423A35C7

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-06-10 17:28:55 +08:00

cabe59a343 [model] add MiniCPM5-1B-Chat (#10558)

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-06-09 08:48:54 +08:00

9ca4026efe [model] handle unsloth model loading fallback during checkpoint resume (#7156) (#10551)

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-06-05 14:58:54 +08:00

0b7aaf8f6a [fix] correctly place new token embeddings when embedding is padded (#10547)

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-06-05 06:48:55 +08:00

8a4f6a3da5 [model] add gemma-4-12B-it (#10549)

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-06-04 22:38:54 +08:00

409e8a477f [model] Patch GDN for NPU (#10504)

423A35C7 synced commits to main at 423A35C7/pytorch3d from mirror 2026-06-03 13:48:53 +08:00

1f7f85c0a3 Revert D107142434: Enable Pyrefly in fbcode/vision/fair

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-06-03 05:48:55 +08:00

053d43c0ac [feat] support HyperParallel PT training and activation optimization (#10370)

423A35C7 synced commits to main at 423A35C7/pytorch3d from mirror 2026-06-02 21:28:55 +08:00

05025bf005 Enable Pyrefly in fbcode/vision/fair

423A35C7 synced commits to main at 423A35C7/pytorch3d from mirror 2026-06-02 05:08:54 +08:00

b73d735ecf Port pytorch3d (#2039)

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-06-01 21:08:55 +08:00

a98a1ef101 [docs] fix README citation typo (#10540)

423A35C7 synced new reference refs/tags/v0.9.5 to 423A35C7/LLaMA-Factory from mirror 2026-05-31 04:18:56 +08:00

423A35C7 synced commits to refs/tags/v0.9.5 at 423A35C7/LLaMA-Factory from mirror 2026-05-31 04:18:56 +08:00

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-05-31 04:18:56 +08:00

8ef7335b6a [misc] set dev version (#10533)

7af909522a [version] release v0.9.5 (#10532)

e016d2480e [fix] Fix NPU FusedMoE and RMSNorm (#10512)

7d719182c9 [model] fix non-packing batch (bsz>1) for Qwen3.5 with flash attention (#10529)

Compare 4 commits »

423A35C7 synced new reference dependabot/npm_and_yarn/demo/frontend/postcss-8.5.10 to 423A35C7/sam2 from mirror 2026-05-30 12:28:56 +08:00

423A35C7 synced commits to dependabot/npm_and_yarn/demo/frontend/postcss-8.5.10 at 423A35C7/sam2 from mirror 2026-05-30 12:28:56 +08:00

423A35C7 synced new reference dependabot/npm_and_yarn/demo/frontend/tmp-0.2.6 to 423A35C7/sam2 from mirror 2026-05-27 10:58:54 +08:00

423A35C7 synced commits to dependabot/npm_and_yarn/demo/frontend/tmp-0.2.6 at 423A35C7/sam2 from mirror 2026-05-27 10:58:54 +08:00

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-05-27 02:18:54 +08:00

01398eb18d [v1] fix padding free with sp (#10513)

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-05-26 01:48:54 +08:00

8e68764b65 [v1] Implement dynamic padding-free stretrgy for batching (#10507)

423A35C7 synced commits to main at 423A35C7/LLaMA-Factory from mirror 2026-05-21 23:48:54 +08:00

16ff5a23cb [fix] use getattr for profiler attrs to support MCA TrainingArguments (#10506)

bdcb92d035 [v1] Add FlashAttention selection and implement normal / padding-free / dynamic batching (#10469)

Compare 2 commits »