Update README.md
Browse files
README.md
CHANGED
|
@@ -26,7 +26,7 @@ pipeline_tag: text-generation
|
|
| 26 |
|
| 27 |
# NuMarkdown-reasoning 📄
|
| 28 |
|
| 29 |
-
**NuMarkdown-reasoning** is the first reasoning vision-language model trained specifically to convert documents into clean GitHub-flavoured Markdown.
|
| 30 |
It is a fine-tune of **Qwen 2.5-VL-7B** using ~10k synthetic Doc-to-Reasoning-to-Markdown pairs, followed by an RL phase (GRPO) with a layout-centric reward.
|
| 31 |
|
| 32 |
*(Note: the number of thinking tokens can vary from 20% to 500% the number of tokens in the final answer)*
|
|
@@ -167,7 +167,7 @@ Pàgina 2 de 2
|
|
| 167 |
|
| 168 |
## vLLM:
|
| 169 |
```
|
| 170 |
-
vllm serve numind/NuMarkdown-reasoning --trust_remote_code --limit-mm-per-prompt image=1
|
| 171 |
```
|
| 172 |
|
| 173 |
```python
|
|
@@ -193,8 +193,8 @@ def encode_image(image_path):
|
|
| 193 |
base64_image = encode_image("invoice.png")
|
| 194 |
|
| 195 |
chat_response = client.chat.completions.create(
|
| 196 |
-
model="numind/NuMarkdown-reasoning",
|
| 197 |
-
temperature=0,
|
| 198 |
messages=[
|
| 199 |
{
|
| 200 |
"role": "user",
|
|
@@ -218,7 +218,7 @@ import torch
|
|
| 218 |
from PIL import Image
|
| 219 |
from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration
|
| 220 |
|
| 221 |
-
model_id = "
|
| 222 |
|
| 223 |
processor = AutoProcessor.from_pretrained(
|
| 224 |
model_id,
|
|
|
|
| 26 |
|
| 27 |
# NuMarkdown-reasoning 📄
|
| 28 |
|
| 29 |
+
**NuMarkdown-8B-reasoning** is the first reasoning vision-language model trained specifically to convert documents into clean GitHub-flavoured Markdown.
|
| 30 |
It is a fine-tune of **Qwen 2.5-VL-7B** using ~10k synthetic Doc-to-Reasoning-to-Markdown pairs, followed by an RL phase (GRPO) with a layout-centric reward.
|
| 31 |
|
| 32 |
*(Note: the number of thinking tokens can vary from 20% to 500% the number of tokens in the final answer)*
|
|
|
|
| 167 |
|
| 168 |
## vLLM:
|
| 169 |
```
|
| 170 |
+
vllm serve numind/NuMarkdown-8B-reasoning --trust_remote_code --limit-mm-per-prompt image=1
|
| 171 |
```
|
| 172 |
|
| 173 |
```python
|
|
|
|
| 193 |
base64_image = encode_image("invoice.png")
|
| 194 |
|
| 195 |
chat_response = client.chat.completions.create(
|
| 196 |
+
model="numind/NuMarkdown-8B-reasoning",
|
| 197 |
+
temperature=0.8,
|
| 198 |
messages=[
|
| 199 |
{
|
| 200 |
"role": "user",
|
|
|
|
| 218 |
from PIL import Image
|
| 219 |
from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration
|
| 220 |
|
| 221 |
+
model_id = "numind/NuMarkdown-8B-reasoning"
|
| 222 |
|
| 223 |
processor = AutoProcessor.from_pretrained(
|
| 224 |
model_id,
|