Update app.py

app.py

@@ -1,35 +1,45 @@
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 from peft import PeftModel, PeftConfig
-from transformers import AutoModelForCausalLM
-
+import torch
 import gradio as gr
 
 # Use the base model's ID
 base_model_id = "mistralai/Mistral-7B-v0.1"
 model_directory = "Tonic/mistralmed"
 
-#instantiate the Models
-
-config = PeftConfig.from_pretrained("Tonic/mistralmed", token="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")
-model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1")
-model = PeftModel.from_pretrained(model, "Tonic/mistralmed", token="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")
+# Instantiate the Models
 tokenizer = AutoTokenizer.from_pretrained(base_model_id, trust_remote_code=True)
 tokenizer.pad_token = tokenizer.eos_token
 tokenizer.padding_side = 'left'
 
+# Load the PEFT model
+peft_config = PeftConfig.from_pretrained("Tonic/mistralmed")
+base_model = AutoModelForSeq2SeqLM.from_pretrained(model_directory)
+peft_model = PeftModel.from_pretrained(base_model, "Tonic/mistralmed")
+
 class ChatBot:
     def __init__(self):
         self.history = []
 
     def predict(self, input):
-        new_user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors="pt")
-        flat_history = [item for sublist in self.history for item in sublist]
-        flat_history_tensor = torch.tensor(flat_history).unsqueeze(dim=0)
-        bot_input_ids = torch.cat([flat_history_tensor, new_user_input_ids], dim=-1) if self.history else new_user_input_ids
-        chat_history_ids = model.generate(bot_input_ids, max_length=512, pad_token_id=tokenizer.eos_token_id)
-        self.history.append(chat_history_ids[:, bot_input_ids.shape[-1]:].tolist()[0])
-        response = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
-        return response
+        # Encode user input
+        user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors="pt")
+
+        # Concatenate the user input with chat history
+        if self.history:
+            chat_history_ids = torch.cat([self.history, user_input_ids], dim=-1)
+        else:
+            chat_history_ids = user_input_ids
+
+        # Generate a response using the PEFT model
+        response = peft_model.generate(chat_history_ids, max_length=512, pad_token_id=tokenizer.eos_token_id)
+
+        # Update chat history
+        self.history = response
+
+        # Decode and return the response
+        response_text = tokenizer.decode(response[0], skip_special_tokens=True)
+        return response_text
 
 bot = ChatBot()