Spaces:
Runtime error
Runtime error
Commit
·
276b8cf
1
Parent(s):
c6872be
#7 talking duck v0.1 committed
Browse files- Dockerfile +1 -1
- app.py +72 -6
- docker-compose.yml +9 -0
- helper.txt +19 -0
Dockerfile
CHANGED
|
@@ -14,4 +14,4 @@ RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
|
|
| 14 |
COPY . .
|
| 15 |
|
| 16 |
# Specify the command to run on container start
|
| 17 |
-
CMD ["python", "app.py"
|
|
|
|
| 14 |
COPY . .
|
| 15 |
|
| 16 |
# Specify the command to run on container start
|
| 17 |
+
CMD ["python", "app.py"]
|
app.py
CHANGED
|
@@ -4,7 +4,7 @@ import json
|
|
| 4 |
import os
|
| 5 |
|
| 6 |
API_TOKEN = os.getenv("HF_API_TOKEN")
|
| 7 |
-
TRANSCRIBE_API_URL = "https://api-inference.huggingface.co/models/
|
| 8 |
LLM_API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-v0.1"
|
| 9 |
|
| 10 |
def transcribe_audio(audio_file):
|
|
@@ -18,23 +18,88 @@ def transcribe_audio(audio_file):
|
|
| 18 |
|
| 19 |
def get_answer(context, question):
|
| 20 |
"""Get an answer from the LLM based on the context and question."""
|
| 21 |
-
prompt =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
headers = {"Authorization": f"Bearer {API_TOKEN}"}
|
| 23 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
answer = json.loads(response.content.decode("utf-8"))[0].get("generated_text", "Answer not available")
|
| 25 |
return answer
|
| 26 |
|
|
|
|
| 27 |
def transcribe_and_answer(audio_file, question):
|
| 28 |
"""Process the audio file for transcription and use the result to get an answer to a question."""
|
| 29 |
transcription = transcribe_audio(audio_file)
|
| 30 |
answer = get_answer(transcription, question)
|
| 31 |
return transcription, answer
|
| 32 |
|
|
|
|
|
|
|
|
|
|
| 33 |
# Create the Gradio app
|
| 34 |
with gr.Blocks() as app:
|
| 35 |
-
gr.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
with gr.Row():
|
| 37 |
-
# Corrected 'type' parameter value to 'filepath'
|
| 38 |
audio_input = gr.Audio(type="filepath", label="Upload your audio question")
|
| 39 |
question_input = gr.Textbox(label="Type your question here")
|
| 40 |
answer_button = gr.Button("Get Answer")
|
|
@@ -45,8 +110,9 @@ with gr.Blocks() as app:
|
|
| 45 |
answer_button.click(transcribe_and_answer, inputs=[audio_input, question_input], outputs=[transcription_output, answer_output])
|
| 46 |
|
| 47 |
|
|
|
|
| 48 |
if __name__ == "__main__":
|
| 49 |
-
app.launch()
|
| 50 |
|
| 51 |
|
| 52 |
|
|
|
|
| 4 |
import os
|
| 5 |
|
| 6 |
API_TOKEN = os.getenv("HF_API_TOKEN")
|
| 7 |
+
TRANSCRIBE_API_URL = "https://api-inference.huggingface.co/models/openai/whisper-base.en"
|
| 8 |
LLM_API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-v0.1"
|
| 9 |
|
| 10 |
def transcribe_audio(audio_file):
|
|
|
|
| 18 |
|
| 19 |
def get_answer(context, question):
|
| 20 |
"""Get an answer from the LLM based on the context and question."""
|
| 21 |
+
prompt = (
|
| 22 |
+
"As an intelligent coding assistant, your task is to provide clear, concise, and accurate answers to coding-related questions. "
|
| 23 |
+
"Below are examples of questions and the kind of direct answers expected:\n\n"
|
| 24 |
+
"Example Question 1: How can I remove duplicates from a list in Python?\n"
|
| 25 |
+
"Example Answer 1: Use the set() function to convert the list to a set, which removes duplicates, then convert it back to a list.\n\n"
|
| 26 |
+
"Example Question 2: What's the difference between '==' and '===' in JavaScript?\n"
|
| 27 |
+
"Example Answer 2: '==' checks for equality of values after type coercion, while '===' checks for both value and type equality without coercion.\n\n"
|
| 28 |
+
"Example Question 3: How to check if a key exists in a dictionary in Python?\n"
|
| 29 |
+
"Example Answer 3: Use the 'in' keyword, like 'if key in my_dict:'.\n\n"
|
| 30 |
+
"Based on the above examples, answer the following question:\n\n"
|
| 31 |
+
f"Question: {question}\n"
|
| 32 |
+
"Answer:"
|
| 33 |
+
)
|
| 34 |
headers = {"Authorization": f"Bearer {API_TOKEN}"}
|
| 35 |
+
|
| 36 |
+
# Adjust generation parameters for more focused and relevant responses
|
| 37 |
+
payload = {
|
| 38 |
+
"inputs": prompt,
|
| 39 |
+
"parameters": {
|
| 40 |
+
"temperature": 0.3, # More deterministic
|
| 41 |
+
"top_p": 0.95, # Consider top 90% probable tokens at each step
|
| 42 |
+
"max_new_tokens": 100, # Limit the response length
|
| 43 |
+
"repetition_penalty": 1.2, # Discourage repetition
|
| 44 |
+
"num_return_sequences": 1, # Number of responses to generate
|
| 45 |
+
"return_full_text": False, # Return only generated text, not the full prompt
|
| 46 |
+
"top_k" : 50,
|
| 47 |
+
"truncate" : 24576,
|
| 48 |
+
"max_new_tokens" : 8192,
|
| 49 |
+
"stop" : ["</s>"]
|
| 50 |
+
},
|
| 51 |
+
"options": {
|
| 52 |
+
"use_cache": True # Use cached responses when available
|
| 53 |
+
}
|
| 54 |
+
}
|
| 55 |
+
|
| 56 |
+
response = requests.post(LLM_API_URL, headers=headers, json=payload)
|
| 57 |
answer = json.loads(response.content.decode("utf-8"))[0].get("generated_text", "Answer not available")
|
| 58 |
return answer
|
| 59 |
|
| 60 |
+
|
| 61 |
def transcribe_and_answer(audio_file, question):
|
| 62 |
"""Process the audio file for transcription and use the result to get an answer to a question."""
|
| 63 |
transcription = transcribe_audio(audio_file)
|
| 64 |
answer = get_answer(transcription, question)
|
| 65 |
return transcription, answer
|
| 66 |
|
| 67 |
+
# Create the Gradio app
|
| 68 |
+
import gradio as gr
|
| 69 |
+
|
| 70 |
# Create the Gradio app
|
| 71 |
with gr.Blocks() as app:
|
| 72 |
+
gr.HTML("""
|
| 73 |
+
<div style="display: flex; align-items: center; justify-content: center; margin-bottom: 20px;">
|
| 74 |
+
<img src="https://huggingface.co/spaces/sinatayebati/Talking-Duck/resolve/main/assets/talking-duck-logo.webp" alt="Talking Duck Logo" style="width: 120px;"/>
|
| 75 |
+
<div style="margin-left: 20px;">
|
| 76 |
+
<h1 style="font-weight: bold; font-size: 32px; margin: 0;">TALKING DUCK</h1>
|
| 77 |
+
<h3 style="margin: 0;">An Audio to Text Q&A Chatbot</h3>
|
| 78 |
+
</div>
|
| 79 |
+
</div>
|
| 80 |
+
<p style="text-align: center;">Your swift coding sidekick. Speak your code queries, and let the duck do the magic.</p>
|
| 81 |
+
""")
|
| 82 |
+
|
| 83 |
+
gr.Markdown("""
|
| 84 |
+
<div style="background-color: #0A192F; color: white; padding: 20px; border-radius: 10px; margin-bottom: 20px;">
|
| 85 |
+
<div style="font-size: 16px; font-weight: bold; text-align: center; margin-bottom: 10px;">Models running on backend</div>
|
| 86 |
+
<div style="display: flex; justify-content: space-around; align-items: center;">
|
| 87 |
+
<div>
|
| 88 |
+
<img src="https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/mistral-logo.png" alt="Mistral Logo" style="width: 40px; margin-bottom: 10px;"/>
|
| 89 |
+
<div style="font-size: 14px;">mistralai/Mistral-7B-v0.1</div>
|
| 90 |
+
<a href="https://huggingface.co/mistralai/Mistral-7B-v0.1" target="_blank" style="color: white; text-decoration: none; font-size: 12px;">Model Page</a>
|
| 91 |
+
</div>
|
| 92 |
+
<div>
|
| 93 |
+
<img src="https://aeiljuispo.cloudimg.io/v7/https://cdn-uploads.huggingface.co/production/uploads/1620805164087-5ec0135ded25d76864d553f1.png?w=200&h=200&f=face" alt="Second Model Logo" style="width: 40px; margin-bottom: 10px;"/>
|
| 94 |
+
<div style="font-size: 14px;">openai/whisper-base.en</div>
|
| 95 |
+
<a href="https://huggingface.co/openai/whisper-base.en" target="_blank" style="color: white; text-decoration: none; font-size: 12px;">Model Page</a>
|
| 96 |
+
</div>
|
| 97 |
+
</div>
|
| 98 |
+
</div>
|
| 99 |
+
""")
|
| 100 |
+
|
| 101 |
+
|
| 102 |
with gr.Row():
|
|
|
|
| 103 |
audio_input = gr.Audio(type="filepath", label="Upload your audio question")
|
| 104 |
question_input = gr.Textbox(label="Type your question here")
|
| 105 |
answer_button = gr.Button("Get Answer")
|
|
|
|
| 110 |
answer_button.click(transcribe_and_answer, inputs=[audio_input, question_input], outputs=[transcription_output, answer_output])
|
| 111 |
|
| 112 |
|
| 113 |
+
|
| 114 |
if __name__ == "__main__":
|
| 115 |
+
app.launch(server_name="0.0.0.0")
|
| 116 |
|
| 117 |
|
| 118 |
|
docker-compose.yml
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Example docker-compose.yml for Gradio app
|
| 2 |
+
version: '3'
|
| 3 |
+
services:
|
| 4 |
+
gradio-app:
|
| 5 |
+
build: .
|
| 6 |
+
ports:
|
| 7 |
+
- "7860:7860"
|
| 8 |
+
environment:
|
| 9 |
+
- HF_API_TOKEN
|
helper.txt
ADDED
|
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Ensure Docker Buildx is used and setup
|
| 2 |
+
docker buildx create --use
|
| 3 |
+
|
| 4 |
+
# Build your Docker image with Buildx (assuming you're now working locally and not pushing to a registry)
|
| 5 |
+
docker buildx build --load --secret id=HF_API_TOKEN,env=HF_API_TOKEN -t gradio-app .
|
| 6 |
+
|
| 7 |
+
# Now run your Docker container, passing the secret environment variable
|
| 8 |
+
docker run -it -p 7860:7860 -e HF_API_TOKEN=$HF_API_TOKEN gradio-app
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
docker build -t gradio-app .
|
| 12 |
+
|
| 13 |
+
docker run -it -p 7860:7860 -e HF_API_TOKEN=$HF_API_TOKEN gradio-app
|
| 14 |
+
|
| 15 |
+
docker-compose up --build
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
"--host", "0.0.0.0", "--port", "7860"
|