chore: bump v0.3.6.post3

[zhyncs]

Motivation

install with one-click, set flashinfer in deps

Modifications

Checklist

• Format your code according to the Contributor Guide.
• Add unit tests as outlined in the Contributor Guide.
• Update documentation as needed, including docstrings or example tutorials.

[zhyncs]

Why not use --extra-index-url ? Because https://github.com/flashinfer-ai/whl is HTML
When I use pip index versions flashinfer --index-url https://flashinfer.ai/whl/cu121/

WARNING: The directory '/root/.cache/pip' or its parent directory is not owned or is not writable by the current user. The cache has been disabled. Check the permissions and owner of that directory. If executing pip with sudo, you should use sudo's -H flag.
WARNING: pip index is currently an experimental command. It may be removed/changed in a future release without prior warning.
flashinfer (0.0.1+cu121)
Available versions: 0.0.1+cu121
  INSTALLED: 0.1.6+cu121torch2.4
  LATEST:    0.0.1+cu121

It can't get the right LATEST

[zhyncs]

Docker v0.3.6.post3 can be overridden, so don't worry about it.

[zhyncs]

Locally verified

[zhyncs]

Another purpose of this PR is to support a Dockerfile.dev #2060 in the future, where we can specify the flashinfer index as nightly in dev.
ref flashinfer-ai/flashinfer-nightly#1 (comment)

[zhyncs]

docker pull lmsysorg/sglang:latest

# need to setup HF_TOKEN
docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v /opt/dlami/nvme/.cache:/root/.cache \
    -v /tmp:/tmp \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server --model-path meta-llama/Llama-3.1-8B-Instruct --host 0.0.0.0 --port 30000 --disable-radix

This works well with v0.3.6.post3

[zhyncs]

ref flashinfer-ai/whl#1

[merrymercy]

Will this break AMD?

[HaiShaw]

It is okay to AMD.