-
Notifications
You must be signed in to change notification settings - Fork 725
Pull requests: pytorch/torchtitan
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
dp2ep Expert Parallel
CLA Signed
This label is managed by the Meta Open Source bot.
#1324
by tianyu-l
was merged Jul 8, 2025
Loading…
GQA without kv repeats
CLA Signed
This label is managed by the Meta Open Source bot.
#2259
by francesco-bertolotti
was merged Jan 27, 2026
Loading…
Implement debug mode for GarbageCollection
CLA Signed
This label is managed by the Meta Open Source bot.
#1230
by fegin
was merged May 28, 2025
Loading…
Add deterministic RL training experiment with vLLM
CLA Signed
This label is managed by the Meta Open Source bot.
#1975
by bwasti
was merged Nov 7, 2025
Loading…
[WIP] Experimental implementation of gpt-oss (grouped GEMM MoE + FlexAttention sink/sliding)
CLA Signed
This label is managed by the Meta Open Source bot.
#1559
by KhoomeiK
was closed Oct 23, 2025
Loading…
Refactor attention and make attention mask an argument to the model
CLA Signed
This label is managed by the Meta Open Source bot.
#1776
by fegin
was merged Oct 10, 2025
Loading…
unify moe implementation for llama4 and deepseek_v3
CLA Signed
This label is managed by the Meta Open Source bot.
#1534
by tianyu-l
was merged Aug 6, 2025
Loading…
Adding Qwen3 model to the experiments folder
CLA Signed
This label is managed by the Meta Open Source bot.
#1429
by HosseinKaviani-H
was merged Aug 18, 2025
Loading…
Enable PP and EP overlap for MoE
CLA Signed
This label is managed by the Meta Open Source bot.
#1721
by H-Huang
was merged Dec 12, 2025
Loading…
Add This label is managed by the Meta Open Source bot.
grad_norm metrics
CLA Signed
#1143
by yzhangcs
was closed Aug 21, 2025
Loading…
improve MoE bias update logic in optimizer
CLA Signed
This label is managed by the Meta Open Source bot.
release blocking
Issues that are blocking the milestone / release completion
#1593
by rakkit
was merged Aug 22, 2025
Loading…
Adding OBELICS DataLoader
CLA Signed
This label is managed by the Meta Open Source bot.
#663
by TJ-Solergibert
was merged Mar 31, 2025
Loading…
[DeepSeek] Move permutation index generation to GPU
CLA Signed
This label is managed by the Meta Open Source bot.
#1062
by kwen2501
was merged Apr 7, 2025
Loading…
temp fix state dict loading: avoid cache_state_dict
CLA Signed
This label is managed by the Meta Open Source bot.
Add config to AC to toggle early-stop and revert A2A autograd.Function workaround
ci-no-td
CLA Signed
This label is managed by the Meta Open Source bot.
#1580
by soulitzer
was merged Aug 29, 2025
Loading…
add infra support for HF checkpoint conversion
CLA Signed
This label is managed by the Meta Open Source bot.
#1404
by tianyu-l
was merged Jul 20, 2025
Loading…
[Not for land] Settings to make Llama3-8B on 8 GPUs faster
CLA Signed
This label is managed by the Meta Open Source bot.
Support gradient accumulation
CLA Signed
This label is managed by the Meta Open Source bot.
#1238
by janEbert
was merged Jun 5, 2025
Loading…
fix: pp grad accumulation is broken
CLA Signed
This label is managed by the Meta Open Source bot.
#1732
by jdinalt
was merged Sep 24, 2025
Loading…
Add DualPipeV
CLA Signed
This label is managed by the Meta Open Source bot.
#1571
by H-Huang
was merged Aug 15, 2025
Loading…
3outeille/transformers backend (Dense model only)
CLA Signed
This label is managed by the Meta Open Source bot.
#2048
by 3outeille
was merged Nov 20, 2025
Loading…
test csv schedule on different runtime
CLA Signed
This label is managed by the Meta Open Source bot.
#707
by H-Huang
was merged Dec 11, 2024
Loading…
some cleanups on docs and parallel_dims
CLA Signed
This label is managed by the Meta Open Source bot.
#729
by tianyu-l
was merged Dec 12, 2024
Loading…
Make This label is managed by the Meta Open Source bot.
betas and weight_decay Adam(W) hyperparameters configurable
CLA Signed
#1282
by runame
was merged Jun 11, 2025
Loading…
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.