Skip to content

Add grouped RaggedShard bucket layout#3407

Closed
weifengpy wants to merge 2 commits into
gh/weifengpy/29/basefrom
gh/weifengpy/29/head
Closed

Add grouped RaggedShard bucket layout#3407
weifengpy wants to merge 2 commits into
gh/weifengpy/29/basefrom
gh/weifengpy/29/head

Conversation

@weifengpy

@weifengpy weifengpy commented May 20, 2026

Copy link
Copy Markdown
Contributor

Stack from ghstack (oldest at bottom):

Introduce an explicit bucket-level layout contract for placements that need grouped planning, and add GroupedRaggedShard with param-major gathered storage for view-in/view-out unshard.

Tests: python -m pytest -q torchtitan/experiments/flex_shard/tests/test_flex_shard_ragged_shard.py; python -m pytest -q torchtitan/experiments/flex_shard/tests/test_flex_shard_buckets.py

[ghstack-poisoned]
@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Meta Open Source bot. label May 20, 2026
[ghstack-poisoned]
weifengpy added a commit that referenced this pull request Jun 3, 2026
Introduce an explicit bucket-level layout contract for placements that need grouped planning, and add GroupedRaggedShard with param-major gathered storage for view-in/view-out unshard.

Tests: python -m pytest -q torchtitan/experiments/flex_shard/tests/test_flex_shard_ragged_shard.py; python -m pytest -q torchtitan/experiments/flex_shard/tests/test_flex_shard_buckets.py
ghstack-source-id: 6336f93
Pull-Request: #3407
@weifengpy weifengpy marked this pull request as draft June 3, 2026 22:29
@weifengpy

Copy link
Copy Markdown
Contributor Author

Squashed into #3239 via ghstack; the grouped RaggedShard layout now lands as part of that PR.

@weifengpy weifengpy closed this Jun 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ciflow/8gpu CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant