Umranz/Shruti-Soft-2.6b
Overwrite chat_template.jinja with automatic Shruti system prompt fallback
Update chat_template in tokenizer_config.json
Inject automatic default Shruti system prompt into chat_template
Fix tokenizer_class compatibility for transformers v4 and v5
Embed graphical sweep results table image into README
Embed graphical sweep results table image into README
Upload sweep results table graphic
Add technical benchmark: Why Shruti-Soft-2.6B outperforms 7B/8B models
Update YAML frontmatter tags for maximum search visibility
Remove markdown table in favor of visual infographic
Embed training metrics infographic into README
Upload Phase 3 training metrics infographic diagram
Add technical benchmark: Why Shruti-Soft-2.6B outperforms 7B/8B models
Add technical benchmark section: Why Shruti-Soft-2.6B outperforms 7B/8B models
Embed graphical persona and pipeline diagrams into README
Embed graphical persona and pipeline diagrams into README
Upload engineering pipeline diagram image
Upload persona specifications diagram image
Update README.md
Update model card with W&B image and pure metrics formatting
Upload WandB training curve chart
Remove emojis and clean model card for professional formatting
Add premium model card with banner, training pipeline, and WandB charts
Upload Shruti banner image
Update model card for full merged release
Phase 3: Full standalone merged model (5.2 GB safetensors)
Add model card
Phase 3: Shruti-Soft fine-tune (source: Umranz/Shruti-Soft-2.6b-run-b)
initial commit