Upload README.md with huggingface_hub

README.md

@@ -21,106 +21,39 @@ base_model: eulogik/TinyDoc-VLM-256M
 library_name: peft
 pipeline_tag: visual-question-answering
 ---
-
+---
 # TinyDoc-VLM LoRA Checkpoint
 
-> **Fine-tuned document AI.** 2.7M trainable params. 15 hours on a Mac. Loss: 43 → 15.
-
-[![GitHub](https://img.shields.io/badge/GitHub-⭐-181717?logo=github)](https://github.com/eulogik/TinyDoc-VLM)
-[![Base Model](https://img.shields.io/badge/Base-eulogik/TinyDoc--VLM--256M-yellow)](https://huggingface.co/eulogik/TinyDoc-VLM-256M)
-[![HF Space](https://img.shields.io/badge/🤗-Demo-yellow)](https://huggingface.co/spaces/eulogik/TinyDoc-VLM)
-
-## What is this?
-
-A **LoRA adapter** for TinyDoc-VLM-256M that fine-tunes the model on document understanding tasks. Only **2.7M params** (0.93% of total) are trained, making it efficient to train and deploy.
-
-## Quick Start
+> ## ⚠️ RETIRED — adapter makes outputs worse
+>
+> This LoRA adapter is part of the failed training line of the 256M checkpoint.
+> In paired evaluation it **degrades generation into literal degeneration**
+> (`"$ "$ "$"`, `22222`) compared even already-broken base weights, which
+> themselves measure **0.0% OCRBench (n=1,000)**. Loss values below are training
+> loss only — they never correlated with output quality. Kept for
+> reproducibility; do not use for extraction.
 
-```python
-from tinydoc_vlm import TinyDocVLMForConditionalGeneration, TinyDocVLMProcessor
-from peft import PeftModel
+## What actually ships
 
-# Load base model
-model = TinyDocVLMForConditionalGeneration.from_pretrained("eulogik/TinyDoc-VLM-256M")
+The working product is the **local grounded-extraction SDK** running on free
+`ollama:qwen2.5vl:3b` — measured **SROIE field F1 0.870 vs 0.376** for the free
+PP-OCR+heuristics competitor (same 100 docs, same scorer):
+[github.com/eulogik/TinyDoc-VLM](https://github.com/eulogik/TinyDoc-VLM).
 
-# Apply LoRA adapter
-model = PeftModel.from_pretrained(model, "eulogik/TinyDoc-VLM-LoRA")
-
-# Merge for inference (optional, but faster)
-model = model.merge_and_unload()
-
-processor = TinyDocVLMProcessor()
-```
-
-## Training Details
+## Historical training record (2026-06)
 
 | Parameter | Value |
-|-----------|-------|
-| Base model | [eulogik/TinyDoc-VLM-256M](https://huggingface.co/eulogik/TinyDoc-VLM-256M) |
-| LoRA rank | 16 |
-| LoRA alpha | 32 |
-| Trainable params | 2,727,936 (0.93% of total) |
+|---|---|
+| Base model | [eulogik/TinyDoc-VLM-256M](https://huggingface.co/eulogik/TinyDoc-VLM-256M) (retired) |
+| LoRA rank / alpha | 16 / 32 |
+| Trainable params | 2,727,936 (0.93%) |
 | Target modules | q_proj, v_proj, k_proj, o_proj |
-| Training data | 3,000 synthetic documents (6,815 QA pairs) |
-| Training steps | 17,000 |
-| Best step | 14,000 (loss: 15.0) |
-| Final loss | 17.2 (from 43.3) |
-| Hardware | Apple M4 Mac |
-| Training time | 15.1 hours |
-
-## Training Curve
-
-```
-Step 25:    43.3  ████████████████████████████████████████████
-Step 500:   25.7  ███████████████████████████
-Step 1000:  20.9  █████████████████████
-Step 5000:  18.6  ██████████████████
-Step 10000: 16.5  ████████████████
-Step 14000: 15.0  ███████████████ ★ Best
-Step 17000: 17.2  █████████████████
-```
-
-## What This Trains
-
-- **Document OCR** — Printed text recognition
-- **Form field extraction** — Key-value pair extraction
-- **Receipt/invoice parsing** — Amount, date, vendor extraction
-- **Table structure understanding** — Cell extraction
-- **Visual question answering** — Answer questions about documents
-
-## How to Reproduce
-
-```bash
-# Clone repo
-git clone https://github.com/eulogik/TinyDoc-VLM.git
-cd TinyDoc-VLM
-pip install -e .
-
-# Generate synthetic docs
-python data/synthetic/generator.py --num-docs 3000 --output-dir data/synthetic/output
-
-# Train LoRA (17K steps, ~15 hours on M4)
-python training/fast_train.py     --manifest data/synthetic/output/manifest.jsonl     --data-root data/synthetic     --steps 17000 --batch-size 1 --grad-accum 4 --device mps
-
-# Or use overnight script
-bash training/overnight_train.sh
-```
-
-## Next Steps
-
-- Scale to 10K+ documents
-- Add public benchmarks (DocVQA, FUNSD, CORD)
-- Train for 50K+ steps
-- Export to ONNX
-
-## Links
+| Data | 3,000 synthetic documents (6,815 QA pairs) |
+| Steps / best step | 17,000 / 14,000 (train loss 15.0) |
+| Hardware / time | Apple M4, 15.1 h |
 
-| Resource | URL |
-|----------|-----|
-| Base Model | [eulogik/TinyDoc-VLM-256M](https://huggingface.co/eulogik/TinyDoc-VLM-256M) |
-| GitHub | [github.com/eulogik/TinyDoc-VLM](https://github.com/eulogik/TinyDoc-VLM) |
-| Live Demo | [huggingface.co/spaces/eulogik/TinyDoc-VLM](https://huggingface.co/spaces/eulogik/TinyDoc-VLM) |
-| Training Script | [training/fast_train.py](https://github.com/eulogik/TinyDoc-VLM/blob/main/training/fast_train.py) |
+Training loss trajectory (loss only — not a quality metric):
+`43.3 → 25.7 → 20.9 → 18.6 → 16.5 → 15.0 (best, step 14k) → 17.2 (final)`
 
 ## License