Visual Question Answering
Transformers
Safetensors
Vietnamese
openvivqa
scene-text-vqa
vietnamese
ocr
t5
Instructions to use Kus669/ViSceT5-ViTextVQA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Kus669/ViSceT5-ViTextVQA with Transformers:
# Use a pipeline as a high-level helper # Warning: Pipeline type "visual-question-answering" is no longer supported in transformers v5. # You must load the model directly (see below) or downgrade to v4.x with: # pip install "transformers<5.0.0" from transformers import pipeline pipe = pipeline("visual-question-answering", model="Kus669/ViSceT5-ViTextVQA")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Kus669/ViSceT5-ViTextVQA", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download config.json from Kus669/ViSceT5-ViTextVQA: direct link, hf CLI and curl.
- Browser
- Download file 2.2 kB
-
https://huggingface.co/Kus669/ViSceT5-ViTextVQA/resolve/main/config.json
- Command line
-
hf download hf://Kus669/ViSceT5-ViTextVQA/config.json
-
curl -L -o config.json https://huggingface.co/Kus669/ViSceT5-ViTextVQA/resolve/main/config.json
2.2 kB
| { | |
| "ablation_use_image": true, | |
| "ablation_use_ocr": true, | |
| "ablation_use_qaclip": true, | |
| "architectures": [ | |
| "OpenViVQAModel" | |
| ], | |
| "char_max_num": 50, | |
| "char_num": 168, | |
| "clip_vision_name": "openai/clip-vit-base-patch16", | |
| "consformer_constituent": "tree", | |
| "consformer_gate": true, | |
| "consformer_geometry": true, | |
| "consformer_residual": true, | |
| "consformer_span": true, | |
| "consformer_span_weight": 0.1, | |
| "consformer_tree_transformer": false, | |
| "consformer_word_link": true, | |
| "decoder_start_token_id": 0, | |
| "do_center_crop": false, | |
| "do_resize": true, | |
| "dove_adv": "ocr_both", | |
| "dove_adv_combine": "joint", | |
| "dove_adv_eps": 0.05, | |
| "dove_adv_rampup": true, | |
| "dove_adv_steps": 2, | |
| "dove_adv_weight": 0.5, | |
| "dove_contrastive": false, | |
| "dove_contrastive_weight": 0.1, | |
| "dove_temperature": 0.1, | |
| "eos_token_id": 1, | |
| "generation_max_new_tokens": 56, | |
| "generation_num_beams": 4, | |
| "image_mean": [ | |
| 0.48145466, | |
| 0.4578275, | |
| 0.40821073 | |
| ], | |
| "image_size": null, | |
| "image_std": [ | |
| 0.26862954, | |
| 0.26130258, | |
| 0.27577711 | |
| ], | |
| "model_type": "openvivqa", | |
| "ocr_cuda_device": "cuda:0", | |
| "ocr_d_det": 256, | |
| "ocr_d_rec": 256, | |
| "ocr_edit_distance": 2, | |
| "ocr_feature_norm": "layernorm", | |
| "ocr_full_features": true, | |
| "ocr_jaccard_dupe": 0.78, | |
| "ocr_lowercase": true, | |
| "ocr_max_2d_position_embeddings": 1024, | |
| "ocr_max_non_alnum_ratio": 0.6, | |
| "ocr_max_scene_text": 180, | |
| "ocr_min_text_len_keep": 2, | |
| "ocr_num_distances": 32, | |
| "ocr_visual_pool": false, | |
| "ocrseq_dropout": 0.1, | |
| "ocrseq_margin_frac": 0.14, | |
| "ocrseq_pool_size": 3, | |
| "ocrseq_use_spatial": true, | |
| "pad_token_id": 0, | |
| "qa_clip_d_text": null, | |
| "qavit_gated_joint": true, | |
| "qavit_ground": true, | |
| "qavit_ground_weight": 0.1, | |
| "text_max_input_length": 32, | |
| "text_max_target_length": 56, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.45.2", | |
| "twc_adv_probability": 0.35, | |
| "twc_contrastive": false, | |
| "twc_eval_related": "none", | |
| "twc_label_list": [ | |
| 0.9, | |
| 0.9 | |
| ], | |
| "twc_siglip": false, | |
| "twc_temperature_init": 0.1, | |
| "twc_text_labels": false, | |
| "twc_text_only": false, | |
| "twc_two_views": true, | |
| "twc_weight": 1.0, | |
| "use_twc": true, | |
| "vit5_name": "VietAI/vit5-base" | |
| } | |