feat: use flashinfer jit package (#5547)

.github/workflows/pr-test.yml

@@ -38,8 +38,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
 
@@ -62,8 +60,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
 
@@ -82,8 +78,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
 
@@ -102,8 +96,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
 
@@ -146,8 +138,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
 
@@ -178,8 +168,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
 
@@ -216,8 +204,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
           git clone https://github.com/merrymercy/human-eval.git
@@ -239,8 +225,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: ${{ inputs.version == 'nightly' && 'https://flashinfer.ai/whl/nightly/cu124/torch2.5/flashinfer-python' || 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python' }}
         run: |
           bash scripts/ci_install_dependency.sh
           git clone https://github.com/merrymercy/human-eval.git

.github/workflows/vllm-dependency-test.yml

@@ -28,8 +28,6 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install dependencies
-        env:
-          FLASHINFER_REPO: 'https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python'
         run: |
           bash scripts/ci_install_dependency.sh
           pip install "vllm>=0.6.4.post1,<=0.7.2"

docs/start/install.md

@@ -11,7 +11,7 @@ It is recommended to use uv to install the dependencies for faster installation:
 ```bash
 pip install --upgrade pip
 pip install uv
-uv pip install "sglang[all]>=0.4.5.post1" --find-links https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python
+uv pip install "sglang[all]>=0.4.5.post1"
 ```
 
 **Quick Fixes to Common Problems**
@@ -23,7 +23,7 @@ uv pip install "sglang[all]>=0.4.5.post1" --find-links https://flashinfer.ai/whl
   1. Use `export CUDA_HOME=/usr/local/cuda-<your-cuda-version>` to set the `CUDA_HOME` environment variable.
   2. Install FlashInfer first following [FlashInfer installation doc](https://docs.flashinfer.ai/installation.html), then install SGLang as described above.
 
-- If you encounter `ImportError; cannot import name 'is_valid_list_of_images' from 'transformers.models.llama.image_processing_llama'`, try to use the specified version of `transformers` in [pyproject.toml](https://github.com/sgl-project/sglang/blob/main/python/pyproject.toml). Currently, just running `pip install transformers==4.48.3`.
+- If you encounter `ImportError; cannot import name 'is_valid_list_of_images' from 'transformers.models.llama.image_processing_llama'`, try to use the specified version of `transformers` in [pyproject.toml](https://github.com/sgl-project/sglang/blob/main/python/pyproject.toml). Currently, just running `pip install transformers==4.51.1`.
 
 ## Method 2: From source
 
@@ -33,7 +33,7 @@ git clone -b v0.4.5.post1 https://github.com/sgl-project/sglang.git
 cd sglang
 
 pip install --upgrade pip
-pip install -e "python[all]" --find-links https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python
+pip install -e "python[all]"
 ```
 
 Note: SGLang currently uses torch 2.5, so you need to install flashinfer for torch 2.5. If you want to install flashinfer separately, please refer to [FlashInfer installation doc](https://docs.flashinfer.ai/installation.html).

scripts/ci_install_dependency.sh

@@ -2,9 +2,6 @@
 # Install the dependency in CI.
 set -euxo pipefail
 
-# Use repo from environment variables, passed from GitHub Actions
-FLASHINFER_REPO="${FLASHINFER_REPO:-https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python}"
-
 SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
 bash "${SCRIPT_DIR}/killall_sglang.sh"
 
@@ -18,12 +15,11 @@ rm -rf /usr/local/lib/python3.10/dist-packages/sgl_kernel*
 # Update pip
 pip install --upgrade pip
 
-# Install flashinfer and sgl-kernel
+# Install sgl-kernel
-pip install flashinfer_python==0.2.3 --find-links ${FLASHINFER_REPO} --no-cache-dir
 pip install sgl-kernel==0.0.9.post2 --no-cache-dir
 
 # Install the main package
-pip install -e "python[all]" --find-links ${FLASHINFER_REPO}
+pip install -e "python[all]"
 
 # Install additional dependencies
 pip install torch_memory_saver