cleanup: remove dead SDPA patches (#3488) [skip ci]

Transformers 5.x routes attention through sdpa_attention.py and no longer
calls the _prepare_4d_causal_attention_mask* or _expand_mask functions that
these patches targeted. This makes the following patches dead code:

- llama_patch_multipack.py (patched _prepare_4d_causal_attention_mask*)
- llama_expand_mask.py (patched _expand_mask, never called)
- Related utility functions in monkeypatch/utils.py

Closes axolotl-ai-cloud/axolotl#3331
This commit is contained in:
Avaya Aggarwal
2026-03-20 15:40:41 +05:30
committed by GitHub
parent c57acef2c7
commit 7ddfb2d8a0
6 changed files with 1 additions and 173 deletions

View File

@@ -128,11 +128,9 @@ quartodoc:
- monkeypatch.mistral_attn_hijack_flash
- monkeypatch.multipack
- monkeypatch.relora
- monkeypatch.llama_expand_mask
- monkeypatch.lora_kernels
- monkeypatch.utils
- monkeypatch.btlm_attn_hijack_flash
- monkeypatch.llama_patch_multipack
- monkeypatch.stablelm_attn_hijack_flash
- monkeypatch.trainer_fsdp_optim
- monkeypatch.transformers_fa_utils