michelecafagna26 commited on
Commit
b32abc6
1 Parent(s): 2327155

Upload 9 files

Browse files
README.md CHANGED
@@ -1,3 +1,66 @@
1
  ---
2
  license: apache-2.0
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
+ tags:
4
+ - image-captioning
5
+ languages:
6
+ - en
7
+ pipeline_tag: image-to-text
8
+ datasets:
9
+ - michelecafagna26/hl
10
+ language:
11
+ - en
12
+ metrics:
13
+ - sacrebleu
14
+ - rouge
15
+ library_name: transformers
16
  ---
17
+ ## GIT-base fine-tuned for Image Captioning on High-Level descriptions of Scenes
18
+
19
+ [GIT](https://arxiv.org/abs/2205.14100) base trained on the [HL dataset](https://huggingface.co/datasets/michelecafagna26/hl) for **scene generation of images**
20
+
21
+ ## Model fine-tuning 🏋️‍
22
+
23
+ - Trained for 10
24
+ - lr: 5e−5
25
+ - Adam optimizer
26
+ - half-precision (fp16)
27
+
28
+ ## Test set metrics 🧾
29
+
30
+ | Cider | SacreBLEU | Rouge-L|
31
+ |--------|------------|--------|
32
+ | 103.00 | 24.67 | 33.90 |
33
+
34
+ ## Model in Action 🚀
35
+
36
+ ```python
37
+ import requests
38
+ from PIL import Image
39
+ from transformers import AutoProcessor, AutoModelForCausalLM
40
+
41
+ processor = AutoProcessor.from_pretrained("git-base-captioning-ft-hl-scenes")
42
+ model = AutoModelForCausalLM.from_pretrained("git-base-captioning-ft-hl-scenes").to("cuda")
43
+
44
+ img_url = 'https://datasets-server.huggingface.co/assets/michelecafagna26/hl/--/default/train/0/image/image.jpg'
45
+ raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
46
+
47
+
48
+ inputs = processor(raw_image, return_tensors="pt").to("cuda")
49
+ pixel_values = inputs.pixel_values
50
+
51
+ generated_ids = model.generate(pixel_values=pixel_values, max_length=50,
52
+ do_sample=True,
53
+ top_k=120,
54
+ top_p=0.9,
55
+ early_stopping=True,
56
+ num_return_sequences=1)
57
+
58
+ processor.batch_decode(generated_ids, skip_special_tokens=True)
59
+
60
+ >>>
61
+ ```
62
+
63
+ ## BibTex and citation info
64
+
65
+ ```BibTeX
66
+ ```
config.json ADDED
@@ -0,0 +1,107 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_commit_hash": "1f7fe8444292beb4a259e3a5b6eba440cd5999d4",
3
+ "_name_or_path": "microsoft/git-base",
4
+ "architectures": [
5
+ "GitForCausalLM"
6
+ ],
7
+ "attention_probs_dropout_prob": 0.1,
8
+ "bos_token_id": 101,
9
+ "classifier_dropout": null,
10
+ "eos_token_id": 102,
11
+ "hidden_act": "gelu",
12
+ "hidden_dropout_prob": 0.1,
13
+ "hidden_size": 768,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 3072,
16
+ "layer_norm_eps": 1e-12,
17
+ "max_position_embeddings": 1024,
18
+ "model_type": "git",
19
+ "num_attention_heads": 12,
20
+ "num_hidden_layers": 6,
21
+ "num_image_with_embedding": null,
22
+ "pad_token_id": 0,
23
+ "position_embedding_type": "absolute",
24
+ "tie_word_embeddings": false,
25
+ "torch_dtype": "float32",
26
+ "transformers_version": null,
27
+ "use_cache": true,
28
+ "vision_config": {
29
+ "_name_or_path": "",
30
+ "add_cross_attention": false,
31
+ "architectures": null,
32
+ "attention_dropout": 0.0,
33
+ "bad_words_ids": null,
34
+ "begin_suppress_tokens": null,
35
+ "bos_token_id": null,
36
+ "chunk_size_feed_forward": 0,
37
+ "cross_attention_hidden_size": null,
38
+ "decoder_start_token_id": null,
39
+ "diversity_penalty": 0.0,
40
+ "do_sample": false,
41
+ "dropout": 0.0,
42
+ "early_stopping": false,
43
+ "encoder_no_repeat_ngram_size": 0,
44
+ "eos_token_id": null,
45
+ "exponential_decay_length_penalty": null,
46
+ "finetuning_task": null,
47
+ "forced_bos_token_id": null,
48
+ "forced_eos_token_id": null,
49
+ "hidden_act": "quick_gelu",
50
+ "hidden_size": 768,
51
+ "id2label": {
52
+ "0": "LABEL_0",
53
+ "1": "LABEL_1"
54
+ },
55
+ "image_size": 224,
56
+ "initializer_factor": 1.0,
57
+ "initializer_range": 0.02,
58
+ "intermediate_size": 3072,
59
+ "is_decoder": false,
60
+ "is_encoder_decoder": false,
61
+ "label2id": {
62
+ "LABEL_0": 0,
63
+ "LABEL_1": 1
64
+ },
65
+ "layer_norm_eps": 1e-05,
66
+ "length_penalty": 1.0,
67
+ "max_length": 20,
68
+ "min_length": 0,
69
+ "model_type": "git_vision_model",
70
+ "no_repeat_ngram_size": 0,
71
+ "num_attention_heads": 12,
72
+ "num_beam_groups": 1,
73
+ "num_beams": 1,
74
+ "num_channels": 3,
75
+ "num_hidden_layers": 12,
76
+ "num_return_sequences": 1,
77
+ "output_attentions": false,
78
+ "output_hidden_states": false,
79
+ "output_scores": false,
80
+ "pad_token_id": null,
81
+ "patch_size": 16,
82
+ "prefix": null,
83
+ "problem_type": null,
84
+ "projection_dim": 512,
85
+ "pruned_heads": {},
86
+ "remove_invalid_values": false,
87
+ "repetition_penalty": 1.0,
88
+ "return_dict": true,
89
+ "return_dict_in_generate": false,
90
+ "sep_token_id": null,
91
+ "suppress_tokens": null,
92
+ "task_specific_params": null,
93
+ "temperature": 1.0,
94
+ "tf_legacy_loss": false,
95
+ "tie_encoder_decoder": false,
96
+ "tie_word_embeddings": true,
97
+ "tokenizer_class": null,
98
+ "top_k": 50,
99
+ "top_p": 1.0,
100
+ "torch_dtype": null,
101
+ "torchscript": false,
102
+ "transformers_version": "4.30.2",
103
+ "typical_p": 1.0,
104
+ "use_bfloat16": false
105
+ },
106
+ "vocab_size": 30522
107
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 101,
4
+ "eos_token_id": 102,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.30.2"
7
+ }
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:fd3b41e34260adfcd1996392bf64fb63164acf1255328547510b0aa342493ef2
3
+ size 706594713
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