Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
base_model: rednote-hilab/dots.ocr
|
| 4 |
+
tags:
|
| 5 |
+
- gguf
|
| 6 |
+
- ocr
|
| 7 |
+
- llama-cpp
|
| 8 |
+
- vision
|
| 9 |
+
- image-to-text
|
| 10 |
+
language:
|
| 11 |
+
- en
|
| 12 |
+
- zh
|
| 13 |
+
- multilingual
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# dots.ocr GGUF
|
| 17 |
+
|
| 18 |
+
GGUF conversions of [rednote-hilab/dots.ocr](https://huggingface.co/rednote-hilab/dots.ocr) for use with [llama.cpp](https://github.com/ggml-org/llama.cpp).
|
| 19 |
+
|
| 20 |
+
## Files
|
| 21 |
+
|
| 22 |
+
| File | Size | Description |
|
| 23 |
+
|---|---|---|
|
| 24 |
+
| Dots.Ocr-1.8B-Q8_0.gguf | 1.8 GB | Text model, 8-bit quantized |
|
| 25 |
+
| Dots.Ocr-1.8B-F16.gguf | 3.4 GB | Text model, float16 |
|
| 26 |
+
| mmproj-Dots.Ocr-F16.gguf | 2.4 GB | Vision encoder (mmproj), float16 |
|
| 27 |
+
|
| 28 |
+
## Architecture
|
| 29 |
+
|
| 30 |
+
dots.ocr = Qwen2 text backbone (1.7B params, 28 layers) + modified Qwen2-VL vision encoder (1.2B params, 42 layers).
|
| 31 |
+
|
| 32 |
+
Key differences from Qwen2-VL:
|
| 33 |
+
- Text model is standard Qwen2 with 1D RoPE (not M-RoPE)
|
| 34 |
+
- Vision uses RMSNorm, SiLU gated MLP, Conv2D patches, no attention bias
|
| 35 |
+
- 2D M-RoPE internal to vision encoder only
|
| 36 |
+
|
| 37 |
+
## Usage with llama.cpp
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
> **Note:** Requires llama.cpp with dots.ocr support (pending upstream merge).
|