Upload model

config.json

@@ -1,9 +1,9 @@
 {
   "architectures": [
-    "MIMICIVEDCXRMultimodalModel"
+    "CXRMateEDModel"
   ],
   "auto_map": {
-    "AutoModel": "modelling_cxrmate_ed.MIMICIVEDCXRMultimodalModel"
+    "AutoModel": "modelling_cxrmate_ed.CXRMateEDModel"
   },
   "decoder": {
     "_name_or_path": "",

configuration_cxrmate_ed.py

@@ -4,9 +4,9 @@ from transformers.utils import logging
 logger = logging.get_logger(__name__)
 
 
-class EncoderDecoderConfig(PretrainedConfig):
+class CXRMateEDConfig(PretrainedConfig):
 
-    model_type = "encoder-decoder"
+    model_type = "cxrmate-ed"
     is_composition = True
 
     def __init__(self, **kwargs):
@@ -25,13 +25,7 @@ class EncoderDecoderConfig(PretrainedConfig):
     def from_encoder_decoder_configs(
         cls, encoder_config: PretrainedConfig, decoder_config: PretrainedConfig, **kwargs
     ) -> PretrainedConfig:
-        r"""
-        Instantiate a [`EncoderDecoderConfig`] (or a derived class) from a pre-trained encoder model configuration and
-        decoder model configuration.
 
-        Returns:
-            [`EncoderDecoderConfig`]: An instance of a configuration object
-        """
         logger.info("Set `config.is_decoder=True` and `config.add_cross_attention=True` for decoder_config")
         decoder_config.is_decoder = True
         decoder_config.add_cross_attention = True

modelling_cxrmate_ed.py

@@ -16,7 +16,7 @@ from transformers.modeling_outputs import Seq2SeqLMOutput
 from transformers.modeling_utils import PreTrainedModel
 from transformers.utils import logging
 
-from .configuration_cxrmate_ed import EncoderDecoderConfig
+from .configuration_cxrmate_ed import CXRMateEDConfig
 from .dataset import PriorsDataset
 from .modelling_uniformer import MultiUniFormerWithProjectionHead
 from .prepare_dataset import prepare_dataset
@@ -46,9 +46,9 @@ class FNNEncoder(torch.nn.Module):
         return self.down_proj(self.act_fn(self.up_proj(x)))
 
 
-class MIMICIVEDCXRMultimodalModel(VisionEncoderDecoderModel):
+class CXRMateEDModel(VisionEncoderDecoderModel):
 
-    config_class = EncoderDecoderConfig
+    config_class = CXRMateEDConfig
     base_model_prefix = "vision_encoder_decoder"
     main_input_name = "input_ids"
     supports_gradient_checkpointing = True
@@ -69,7 +69,7 @@ class MIMICIVEDCXRMultimodalModel(VisionEncoderDecoderModel):
         if config is None and (encoder is None or decoder is None):
             raise ValueError("Either a configuration or an encoder and a decoder has to be provided.")
         if config is None:
-            config = EncoderDecoderConfig.from_encoder_decoder_configs(encoder.config, decoder.config)
+            config = CXRMateEDConfig.from_encoder_decoder_configs(encoder.config, decoder.config)
         else:
             if not isinstance(config, self.config_class):
                 raise ValueError(f"Config: {config} has to be of type {self.config_class}")
@@ -301,7 +301,7 @@ class MIMICIVEDCXRMultimodalModel(VisionEncoderDecoderModel):
             decoder = transformers.AutoModelForCausalLM.from_pretrained(decoder_pretrained_model_name_or_path, **kwargs_decoder)
 
         # instantiate config with corresponding kwargs
-        config = EncoderDecoderConfig.from_encoder_decoder_configs(encoder.config, decoder.config, **kwargs)
+        config = CXRMateEDConfig.from_encoder_decoder_configs(encoder.config, decoder.config, **kwargs)
 
         # make sure input & output embeddings is not tied
         config.tie_word_embeddings = False