Spaces:
Build error
Build error
Yongdong Wang
commited on
Commit
·
a1f2008
1
Parent(s):
9d3765e
Add JSON processing support for Python dict format in model responses
Browse files- Added JsonProcessor module to handle Python dictionary to JSON conversion
- Enhanced response processing to automatically format single quotes and
None values
- Updated requirements.txt to include loguru dependency
- Model responses now support both Python dict and JSON formats seamlessly
- app.py +31 -0
- json_processor.py +46 -0
- requirements.txt +1 -0
app.py
CHANGED
|
@@ -2,6 +2,8 @@ import gradio as gr
|
|
| 2 |
import spaces # Import spaces module for ZeroGPU
|
| 3 |
from huggingface_hub import login
|
| 4 |
import os
|
|
|
|
|
|
|
| 5 |
|
| 6 |
# 1) Read Secrets
|
| 7 |
hf_token = os.getenv("HUGGINGFACE_TOKEN")
|
|
@@ -93,6 +95,32 @@ def load_model_on_gpu():
|
|
| 93 |
print(f"❌ Model loading failed: {load_error}")
|
| 94 |
raise load_error
|
| 95 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 96 |
@spaces.GPU(duration=60) # GPU inference
|
| 97 |
def generate_response_gpu(prompt, max_tokens=512):
|
| 98 |
"""Generate response - executed on GPU"""
|
|
@@ -148,6 +176,9 @@ def generate_response_gpu(prompt, max_tokens=512):
|
|
| 148 |
elif len(response) > len(formatted_prompt):
|
| 149 |
response = response[len(formatted_prompt):].strip()
|
| 150 |
|
|
|
|
|
|
|
|
|
|
| 151 |
return response if response else "❌ No response generated. Please try again with a different prompt."
|
| 152 |
|
| 153 |
except Exception as generation_error:
|
|
|
|
| 2 |
import spaces # Import spaces module for ZeroGPU
|
| 3 |
from huggingface_hub import login
|
| 4 |
import os
|
| 5 |
+
from json_processor import JsonProcessor
|
| 6 |
+
import json
|
| 7 |
|
| 8 |
# 1) Read Secrets
|
| 9 |
hf_token = os.getenv("HUGGINGFACE_TOKEN")
|
|
|
|
| 95 |
print(f"❌ Model loading failed: {load_error}")
|
| 96 |
raise load_error
|
| 97 |
|
| 98 |
+
def process_json_in_response(response):
|
| 99 |
+
"""Process and format JSON content in the response"""
|
| 100 |
+
try:
|
| 101 |
+
# Check if response contains JSON-like content
|
| 102 |
+
if '{' in response and '}' in response:
|
| 103 |
+
processor = JsonProcessor()
|
| 104 |
+
|
| 105 |
+
# Try to process the response for JSON content
|
| 106 |
+
processed_json = processor.process_response(response)
|
| 107 |
+
|
| 108 |
+
if processed_json:
|
| 109 |
+
# Format the JSON nicely
|
| 110 |
+
formatted_json = json.dumps(processed_json, indent=2, ensure_ascii=False)
|
| 111 |
+
# Replace the JSON part in the response
|
| 112 |
+
import re
|
| 113 |
+
json_pattern = r'\{.*\}'
|
| 114 |
+
match = re.search(json_pattern, response, re.DOTALL)
|
| 115 |
+
if match:
|
| 116 |
+
# Replace the matched JSON with the formatted version
|
| 117 |
+
response = response.replace(match.group(), formatted_json)
|
| 118 |
+
|
| 119 |
+
return response
|
| 120 |
+
except Exception:
|
| 121 |
+
# If processing fails, return original response
|
| 122 |
+
return response
|
| 123 |
+
|
| 124 |
@spaces.GPU(duration=60) # GPU inference
|
| 125 |
def generate_response_gpu(prompt, max_tokens=512):
|
| 126 |
"""Generate response - executed on GPU"""
|
|
|
|
| 176 |
elif len(response) > len(formatted_prompt):
|
| 177 |
response = response[len(formatted_prompt):].strip()
|
| 178 |
|
| 179 |
+
# Process JSON if present in response
|
| 180 |
+
response = process_json_in_response(response)
|
| 181 |
+
|
| 182 |
return response if response else "❌ No response generated. Please try again with a different prompt."
|
| 183 |
|
| 184 |
except Exception as generation_error:
|
json_processor.py
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import json
|
| 2 |
+
import re
|
| 3 |
+
import ast
|
| 4 |
+
from loguru import logger
|
| 5 |
+
|
| 6 |
+
class JsonProcessor:
|
| 7 |
+
def process_response(self, response):
|
| 8 |
+
try:
|
| 9 |
+
# Search for JSON string in the response
|
| 10 |
+
json_str_match = re.search(r'\{.*\}', response, re.DOTALL)
|
| 11 |
+
if json_str_match:
|
| 12 |
+
# Get the matched JSON string
|
| 13 |
+
json_str = json_str_match.group()
|
| 14 |
+
logger.debug(f"Full JSON string: {json_str}")
|
| 15 |
+
|
| 16 |
+
# Try to parse as Python literal first, then convert to JSON
|
| 17 |
+
try:
|
| 18 |
+
# First try to evaluate as Python literal
|
| 19 |
+
python_obj = ast.literal_eval(json_str)
|
| 20 |
+
# Convert to proper JSON
|
| 21 |
+
response_json = json.loads(json.dumps(python_obj))
|
| 22 |
+
except (ValueError, SyntaxError):
|
| 23 |
+
# Fall back to string replacement method
|
| 24 |
+
# Replace escape characters and remove trailing commas
|
| 25 |
+
json_str = json_str.replace("\\", "")
|
| 26 |
+
json_str = json_str.replace(r'\\_', '_')
|
| 27 |
+
json_str = re.sub(r',\s*}', '}', json_str)
|
| 28 |
+
json_str = re.sub(r',\s*\]', ']', json_str)
|
| 29 |
+
|
| 30 |
+
# Convert Python format to JSON format
|
| 31 |
+
json_str = json_str.replace("'", '"') # Single quotes to double quotes
|
| 32 |
+
json_str = json_str.replace('None', 'null') # Python None to JSON null
|
| 33 |
+
|
| 34 |
+
# Parse the JSON string
|
| 35 |
+
response_json = json.loads(json_str)
|
| 36 |
+
return response_json
|
| 37 |
+
else:
|
| 38 |
+
logger.error("No JSON string match found in response.")
|
| 39 |
+
return None
|
| 40 |
+
|
| 41 |
+
except json.JSONDecodeError as e:
|
| 42 |
+
logger.error(f"JSONDecodeError: {e}")
|
| 43 |
+
except Exception as e:
|
| 44 |
+
logger.error(f"Unexpected error: {e}")
|
| 45 |
+
|
| 46 |
+
return None
|
requirements.txt
CHANGED
|
@@ -9,3 +9,4 @@ scipy
|
|
| 9 |
sentencepiece
|
| 10 |
protobuf
|
| 11 |
spaces
|
|
|
|
|
|
| 9 |
sentencepiece
|
| 10 |
protobuf
|
| 11 |
spaces
|
| 12 |
+
loguru
|