moving monkeypatch
This commit is contained in:
@@ -3,7 +3,7 @@
|
|||||||
from transformers import PreTrainedModel
|
from transformers import PreTrainedModel
|
||||||
from transformers.models.llama.modeling_llama import LLAMA_ATTENTION_CLASSES
|
from transformers.models.llama.modeling_llama import LLAMA_ATTENTION_CLASSES
|
||||||
|
|
||||||
from .multihead_diffattn import (
|
from axolotl.integrations.diff_transformer.multihead_diffattn import (
|
||||||
LlamaDifferentialAttention,
|
LlamaDifferentialAttention,
|
||||||
LlamaDifferentialSdpaAttention,
|
LlamaDifferentialSdpaAttention,
|
||||||
)
|
)
|
||||||
@@ -445,7 +445,7 @@ class ModelLoader:
|
|||||||
patch_mistral_cross_entropy()
|
patch_mistral_cross_entropy()
|
||||||
|
|
||||||
if self.cfg.diff_attention:
|
if self.cfg.diff_attention:
|
||||||
from axolotl.integrations.diff_transformer.patches import (
|
from axolotl.monkeypatch.attention.differential import (
|
||||||
patch_llama_attention_classes,
|
patch_llama_attention_classes,
|
||||||
)
|
)
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user