Upload model

config.json

@@ -1,5 +1,4 @@
 {
-  "_commit_hash": null,
   "architectures": [
     "MultiCXREncoderDecoderModel"
   ],
@@ -78,7 +77,6 @@
     "top_p": 1.0,
     "torch_dtype": null,
     "torchscript": false,
-    "transformers_version": "4.31.0",
     "type_vocab_size": 2,
     "typical_p": 1.0,
     "use_bfloat16": false,
@@ -2243,7 +2241,6 @@
     "top_p": 1.0,
     "torch_dtype": "float32",
     "torchscript": false,
-    "transformers_version": "4.31.0",
     "typical_p": 1.0,
     "use_bfloat16": false
   },
@@ -2251,5 +2248,5 @@
   "model_type": "vision-encoder-decoder",
   "tie_word_embeddings": false,
   "torch_dtype": "float32",
-  "transformers_version": null
+  "transformers_version": "4.36.2"
 }

generation_config.json

@@ -1,5 +1,5 @@
 {
   "_from_model_config": true,
   "pad_token_id": 0,
-  "transformers_version": "4.31.0"
+  "transformers_version": "4.36.2"
 }

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6640d7d1687ce14e46f620f37c7df0eaf57bc9534a8314950e41e5faf08eed3e
+size 449521072

modelling_multi.py

@@ -6,7 +6,8 @@ import transformers
 from torch.nn import CrossEntropyLoss
 from transformers import PreTrainedTokenizerFast, VisionEncoderDecoderModel
 from transformers.configuration_utils import PretrainedConfig
-from transformers.modeling_outputs import BaseModelOutput, Seq2SeqLMOutput
+from transformers.modeling_outputs import (BaseModelOutput, ModelOutput,
+                                           Seq2SeqLMOutput)
 from transformers.modeling_utils import PreTrainedModel
 from transformers.models.vision_encoder_decoder.configuration_vision_encoder_decoder import \
     VisionEncoderDecoderConfig
@@ -21,11 +22,6 @@ class CvtWithProjectionHeadConfig(transformers.CvtConfig):
         self.projection_size = projection_size
 
 
-class ModelOutputWithProjectionEmbedding(transformers.modeling_outputs.ModelOutput):
-    last_hidden_state: torch.FloatTensor
-    attention_mask: torch.FloatTensor
-
-
 class CvtProjectionHead(torch.nn.Module):
 
     def __init__(self, config) -> None:
@@ -59,7 +55,7 @@ class MultiCvtWithProjectionHead(transformers.CvtPreTrainedModel):
         pixel_values: Optional[torch.Tensor] = None,
         output_hidden_states: Optional[bool] = None,
         return_dict: Optional[bool] = None,
-    ) -> Union[Tuple, ModelOutputWithProjectionEmbedding]:
+    ) -> Union[Tuple, ModelOutput]:
 
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
@@ -85,7 +81,7 @@ class MultiCvtWithProjectionHead(transformers.CvtPreTrainedModel):
         if not return_dict:
             return projection
 
-        return ModelOutputWithProjectionEmbedding(
+        return ModelOutput(
             last_hidden_state=projection, attention_mask=attention_mask,
         )