Add Qwen3 from_pretrained, generation example, and DeepseekV3 model support #92

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Open

danikhan632 wants to merge 1 commit into google:main from danikhan632:qwen3_deepseek

danikhan632 commented Jul 3, 2025

✅ Qwen3 Enhancements

New from_pretrained() method in tunix.models.qwen3.params:
- Supports loading Qwen3 models directly from the Hugging Face Hub.
- Automatically maps Hugging Face config (AutoConfig) to Tunix's ModelConfig.
- Loads weights using snapshot_download and create_model_from_safe_tensors.

🧪 Qwen3 Generation Example

Adds a standalone example script demonstrating:
- Loading Qwen/Qwen3-0.6B from Hugging Face.
- Using AutoTokenizer with chat templating (enable_thinking=True).
- Running inference using Sampler with KV cache and token generation.

🚀 DeepseekV3 Model Support

Adds full model definition for DeepseekV3:
Supports from_pretrained() loading from Hugging Face safetensors.


          Adding Qwen3 Flax and Deepseek

04572ed

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet