ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Updated Home (markdown)

Daniel Han 2024-10-26 13:44:45 -07:00
commit 68253b1324
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
Home.md

@ -1005,7 +1005,7 @@ We value engineers who are proactive, independent and who ship features and idea
* Somewhat optimized QLoRA for FLUX / stable diffusion models. Diffusers just added 4bit QLoRA support - make this faster. Provide a Colab notebook tutorial on how to use it. Must fit in 1 x 16GB Tesla T4.
* Unoptimized Apple Silicon / Metal support LoRA - MLX, Core ML Tools etc support.
* Utilies to export Unsloth finetunes to vLLM, SGLang & Ollama - LoRA adapters only. Provide a standalone serving interface to vLLM, SGLang
* Add float8 + QLoRA support via Torch AO into Unsloth. Show training losses match. Reproducible Colab with L4 GPU. Do not use `use torch.compile`.
* Add float8 + QLoRA support via Torch AO into Unsloth. Show training losses match. Reproducible Colab with L4 GPU. Do not use use `torch.compile`.
* Bitsandbytes 4bit QLoRA dequant Triton kernel - must be faster than CUDA version.
* Add TPU & AMD support into Unsloth. TPU - Colab & Kaggle notebooks. AMD - Runpod equivalent.
* Use torch AO and add MXFP4 support in preparation for Blackwell.