Skip to content

[https://nvbugs/5437106][fix] Add L4 Scout benchmarking WAR option in deploy guide #6829

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
litaotju merged 1 commit into from
Aug 15, 2025
Merged

[https://nvbugs/5437106][fix] Add L4 Scout benchmarking WAR option in deploy guide #6829

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions docs/source/deployment-guide/quick-start-recipe-for-llama4-scout-on-trtllm.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,9 +65,14 @@ cuda_graph_config:
max_batch_size: 1024
kv_cache_config:
dtype: fp8
use_torch_sampler: true
Comment thread
litaotju marked this conversation as resolved.
Resolved
EOF
```

> Here `use_torch_sampler: true` is added as a temporary WAR to solve illegal memory access issue when using trtllm native sampler.
>
> TODO: Remove this after the issue is resolved

### Launch the TRT-LLM Server

Below is an example command to launch the TRT-LLM server with the Llama-4-Scout-17B-16E-Instruct-FP8 model from within the container. The command is specifically configured for the 1024/1024 Input/Output Sequence Length test. The explanation of each flag is shown in the “Configs and Parameters” section.
Expand Down