Spaces:
Running
Running
Commit History
-
delete debug and fix files for audio before moving onto YuE Running
futurespyhi73d03a5 -
Update lyrics with improved format recognition - extract only segments
futurespyhi8b5e5aa -
add --rescale to enable audio rescaling to proper volume
futurespyhi326af3b -
Enhance audio generation debugging and error handling
futurespyhi- Add comprehensive debugging output for audio decoding pipeline - Implement silent audio detection with amplitude checking - Improve error handling with try-catch blocks for robustness - Expand critical file validation in model download process - Optimize audio file search priority for better quality output - Update gitignore to exclude test files and logs3dfa434
-
use absolute path for xcodec_base
futurespyhi28df979 -
add comprehensive debug printing for downloaded files check
futurespyhi64ea52e -
Set environment variable to skip LFS files during clone
futurespyhi95bb127 -
1.try using git clone for structure and use wget for critical lfs files
futurespyhi06059a8 -
1.modify the logic of downloading_xcodec_models: using git clone instead of huggingface_hub 2. modify .gitignore
futurespyhi297dc31 -
modify GPU time for compatibility on A10G large and L40s
futurespyhic87f4fd -
1. modify generated files search logic 2.modify time to 12 mins for H200
futurespyhi47f310c -
modify the logic of response_audio and output binding to make sure current lyrics to be correctly set up
futurespyhi2e48849 -
add GPU running time for less powerful GPUs
futurespyhi1f93c91 -
update lyrics representing logic
futurespyhi7e55e4f -
modify conversation lyrics extraction for compatibility
futurespyhi51c37c6 -
1.modify flash-attn2 for other spaces 2.update .gitignore
futurespyhi2ce587d -
revise expected time
futurespyhi77da414 -
improve state.lyrics of text input
futurespyhi26a0c39
-
Improve app.py: unified text input experience and performance optimizations
futurespyhi- Unify text and voice input processing to use conversational generation logic - Optimize flash-attn compilation with dynamic CPU core allocation using min(4, cpu_count()) - Update performance notice to reflect realistic generation time (6-8 minutes) - Add multiprocessing import for CPU count detection966b1a5
-
updata gradio logic: trying to solve generation error of current lyrics
futurespyhi6cfc4a3 -
increase quality
futurespyhi921388c -
reduce time for GPU use
futurespyhibe99110 -
decrease quality for fast generation
futurespyhi74aed6e -
add GPU processing time for full model
futurespyhi035fdcd -
Fix split_lyrics regex to properly detect all lyric segments
futurespyhi +1: Claude- Update regex pattern from \[(\w+)\] to \[([^]]+)\] to handle numbered verses - Previous pattern only matched CHORUS, missing VERSE 1, VERSE 2, etc. - Now correctly identifies all segments for proper YuEGP music generation - Add tests/ to .gitignore to exclude debug scripts 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>a128360 -
Fix undefined variables by extracting parameters from cmd
futurespyhi- Extract run_n_segments and max_new_tokens from cmd list for logging - Add safe fallback values if parameters not found in cmd - Resolve variable not defined warnings while maintaining functionality2ab39a8 -
Add detailed progress updates and extend timeouts for music generation
futurespyhi- Increase GPU timeout to 5 minutes for complex generation - Add background thread for progress updates every 15 seconds - Show detailed stages: Stage 1 (7B model), Stage 2 (1B model), audio decoding - Extend subprocess timeout to 30 minutes for safety - Add generation settings and time estimates in logs - Improve user feedback during long generation processe3adaee -
Fix lyrics format conversion for YuEGP compatibility
futurespyhi- Convert **VERSE**, **CHORUS** format to [VERSE], [CHORUS] format - Remove AI commentary and keep only actual lyrics content - Add regex processing to clean lyrics for YuE model parsing - Add debugging output to track conversion processce7990e -
Add debugging output for lyrics processing
futurespyhi- Print raw lyrics input from UI - Print formatted lyrics content and length - Print lyrics file path and written content - Help identify where lyrics are being lost in the pipeline9853042 -
Fix syntax errors in infer.py
futurespyhi- Remove incomplete try-except blocks that caused syntax errors - Fix indentation issues in generation loop - Clean up orphaned except clauses - Restore proper code structure while keeping debug informationb81feb7 -
Add simple debug check for raw_output variable
futurespyhi- Add debug information when raw_output is not defined - Print generation loop parameters for troubleshooting - Provide clearer error message for generation failures0198d75 -
Add exception handling to YuEGP generation loop
futurespyhi- Add try-catch around generation loop to capture actual errors - Provide detailed error messages and stack traces for debugging - Handle partial output recovery when possible - Prevent undefined raw_output variable errora4c1ace -
Fix huggingface_hub version conflict
futurespyhi- Update huggingface_hub to >=0.28.1 to satisfy gradio 5.25.0 requirements - Resolve dependency conflict between transformers, gradio, and huggingface_hub57403ec -
Add runtime model downloading to avoid storage limits
futurespyhi- Add download_models.py script to download xcodec_mini_infer models at runtime - Modify app.py to automatically download models on startup - Add huggingface_hub dependency for model downloads - Remove large model files from repository to stay under 1GB limit3bfebb6 -
add debugging for generation failure
futurespyhief411bc -
avoid using format_lyrics_for_yue to simplify the logic
futurespyhi9362c34 -
Fix lyrics generation function call and response formatting
futurespyhi- Correct parameter order for generate_structured_lyrics() call in respond() function - Add format_lyrics import to properly format SongStructure objects for display - Fix 'takes 4 positional arguments but 5 were given' error in chat interface7b364b8 -
Fix API key validation and Gradio chatbot format issues
futurespyhi- Remove unused OPENAI_API_KEY requirement and update to use GEMINI_API_KEY - Fix Gradio chatbot format errors by ensuring all functions return proper [user, assistant] tuple format - Resolve 'Data incompatible with tuples format' errors in respond() and response_audio() functions1fba260 -
Fix packages.txt format - remove comments for proper apt-get parsing
futurespyhi7ef5bb3
-
Complete MiloMusic implementation with voice-to-song generation
futurespyhi🎵 Features: - Voice recording and transcription (Whisper via Groq) - AI lyrics generation (LLaMA + Gemini) - High-quality music synthesis (YuE model) - Optimized for Hugging Face Spaces deployment 🔧 Technical: - Transformers patches for 3x performance boost - Flash Attention support - Comprehensive dependency management - Smart caching and optimization658e790 -
initial commit
Verified signature10e24ec