Spaces:
Running
Running
File size: 8,742 Bytes
038f313 fab24df c5a20a4 038f313 880ced6 e13eb1b 038f313 a8fc89d 038f313 e13eb1b 038f313 30153c5 038f313 27c8b8d 038f313 3a64d68 98674ca c5a20a4 038f313 e13eb1b 901bafe e13eb1b 901bafe 27c8b8d be3f346 f7c4208 901bafe 52ad57a 038f313 901bafe c5a20a4 901bafe 27c8b8d 901bafe 27c8b8d 30153c5 901bafe 27c8b8d 30153c5 901bafe 27c8b8d 901bafe 27c8b8d 901bafe 27c8b8d 901bafe c5a20a4 901bafe a8fc89d 901bafe 27c8b8d 901bafe 27c8b8d 30153c5 27c8b8d 901bafe 27c8b8d 901bafe a8fc89d 542c2ac 901bafe f7c4208 a8fc89d be3f346 901bafe a8fc89d 901bafe 30153c5 901bafe a8fc89d 901bafe a8fc89d 30153c5 a8fc89d 901bafe a8fc89d 30153c5 a8fc89d 30153c5 901bafe a8fc89d 30153c5 a8fc89d 901bafe a8fc89d 30153c5 a8fc89d 30153c5 a8fc89d 901bafe a8fc89d 901bafe a8fc89d 30153c5 a8fc89d 901bafe a8fc89d 901bafe 30153c5 901bafe a8fc89d 30153c5 a8fc89d 901bafe a8fc89d 30153c5 a8fc89d 901bafe a8fc89d be3f346 769901b 77298b9 27c8b8d 77298b9 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 |
import gradio as gr
from openai import OpenAI
import os
# Retrieve the access token from the environment variable
ACCESS_TOKEN = os.getenv("HF_TOKEN")
print("Access token loaded.")
# Initialize the OpenAI client with the Hugging Face Inference API endpoint
client = OpenAI(
base_url="https://api-inference.huggingface.co/v1/",
api_key=ACCESS_TOKEN,
)
print("OpenAI client initialized.")
def respond(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
frequency_penalty,
seed,
custom_model
):
"""
This function handles the chatbot response. It takes in:
- message: the user's new message
- history: the list of previous messages, each as a tuple (user_msg, assistant_msg)
- system_message: the system prompt
- max_tokens: the maximum number of tokens to generate in the response
- temperature: sampling temperature
- top_p: top-p (nucleus) sampling
- frequency_penalty: penalize repeated tokens in the output
- seed: a fixed seed for reproducibility; -1 will mean 'random'
- custom_model: the final model name in use, which may be set by selecting from the Featured Models radio or by typing a custom model
"""
print(f"Received message: {message}")
print(f"History: {history}")
print(f"System message: {system_message}")
print(f"Max tokens: {max_tokens}, Temperature: {temperature}, Top-P: {top_p}")
print(f"Frequency Penalty: {frequency_penalty}, Seed: {seed}")
print(f"Selected model (custom_model): {custom_model}")
# Convert seed to None if -1 (meaning random)
if seed == -1:
seed = None
# Construct the messages array required by the API
messages = [{"role": "system", "content": system_message}]
print("Initial messages array constructed.")
# Add conversation history to the context
for val in history:
user_part = val[0] # Extract user message from the tuple
assistant_part = val[1] # Extract assistant message from the tuple
if user_part:
messages.append({"role": "user", "content": user_part})
print(f"Added user message to context: {user_part}")
if assistant_part:
messages.append({"role": "assistant", "content": assistant_part})
print(f"Added assistant message to context: {assistant_part}")
# Append the latest user message
messages.append({"role": "user", "content": message})
print("Latest user message appended.")
# If user provided a model, use that; otherwise, fall back to a default model
model_to_use = custom_model.strip() if custom_model.strip() != "" else "meta-llama/Llama-3.3-70B-Instruct"
print(f"Model selected for inference: {model_to_use}")
# Start with an empty string to build the response as tokens stream in
response = ""
print("Sending request to OpenAI API.")
# Make the streaming request to the HF Inference API via openai-like client
for message_chunk in client.chat.completions.create(
model=model_to_use,
max_tokens=max_tokens,
stream=True,
temperature=temperature,
top_p=top_p,
frequency_penalty=frequency_penalty,
seed=seed,
messages=messages,
):
# Extract the token text from the response chunk
token_text = message_chunk.choices[0].delta.content
print(f"Received token: {token_text}")
response += token_text
yield response
print("Completed response generation.")
# -------------------------
# GRADIO UI CONFIGURATION
# -------------------------
# Create a Chatbot component with a specified height
chatbot = gr.Chatbot(height=600, show_copy_button=True, placeholder="Select a model and begin chatting", likeable=True, layout="panel")
print("Chatbot interface created.")
# Create textboxes and sliders for system prompt, tokens, and other parameters
system_message_box = gr.Textbox(value="", label="System message")
max_tokens_slider = gr.Slider(
minimum=1,
maximum=4096,
value=512,
step=1,
label="Max new tokens"
)
temperature_slider = gr.Slider(
minimum=0.1,
maximum=4.0,
value=0.7,
step=0.1,
label="Temperature"
)
top_p_slider = gr.Slider(
minimum=0.1,
maximum=1.0,
value=0.95,
step=0.05,
label="Top-P"
)
frequency_penalty_slider = gr.Slider(
minimum=-2.0,
maximum=2.0,
value=0.0,
step=0.1,
label="Frequency Penalty"
)
seed_slider = gr.Slider(
minimum=-1,
maximum=65535,
value=-1,
step=1,
label="Seed (-1 for random)"
)
# The custom_model_box is what the respond function sees as "custom_model"
custom_model_box = gr.Textbox(
value="",
label="Custom Model",
info="(Optional) Provide a custom Hugging Face model path. Overrides any selected featured model."
)
def set_custom_model_from_radio(selected):
"""
This function will get triggered whenever someone picks a model from the 'Featured Models' radio.
We will update the Custom Model text box with that selection automatically.
"""
print(f"Featured model selected: {selected}")
return selected
# IMPORTANT: Because we have 1 main user input + 7 additional inputs,
# each example should be an 8-item list [user_text, system_prompt, max_tokens, temperature,
# top_p, frequency_penalty, seed, custom_model].
# You can adjust the default parameter values if desired.
demo = gr.ChatInterface(
fn=respond,
additional_inputs=[
system_message_box,
max_tokens_slider,
temperature_slider,
top_p_slider,
frequency_penalty_slider,
seed_slider,
custom_model_box,
],
fill_height=True,
chatbot=chatbot,
theme="Nymbo/Nymbo_Theme",
)
print("ChatInterface object created.")
# Add examples to the interface
demo.add_examples(
examples=[
["Howdy, partner!", "You are a friendly assistant.", 512, 0.7, 0.95, 0.0, -1, ""],
["What's your model name and who trained you?", "You are a factual assistant.", 512, 0.7, 0.95, 0.0, -1, ""],
["How many R's are there in 'Strawberry'?", "You are a playful assistant.", 512, 0.7, 0.95, 0.0, -1, ""],
],
inputs=[
chatbot,
system_message_box,
max_tokens_slider,
temperature_slider,
top_p_slider,
frequency_penalty_slider,
seed_slider,
custom_model_box,
],
)
print("Examples added to the interface.")
with demo:
with gr.Accordion("Featured Models", open=False):
model_search_box = gr.Textbox(
label="Filter Models",
placeholder="Search for a featured model...",
lines=1
)
print("Model search box created.")
models_list = [
"meta-llama/Llama-3.3-70B-Instruct",
"meta-llama/Llama-3.2-3B-Instruct",
"meta-llama/Llama-3.2-1B-Instruct",
"meta-llama/Llama-3.1-8B-Instruct",
"NousResearch/Hermes-3-Llama-3.1-8B",
"google/gemma-2-27b-it",
"google/gemma-2-9b-it",
"google/gemma-2-2b-it",
"mistralai/Mistral-Nemo-Instruct-2407",
"mistralai/Mixtral-8x7B-Instruct-v0.1",
"mistralai/Mistral-7B-Instruct-v0.3",
"Qwen/Qwen2.5-72B-Instruct",
"Qwen/QwQ-32B-Preview",
"PowerInfer/SmallThinker-3B-Preview",
"HuggingFaceTB/SmolLM2-1.7B-Instruct",
"TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"microsoft/Phi-3.5-mini-instruct",
]
print("Models list initialized.")
featured_model_radio = gr.Radio(
label="Select a model below",
choices=models_list,
value="meta-llama/Llama-3.3-70B-Instruct",
interactive=True
)
print("Featured models radio button created.")
def filter_models(search_term):
print(f"Filtering models with search term: {search_term}")
filtered = [m for m in models_list if search_term.lower() in m.lower()]
print(f"Filtered models: {filtered}")
return gr.update(choices=filtered)
model_search_box.change(
fn=filter_models,
inputs=model_search_box,
outputs=featured_model_radio
)
print("Model search box change event linked.")
featured_model_radio.change(
fn=set_custom_model_from_radio,
inputs=featured_model_radio,
outputs=custom_model_box
)
print("Featured model radio button change event linked.")
print("Gradio interface initialized.")
if __name__ == "__main__":
print("Launching the demo application.")
demo.launch() |