docs: full documentation update v3.13.1

README.md

@@ -8,9 +8,9 @@ tags:
 
 [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.20279248.svg)](https://doi.org/10.5281/zenodo.20279248)
 
-# Phi-3.5 Mini Instruct — BigSmall Compressed
+# Phi-3.5 Mini Instruct — Lossless Compressed
 
-**65.6% smaller than [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct). Bit-identical weights. Drop-in replacement.**
+> **7.12 GB → 4.67 GB (34% smaller). Bit-identical weights. Drop-in replacement.**
 
 ## Use it in 2 lines
 
@@ -19,27 +19,29 @@ pip install bigsmall
 ```
 
 ```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
+from transformers import AutoModelForCausalLM
 model = AutoModelForCausalLM.from_pretrained("wpferrell/phi-3.5-mini-instruct-bigsmall")
-tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3.5-mini-instruct")
 ```
 
+It works exactly like loading the original model. No code changes needed.
+
 ## Size comparison
 
-|  | Size |
+| | Size |
 |---|---|
-| Original ([microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)) | 7.2 GB |
+| Original ([microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)) | 7.12 GB |
 | This compressed version | 4.67 GB |
-| Compression ratio | 65.6% of original |
+| Saved | 2.45 GB (34%) |
 
 ## What "lossless" means
 
-Every weight is mathematically identical to the original. Not quantized. Not pruned. Identical model behaviour. Compressed using entropy coding — like ZIP but designed for neural weights.
+Every weight is mathematically identical to the original model.
 
-Every tensor's md5 is stored in the file header; `bigsmall verify` decompresses and verifies bit-for-bit.
+- **Not quantized.** Quantization rounds weights and changes model behaviour.
+- **Not pruned.** Pruning removes parts of the model.
+- **Bit-for-bit identical.** md5 is verified on every tensor at decompression.
 
-## Low VRAM streaming
+## Low-VRAM streaming
 
 ```python
 from bigsmall import BigSmallStreamingModel
@@ -47,43 +49,50 @@ from bigsmall import BigSmallStreamingModel
 model = BigSmallStreamingModel.from_pretrained(
     "wpferrell/phi-3.5-mini-instruct-bigsmall",
     device="cuda",
-    lru_max_vram_gb=2.0,    # cache 2 GB of decoded layers (v3.13.0+)
+    lru_max_vram_gb=2.0,
 )
-out = model.generate(input_ids, max_new_tokens=100)
 ```
 
-Streams layers on demand — uses ~12× less VRAM than a full load.
+Uses up to ~12× less VRAM than standard loading by streaming layers on demand.
 
-## New in BigSmall v3.13.0
+## Decompress to safetensors
 
-- **Delta compression** — if you have the base model already, store a fine-tune as a tiny patch on top:
-  ```bash
-  bigsmall compress my_finetune.safetensors --delta-from base.safetensors -o patch.bs
-  ```
-- **`bigsmall scan`** — analyse a model before compressing.
-- **`bigsmall apply`** — reconstruct a fine-tune from base + patch.
-- **`bigsmall verify --sample 0.001`** — fast probabilistic integrity check.
-- **`bigsmall compress --ecc`** — optional Reed-Solomon error recovery.
-- **`bigsmall compress --resume`** — resumable compression.
+```bash
+pip install bigsmall
+bigsmall decompress wpferrell/phi-3.5-mini-instruct-bigsmall -o phi-3.5-mini-instruct-bigsmall/
+```
 
-[Full release notes →](https://github.com/wpferrell/Bigsmall/releases/tag/v3.13.0)
+## Original model
 
-## All pre-compressed models
+This is a lossless-compressed copy of [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct). All credit to the original authors. The weights are unchanged.
 
-See [wpferrell on HuggingFace](https://huggingface.co/wpferrell) for all available models.
+## Want to compress your own model?
 
-## License
+```bash
+pip install bigsmall
+bigsmall compress my-model/ -o my-model.bs
+```
 
-Model weights: same license as the original model ([microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)).
-BigSmall format: [Elastic License 2.0](https://www.elastic.co/licensing/elastic-license) — free for personal, research, and internal commercial use.
-Commercial/SaaS licensing: wpferrell@gmail.com
+See [github.com/wpferrell/Bigsmall](https://github.com/wpferrell/Bigsmall) for the full docs.
 
-## Requires
+## License
 
-`bigsmall >= 3.13.0` to access the latest features. Older versions (>= 3.0.0) can still decode this model.
+- **Model weights:** mit — same as [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct).
+- **BigSmall format:** [Elastic License 2.0](https://github.com/wpferrell/Bigsmall/blob/main/LICENSE) — free for personal, research, and commercial use.
+- **Commercial SaaS licensing:** wpferrell@gmail.com
 
 ## Citation
 
-If you use BigSmall in research:
+```bibtex
+@misc{bigsmall2026,
+  title={BigSmall: Lossless Neural Network Weight Compression},
+  author={Ferrell, Will},
+  year={2026},
+  doi={10.5281/zenodo.20279248},
+  url={https://doi.org/10.5281/zenodo.20279248}
+}
+```
+
+## Requires
 
-> Ferrell, W. (2026). *BigSmall: Lossless Compression for Neural Network Weights.* Zenodo. https://doi.org/10.5281/zenodo.20279248
+`bigsmall >= 3.13.0` for the latest features. Earlier versions (>= 3.0.0) can still decode this model.