Activity
Update to vllm 0.7.1 for serving Deepseek on a 10xGH200 cluster
Update to vllm 0.7.1 for serving Deepseek on a 10xGH200 cluster
Add an example file of training input
Add an example file of training input
New training scripts which for Lllama 3.1 405B training
New training scripts which for Lllama 3.1 405B training
Bump typer==0.12.3 to support fastapi version
Bump typer==0.12.3 to support fastapi version
Pin fastapi==0.111.0 for vllm compat
Pin fastapi==0.111.0 for vllm compat
Fix deepspeed triton, xformers update
Fix deepspeed triton, xformers update
Update to vllm 0.5.5 and rebuild correctly
Update to vllm 0.5.5 and rebuild correctly
Latest docker with new pytorch; vllm and xformers
Latest docker with new pytorch; vllm and xformers
More LLama3.1 405b compatibility changes
More LLama3.1 405b compatibility changes