Commit History
-
fix: pin transformers stack and force slow tokenizer by default to avoid fast-tokenizer errors
Verified signature9cffd85 -
fix: pin transformers stack and force slow tokenizer by default to avoid fast-tokenizer errors
Verified signature4d8db9e -
fix: pin transformers stack and force slow tokenizer by default to avoid fast-tokenizer errors
Verified signaturedea3286 -
fix: pin transformers stack and force slow tokenizer by default to avoid fast-tokenizer errors
Verified signaturee752fd3 -
fix: pin transformers stack and force slow tokenizer by default to avoid fast-tokenizer errors
Verified signature7828436 -
fix: pin transformers stack and force slow tokenizer by default to avoid fast-tokenizer errors
Verified signaturec7aaa20 -
PROPER MISTRAL 7B TRAINING: v0.3 with LoRA, proper chat format, no logging issues
Verified signature543b060 -
PRODUCTION TRAINING: 150 perfect conversations, no wandb issues, production-ready
Verified signature06e4891 -
SIMPLE WORKING TRAINING: 45 perfect conversations, clean responses, no broken text - QUALITY OVER QUANTITY
Verified signature960b372 -
FINAL AUTHENTIC COMPREHENSIVE: Scripture-based divine personas + 2GB spiritual library integration + all functional features
Verified signature8eaa429 -
FINAL WARM FUNCTIONAL: Updated to use warm divine personas dataset - Personal, loving responses like "Dear child, just as I told Arjuna..."
Verified signature754fb45 -
FINAL FUNCTIONAL: Updated to use complete functional dataset with ALL features having REAL FUNCTIONALITY
Verified signatured7c9e7e
-
CRITICAL FIX: Dynamic LoRA target module detection - resolves ValueError for different model architectures
Verified signaturef9905c2 -
FINAL CLEAN: Comprehensive training notebook - 2925+ conversations, proper divine names, advanced numerology (no trademark names), all fixes applied
Verified signature0488487 -
UPDATE: Replace with CORRECT comprehensive training notebook - 2925+ conversations, proper divine names, advanced numerology, all fixes
Verified signature8275b1f -
Clean up: Remove unnecessary file SPEED_OPTIMIZATION_GUIDE.md
Verified signaturebf0d8fd -
Clean up: Remove unnecessary file CELESTIAL_Comprehensive_Training.ipynb
Verified signaturea6bdad6 -
Add comprehensive integration blueprint for production deployment
Verified signature3dce6bc -
FINAL FIX: All training errors resolved - LoRA adapters properly attached to quantized model
Verified signature1c18b2c -
Update CELESTIAL_Training_Notebook.ipynb
Verified signature8a38293 -
FIXED: Update training arguments - eval_strategy parameter compatibility issue resolved
Verified signaturee3cb044 -
Update README with Sanjay Jumaani numerology, speed optimizations, and comprehensive feature list
Verified signature3faf168 -
Add comprehensive speed optimization guide
Verified signature44ab159 -
Add speed-optimized training notebook (45-90 min) with HF dataset integration
Verified signature93554f7
-
Upload CELESTIAL_Comprehensive_Training.ipynb with huggingface_hub
Verified signaturebe3a48c -
Upload CELESTIAL_Comprehensive_Training.ipynb with huggingface_hub
Verified signaturef8aaaa2
-
Upload CELESTIAL_Comprehensive_Training.ipynb with huggingface_hub
Verified signature587afe3
-
Upload CELESTIAL_Comprehensive_Training.ipynb with huggingface_hub
Verified signature27a50f2 -
Upload CELESTIAL_Comprehensive_Training.ipynb with huggingface_hub
Verified signature8bf9ede -
Upload CELESTIAL_Comprehensive_Training.ipynb with huggingface_hub
Verified signature1f2abdb -
Upload CELESTIAL_Comprehensive_Training.ipynb with huggingface_hub
Verified signature6c391f9 -
initial commit
Verified signature000e682