Visual Document Retrieval
Transformers
Safetensors
multilingual
qwen3_5
feature-extraction
text
image
multimodal-embedding
vidore
colbert
colqwen3_5
multilingual-embedding
custom_code
Instructions to use webAI-Official/webAI-ColVec1.1-4b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use webAI-Official/webAI-ColVec1.1-4b with Transformers:
# Load model directly from transformers import AutoProcessor, AutoModel processor = AutoProcessor.from_pretrained("webAI-Official/webAI-ColVec1.1-4b", trust_remote_code=True) model = AutoModel.from_pretrained("webAI-Official/webAI-ColVec1.1-4b", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Accept forwarded kwargs in replace_image_token (e.g. return_tensors from transformers 5.15 ProcessorMixin)
Browse files
processing_colqwen35_bidirection.py
CHANGED
|
@@ -166,7 +166,7 @@ class ColQwen35BidirectionProcessor(ProcessorMixin):
|
|
| 166 |
if cur_min_pixels > max_pixels:
|
| 167 |
self.image_processor.min_pixels = max_pixels
|
| 168 |
|
| 169 |
-
def replace_image_token(self, image_inputs: dict, image_idx: int) -> str:
|
| 170 |
"""Expand one ``<|image_pad|>`` placeholder into its per-image token run.
|
| 171 |
|
| 172 |
``ProcessorMixin.__call__`` delegates placeholder expansion here, and
|
|
|
|
| 166 |
if cur_min_pixels > max_pixels:
|
| 167 |
self.image_processor.min_pixels = max_pixels
|
| 168 |
|
| 169 |
+
def replace_image_token(self, image_inputs: dict, image_idx: int, **kwargs) -> str:
|
| 170 |
"""Expand one ``<|image_pad|>`` placeholder into its per-image token run.
|
| 171 |
|
| 172 |
``ProcessorMixin.__call__`` delegates placeholder expansion here, and
|