h7878778h 09dedf144f [npu] Redirect SDPA to torch_npu.npu_fusion_attention (opt-in, ZeRO-3 safe, no impact off NPU) (#8972)

2025-09-30 18:11:31 +08:00

..

__init__.py

rename files

2024-06-07 00:09:06 +08:00

attention.py

[model] switch to gptqmodel (#8108)

2025-05-19 22:25:40 +08:00

checkpointing.py

[model] fix vit gradient checkpointing (#7830)

2025-04-23 22:48:48 +08:00

embedding.py

[train] fix adjusting logits size after adding special tokens (#8823)

2025-08-05 20:35:07 +08:00

kv_cache.py

[model] fix kv cache (#7564)

2025-04-01 23:07:46 +08:00

liger_kernel.py

[misc] update glm4v ligerkernel (#8978)

2025-08-20 23:39:56 +08:00

longlora.py

[trainer] fix pt loss (#7748)

2025-04-17 03:15:35 +08:00

misc.py

[misc] upgrade format to py39 (#7256)

2025-03-12 00:08:41 +08:00

mod.py

[misc] update license year & fix llama pro (#6814)

2025-02-05 01:53:33 +08:00

moe.py

[model] add qwen3-vl/qwen3-omni (#9196)

2025-09-27 01:21:47 +08:00

packing.py

[trainer] fix pt loss (#7748)

2025-04-17 03:15:35 +08:00

quantization.py

[data] fix template (#8827)

2025-08-06 06:58:09 +08:00

rope.py

[model] switch to gptqmodel (#8108)

2025-05-19 22:25:40 +08:00

sdpa_npu_redirect.py

[npu] Redirect SDPA to torch_npu.npu_fusion_attention (opt-in, ZeRO-3 safe, no impact off NPU) (#8972)

2025-09-30 18:11:31 +08:00

unsloth.py

Revert "[model] add lora dropout to unsloth" - requested feature already exists (#8554)

2025-07-05 11:25:31 +08:00

valuehead.py

Merge commit from fork

2025-06-26 13:55:42 +08:00

visual.py

[data] fix qwen omni plugin (#9204)

2025-09-28 01:02:29 +08:00