Spaces:
Sleeping
Sleeping
Commit History
-
Fix architecture flowchart — consistent KV Cache styling, responsive layout
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>9fbf393 -
Add KV Cache block to architecture flowchart
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>9d565ee -
Restore deploy files to root for HF Spaces compatibility
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>ce00f04 -
Add HF Spaces frontmatter to README
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>faf75e0 -
Move KV Cache into Transformer Block section on architecture page
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>bd49299 -
Organize project structure and rewrite README
- Move QA data to data/, deploy files to deploy/ - Remove adapter weights from repo (hosted on HF Hub) - Comprehensive README with architecture details, setup, and references Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>11b8915 -
Add KV Cache section to architecture page, reorder bibliography
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>98e2ecc -
Add KV caching to from-scratch model for faster inference
Prefill phase processes the full prompt and caches K/V tensors. Decode phase only computes the new token and reuses the cache. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>a7c26c0 -
Deploy StoicGPT to HF Spaces
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>842c608 -
Switch to Trainer — handle tokenization manually to avoid trl API issues
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>c035927 -
Fix trl compatibility — use TrainingArguments + older SFTTrainer API
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>18cd34b -
Fix SFTTrainer API for newer trl — tokenizer → processing_class
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>1c69f33 -
Add QA dataset generation, fine-tuning, and inference scripts
- generate_qa.py: generates 7k+ Stoic Q&A pairs via OpenAI API - stoic_qa.jsonl / stoic_qa_chat.jsonl: generated dataset - finetune.py: QLoRA fine-tuning on Qwen 2.5 1.5B - inference.py: CLI inference for the fine-tuned model Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>9929072
-
Clean up root — remove duplicate txt files and corpus.txt
Texts already exist in Stoic Knowledge/. Also ignore .vscode and .claude folders. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>4480722 -
Add README with setup, training, and usage instructions
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>661a829 -
Resolve merge conflict — keep updated data.py
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>09a427e -
Initial commit — StoicGPT
15M parameter GPT trained on Stoic texts (Marcus Aurelius, Seneca, Epictetus). Includes model architecture, training loop, Flask web app with Stoic Oracle UI. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>8257f60
-
Add files via upload
Unverified signature22486ea -
Initial commit
Unverified signature1d63fa8