JV A commited on
Commit
f67dc8a
1 Parent(s): b12b1f9

Upload 8 files

Browse files
README.md ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ license: mit
4
+ tags:
5
+ - vision
6
+ model_name: microsoft/git-large-textcaps
7
+ pipeline_tag: image-to-text
8
+ ---
9
+
10
+ # GIT (GenerativeImage2Text), large-sized, fine-tuned on TextCaps, R*
11
+
12
+ R = re-trained by removing some offensive captions in cc12m dataset
13
+
14
+ GIT (short for GenerativeImage2Text) model, large-sized version, fine-tuned on TextCaps. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
15
+
16
+ Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
17
+
18
+ ## Model description
19
+
20
+ GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
21
+
22
+ The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
23
+
24
+ The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
25
+
26
+ ![GIT architecture](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/git_architecture.jpg)
27
+
28
+ This allows the model to be used for tasks like:
29
+
30
+ - image and video captioning
31
+ - visual question answering (VQA) on images and videos
32
+ - even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
33
+
34
+ ## Intended uses & limitations
35
+
36
+ You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
37
+ fine-tuned versions on a task that interests you.
38
+
39
+ ### How to use
40
+
41
+ For code examples, we refer to the [documentation](https://huggingface.co/transformers/main/model_doc/git.html).
42
+
43
+ ## Training data
44
+
45
+ From the paper:
46
+
47
+ > We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
48
+ (CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
49
+ Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
50
+ data following a similar collection procedure in Hu et al. (2021a).
51
+
52
+ => however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
53
+
54
+ This checkpoint is "GIT-large", which is a smaller variant of GIT trained on 20 million image-text pairs.
55
+
56
+ Next, the model was fine-tuned on TextCaps.
57
+
58
+ See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
59
+
60
+ ### Preprocessing
61
+
62
+ We refer to the original repo regarding details for preprocessing during training.
63
+
64
+ During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
65
+
66
+ ## Evaluation results
67
+
68
+ For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100).
config.json ADDED
@@ -0,0 +1,106 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_commit_hash": null,
3
+ "architectures": [
4
+ "GitForCausalLM"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": 101,
8
+ "classifier_dropout": null,
9
+ "eos_token_id": 102,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout_prob": 0.1,
12
+ "hidden_size": 768,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 3072,
15
+ "layer_norm_eps": 1e-12,
16
+ "max_position_embeddings": 1024,
17
+ "model_type": "git",
18
+ "num_attention_heads": 12,
19
+ "num_hidden_layers": 6,
20
+ "num_image_with_embedding": null,
21
+ "pad_token_id": 0,
22
+ "position_embedding_type": "absolute",
23
+ "tie_word_embeddings": false,
24
+ "torch_dtype": "float32",
25
+ "transformers_version": null,
26
+ "use_cache": true,
27
+ "vision_config": {
28
+ "_name_or_path": "",
29
+ "add_cross_attention": false,
30
+ "architectures": null,
31
+ "attention_dropout": 0.0,
32
+ "bad_words_ids": null,
33
+ "begin_suppress_tokens": null,
34
+ "bos_token_id": null,
35
+ "chunk_size_feed_forward": 0,
36
+ "cross_attention_hidden_size": null,
37
+ "decoder_start_token_id": null,
38
+ "diversity_penalty": 0.0,
39
+ "do_sample": false,
40
+ "dropout": 0.0,
41
+ "early_stopping": false,
42
+ "encoder_no_repeat_ngram_size": 0,
43
+ "eos_token_id": null,
44
+ "exponential_decay_length_penalty": null,
45
+ "finetuning_task": null,
46
+ "forced_bos_token_id": null,
47
+ "forced_eos_token_id": null,
48
+ "hidden_act": "quick_gelu",
49
+ "hidden_size": 1024,
50
+ "id2label": {
51
+ "0": "LABEL_0",
52
+ "1": "LABEL_1"
53
+ },
54
+ "image_size": 224,
55
+ "initializer_factor": 1.0,
56
+ "initializer_range": 0.02,
57
+ "intermediate_size": 4096,
58
+ "is_decoder": false,
59
+ "is_encoder_decoder": false,
60
+ "label2id": {
61
+ "LABEL_0": 0,
62
+ "LABEL_1": 1
63
+ },
64
+ "layer_norm_eps": 1e-05,
65
+ "length_penalty": 1.0,
66
+ "max_length": 20,
67
+ "min_length": 0,
68
+ "model_type": "git_vision_model",
69
+ "no_repeat_ngram_size": 0,
70
+ "num_attention_heads": 16,
71
+ "num_beam_groups": 1,
72
+ "num_beams": 1,
73
+ "num_channels": 3,
74
+ "num_hidden_layers": 24,
75
+ "num_return_sequences": 1,
76
+ "output_attentions": false,
77
+ "output_hidden_states": false,
78
+ "output_scores": false,
79
+ "pad_token_id": null,
80
+ "patch_size": 14,
81
+ "prefix": null,
82
+ "problem_type": null,
83
+ "projection_dim": 512,
84
+ "pruned_heads": {},
85
+ "remove_invalid_values": false,
86
+ "repetition_penalty": 1.0,
87
+ "return_dict": true,
88
+ "return_dict_in_generate": false,
89
+ "sep_token_id": null,
90
+ "suppress_tokens": null,
91
+ "task_specific_params": null,
92
+ "temperature": 1.0,
93
+ "tf_legacy_loss": false,
94
+ "tie_encoder_decoder": false,
95
+ "tie_word_embeddings": true,
96
+ "tokenizer_class": null,
97
+ "top_k": 50,
98
+ "top_p": 1.0,
99
+ "torch_dtype": null,
100
+ "torchscript": false,
101
+ "transformers_version": "4.26.0.dev0",
102
+ "typical_p": 1.0,
103
+ "use_bfloat16": false
104
+ },
105
+ "vocab_size": 30522
106
+ }
preprocessor_config.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "crop_size": {
3
+ "height": 224,
4
+ "width": 224
5
+ },
6
+ "do_center_crop": true,
7
+ "do_convert_rgb": true,
8
+ "do_normalize": true,
9
+ "do_rescale": true,
10
+ "do_resize": true,
11
+ "image_mean": [
12
+ 0.48145466,
13
+ 0.4578275,
14
+ 0.40821073
15
+ ],
16
+ "image_processor_type": "CLIPImageProcessor",
17
+ "image_std": [
18
+ 0.26862954,
19
+ 0.26130258,
20
+ 0.27577711
21
+ ],
22
+ "processor_class": "GitProcessor",
23
+ "resample": 3,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "shortest_edge": 224
27
+ }
28
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb3f55213b6a6e8d1e451705c911c6c2e8c2dcaa46027176dd20b57438eb8a2a
3
+ size 1576966105
special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "do_lower_case": true,
4
+ "mask_token": "[MASK]",
5
+ "model_input_names": [
6
+ "input_ids",
7
+ "attention_mask"
8
+ ],
9
+ "model_max_length": 512,
10
+ "name_or_path": "bert-base-uncased",
11
+ "pad_token": "[PAD]",
12
+ "processor_class": "GitProcessor",
13
+ "sep_token": "[SEP]",
14
+ "special_tokens_map_file": null,
15
+ "strip_accents": null,
16
+ "tokenize_chinese_chars": true,
17
+ "tokenizer_class": "BertTokenizer",
18
+ "unk_token": "[UNK]"
19
+ }
vocab.txt ADDED
The diff for this file is too large to render. See raw diff