Skip to content

Conversation

@sudhu2k
Copy link
Collaborator

@sudhu2k sudhu2k commented Sep 10, 2025

Motivation

To update the stable branch and to get latest features from the rocm_dev branch.

Test Plan

Unit tests will be run before PR Merge.

Test Result

N/A

Submission Checklist

lmcafee-nvidia and others added 30 commits November 13, 2024 22:06
Rename optimizer's model_parallel_group -> grad_stats_parallel_group.

See merge request ADLR/megatron-lm!2240
Co-authored-by: Deepak Narayanan <[email protected]>
Co-authored-by: Oliver Koenig <[email protected]>
Co-authored-by: James Shen <[email protected]>
Co-authored-by: Kirthi Shankar Sivamani <[email protected]>
Co-authored-by: Keshav Santhanam <[email protected]>
Co-authored-by: jasonwan <[email protected]>
Add support for PyTorch FSDP-2

See merge request ADLR/megatron-lm!2150
Update simple_text_generation_controller.py

See merge request ADLR/megatron-lm!2345
…oder, encoder-decoder) to be compatible with all 3 TE backends

Co-authored-by: Huy Vu2 <[email protected]>
Co-authored-by: root <[email protected]>
Updating all T5 attention masks (encoder, decoder, encoder-decoder) to be compatible with all 3 TE backends

See merge request ADLR/megatron-lm!2273
Co-authored-by: root <[email protected]>
Co-authored-by: root <[email protected]>
Co-authored-by: root <[email protected]>
Co-authored-by: root <[email protected]>
Add hierarchical cp comm group

See merge request ADLR/megatron-lm!2279
Add missing arg to save_checkpoint call

See merge request ADLR/megatron-lm!2351
NVLM example scripts

See merge request ADLR/megatron-lm!2306
ci: Re-enable llava tests

See merge request ADLR/megatron-lm!2348
ci: Retry download assets

See merge request ADLR/megatron-lm!2357
Support etp==tp when epp==0 and enforce torch ckpt-format when epp>1

See merge request ADLR/megatron-lm!2260
QKNorm to work with TENorm

See merge request ADLR/megatron-lm!2347
Support RMSNorm when TE and Apex are not installed

See merge request ADLR/megatron-lm!2015
Clarifications for batch x pipeline parallel logic

See merge request ADLR/megatron-lm!2343
Add attention bias arg in MCore transformer for TE cuDNN FusedAttention

See merge request ADLR/megatron-lm!2293
chore: Add mypy optionally

See merge request ADLR/megatron-lm!2360
mpashkovskii and others added 24 commits May 6, 2025 09:05
…_rocm

[Fix] LayerNorm and RMSNorm not compatible on TE 1.13
…on_rocm

Revert "[Fix] LayerNorm and RMSNorm not compatible on TE 1.13"
tests: add TENorm constructor tests
fix: support MoE models serving with EP
pinned TE and added pytest arg to make distributed test more stable
…ity (#84)

* Updated iter count to ensure pytorch profiler stability

* Included change into `train_llama2.sh`
* add keep_fp8_weight_transpose_cache to control memory allocation in TE

* check TE class signature

* update te min version for weight cache transpose

* updated fsdp args in examples/train_llama
feat: add LoRA adapter layer and Mixtral LoRA training
@sudhu2k sudhu2k self-assigned this Sep 10, 2025
@sudhu2k sudhu2k closed this Oct 14, 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.