Create SMOLLM_VisionModel.py

SMOLLM_VisionModel.py

@@ -0,0 +1,41 @@
+import torch 
+
+class SMOLLm_VISION_ImageCaptioning(torch.nn.Module):
+    def __init__(self, llm_model, hidden_dim):
+        super(ImageCaptioningModel, self).__init__()
+        self.llm_model = llm_model
+        self.fc = torch.nn.Linear(768, 960)
+        self.relu=torch.nn.GELU()
+    def forward(self, images, input_ids,att):
+        # Encode images
+        image_features = self.relu(self.fc(images))
+        #image_att=torch.zeros([images.shape[0],]).view(-1,1).to('cuda:0')
+
+        # Prepare text inputs for LLaMA2
+        llama_inputs = self.llm_model.prepare_inputs_for_generation(input_ids)
+        with torch.no_grad():
+            llama_embeds=self.llm_model.get_input_embeddings()(llama_inputs['input_ids'])
+
+        # Concatenate image features with LLaMA2 text inputs
+        combined_inputs = torch.cat([image_features.unsqueeze(1).float(),llama_embeds], dim=1)
+        #attention_mask=torch.cat((image_att,att),dim=-1)
+        outputs = self.llm_model(inputs_embeds=combined_inputs,attention_mask=att)
+
+        return outputs.logits[:,1:,:],combined_inputs
+        #return
+
+class SmoLLM_processor():
+    def __init__(self,image_model,image_processor):
+        self.image_model=image_model
+        self.image_processor=image_processor
+
+    def get_features(self,image):
+        inputs = clip_processor(images=image, return_tensors="pt")
+        with torch.no_grad():
+            image_features = clip_model.get_image_features(**inputs.to('cuda:0')).squeeze()
+        #tokenized=tokenizer(prompt,padding=True, return_tensors='pt', max_length=50)
+        #input_ids=tokenized['input_ids'].squeeze()        #image=tfms(image/255.)
+        #attention_mask=tokenized['attention_mask'].squeeze()
+        #x=input_ids[:-1]
+        #y=input_ids[1:]
+        return image_features