bugfix: remove spurious backslash breaking output redirection in launch scripts.

[kuishou68]

Closes #1247

Problem

In docs/en/getting_started/launch_xllm.md, all three platform launch scripts (NPU, NVIDIA GPU, MLU) contain a shell syntax error on the final argument line:

--node_rank=$i \ > $LOG_FILE 2>&1 &

The \ > sequence causes the shell to interpret > as a literal escaped character rather than a redirection operator. As a result:

• The log file is never created
• All xllm stdout/stderr prints to the terminal instead of being captured
• Users running the launch script lose all logs

Fix

Remove the spurious \ before > in all three script blocks (lines 46, 87, 123):

# Before (broken):
    --node_rank=$i \ > $LOG_FILE 2>&1 &

# After (correct):
    --node_rank=$i > $LOG_FILE 2>&1 &

Testing

# Demonstrate the bug (no file created):
echo hello \ > /tmp/test_broken.log
cat /tmp/test_broken.log  # file empty or missing

# Correct form:
echo hello > /tmp/test_fixed.log
cat /tmp/test_fixed.log   # prints: hello

[gemini-code-assist]

Code Review

This pull request cleans up shell command examples in the documentation by removing unnecessary backslashes before output redirection in docs/en/getting_started/launch_xllm.md. I have no feedback to provide as there were no review comments.

[kuishou68]

Hi! Just following up on this PR — it removes the spurious backslash that was breaking output redirection in the launch scripts. The automated review (Gemini) found no issues. Would appreciate a maintainer review when you get a chance. Thanks!

[XuZhang99]

Hi @kuishou68, we couldn't reproduce this issue in our all docker image.

[kuishou68]

Thanks for checking. The issue I observed was on a bare Ubuntu 22.04 host running the launch scripts directly (not via Docker image). The backslash on line N breaks the shell redirection when the script is run with bash outside a container, since bash interprets the backslash differently than the Docker entrypoint. If the Docker image wraps the script in a way that skips the problematic redirection path, it would explain why it's not reproducible there. Happy to add a test case or provide the exact environment if that helps narrow it down.

[XuZhang99]

Thanks for checking. The issue I observed was on a bare Ubuntu 22.04 host running the launch scripts directly (not via Docker image). The backslash on line N breaks the shell redirection when the script is run with bash outside a container, since bash interprets the backslash differently than the Docker entrypoint. If the Docker image wraps the script in a way that skips the problematic redirection path, it would explain why it's not reproducible there. Happy to add a test case or provide the exact environment if that helps narrow it down.

Currently, xllm only can run on docker images provided by us.