| import streamlit as st | |
| import transformers | |
| from transformers import AutoTokenizer | |
| import torch | |
| model = "PY007/TinyLlama-1.1B-Chat-v0.1" | |
| tokenizer = transformers.AutoTokenizer.from_pretrained(model) | |
| pipeline = transformers.pipeline( | |
| "text-generation", | |
| model=model, | |
| torch_dtype=torch.float32, | |
| ) | |
| prompt = "What is 6534+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']}") | |