Youngeun Kwon

This MR provides an interface to enable the UCC backend for PP communication. To enable the UCC backend, set the following argument: `training.model.pipeline_model_parallel_comm_backend=ucc` Requires a related NeMo PR ([link](https://github.com/NVIDIA/NeMo/pull/10531)).

Long context performance doc hot fix

1

# What does this PR do ? Hotfix for table style in long context performance documentation. # Changelog - Add specific line by line info of high level changes in...

Updating the long context performance number for B200

1

- Add the without CP numbers for the B200 - Keep the consistent format with the H100 table - Merge the captioning text of both H100 and B200 tables. The...

> [!IMPORTANT] > The `Update branch` button must only be pressed in very rare occassions. > An outdated branch is never blocking the merge of a PR. > Please reach...

> [!IMPORTANT] > The `Update branch` button must only be pressed in very rare occassions. > An outdated branch is never blocking the merge of a PR. > Please reach...

Adding interface to use mcore_fsdp in perf script

1

> [!IMPORTANT] > The `Update branch` button must only be pressed in very rare occassions. > An outdated branch is never blocking the merge of a PR. > Please reach...

feat: plot vllm internal metrics to the wandb log

2

# What does this PR do ? - Builds upon [#1534] by tracking two additional vLLM metrics:` kv_cache_usage_perc` and `generation_tokens`. - Adds W&B plotting for all vLLM metrics introduced in...

feat: pipeline-rl style # of inflight prompt regulation

1

# What does this PR do ? **Add a one line overview of what this PR aims to accomplish.** # Issues List issues that this PR closes ([syntax](https://docs.github.com/en/issues/tracking-your-work-with-issues/using-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword)): # Usage...