docs: professional README — Feb 2026
This commit is contained in:
parent
f1fc93f322
commit
ae0c2e31c3
64
README.md
64
README.md
@ -1,44 +1,44 @@
|
||||
<div align="center">
|
||||
|
||||
# Organ Store
|
||||
*Part of the Inference-X Ecosystem*
|
||||
|
||||
[](LICENSE)
|
||||
**Extract, share and transplant AI model components**
|
||||
|
||||
[]()
|
||||
|
||||
Neural networks have anatomy: attention heads, FFN blocks, expert layers, embeddings.
|
||||
The Organ Store is the community marketplace for these components.
|
||||
[inference-x.com](https://inference-x.com)
|
||||
|
||||
</div>
|
||||
|
||||
---
|
||||
|
||||
## What is Organ Store?
|
||||
|
||||
Organ Store is a community marketplace for AI model "organs" — specific layers, attention heads, or MLP blocks that can be transplanted between compatible models.
|
||||
|
||||
## Why organs?
|
||||
|
||||
Fine-tuning entire models is expensive. Fine-tuning specific organs (e.g., "the math reasoning layers") is fast and cheap. Organ Store lets the community share these improvements.
|
||||
|
||||
## How it works
|
||||
|
||||
```
|
||||
Source Model → Extract Organ → Publish → Transplant into Target
|
||||
🧠 ⚗️ 📦 💉
|
||||
1. Extract → use Organ Architecture to pull specific layers
|
||||
2. Fine-tune → train on your dataset
|
||||
3. Share → publish as an organ with metadata
|
||||
4. Transplant → apply to any compatible base model
|
||||
```
|
||||
|
||||
## Deploy
|
||||
```bash
|
||||
pip install fastapi uvicorn
|
||||
python organ_store.py
|
||||
# → API at http://localhost:7939
|
||||
```
|
||||
## Example use cases
|
||||
|
||||
## Organ Types
|
||||
| Type | Description |
|
||||
|------|-------------|
|
||||
| `attention_head` | Specialized attention mechanisms |
|
||||
| `ffn_block` | Feed-forward network layers |
|
||||
| `expert_block` | MoE expert blocks (Mixtral, DeepSeek) |
|
||||
| `embedding` | Domain-specific embedding layers |
|
||||
| `lm_head` | Language model heads |
|
||||
| `custom` | Any other component |
|
||||
- Share "code generation" layers trained on your proprietary codebase
|
||||
- Download and apply "French language" organs to improve a base model
|
||||
- Mix organs from different models to create hybrids
|
||||
|
||||
## API
|
||||
```
|
||||
GET /catalog — Browse organs (filter by type)
|
||||
POST /publish — Publish organ to community
|
||||
GET /organ/{id} — Get organ details + download
|
||||
GET /health — Service status
|
||||
```
|
||||
## Status
|
||||
|
||||
Built in Morocco for the world. 🇲🇦
|
||||
Alpha — concept validated, marketplace in development.
|
||||
|
||||
## License
|
||||
[Business Source License 1.1](LICENSE)
|
||||
**Author:** Salka Elmadani
|
||||
---
|
||||
|
||||
*See also: [organ-architecture](https://git.inference-x.com/inference-x/organ-architecture) · [organ-architect](https://git.inference-x.com/inference-x-community/organ-architect)*
|
||||
|
||||
Loading…
Reference in New Issue
Block a user