Skip to content

Conversation

@yurekami
Copy link
Contributor

Summary

Addresses #148

Added a prominent note at the top of the Qwen3-4B documentation warning users about GPU configuration requirements when their hardware differs from the default 8-GPU setup.

Problem

Users with different GPU counts (e.g., 4 GPUs instead of 8) would run the example scripts without adjusting configuration, causing:

  • Ray jobs to hang indefinitely
  • Confusing/misleading error messages that don't indicate the root cause

Solution

Added a callout box to both English and Chinese Qwen3-4B documentation that:

  1. Clearly states the example assumes 8 GPUs
  2. Lists the specific parameters that need adjustment (ray start --num-gpus, --actor-num-gpus-per-node)
  3. Points users to the pre-configured 4-GPU script (scripts/run-qwen3-4B_4xgpu.sh)
  4. Warns that mismatched configs cause confusing errors

Files Changed

  • docs/en/examples/qwen3-4B.md
  • docs/zh/examples/qwen3-4B.md

🤖 Generated with Claude Code

Addresses THUDM#148

Added a prominent note at the top of the Qwen3-4B documentation (both
English and Chinese) warning users about GPU configuration requirements:

- Explains which parameters need adjustment for different GPU counts
- Points users to the pre-configured 4-GPU script as an alternative
- Warns that mismatched GPU configs cause confusing errors

This helps users avoid the common pitfall of running 8-GPU scripts on
machines with fewer GPUs, which causes Ray jobs to hang with misleading
error messages.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
@zhuzilin zhuzilin closed this Dec 30, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants