Spaces:
Running
on
Zero
Running
on
Zero
File size: 3,437 Bytes
7403df3 c464ed4 7403df3 c464ed4 7403df3 4aba831 7403df3 9f18ec6 c464ed4 7403df3 9f18ec6 c464ed4 9f18ec6 7403df3 df6b11b 7403df3 c464ed4 8212759 7403df3 c464ed4 7403df3 |
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 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 |
import spaces
import json
import subprocess
from llama_cpp import Llama
from llama_cpp_agent import LlamaCppAgent, MessagesFormatterType
from llama_cpp_agent.providers import LlamaCppPythonProvider
from llama_cpp_agent.chat_history import BasicChatHistory
from llama_cpp_agent.chat_history.messages import Roles
import gradio as gr
from huggingface_hub import hf_hub_download
llm = None
llm_model = None
hf_hub_download(
repo_id="bartowski/Llama-3.2-1B-Instruct-GGUF",
filename="Llama-3.2-1B-Instruct-Q5_K_M.gguf",
local_dir = "./models"
)
def get_messages_formatter_type(model_name):
return MessagesFormatterType.LLAMA_3
@spaces.GPU
def respond(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
top_k,
repeat_penalty,
):
global llm
global llm_model
model = "Llama-3.2-1B-Instruct-Q5_K_M.gguf"
chat_template = get_messages_formatter_type(model)
if llm is None or llm_model != model:
llm = Llama(
model_path=f"models/{model}",
flash_attn=True,
n_gpu_layers=81,
n_batch=1024,
n_ctx=8192,
)
llm_model = model
provider = LlamaCppPythonProvider(llm)
agent = LlamaCppAgent(
provider,
system_prompt=f"{system_message}",
predefined_messages_formatter_type=chat_template,
debug_output=True
)
settings = provider.get_provider_default_settings()
settings.temperature = temperature
settings.top_k = top_k
settings.top_p = top_p
settings.max_tokens = max_tokens
settings.repeat_penalty = repeat_penalty
settings.stream = True
messages = BasicChatHistory()
for msn in history:
user = {
'role': Roles.user,
'content': msn[0]
}
assistant = {
'role': Roles.assistant,
'content': msn[1]
}
messages.add_message(user)
messages.add_message(assistant)
stream = agent.get_chat_response(
message,
llm_sampling_settings=settings,
chat_history=messages,
returns_streaming_generator=True,
print_output=False
)
outputs = ""
for output in stream:
outputs += output
yield outputs
demo = gr.ChatInterface(
fn=respond,
additional_inputs=[
gr.Textbox(value="You are a helpful assistant.", label="System message"),
gr.Slider(minimum=1, maximum=8192, value=2048, step=1, label="Max tokens"),
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
gr.Slider(
minimum=0.1,
maximum=1.0,
value=0.95,
step=0.05,
label="Top-p",
),
gr.Slider(
minimum=0,
maximum=100,
value=40,
step=1,
label="Top-k",
),
gr.Slider(
minimum=0.0,
maximum=2.0,
value=1.1,
step=0.1,
label="Repetition penalty",
),
],
retry_btn="Retry",
undo_btn="Undo",
clear_btn="Clear",
submit_btn="Send",
title="Llama 3 Lightning",
description="Chat with Llama 3 Lightning",
chatbot=gr.Chatbot(
scale=1,
likeable=False,
show_copy_button=True
)
)
if __name__ == "__main__":
demo.launch() |