Spaces:
Runtime error
Runtime error
File size: 2,134 Bytes
db328d1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 |
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel
from langchain.memory import ConversationBufferWindowMemory
import gradio as gr
bnb_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_use_double_quant=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_compute_dtype=torch.bfloat16,
)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
base_model = "mistralai/Mistral-7B-Instruct-v0.2"
tokenizer = AutoTokenizer.from_pretrained(base_model, pad_token="[PAD]")
model = AutoModelForCausalLM.from_pretrained(
base_model,
quantization_config=bnb_config,
device_map="auto",
trust_remote_code=True,
)
ft_model = PeftModel.from_pretrained(model, "nuratamton/story_sculptor_mistral").eval()
memory = ConversationBufferWindowMemory(k=10)
def generate_text(message):
user_in = message
if user_in.lower() in ["adventure", "mystery", "horror", "sci-fi"]:
memory.clear()
if user_in.lower() == "quit":
raise ValueError("User requested to quit")
memory_context = memory.load_memory_variables({})["history"]
user_input = f"{memory_context}[INST] Continue the game and maintain context: {user_in}[/INST]"
encodings = tokenizer(user_input, return_tensors="pt", padding=True).to(device)
input_ids, attention_mask = encodings["input_ids"], encodings["attention_mask"]
output_ids = ft_model.generate(
input_ids,
attention_mask=attention_mask,
max_new_tokens=1000,
num_return_sequences=1,
do_sample=True,
temperature=1.1,
top_p=0.9,
repetition_penalty=1.2,
)
generated_ids = output_ids[0, input_ids.shape[-1] :]
response = tokenizer.decode(generated_ids, skip_special_tokens=True)
memory.save_context({"input": user_in}, {"output": response})
response = response.replace("AI: ", "")
return response
iface = gr.Interface(
fn=generate_text,
inputs="text",
outputs="text",
title="Text Generation",
description="Enter a message to generate text.",
)
iface.launch(share=True)
|