma 2, Phi-3, GLM-4-MoE), and a chunky SFT entropy bug fix. Full release notes: https://github.com/huggingface/trl/releases/tag/v1.3.0
published a bucket about 23 hours ago
mrs83/huggingface-static-62f6b6-bucket
published a bucket about 23 hours ago
mrs83/huggingface-static-0eb09e-bucket
View all activity

Organizations

Blog-explorers's profile picture ethicalabs.ai's profile picture BlossomTune's profile picture ML intern explorers's profile picture