Image-Text-to-Text
Transformers.js
ONNX
English
llava
text-generation
multimodal
qwen
conversational
custom_code
Instructions to use Xenova/nanoLLaVA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers.js
How to use Xenova/nanoLLaVA with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('image-text-to-text', 'Xenova/nanoLLaVA');
Add `num_image_tokens` to config.json
#2
by Xenova HF Staff - opened
- config.json +3 -0
config.json
CHANGED
|
@@ -40,6 +40,7 @@
|
|
| 40 |
"model_type": "qwen2",
|
| 41 |
"num_attention_heads": 16,
|
| 42 |
"num_hidden_layers": 24,
|
|
|
|
| 43 |
"num_key_value_heads": 16,
|
| 44 |
"rms_norm_eps": 1e-06,
|
| 45 |
"rope_theta": 1000000.0,
|
|
@@ -58,10 +59,12 @@
|
|
| 58 |
"model_type": "siglip_vision_model",
|
| 59 |
"num_attention_heads": 16,
|
| 60 |
"num_hidden_layers": 27,
|
|
|
|
| 61 |
"patch_size": 14
|
| 62 |
},
|
| 63 |
"num_attention_heads": 16,
|
| 64 |
"num_hidden_layers": 24,
|
|
|
|
| 65 |
"num_key_value_heads": 16,
|
| 66 |
"rms_norm_eps": 1e-06,
|
| 67 |
"rope_theta": 1000000.0,
|
|
|
|
| 40 |
"model_type": "qwen2",
|
| 41 |
"num_attention_heads": 16,
|
| 42 |
"num_hidden_layers": 24,
|
| 43 |
+
"num_image_tokens": 729,
|
| 44 |
"num_key_value_heads": 16,
|
| 45 |
"rms_norm_eps": 1e-06,
|
| 46 |
"rope_theta": 1000000.0,
|
|
|
|
| 59 |
"model_type": "siglip_vision_model",
|
| 60 |
"num_attention_heads": 16,
|
| 61 |
"num_hidden_layers": 27,
|
| 62 |
+
"num_image_tokens": 729,
|
| 63 |
"patch_size": 14
|
| 64 |
},
|
| 65 |
"num_attention_heads": 16,
|
| 66 |
"num_hidden_layers": 24,
|
| 67 |
+
"num_image_tokens": 729,
|
| 68 |
"num_key_value_heads": 16,
|
| 69 |
"rms_norm_eps": 1e-06,
|
| 70 |
"rope_theta": 1000000.0,
|