| import streamlit as st | |
| import transformers | |
| from transformers import LLaMATokenizer | |
| import torch | |
| model = "PY007/TinyLlama-1.1B-Chat-v0.1" | |
| tokenizer = transformers.LLaMATokenizer.from_pretrained(model) | |
| pipeline = transformers.pipeline("text-generation",model=model,torch_dtype=torch.bfloat16) | |
| prompt = "What is 653+2343?" | |
| formatted_prompt = ( | |
| f"### Human: {prompt}### Assistant:" | |
| ) | |
| sequences = pipeline(formatted_prompt,do_sample=True,top_k=50,top_p = 0.7,num_return_sequences=1,repetition_penalty=1.1,max_new_tokens=500) | |
| for seq in sequences: | |
| st.write(f"Result: {seq['generated_text']}") | |