| |
|
| | import torch
|
| | from transformers import AutoTokenizer, AutoModelForCausalLM
|
| |
|
| |
|
| | model_dir = "./gpt2-finetuned"
|
| |
|
| |
|
| | tokenizer = AutoTokenizer.from_pretrained(model_dir)
|
| | model = AutoModelForCausalLM.from_pretrained(model_dir)
|
| |
|
| |
|
| | device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| | model.to(device)
|
| |
|
| | print("Chat with the model! Type 'exit' or 'quit' to end the conversation.")
|
| |
|
| | while True:
|
| |
|
| | user_input = input("You: ")
|
| | if user_input.lower() in ["exit", "quit"]:
|
| | print("Exiting chat.")
|
| | break
|
| |
|
| |
|
| | inputs = tokenizer(user_input, return_tensors="pt", padding=True, truncation=True)
|
| | input_ids = inputs["input_ids"].to(device)
|
| | attention_mask = inputs["attention_mask"].to(device)
|
| |
|
| |
|
| | output_ids = model.generate(
|
| | input_ids,
|
| | attention_mask=attention_mask,
|
| | max_length=100,
|
| | do_sample=True,
|
| | top_p=0.95,
|
| | top_k=50,
|
| | pad_token_id=tokenizer.eos_token_id
|
| | )
|
| |
|
| |
|
| | response = tokenizer.decode(output_ids[0], skip_special_tokens=True)
|
| | print("Bot:", response)
|
| |
|