Update README.md

README.md

@@ -1,65 +1,83 @@
 ---
 license: mit
 ---
-
 # ZipPlus Model Card
-**A pre-trained 4-layer GRU model for neural file compression. Compress any file into a PNG image using a neural network and extract it back.**
-This is a pre-trained ByteGRU model for [Zip+](https://github.com/CompactAIOfficial/ZipPlus) - use this instead of training your own.
+
+**A pre-trained 4-layer GRU model for neural file compression. Each compressed file contains its own adapted model — no external model needed to decompress.**
+
+This is a pre-trained ByteGRU model for [Zip+](https://github.com/CompactAIOfficial/ZipPlus).
 
 ## What is this
-This model compresses any file into a PNG image using a neural network (GRU + range coding) and extracts it back. The compressed files look like weird colorful static - perfect for confusing anyone who peeks at your folder.
+
+Zip+ compresses any file into a PNG image using a neural network (GRU + range coding). Each compressed file embeds its own adapted model:
+
 ```
-file.txt → [ByteGRU + Range Coding] → file.txt.zpng.png → [ByteGRU + Range Coding] → file.txt
+file.txt → [ByteGRU + Range Coding] → file.txt.zpng.png → [embedded model] → file.txt
 ```
-**The PNG contains a special `ZPNG` magic header**, so random images won't decompress. Your cat photos are safe. Mostly.
+
+**Every PNG is self-contained** — decompress even if you lose the original model file!
 
 ## Model Details
-- **Architecture**: 2-layer GRU over byte embeddings
-- **Embedding dim**: 64 → Hidden dim: 256
-- **Trained on**: A variety of file types
+
+- **Architecture**: 4-layer GRU over byte embeddings
+- **Embedding dim**: 64 → Hidden dim: 512
+- **Trained on**: FineWeb-Edu (10BT of educational web text) + adaptive per-file training
 - **Entropy coding**: Range coding via Constriction
-- **Output format**: PNG where payload lives in RGB pixel bytes
+- **Output format**: PNG where payload + model live in RGB pixel bytes
 - **Magic header**: `ZPNG` (first 4 bytes)
-- 
+
 ## Requirements
+
 - Python 3.10+
-- PyTorch
+- PyTorch (CUDA recommended)
 - Constriction (`pip install constriction`)
 - Pillow
 - numpy
+- huggingface_hub
+
 ```bash
-pip install torch constriction pillow numpy
+pip install torch constriction pillow numpy huggingface_hub
 ```
+
 ## Quick Start
 
-### Compress a file
+### Compress a file (with auto-adaptation)
 
 ```bash
-python inference.py compress myfile.txt -o myfile.zpng.png -m model.pt
+python inference.py compress myfile.txt -o myfile.zpng.png
 ```
+
+- Automatically adapts model to your file (50 steps)
+- Embeds adapted model in PNG for self-contained decoding
+
 ### Decompress
 
 ```bash
-python inference.py decompress myfile.zpng.png -o restored.txt -m model.pt
+python inference.py decompress myfile.zpng.png -o restored.txt
 ```
-Done. Your file is back. Hopefully.
 
-## Interactive Menu
+Loads the model embedded in the PNG — no external files needed!
 
-Just run `python compressor.py -m model.pt` for the menu. It's vaguely intuitive if you squint.
+### Training (optional)
+
+```bash
+python train.py --grid 128 --steps 10000
+```
+
+Auto-downloads FineWeb-Edu if no corpus specified.
 
 ## Performance
 
-Compression ratio varies. Text files compress okay. Binary files? Less okay. Random data? It might actually grow. That's the fun part.
+- Text files: ~5-20% of original size
+- Works best on files > 10KB
+- Smaller files: embedding overhead (~21MB) may exceed compression gains
 
 ## Warnings
 
-- **Don't lose this model**. Without the model file, your `.zpng.png` files are colorful but useless.
-- **Lossiness is possible**. If the compression produces artifacts, restored files may differ. Check with checksums.
-- **GPU recommended**. CPU inference is tolerable.
-- 
+- **Embedding adds ~21MB** to output — worth it for large files
+- **GPU recommended** for training and compression
+- **Lossless** — verified via SHA256 checksums
+
 ## License
 
-MIT. I'm not liable if this eats your thesis/pixels/anything.
----
-Use it because it's amusing. Or don't
+MIT. I'm not liable if this eats your thesis/pixels/anything.