NVIDIA/NeMo-Framework-Launcher

Python 510 stars

Provides end-to-end model development pipelines for LLMs and Multimodal models that can be launched on-prem or cloud-native.

✓ Synced 1h ago Share on X →

README badge:

[![ngmi](https://ngmi.review/badge/NVIDIA/NeMo-Framework-Launcher.svg)](https://ngmi.review/repo/NVIDIA/NeMo-Framework-Launcher)

350 Merged PRs

5 days Avg Merge Time

0m Fastest PR

3 months Slowest PR

      #925
      Global Speed Rank
    

PR Size Analysis

Lines changed (additions + deletions) vs review outcomes. Re-sync to populate data for older PRs.

PRs by size

Avg review time (hrs)

Clean approval rate (%)

#	Title	Author	Time	Reviews	Blocks
#460	Remove sqlitedict from requirements and import guard	@thomasdhc	40m	5	✓
#459	Remove unneeded dask env variable.	@ayushdg	18.5h	1	✓
#456	Add GPU metrics support for nsight profiling	@zswerth	5 days	2	✓
#457	bump hydra-core	@akoumpa	7m	2	✓
#454	remove duplicate 340b params	@malay-nagda	23.4h	4	✓
#450	Update nemotron_340b.yaml	@zhenghax	1 day	2	✓
#449	Revert vit optimizer	@yaoyu-33	30m	1	✓
#448	switch to mcore's optimizer	@dimapihtar	9 days	1	✓
#441	Add --cuda_graph_trace flag to the generated nsys ccommand	@guyueh1	1 month	1	✓
#443	Allow LN overlap for cc >=9 instead of ==9	@guyueh1	29 days	1	✓
#433	Add siglip_config.yaml as a SigLIP training template	@li-li-github	1 month	1	✓
#447	Update PTL import	@maanug-nv	1.5h	1	✓
#446	Updates for new container 24.09	@pablo-garay	40m	1	✓
#440	update mixtral dist_ckpt format to zarr for PP>1 issue	@paul-gibbons	1 month	1	✓
#439	add path access checking	@akoumpa	1 day	1	✓
#438	[MoE] Add mixtral 3B/7B fp8 configs and grok bf16/fp8 configs	@gdengk	11m	1	✓
#437	timeout only on first node	@akoumpa	2 days	1	✓
#435	Add a new tp overlap cfg for tp2 llama2 70B	@guyueh1	2 days	1	✓
#436	Update conditional None handling	@thomasdhc	1.7h	1	✓
#434	Pass dist_ckpt_load_strictness for evalharness	@thomasdhc	4.2h	1	✓