microsoft
/

Magma-8B

Image-Text-to-Text

text-generation

Model card Files Files and versions Community

jw2yang commited on 7 days ago

Commit

6a6bb9b

·

1 Parent(s): fb92d9b

fix _no_split_modules

Files changed (1) hide show

modeling_magma.py +1 -1

modeling_magma.py CHANGED Viewed

@@ -141,7 +141,7 @@ class MagmaPreTrainedModel(PreTrainedModel):
     config_class = MagmaConfig
     base_model_prefix = "model"
     supports_gradient_checkpointing = True
-    _no_split_modules = ["MagmaVisionAttention"]
     _skip_keys_device_placement = "past_key_values"
     _supports_flash_attn_2 = True

     config_class = MagmaConfig
     base_model_prefix = "model"
     supports_gradient_checkpointing = True
+    _no_split_modules = ["MagmaImageTower"]
     _skip_keys_device_placement = "past_key_values"
     _supports_flash_attn_2 = True