Spaces:
Runtime error
Runtime error
Commit History
-
share=True
Ерёменко Иванf9f9f66 -
Add better token detection and error message
Ерёменко Иван- Try multiple token environment variables (HUGGINGFACE_API_TOKEN, HF_TOKEN, HUGGINGFACE_HUB_TOKEN) - Add clear error message if no token is found - Help user understand they need to set token in Space settings37991fe -
HUGGINGFACE_API_TOKEN added
Ерёменко Иван9275aec -
Follow official Hugging Face MCP Course example exactly
Ерёменко Иван- Use InferenceClientModel with HUGGINGFACE_API_TOKEN (required for API access) - Simplify code structure to match official example - Remove custom error handling and logging - Use exact same pattern as official documentation - Should resolve 'Error in generating model output' by providing proper API token6bfcd4b -
Switch to InferenceClientModel to avoid TransformersModel issues
Ерёменко Иван- Replace TransformersModel with InferenceClientModel - Use microsoft/DialoGPT-medium through Inference API - Avoid local model loading and tokenization issues - InferenceClientModel is designed to work with smolagents - Should resolve all 'index out of range' and tokenization errors5029991 -
Try distilgpt2 as lighter alternative and simplify initialization
Ерёменко Иван- Add distilgpt2 (82MB) as second option - much lighter than DialoGPT - Remove complex GPT-2 tokenizer configuration that didn't help - Keep DialoGPT-small as primary option - Simplify model initialization for all models08e8abd -
Reorder models to prioritize smaller models for memory constraints
Ерёменко Иван- Put DialoGPT-small first (351MB) - fits in 16Gi memory - Put gpt2 second (548MB) - also fits in memory - Move zephyr-7b-beta to last (too large for 16Gi) - Should resolve 'Memory limit exceeded' error on Hugging Face Spacescb170ae -
Simplify model initialization and try zephyr-7b-beta
Ерёменко Иван- Remove complex DialoGPT parameters that may not be supported - Add zephyr-7b-beta as primary model (modern chat model) - Use simple TransformersModel initialization for all models - Should resolve 'index out of range' errors388c175 -
Fix DialoGPT tokenization issues and add fallback
Ерёменко Иван- Configure proper parameters for DialoGPT models - Set max_new_tokens=100 to prevent sequence length errors - Add pad_token_id=50256 for proper tokenization - Implement fallback sentiment analysis when agent fails - Add temperature and sampling parameters for better generation8c4aa8f -
Fix chat template error by using DialoGPT model
Ерёменко Иван- Replace GPT-2 with DialoGPT-small which supports chat format - Add fallback mechanism for multiple models - DialoGPT models have built-in chat template support - Should resolve 'tokenizer.chat_template is not set' errorfd41aea -
Fix Gradio launch parameters
Ерёменко Иван- Remove unsupported cache_examples parameter - Keep show_error=True for debugging - App should now launch successfully1226cf0
-
Fix app structure - move Gradio launch outside try-except blocks
Ерёменко Иван- Move demo.launch() outside exception handling blocks - Ensure Gradio interface is always created and launched - Fix runtime error where app initialized but didn't start web server - Maintain proper error handling for MCP connection9337a4c -
Improve app stability and debugging
Ерёменко Иван- Add detailed logging for troubleshooting - Separate MCP connection from core app startup - Graceful fallback when MCP server unavailable - Enhanced error handling with emoji indicators - Disable example caching for faster startupbfb7137 -
update app.py
Ерёменко Иванd7c4d56 -
Replace InferenceClientModel with local TransformersModel
Ерёменко Иван- Remove dependency on Hugging Face Inference API keys - Use local DialoGPT-medium model through transformers - Add proper tokenizer configuration for DialoGPT - Add necessary dependencies: transformers, torch, accelerate - App now works without external API keys while preserving MCP functionalityd129ef0 -
Fix API key issue and improve error handling
Ерёменко Иван- Replace problematic meta-llama model with fallback models - Add robust error handling for MCP server connection - Implement model fallback mechanism (zephyr-7b-beta, DialoGPT, gpt2) - Improve logging and error messages - Allow app to run without MCP tools if server is unavailableb0e801a -
Initial commit: MCP Client with Gradio interface
Ерёменко Иван79016d5