update

chatbot.py

@@ -0,0 +1,47 @@
+from transformers import GPT2Tokenizer, GPT2LMHeadModel
+import torch
+from datasets import load_dataset
+import pandas as pd
+import re
+
+
+class ChatBot:
+    def __init__(self):
+        self.directory = 'models/fine-tuned-gpt2'
+        self.tokenizer = GPT2Tokenizer.from_pretrained(self.directory)
+        self.model = GPT2LMHeadModel.from_pretrained(self.directory)
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        self.model.to(self.device)
+
+    def generate_response(self, history):
+        combined_prompt = ""
+        # self.tokenizer.eos_token_id = '<|endoftext|>'
+        if len(history.user) > 7:
+            history.user = history.user[-7:]
+            history.ai = history.ai[-6:] 
+        # Iterate over user and AI messages
+        for user_message, ai_message in zip(history.user, history.ai):
+            combined_prompt += f"<user> {user_message}{self.tokenizer.eos_token_id}<AI> {ai_message}{self.tokenizer.eos_token_id}"
+        # Include the last user message in the prompt for response generation
+        if history.user:
+            combined_prompt += f"<user> {history.user[-1]}{self.tokenizer.eos_token_id}<AI>"
+        # Tokenize and generate response
+        inputs = self.tokenizer.encode(combined_prompt, return_tensors="pt").to(self.device)
+        attention_mask = torch.ones(inputs.shape, device=self.device)
+        outputs = self.model.generate(
+            inputs,
+            max_length=500,  # Adjust length as needed
+            num_beams=5,
+            early_stopping=True,
+            no_repeat_ngram_size=2,
+            temperature=0.7,
+            top_k=50,
+            top_p=0.95,
+            pad_token_id=self.tokenizer.eos_token_id,
+            attention_mask=attention_mask,
+            repetition_penalty=1.2
+        )
+        response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
+        # response = response.replace(combined_prompt, "").split(".")[0]#.replace("(user 1's name)",'AI').replace("(user 2's name)",'AI').replace("[user 1's name]",'AI').replace('<user>','')
+        # print('here:\n', combined_prompt,'\n\n response:\n', response,'\n\n edit-resposne: \n', response.replace(combined_prompt, "").replace('(name)','AI').split(".")[0],'\n\n')
+        return response.replace(combined_prompt, "").split(".")[0]

main.py

@@ -1,22 +1,11 @@
-from transformers import GPT2Tokenizer, GPT2LMHeadModel
-import torch
-from datasets import load_dataset
-import pandas as pd
-import re
-from transformers import GPT2LMHeadModel, GPT2Tokenizer
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
-from pydantic import BaseModel
+import chatbot
 
 
 app = FastAPI()
 
-# Load your fine-tuned model and tokenizer
-dir = 'models/fine-tuned-gpt2'
-tokenizer = GPT2Tokenizer.from_pretrained(dir)
-model = GPT2LMHeadModel.from_pretrained(dir)
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-model.to(device)
+model = None
 
 # Add CORS middleware to allow any origin
 app.add_middleware(
@@ -29,38 +18,24 @@ app.add_middleware(
 
 @app.get("/")
 def root():
-    return {"Hello": "World"}
+    return "Hello World"
 
 # Define the Pydantic model to parse JSON input
+
+from pydantic import BaseModel
 class HistoryRequest(BaseModel):
     user: list[str]
     ai: list[str]
 
 @app.post("/generate")
 def generate_response(history: HistoryRequest):
-    combined_prompt = ""
-    
-    # Iterate over user and AI messages
-    for user_message, ai_message in zip(history.user, history.ai):
-        combined_prompt += f"<user> {user_message}\n<AI> {ai_message}\n"
+    try:
+        model
+    except:
+        model = chatbot.ChatBot()
 
-    # Include the last user message in the prompt for response generation
-    if history.user:
-        combined_prompt += f"<user> {history.user[-1]}\n<AI>"
-
-    # Tokenize and generate response
-    inputs = tokenizer.encode(combined_prompt, return_tensors="pt").to(device)
-    outputs = model.generate(
-        inputs,
-        max_length=150,  # Adjust length as needed
-        num_beams=5,
-        early_stopping=True,
-        no_repeat_ngram_size=2,
-        temperature=0.7,
-        top_k=50,
-        top_p=0.95
-    )
-
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    response = response.replace(combined_prompt, "").split(".")[0]
+    if type(model) != type(chatbot.ChatBot()):
+        model = chatbot.ChatBot()
+    
+    response = model.generate_response(history)
     return response

models/fine-tuned-gpt2/config.json

@@ -33,7 +33,7 @@
     }
   },
   "torch_dtype": "float32",
-  "transformers_version": "4.42.4",
+  "transformers_version": "4.44.0",
   "use_cache": true,
   "vocab_size": 50257
 }

models/fine-tuned-gpt2/config.json:Zone.Identifier

@@ -0,0 +1,3 @@
+[ZoneTransfer]
+ZoneId=3
+HostUrl=https://www.kaggle.com/

models/fine-tuned-gpt2/generation_config.json

@@ -2,5 +2,5 @@
   "_from_model_config": true,
   "bos_token_id": 50256,
   "eos_token_id": 50256,
-  "transformers_version": "4.42.4"
+  "transformers_version": "4.44.0"
 }

models/fine-tuned-gpt2/generation_config.json:Zone.Identifier

@@ -0,0 +1,3 @@
+[ZoneTransfer]
+ZoneId=3
+HostUrl=https://www.kaggle.com/

models/fine-tuned-gpt2/merges.txt:Zone.Identifier

@@ -0,0 +1,3 @@
+[ZoneTransfer]
+ZoneId=3
+HostUrl=https://www.kaggle.com/

models/fine-tuned-gpt2/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7643d1aa99a0912a7761281c7fa219c490a0be957e054e0b2affca807a49f484
+oid sha256:8be8247018b9ae965bcf6d6e3edaa797753fcf42623b65efa34973d31dae6aa3
 size 497774208

models/fine-tuned-gpt2/model.safetensors:Zone.Identifier

@@ -0,0 +1,3 @@
+[ZoneTransfer]
+ZoneId=3
+HostUrl=https://www.kaggle.com/

models/fine-tuned-gpt2/special_tokens_map.json:Zone.Identifier

@@ -0,0 +1,3 @@
+[ZoneTransfer]
+ZoneId=3
+HostUrl=https://www.kaggle.com/

models/fine-tuned-gpt2/tokenizer_config.json:Zone.Identifier

@@ -0,0 +1,3 @@
+[ZoneTransfer]
+ZoneId=3
+HostUrl=https://www.kaggle.com/

models/fine-tuned-gpt2/vocab.json:Zone.Identifier

@@ -0,0 +1,3 @@
+[ZoneTransfer]
+ZoneId=3
+HostUrl=https://www.kaggle.com/