lm-wheels / README.md
edoarc's picture
Move custom vLLM cu129 wheel into env folder
5ead5e7 verified
metadata
pretty_name: lm-wheels
tags:
  - wheels
  - vllm
  - transformer-engine
  - apex

Prebuilt LM Wheels

This repo contains metadata and installation instructions for prebuilt wheels for parts of the LLM training and deployment stack.

Tree

.
├── apex
│   └── py312torch29cu128
├── causal-conv1d
│   └── py312torch29cu128
├── transformer-engine-cu12
│   └── py312torch29cu128
├── transformer-engine-torch
│   └── py312torch29cu128
└── vllm
    ├── py312torch29cu128
    └── py312torch211cu129

Structure

Each subfolder encodes the package and environment the wheel was built for.

Example, the first wheel pushed for vLLM:

  • apex/py312torch29cu128
  • vllm/py312torch29cu128
  • vllm/py312torch211cu129
  • causal-conv1d/py312torch29cu128
  • transformer-engine-cu12/py312torch29cu128
  • transformer-engine-torch/py312torch29cu128
  • py312 -> python 3.12
  • torch29 -> torch 2.9
  • cu128 -> cuda 12.8

Install

Go to the subfolder matching your environment and see the README for installation instruction. For instance, for the vllm/py312torch29cu128 wheel:

python -m pip install \
  "https://huggingface.co/datasets/SakanaAI/lm-wheels/resolve/main/vllm/py312torch29cu128/vllm-0.1.dev1%2Bg2488d1dca.d20260408.cu128-cp312-cp312-linux_x86_64.whl"

For the vllm/py312torch211cu129 wheel:

python -m pip install \
  "https://huggingface.co/datasets/SakanaAI/lm-wheels/resolve/main/vllm/py312torch211cu129/vllm-0.21.0%2Bcu129.sakana.toolpreamble.11bdaf9-cp312-cp312-linux_x86_64.whl"

For the causal-conv1d/py312torch29cu128 wheel:

python -m pip install \
  "https://huggingface.co/datasets/SakanaAI/lm-wheels/resolve/main/causal-conv1d/py312torch29cu128/causal_conv1d-1.6.1-cp312-cp312-linux_x86_64.whl"

For the Transformer Engine py312torch29cu128 wheels:

python -m pip install \
  "https://huggingface.co/datasets/SakanaAI/lm-wheels/resolve/main/transformer-engine-cu12/py312torch29cu128/transformer_engine_cu12-2.15.0-py3-none-manylinux_2_28_x86_64.whl" \
  "https://huggingface.co/datasets/SakanaAI/lm-wheels/resolve/main/transformer-engine-torch/py312torch29cu128/transformer_engine_torch-2.15.0-cp312-cp312-linux_x86_64.whl"

For the apex/py312torch29cu128 wheel:

python -m pip install \
  "https://huggingface.co/datasets/SakanaAI/lm-wheels/resolve/main/apex/py312torch29cu128/apex-0.1-cp312-cp312-linux_x86_64.whl"