Datasets:

rbiswasfc

/

gorilla-bench-apis

like 0

Reinforcement Learning

Unity ML-Agents

Train and play SoccerTwos

Raiden-1001/poca-Soccerv7.1

mlagents-load-from-hf --repo-id='Raiden-1001/poca-Soccerv7.1' --local-dir='./downloads'

['your_configuration_file_path.yaml', 'run_id']

['ml-agents']

mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume

{'dataset': 'SoccerTwos', 'accuracy': 'Not provided'}

A trained model of a poca agent playing SoccerTwos using the Unity ML-Agents Library.

Reinforcement Learning

Stable-Baselines3

CartPole-v1

sb3/ppo-CartPole-v1

load_from_hub(repo_id='sb3/ppo-CartPole-v1')

['algo', 'env', 'f']

['rl_zoo3', 'stable-baselines3', 'stable-baselines3-contrib']

python -m rl_zoo3.load_from_hub --algo ppo --env CartPole-v1 -orga sb3 -f logs/

{'dataset': 'CartPole-v1', 'accuracy': '500.00 +/- 0.00'}

This is a trained model of a PPO agent playing CartPole-v1 using the stable-baselines3 library and the RL Zoo. The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.

Reinforcement Learning

Stable-Baselines3

LunarLander-v2

araffin/dqn-LunarLander-v2

DQN.load('araffin/dqn-LunarLander-v2')

{'checkpoint': 'araffin/dqn-LunarLander-v2', 'kwargs': {'target_update_interval': 30}}

['huggingface_sb3', 'stable_baselines3']

{'load_model': 'from huggingface_sb3 import load_from_hub\nfrom stable_baselines3 import DQN\nfrom stable_baselines3.common.env_util import make_vec_env\nfrom stable_baselines3.common.evaluation import evaluate_policy\n\ncheckpoint = load_from_hub(araffin/dqn-LunarLander-v2, dqn-LunarLander-v2.zip)\n\nkwargs = dict(target_update_interval=30)\n\nmodel = DQN.load(checkpoint, **kwargs)\nenv = make_vec_env(LunarLander-v2, n_envs=1)', 'evaluate': 'mean_reward, std_reward = evaluate_policy(\n model,\n env,\n n_eval_episodes=20,\n deterministic=True,\n)\nprint(fMean reward = {mean_reward:.2f} +/- {std_reward:.2f})'}

{'dataset': 'LunarLander-v2', 'accuracy': '280.22 +/- 13.03'}

This is a trained model of a DQN agent playing LunarLander-v2 using the stable-baselines3 library.

Reinforcement Learning

Stable-Baselines3

CartPole-v1

dqn-CartPole-v1

load_from_hub(repo_id='sb3/dqn-CartPole-v1')

['algo', 'env', 'logs']

['rl_zoo3', 'stable-baselines3', 'stable-baselines3-contrib']

python train.py --algo dqn --env CartPole-v1 -f logs/

{'dataset': 'CartPole-v1', 'accuracy': '500.00 +/- 0.00'}

This is a trained model of a DQN agent playing CartPole-v1 using the stable-baselines3 library and the RL Zoo. The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.

Reinforcement Learning

Stable-Baselines3

deep-reinforcement-learning

td3-Ant-v3

load_from_hub(repo_id='sb3/td3-Ant-v3')

['algo', 'env', 'f']

['RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo', 'SB3: https://github.com/DLR-RM/stable-baselines3', 'SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib']

['python -m rl_zoo3.load_from_hub --algo td3 --env Ant-v3 -orga sb3 -f logs/', 'python enjoy.py --algo td3 --env Ant-v3 -f logs/', 'python train.py --algo td3 --env Ant-v3 -f logs/', 'python -m rl_zoo3.push_to_hub --algo td3 --env Ant-v3 -f logs/ -orga sb3']

{'dataset': 'Ant-v3', 'accuracy': '5822.96 +/- 93.33'}

This is a trained model of a TD3 agent playing Ant-v3 using the stable-baselines3 library and the RL Zoo. The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.

Reinforcement Learning

Hugging Face Transformers

Transformers

edbeeching/decision-transformer-gym-hopper-expert

AutoModel.from_pretrained('edbeeching/decision-transformer-gym-hopper-expert')

{'mean': [1.3490015, -0.11208222, -0.5506444, -0.13188992, -0.00378754, 2.6071432, 0.02322114, -0.01626922, -0.06840388, -0.05183131, 0.04272673], 'std': [0.15980862, 0.0446214, 0.14307782, 0.17629202, 0.5912333, 0.5899924, 1.5405099, 0.8152689, 2.0173461, 2.4107876, 5.8440027]}

['transformers', 'torch']

See our Blog Post, Colab notebook or Example Script for usage.

{'dataset': 'Gym Hopper environment', 'accuracy': 'Not provided'}

Decision Transformer model trained on expert trajectories sampled from the Gym Hopper environment

Reinforcement Learning

Hugging Face Transformers

Transformers

edbeeching/decision-transformer-gym-halfcheetah-expert

AutoModel.from_pretrained('edbeeching/decision-transformer-gym-halfcheetah-expert')

{'mean': [-0.04489148, 0.03232588, 0.06034835, -0.17081226, -0.19480659, -0.05751596, 0.09701628, 0.03239211, 11.047426, -0.07997331, -0.32363534, 0.36297753, 0.42322603, 0.40836546, 1.1085187, -0.4874403, -0.0737481], 'std': [0.04002118, 0.4107858, 0.54217845, 0.41522816, 0.23796624, 0.62036866, 0.30100912, 0.21737163, 2.2105937, 0.572586, 1.7255033, 11.844218, 12.06324, 7.0495934, 13.499867, 7.195647, 5.0264325]}

['transformers']

See our Blog Post, Colab notebook or Example Script for usage.

{'dataset': 'Gym HalfCheetah environment', 'accuracy': 'Not specified'}

Decision Transformer model trained on expert trajectories sampled from the Gym HalfCheetah environment

Reinforcement Learning

ML-Agents

SoccerTwos

Raiden-1001/poca-Soccerv7

mlagents-load-from-hf --repo-id='Raiden-1001/poca-Soccerv7.1' --local-dir='./downloads'

['your_configuration_file_path.yaml', 'run_id']

['unity-ml-agents', 'deep-reinforcement-learning', 'ML-Agents-SoccerTwos']

Step 1: Write your model_id: Raiden-1001/poca-Soccerv7 Step 2: Select your .nn /.onnx file Click on Watch the agent play 👀

{'dataset': 'SoccerTwos', 'accuracy': 'Not provided'}

This is a trained model of a poca agent playing SoccerTwos using the Unity ML-Agents Library.

Reinforcement Learning

Unity ML-Agents Library

Train and play SoccerTwos

poca-SoccerTwosv2

mlagents-load-from-hf --repo-id='Raiden-1001/poca-SoccerTwosv2' --local-dir='./downloads'

['your_configuration_file_path.yaml', 'run_id']

['ml-agents']

mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume

{'dataset': 'SoccerTwos', 'accuracy': 'Not provided'}

A trained model of a poca agent playing SoccerTwos using the Unity ML-Agents Library.

Reinforcement Learning Robotics

Hugging Face

Inference API

Antheia/Hanna

pipeline('robotics', model='Antheia/Hanna')

model

transformers

{'dataset': 'openai/webgpt_comparisons', 'accuracy': ''}

Antheia/Hanna is a reinforcement learning model for robotics tasks, trained on the openai/webgpt_comparisons dataset.

Reinforcement Learning Robotics

Hugging Face Transformers

EmbodiedAI tasks

VC1_BASE_NAME

model_utils.load_model('model_utils.VC1_BASE_NAME')

img

from vc_models.models.vit import model_utils

model,embd_size,model_transforms,model_info = model_utils.load_model(model_utils.VC1_BASE_NAME) img = your_function_here ... transformed_img = model_transforms(img) embedding = model(transformed_img)

{'dataset': 'CortexBench', 'accuracy': 'Mean Success: 68.7%'}

The VC-1 model is a vision transformer (ViT) pre-trained on over 4,000 hours of egocentric videos from 7 different sources, together with ImageNet. The model is trained using Masked Auto-Encoding (MAE) and is available in two sizes: ViT-B and ViT-L. The model is intended for use for EmbodiedAI tasks, such as object manipulation and indoor navigation.

Reinforcement Learning Robotics

Hugging Face

6D grasping

camusean/grasp_diffusion

AutoModel.from_pretrained('camusean/grasp_diffusion')

N/A

transformers

N/A

{'dataset': 'N/A', 'accuracy': 'N/A'}

Trained Models for Grasp SE(3) DiffusionFields. Check SE(3)-DiffusionFields: Learning smooth cost functions for joint grasp and motion optimization through diffusion for additional details.

Reinforcement Learning Robotics

Hugging Face Transformers

EmbodiedAI tasks, such as object manipulation and indoor navigation

facebook/vc1-large

model_utils.load_model('model_utils.VC1_BASE_NAME')

img

from vc_models.models.vit import model_utils

model,embd_size,model_transforms,model_info = model_utils.load_model(model_utils.VC1_BASE_NAME) img = your_function_here ... transformed_img = model_transforms(img) embedding = model(transformed_img)

{'dataset': 'CortexBench', 'accuracy': '68.7 (Mean Success)'}

The VC-1 model is a vision transformer (ViT) pre-trained on over 4,000 hours of egocentric videos from 7 different sources, together with ImageNet. The model is trained using Masked Auto-Encoding (MAE) and is available in two sizes: ViT-B and ViT-L. The model is intended for use for EmbodiedAI tasks, such as object manipulation and indoor navigation.

Reinforcement Learning

Stable-Baselines3

deep-reinforcement-learning

ppo-BreakoutNoFrameskip-v4

load_from_hub(repo_id='sb3/ppo-BreakoutNoFrameskip-v4')

['algo', 'env', 'orga', 'f']

['rl_zoo3', 'stable-baselines3', 'stable-baselines3-contrib']

['python -m rl_zoo3.load_from_hub --algo ppo --env BreakoutNoFrameskip-v4 -orga sb3 -f logs/', 'python enjoy.py --algo ppo --env BreakoutNoFrameskip-v4 -f logs/', 'python train.py --algo ppo --env BreakoutNoFrameskip-v4 -f logs/', 'python -m rl_zoo3.push_to_hub --algo ppo --env BreakoutNoFrameskip-v4 -f logs/ -orga sb3']

{'dataset': 'BreakoutNoFrameskip-v4', 'accuracy': '398.00 +/- 16.30'}

This is a trained model of a PPO agent playing BreakoutNoFrameskip-v4 using the stable-baselines3 library and the RL Zoo. The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.

Natural Language Processing Text Classification

Hugging Face Transformers

Sentiment Analysis

bert-base-multilingual-uncased-sentiment

pipeline('sentiment-analysis')

['text']

['transformers']

result = sentiment_pipeline('I love this product!')

{'dataset': [{'language': 'English', 'accuracy': {'exact': '67%', 'off-by-1': '95%'}}, {'language': 'Dutch', 'accuracy': {'exact': '57%', 'off-by-1': '93%'}}, {'language': 'German', 'accuracy': {'exact': '61%', 'off-by-1': '94%'}}, {'language': 'French', 'accuracy': {'exact': '59%', 'off-by-1': '94%'}}, {'language': 'Italian', 'accuracy': {'exact': '59%', 'off-by-1': '95%'}}, {'language': 'Spanish', 'accuracy': {'exact': '58%', 'off-by-1': '95%'}}]}

This a bert-base-multilingual-uncased model finetuned for sentiment analysis on product reviews in six languages: English, Dutch, German, French, Spanish and Italian. It predicts the sentiment of the review as a number of stars (between 1 and 5).

Natural Language Processing Text Classification

Hugging Face Transformers

Transformers

sentiment_analysis_generic_dataset

pipeline('text-classification')

[]

['transformers']

sentiment_analysis('I love this product!')

{'dataset': 'generic_dataset', 'accuracy': 'Not specified'}

This is a fine-tuned downstream version of the bert-base-uncased model for sentiment analysis, this model is not intended for further downstream fine-tuning for any other tasks. This model is trained on a classified dataset for text classification.

Natural Language Processing Text Generation

Transformers

Text Generation

distilgpt2

pipeline('text-generation')

['model']

['from transformers import pipeline, set_seed']

set_seed(42) generator(Hello, I’m a language model, max_length=20, num_return_sequences=5)

{'dataset': 'WikiText-103', 'accuracy': '21.100'}

DistilGPT2 is an English-language model pre-trained with the supervision of the 124 million parameter version of GPT-2. With 82 million parameters, it was developed using knowledge distillation and designed to be a faster, lighter version of GPT-2. It can be used for text generation, writing assistance, creative writing, entertainment, and more.

Natural Language Processing Zero-Shot Classification

Hugging Face Transformers

Zero-Shot Image Classification

laion/CLIP-ViT-B-32-laion2B-s34B-b79K

pipeline('zero-shot-classification')

{'image': 'path/to/image', 'class_names': ['class1', 'class2', 'class3']}

{'transformers': '>=4.0.0'}

from transformers import pipeline; classifier = pipeline('zero-shot-classification', model='laion/CLIP-ViT-B-32-laion2B-s34B-b79K'); classifier(image='path/to/image', class_names=['class1', 'class2', 'class3'])

{'dataset': 'ImageNet-1k', 'accuracy': 66.6}

A CLIP ViT-B/32 model trained with the LAION-2B English subset of LAION-5B using OpenCLIP. It enables researchers to better understand and explore zero-shot, arbitrary image classification. The model can be used for zero-shot image classification, image and text retrieval, among others.

Natural Language Processing Translation

Transformers

Text-to-Text Generation

optimum/t5-small

pipeline('translation')

['text']

['transformers', 'optimum.onnxruntime']

from transformers import AutoTokenizer, pipeline from optimum.onnxruntime import ORTModelForSeq2SeqLM tokenizer = AutoTokenizer.from_pretrained(optimum/t5-small) model = ORTModelForSeq2SeqLM.from_pretrained(optimum/t5-small) translator = pipeline(translation_en_to_fr, model=model, tokenizer=tokenizer) results = translator(My name is Eustache and I have a pet raccoon) print(results)

{'dataset': 'c4', 'accuracy': 'N/A'}

T5 is an encoder-decoder model pre-trained on a multi-task mixture of unsupervised and supervised tasks and for which each task is converted into a text-to-text format. It can be used for translation, text-to-text generation, and summarization.

Audio Audio Classification

Hugging Face Transformers

Audio Classification

ast-finetuned-audioset-10-10-0.4593

pipeline('audio-classification')

transformers

{'dataset': 'AudioSet', 'accuracy': ''}

Audio Spectrogram Transformer (AST) model fine-tuned on AudioSet. It was introduced in the paper AST: Audio Spectrogram Transformer by Gong et al. and first released in this repository. The Audio Spectrogram Transformer is equivalent to ViT, but applied on audio. Audio is first turned into an image (as a spectrogram), after which a Vision Transformer is applied. The model gets state-of-the-art results on several audio classification benchmarks.

Audio Automatic Speech Recognition

Hugging Face Transformers

Transformers

tiny-wav2vec2-stable-ln

pipeline('automatic-speech-recognition')

None

['transformers']

None

{'dataset': None, 'accuracy': None}

A tiny wav2vec2 model for Automatic Speech Recognition

Natural Language Processing Text Generation

Hugging Face Transformers

Conversational

pygmalion-350m

pipeline('conversational')

N/A

transformers

N/A

{'dataset': 'The Pile', 'accuracy': 'N/A'}

This is a proof-of-concept fine-tune of Facebook's OPT-350M model optimized for dialogue, to be used as a stepping stone to higher parameter models. Disclaimer: NSFW data was included in the fine-tuning of this model. Although SFW inputs will usually result in SFW outputs, you are advised to chat at your own risk. This model is not suitable for use by minors.

Computer Vision Depth Estimation

Hugging Face Transformers

Depth Estimation

glpn-nyu-finetuned-diode

pipeline('depth-estimation')

[]

['transformers']

{'dataset': 'diode-subset', 'accuracy': {'Loss': 0.4359, 'Rmse': 0.4276}}

This model is a fine-tuned version of vinvino02/glpn-nyu on the diode-subset dataset.

Multimodal Document Question Answer

Hugging Face Transformers

vision-encoder-decoder

naver-clova-ix/donut-base-finetuned-docvqa

pipeline('document-question-answering')

{'image': 'path_to_image', 'question': 'your_question'}

Transformers

from transformers import pipeline # Initialize the pipeline doc_qa = pipeline('document-question-answering', model='naver-clova-ix/donut-base-finetuned-docvqa') # Load an image and ask a question image_path = 'path_to_image' question = 'your_question' # Get the answer answer = doc_qa({'image': image_path, 'question': question}) print(answer)

{'dataset': 'DocVQA', 'accuracy': 'Not provided'}

Donut model fine-tuned on DocVQA. It was introduced in the paper OCR-free Document Understanding Transformer by Geewok et al. and first released in this repository. Donut consists of a vision encoder (Swin Transformer) and a text decoder (BART). Given an image, the encoder first encodes the image into a tensor of embeddings (of shape batch_size, seq_len, hidden_size), after which the decoder autoregressively generates text, conditioned on the encoding of the encoder.

Natural Language Processing Fill-Mask

Transformers

Masked Language Modeling, Next Sentence Prediction

bert-base-uncased

pipeline('fill-mask')

['text']

['transformers']

from transformers import pipeline unmasker = pipeline('fill-mask', model='bert-base-uncased') unmasker(Hello I'm a [MASK] model.)

{'dataset': 'GLUE', 'accuracy': 79.6}

BERT base model (uncased) is a transformer model pretrained on a large corpus of English data using a masked language modeling (MLM) objective. It can be used for masked language modeling, next sentence prediction, and fine-tuning on downstream tasks such as sequence classification, token classification, or question answering.

Computer Vision Image Classification

Hugging Face Transformers

Image Classification

martinezomg/vit-base-patch16-224-diabetic-retinopathy

pipeline('image-classification')

{'model_name': 'martinezomg/vit-base-patch16-224-diabetic-retinopathy'}

{'transformers': '4.28.1', 'pytorch': '2.0.0+cu118', 'datasets': '2.11.0', 'tokenizers': '0.13.3'}

from transformers import pipeline image_classifier = pipeline('image-classification', 'martinezomg/vit-base-patch16-224-diabetic-retinopathy') result = image_classifier('path/to/image.jpg')

{'dataset': 'None', 'accuracy': 0.7744}

This model is a fine-tuned version of google/vit-base-patch16-224 on the None dataset. It is designed for image classification tasks, specifically for diabetic retinopathy detection.

Computer Vision Image Segmentation

Hugging Face Transformers

Transformers

clipseg-rd64-refined

pipeline('image-segmentation')

{'model': 'CIDAS/clipseg-rd64-refined'}

transformers

{'dataset': '', 'accuracy': ''}

CLIPSeg model with reduce dimension 64, refined (using a more complex convolution). It was introduced in the paper Image Segmentation Using Text and Image Prompts by Lüddecke et al. and first released in this repository. This model is intended for zero-shot and one-shot image segmentation.

Multimodal Image-to-Text

Hugging Face Transformers

Image Captioning

microsoft/git-base

pipeline('image-to-text')

image

transformers

git_base(image)

{'dataset': ['COCO', 'Conceptual Captions (CC3M)', 'SBU', 'Visual Genome (VG)', 'Conceptual Captions (CC12M)', 'ALT200M'], 'accuracy': 'Refer to the paper for evaluation results'}

GIT (short for GenerativeImage2Text) model, base-sized version. It was introduced in the paper GIT: A Generative Image-to-text Transformer for Vision and Language by Wang et al. and first released in this repository. The model is trained using 'teacher forcing' on a lot of (image, text) pairs. The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens. This allows the model to be used for tasks like image and video captioning, visual question answering (VQA) on images and videos, and even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).

Natural Language Processing Token Classification

Transformers

Named Entity Recognition

dslim/bert-base-NER-uncased

pipeline('ner')

{}

{'transformers': '>=4.0.0'}

nlp('My name is John and I live in New York.')

{'dataset': '', 'accuracy': ''}

A pretrained BERT model for Named Entity Recognition (NER) on uncased text. It can be used to extract entities such as person names, locations, and organizations from text.

Computer Vision Object Detection

Hugging Face Transformers

Transformers

microsoft/table-transformer-structure-recognition

pipeline('object-detection')

transformers

{'dataset': 'PubTables1M', 'accuracy': ''}

Table Transformer (DETR) model trained on PubTables1M for detecting the structure (like rows, columns) in tables.

Multimodal Document Question Answer

Hugging Face Transformers

Document Question Answering

impira/layoutlm-document-qa

pipeline('question-answering')

['image_url', 'question']

['PIL', 'pytesseract', 'PyTorch', 'transformers']

nlp(https://templates.invoicehome.com/invoice-template-us-neat-750px.png, What is the invoice number?)

{'dataset': ['SQuAD2.0', 'DocVQA'], 'accuracy': 'Not provided'}

A fine-tuned version of the multi-modal LayoutLM model for the task of question answering on documents.

Natural Language Processing Summarization

Hugging Face Transformers

Text Summarization

facebook/bart-large-cnn

pipeline('summarization')

['ARTICLE', 'max_length', 'min_length', 'do_sample']

['transformers']

from transformers import pipeline summarizer = pipeline(summarization, model=facebook/bart-large-cnn) ARTICLE = ... print(summarizer(ARTICLE, max_length=130, min_length=30, do_sample=False))

{'dataset': 'cnn_dailymail', 'accuracy': {'ROUGE-1': 42.949, 'ROUGE-2': 20.815, 'ROUGE-L': 30.619, 'ROUGE-LSUM': 40.038}}

BART (large-sized model), fine-tuned on CNN Daily Mail. BART is a transformer encoder-encoder (seq2seq) model with a bidirectional (BERT-like) encoder and an autoregressive (GPT-like) decoder. BART is pre-trained by (1) corrupting text with an arbitrary noising function, and (2) learning a model to reconstruct the original text. BART is particularly effective when fine-tuned for text generation (e.g. summarization, translation) but also works well for comprehension tasks (e.g. text classification, question answering). This particular checkpoint has been fine-tuned on CNN Daily Mail, a large collection of text-summary pairs.

Natural Language Processing Table Question Answering

Transformers

Table Question Answering

google/tapas-large-finetuned-wtq

pipeline('table-question-answering')

{'model': 'google/tapas-large-finetuned-wtq', 'task': 'table-question-answering'}

transformers

from transformers import pipeline qa_pipeline = pipeline('table-question-answering', model='google/tapas-large-finetuned-wtq') result = qa_pipeline(table=table, query=query)

{'dataset': 'wikitablequestions', 'accuracy': 0.5097}

TAPAS large model fine-tuned on WikiTable Questions (WTQ). This model was pre-trained on MLM and an additional step which the authors call intermediate pre-training, and then fine-tuned in a chain on SQA, WikiSQL and finally WTQ. It uses relative position embeddings (i.e. resetting the position index at every cell of the table).

Natural Language Processing Text2Text Generation

Transformers

Text Generation

google/t5-v1_1-base

pipeline('text2text-generation')

{'model': 'google/t5-v1_1-base'}

{'transformers': '>=4.0.0'}

from transformers import pipeline t5 = pipeline('text2text-generation', model='google/t5-v1_1-base') t5('translate English to French: Hugging Face is a great company')

{'dataset': 'c4', 'accuracy': 'Not provided'}

Google's T5 Version 1.1 is a state-of-the-art text-to-text transformer model that achieves high performance on various NLP tasks such as summarization, question answering, and text classification. It is pre-trained on the Colossal Clean Crawled Corpus (C4) and fine-tuned on downstream tasks.

Natural Language Processing Token Classification

Hugging Face Transformers

Transformers

kredor/punctuate-all

pipeline('token-classification')

[]

['transformers']

{'dataset': 'multilingual', 'accuracy': 0.98}

A finetuned xlm-roberta-base model for punctuation prediction on twelve languages: English, German, French, Spanish, Bulgarian, Italian, Polish, Dutch, Czech, Portugese, Slovak, Slovenian.

Multimodal Visual Question Answering

Hugging Face Transformers

Transformers

microsoft/git-base-vqav2

pipeline('visual-question-answering')

image, question

['transformers']

vqa(image='path/to/image.jpg', question='What is in the image?')

{'dataset': 'VQAv2', 'accuracy': 'Refer to the paper for evaluation results'}

GIT (short for GenerativeImage2Text) model, base-sized version, fine-tuned on VQAv2. It was introduced in the paper GIT: A Generative Image-to-text Transformer for Vision and Language by Wang et al. and first released in this repository.