Spaces:
Runtime error
Runtime error
File size: 1,166 Bytes
06cf9c4 ed5c4cc 06cf9c4 ed5c4cc 06cf9c4 ed5c4cc 8833e69 360ead8 5f6c3f6 06cf9c4 3402c51 88d45e0 b0e95e2 d77d9c9 f7ad5f0 d77d9c9 793eafb 106ae96 2c5e4eb f7ad5f0 36864b0 5da22d1 ce356a3 601f061 0329016 00b813c 0329016 ec04b94 ef70bbb 06cf9c4 117600f |
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 |
import gradio as gr
import time
import ctypes #to run on C api directly
import llama_cpp
from llama_cpp import Llama
from huggingface_hub import hf_hub_download #load from huggingfaces
llm = Llama(model_path= hf_hub_download(repo_id="TheBloke/Vigogne-2-7B-Chat-GGML", filename="vigogne-2-7b-chat.ggmlv3.q4_1.bin"), n_ctx=2048) #download model from hf/ n_ctx=2048 for high ccontext length
history = []
def generate_text(input_text, history):
print("history ",history)
print("input ", input_text)
if history == []:
input_text_with_history = f"Q: {input_text} \n A:"
else:
input_text_with_history = history[-1][1]+ "\n"
input_text_with_history += f"Q: {input_text} \n A:"
print("new input", input_text_with_history)
output = llm(input_text_with_history, max_tokens=1024, stop=["Q:", "\n"], echo=True)
response = output['choices'][0]['text'] + "\n"
print("response", response)
history =["init",input_text_with_history]
for character in history:
time.sleep(0.05)
yield history
demo = gr.ChatInterface(generate_text)
demo.queue(concurrency_count=1, max_size=5)
demo.launch()
|