File size: 3,542 Bytes
21b8ce0 b5c263a d8a3c53 606c0ce d8a3c53 ae70ddf 7418606 ec06a49 98758c3 d8a3c53 3b38821 d8a3c53 7418606 ec06a49 7418606 6c9b3f7 7418606 63c66b0 7418606 606c0ce d8a3c53 606c0ce d8a3c53 |
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 |
import spaces
import subprocess
import gradio as gr
from huggingface_hub import hf_hub_download
from llama_cpp import Llama
from llama_cpp_agent import LlamaCppAgent
from llama_cpp_agent import MessagesFormatterType
from llama_cpp_agent.providers import LlamaCppPythonProvider
from llama_index.core.llms import ChatMessage, MessageRole
from llama_index.llms.llama_cpp import LlamaCPP
from llama_index.llms.llama_cpp.llama_utils import (
messages_to_prompt,
completion_to_prompt,
)
from llama_index.core.memory import ChatMemoryBuffer
subprocess.run('pip install llama-cpp-python==0.2.75 --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu124', shell=True)
subprocess.run('pip install llama-cpp-agent', shell=True)
hf_hub_download(repo_id="TheBloke/Mistral-7B-Instruct-v0.2-GGUF", filename="mistral-7b-instruct-v0.2.Q6_K.gguf", local_dir = "./models")
@spaces.GPU(duration=120)
def respond(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
):
llama_model = Llama(r"models/mistral-7b-instruct-v0.2.Q6_K.gguf", n_batch=1024, n_threads=0, n_gpu_layers=33, n_ctx=8192, verbose=False)
provider = LlamaCppPythonProvider(llama_model)
agent = LlamaCppAgent(
provider,
system_prompt=f"{system_message}",
predefined_messages_formatter_type=MessagesFormatterType.MISTRAL,
debug_output=True
)
settings = provider.get_provider_default_settings()
settings.stream = True
settings.max_tokens = max_tokens
settings.temperature = temperature
settings.top_p = top_p
yield agent.get_chat_response(message, llm_sampling_settings=settings, returns_streaming_generator=True)
# stop_tokens = ["</s>", "[INST]", "[INST] ", "<s>", "[/INST]", "[/INST] "]
# chat_template = '<s>[INST] ' + system_message
# # for human, assistant in history:
# # chat_template += human + ' [/INST] ' + assistant + '</s>[INST]'
# chat_template += ' ' + message + ' [/INST]'
# print(chat_template)
# llm = LlamaCPP(
# model_path="models/mistral-7b-instruct-v0.2.Q6_K.gguf",
# temperature=temperature,
# max_new_tokens=max_tokens,
# context_window=2048,
# generate_kwargs={
# "top_k": 50,
# "top_p": top_p,
# "repeat_penalty": 1.3
# },
# model_kwargs={
# "n_threads": 0,
# "n_gpu_layers": 33
# },
# messages_to_prompt=messages_to_prompt,
# completion_to_prompt=completion_to_prompt,
# verbose=True,
# )
# # response = ""
# # for chunk in llm.stream_complete(message):
# # print(chunk.delta, end="", flush=True)
# # response += str(chunk.delta)
# # yield response
# outputs = []
# for chunk in llm.stream_complete(message):
# outputs.append(chunk.delta)
# if chunk.delta in stop_tokens:
# break
# yield "".join(outputs)
demo = gr.ChatInterface(
respond,
additional_inputs=[
gr.Textbox(value="You are a helpful assistant.", label="System message"),
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new 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 (nucleus sampling)",
),
],
)
if __name__ == "__main__":
demo.launch() |