# Florence-2-base-ft

Hugging Face: `microsoft/Florence-2-base-ft`
Reconstructed from `config.json` alone. No weights were read.

| | |
|---|---|
| Derived from structure | 82,694,400 |
| In the published checkpoint | 231,567,705 (safetensors.total, 2025-08-04) |
| Delta | -64.3% |
| Comparability | custom-code |
| Layers | 26 |
| Priced on | A10G (24GB) |
| Est. one run | $1.85 |

> This repository ships its own modeling code (`auto_map`, e.g. `configuration_florence2.py`), so `config.json` names a class in the repo rather than an architecture `transformers` defines. The graph below is what those config keys mean under `transformers` semantics, which is not necessarily what the repo's own file builds. A gap here is a statement about what we read, not about the checkpoint.

## Structure

| # | Layer | Type | Output shape |
|---|---|---|---|
| 1 | Input | Input | 1 × 1024 |
| 2 | Embedding | Embedding | 1 × 1024 × 768 |
| 3 | Positional_Embedding | Learned Pos Embed | 1 × 1024 × 768 |
| 4 | Attention_1 | Multi-Head Attention | 1 × 1024 × 768 |
| 5 | Add_1 | Add | 1 × 1024 × 768 |
| 6 | LayerNorm_1_1 | LayerNorm | 1 × 1024 × 768 |
| 7 | FFN_1 | Feed Forward | 1 × 1024 × 768 |
| 8 | Attention_2 | Multi-Head Attention | 1 × 1024 × 768 |
| 9 | Add_2 | Add | 1 × 1024 × 768 |
| 10 | LayerNorm_2_1 | LayerNorm | 1 × 1024 × 768 |
| 11 | FFN_2 | Feed Forward | 1 × 1024 × 768 |
| 12 | Attention_3 | Multi-Head Attention | 1 × 1024 × 768 |
| 13 | Add_3 | Add | 1 × 1024 × 768 |
| 14 | LayerNorm_3_1 | LayerNorm | 1 × 1024 × 768 |
| 15 | FFN_3 | Feed Forward | 1 × 1024 × 768 |
| 16 | Attention_4 | Multi-Head Attention | 1 × 1024 × 768 |
| 17 | Add_4 | Add | 1 × 1024 × 768 |
| 18 | LayerNorm_4_1 | LayerNorm | 1 × 1024 × 768 |
| 19 | FFN_4 | Feed Forward | 1 × 1024 × 768 |
| 20 | Attention_5 | Multi-Head Attention | 1 × 1024 × 768 |
| 21 | Add_5 | Add | 1 × 1024 × 768 |
| 22 | LayerNorm_5_1 | LayerNorm | 1 × 1024 × 768 |
| 23 | FFN_5 | Feed Forward | 1 × 1024 × 768 |
| 24 | Attention_6 | Multi-Head Attention | 1 × 1024 × 768 |
| 25 | Add_6 | Add | 1 × 1024 × 768 |
| 26 | LayerNorm_6_1 | LayerNorm | 1 × 1024 × 768 |
| 27 | FFN_6 | Feed Forward | 1 × 1024 × 768 |
| 28 | Output | Output | 1 × 1024 × 768 |

## Do this to your own model

```
pip install neurarch-trace
neurarch-trace microsoft/Florence-2-base-ft --plan --share
```

- Page: https://neurarch.com/m/florence-2-base-ft.html
- All models: https://neurarch.com/m/index.json
- The checks: https://neurarch.com/r/index.json