bhavitvyamalik
commited on
Commit
•
ccc24a7
1
Parent(s):
0ec8765
first commit
Browse files- README.md +103 -0
- config.json +316 -0
- logs/events.out.tfevents.1626777663.t1v-n-e32efd45-w-0.1846718.3.v2 +3 -0
README.md
ADDED
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# CLIP-Vision-mBART50 Seq2Seq Encoder-Decoder Model
|
2 |
+
|
3 |
+
Pretrained CLIP-Vision-mBART50 pre-trained on subset of translated Conceptual-12M image-text pairs using a seq2seq model training objective. 2.5M cleaned English image-text pairs are translated using Marian Model for respective languages to 2.5M examples each in English, French, German and Spanish. We trained CLIP-Vision-mBART50 model during community week hosted by Huggingface 🤗 using JAX/Flax.
|
4 |
+
|
5 |
+
## Model description
|
6 |
+
CLIP-Vision-mBART50 is a modified mBART50 model which takes in visual embeddings from CLIP-Vision transformer and concatenates them with mBART textual embeddings before passing them to the self-attention layers of mBART. This is done for deep cross-modal interaction between the two modes.
|
7 |
+
|
8 |
+
## Intended uses & limitations❗️
|
9 |
+
You can use the raw model for encoder decoder network where you want the encoder to encode images and decoder to decode text.
|
10 |
+
|
11 |
+
Note that this model is primarily aimed at being fine-tuned on tasks like multi-lingual/mono-lingual image captioning.
|
12 |
+
|
13 |
+
### How to use❓
|
14 |
+
You will need to clone the model from [here](https://github.com/gchhablani/multilingual-image-captioning). An example of usage is shown below:
|
15 |
+
```python
|
16 |
+
>>> from torchvision.io import read_image
|
17 |
+
>>> import numpy as np
|
18 |
+
>>> import os
|
19 |
+
>>> from transformers import CLIPProcessor, MBart50TokenizerFast
|
20 |
+
>>> from model.flax_clip_vision_mbart.modeling_clip_vision_mbart import FlaxCLIPVisionMBartForConditionalGeneration
|
21 |
+
>>> image_path = os.path.join('images/val2014', os.listdir('images/val2014')[0])
|
22 |
+
>>> img = read_image(image_path) # reading image
|
23 |
+
>>> clip_processor = CLIPProcessor.from_pretrained('openai/clip-vit-base-patch32')
|
24 |
+
>>> clip_outputs = clip_processor(images=img)
|
25 |
+
>>> clip_outputs['pixel_values'][0] = clip_outputs['pixel_values'][0].transpose(1,2,0) # Need to transpose images as model expected channel last images.
|
26 |
+
>>> tokenizer = MBart50TokenizerFast.from_pretrained('facebook/mbart-large-50"')
|
27 |
+
>>> model = FlaxCLIPVisionBertForMaskedLM.from_pretrained('flax-community/clip-vit-base-patch32_mbart-large-50')
|
28 |
+
>>> output_ids = model.generate(batch["pixel_values"], forced_bos_token_id=tokenizer.lang_code_to_id["en_XX"], num_beams=4, max_length=64).sequences # "en_XX is the language code in which you want the translation
|
29 |
+
# en_XX: English, fr_XX: French, es_XX: Spanish, de_DE: Deutsch
|
30 |
+
>>> output_string = tokenizer.batch_decode(output_ids.reshape(-1, 64), skip_special_tokens=True, max_length=64)
|
31 |
+
>>> output_string # relevant caption
|
32 |
+
```
|
33 |
+
|
34 |
+
## Training data 🏋🏻♂️
|
35 |
+
The Multi-lingual image captioning model was trained on a subset of Conceptual 12M dataset by Google:
|
36 |
+
<br>
|
37 |
+
<br>
|
38 |
+
[Conceptual 12M](https://github.com/google-research-datasets/conceptual-12m), Introduced by Changpinyo et al. in [Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts](https://arxiv.org/abs/2102.08981).
|
39 |
+
|
40 |
+
The translated dataset can be downloaded from [conceptual-12m-multilingual-marian](https://huggingface.co/datasets/flax-community/conceptual-12m-multilingual-marian). We do not provide images as we do not own any of them. One can download images from the `image_url` section of the original Conceptual 12M dataset.
|
41 |
+
|
42 |
+
## Data Cleaning 🧹
|
43 |
+
Though the original dataset contains 12M image-text pairs, a lot of the URLs are invalid now, and in some cases, images are corrupt or broken. We remove such examples from our data, which leaves us with approximately 10M image-text pairs.
|
44 |
+
|
45 |
+
#### **Train set:**
|
46 |
+
Total data: <br>
|
47 |
+
10010625 captions <br>
|
48 |
+
2502656 images br><br>
|
49 |
+
|
50 |
+
Language-wise distribution: <br>
|
51 |
+
English: 2502656 captions<br>
|
52 |
+
Spanish: 2502656 captions<br>
|
53 |
+
Deutsch: 2502656 captions<br>
|
54 |
+
French: 2502656 captions<br>
|
55 |
+
|
56 |
+
#### **Validation set**
|
57 |
+
Total data: <br>
|
58 |
+
110592 captions <br>
|
59 |
+
27648 images <br><br>
|
60 |
+
|
61 |
+
Language-wise distribution: <br>
|
62 |
+
English: 27648 captions<br>
|
63 |
+
Spanish: 27648 captions<br>
|
64 |
+
Deutsch: 27648 captions<br>
|
65 |
+
French: 27648 captions<br>
|
66 |
+
|
67 |
+
## Training procedure 👨🏻💻
|
68 |
+
### Training
|
69 |
+
The model was trained on Google Cloud Engine TPUv3-8 machine (with 335 GB of RAM, 1000 GB of hard drive, 96 CPU cores) **8 v3 TPU cores** for 42K steps with a batch size of 128 and a sequence length of 128. The
|
70 |
+
optimizer used is Adam with a learning rate of 3e-4, β1 = 0.9, β2 = 0.98 and
|
71 |
+
ε = 1e-8, a weight decay of 0.01, learning rate warmup for 1,000 steps and linear decay of the learning
|
72 |
+
rate after.
|
73 |
+
|
74 |
+
We tracked experiments using Tensorboard which can be found in `Training Metrics` tab.
|
75 |
+
|
76 |
+
#### **Pretraining Results 📊**
|
77 |
+
|
78 |
+
Our model reached **eval loss of ~2.6** around ~70K steps. Here are the BLEU^ scores for different languages:
|
79 |
+
|
80 |
+
|Language |BLEU-1|BLEU-2|BLEU-3|BLEU-4|
|
81 |
+
|--------------------------|------|------|------|------|
|
82 |
+
|English | 0.163| 0.127| 0.10 | 0.081|
|
83 |
+
|Spanish | 0.171| 0.133| 0.114| 0.082|
|
84 |
+
|German | 0.165| 0.129| 0.103| 0.077|
|
85 |
+
|French | 0.162| 0.124| 0.104| 0.073|
|
86 |
+
|
87 |
+
^BLEU scores are out of 1
|
88 |
+
|
89 |
+
|
90 |
+
## **App Demo**
|
91 |
+
|
92 |
+
You can try out our model on 🤗 Huggingface's spaces 🪐 :
|
93 |
+
[Streamlit app of Multi-lingual Image Captioning model on Huggingface Spaces](https://huggingface.co/spaces/flax-community/multilingual-image-captioning)
|
94 |
+
|
95 |
+
|
96 |
+
## Team Members
|
97 |
+
- Bhavitvya Malik [@bhavitvyamalik](https://github.com/bhavitvyamalik)
|
98 |
+
- Gunjan Chhablani [@gchhablani](https://github.com/gchhablani)
|
99 |
+
|
100 |
+
## Credits
|
101 |
+
Thanks to Huggingface 🤗 & Google JAX/FLAX team for such a wonderful community week. Big thanks to [@patrickvonplaten](https://github.com/patrickvonplaten) and [@patil-suraj](https://github.com/patil-suraj) for helping us with our solution during the community week.
|
102 |
+
|
103 |
+
<img src=https://pbs.twimg.com/media/E443fPjX0AY1BsR.jpg:large>
|
config.json
ADDED
@@ -0,0 +1,316 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"CLIPVisionMBartForConditionalGeneration"
|
4 |
+
],
|
5 |
+
"clip_vision_config": {
|
6 |
+
"_name_or_path": "",
|
7 |
+
"add_cross_attention": false,
|
8 |
+
"architectures": [
|
9 |
+
"CLIPModel"
|
10 |
+
],
|
11 |
+
"attention_dropout": 0.0,
|
12 |
+
"bad_words_ids": null,
|
13 |
+
"bos_token_id": null,
|
14 |
+
"chunk_size_feed_forward": 0,
|
15 |
+
"decoder_start_token_id": null,
|
16 |
+
"diversity_penalty": 0.0,
|
17 |
+
"do_sample": false,
|
18 |
+
"dropout": 0.0,
|
19 |
+
"early_stopping": false,
|
20 |
+
"encoder_no_repeat_ngram_size": 0,
|
21 |
+
"eos_token_id": null,
|
22 |
+
"finetuning_task": null,
|
23 |
+
"forced_bos_token_id": null,
|
24 |
+
"forced_eos_token_id": null,
|
25 |
+
"gradient_checkpointing": false,
|
26 |
+
"hidden_act": "quick_gelu",
|
27 |
+
"hidden_size": 768,
|
28 |
+
"id2label": {
|
29 |
+
"0": "LABEL_0",
|
30 |
+
"1": "LABEL_1"
|
31 |
+
},
|
32 |
+
"image_size": 224,
|
33 |
+
"initializer_factor": 1.0,
|
34 |
+
"initializer_range": 0.02,
|
35 |
+
"intermediate_size": 3072,
|
36 |
+
"is_decoder": false,
|
37 |
+
"is_encoder_decoder": false,
|
38 |
+
"label2id": {
|
39 |
+
"LABEL_0": 0,
|
40 |
+
"LABEL_1": 1
|
41 |
+
},
|
42 |
+
"layer_norm_eps": 1e-05,
|
43 |
+
"length_penalty": 1.0,
|
44 |
+
"max_length": 20,
|
45 |
+
"min_length": 0,
|
46 |
+
"model_type": "clip_vision_model",
|
47 |
+
"no_repeat_ngram_size": 0,
|
48 |
+
"num_attention_heads": 12,
|
49 |
+
"num_beam_groups": 1,
|
50 |
+
"num_beams": 1,
|
51 |
+
"num_hidden_layers": 12,
|
52 |
+
"num_return_sequences": 1,
|
53 |
+
"output_attentions": false,
|
54 |
+
"output_hidden_states": false,
|
55 |
+
"output_scores": false,
|
56 |
+
"pad_token_id": null,
|
57 |
+
"patch_size": 32,
|
58 |
+
"prefix": null,
|
59 |
+
"problem_type": null,
|
60 |
+
"projection_dim": 512,
|
61 |
+
"pruned_heads": {},
|
62 |
+
"remove_invalid_values": false,
|
63 |
+
"repetition_penalty": 1.0,
|
64 |
+
"return_dict": true,
|
65 |
+
"return_dict_in_generate": false,
|
66 |
+
"sep_token_id": null,
|
67 |
+
"task_specific_params": null,
|
68 |
+
"temperature": 1.0,
|
69 |
+
"text_config": {
|
70 |
+
"_name_or_path": "",
|
71 |
+
"add_cross_attention": false,
|
72 |
+
"architectures": null,
|
73 |
+
"attention_dropout": 0.0,
|
74 |
+
"bad_words_ids": null,
|
75 |
+
"bos_token_id": 0,
|
76 |
+
"chunk_size_feed_forward": 0,
|
77 |
+
"decoder_start_token_id": null,
|
78 |
+
"diversity_penalty": 0.0,
|
79 |
+
"do_sample": false,
|
80 |
+
"dropout": 0.0,
|
81 |
+
"early_stopping": false,
|
82 |
+
"encoder_no_repeat_ngram_size": 0,
|
83 |
+
"eos_token_id": 2,
|
84 |
+
"finetuning_task": null,
|
85 |
+
"forced_bos_token_id": null,
|
86 |
+
"forced_eos_token_id": null,
|
87 |
+
"gradient_checkpointing": false,
|
88 |
+
"hidden_act": "quick_gelu",
|
89 |
+
"hidden_size": 512,
|
90 |
+
"id2label": {
|
91 |
+
"0": "LABEL_0",
|
92 |
+
"1": "LABEL_1"
|
93 |
+
},
|
94 |
+
"initializer_factor": 1.0,
|
95 |
+
"initializer_range": 0.02,
|
96 |
+
"intermediate_size": 2048,
|
97 |
+
"is_decoder": false,
|
98 |
+
"is_encoder_decoder": false,
|
99 |
+
"label2id": {
|
100 |
+
"LABEL_0": 0,
|
101 |
+
"LABEL_1": 1
|
102 |
+
},
|
103 |
+
"layer_norm_eps": 1e-05,
|
104 |
+
"length_penalty": 1.0,
|
105 |
+
"max_length": 20,
|
106 |
+
"max_position_embeddings": 77,
|
107 |
+
"min_length": 0,
|
108 |
+
"model_type": "clip_text_model",
|
109 |
+
"no_repeat_ngram_size": 0,
|
110 |
+
"num_attention_heads": 8,
|
111 |
+
"num_beam_groups": 1,
|
112 |
+
"num_beams": 1,
|
113 |
+
"num_hidden_layers": 12,
|
114 |
+
"num_return_sequences": 1,
|
115 |
+
"output_attentions": false,
|
116 |
+
"output_hidden_states": false,
|
117 |
+
"output_scores": false,
|
118 |
+
"pad_token_id": 1,
|
119 |
+
"prefix": null,
|
120 |
+
"problem_type": null,
|
121 |
+
"pruned_heads": {},
|
122 |
+
"remove_invalid_values": false,
|
123 |
+
"repetition_penalty": 1.0,
|
124 |
+
"return_dict": true,
|
125 |
+
"return_dict_in_generate": false,
|
126 |
+
"sep_token_id": null,
|
127 |
+
"task_specific_params": null,
|
128 |
+
"temperature": 1.0,
|
129 |
+
"tie_encoder_decoder": false,
|
130 |
+
"tie_word_embeddings": true,
|
131 |
+
"tokenizer_class": null,
|
132 |
+
"top_k": 50,
|
133 |
+
"top_p": 1.0,
|
134 |
+
"torchscript": false,
|
135 |
+
"transformers_version": "4.7.0.dev0",
|
136 |
+
"use_bfloat16": false,
|
137 |
+
"vocab_size": 49408
|
138 |
+
},
|
139 |
+
"text_config_dict": null,
|
140 |
+
"tie_encoder_decoder": false,
|
141 |
+
"tie_word_embeddings": true,
|
142 |
+
"tokenizer_class": null,
|
143 |
+
"top_k": 50,
|
144 |
+
"top_p": 1.0,
|
145 |
+
"torch_dtype": null,
|
146 |
+
"torchscript": false,
|
147 |
+
"transformers_version": "4.9.0.dev0",
|
148 |
+
"use_bfloat16": false,
|
149 |
+
"vision_config": {
|
150 |
+
"_name_or_path": "",
|
151 |
+
"add_cross_attention": false,
|
152 |
+
"architectures": null,
|
153 |
+
"attention_dropout": 0.0,
|
154 |
+
"bad_words_ids": null,
|
155 |
+
"bos_token_id": null,
|
156 |
+
"chunk_size_feed_forward": 0,
|
157 |
+
"decoder_start_token_id": null,
|
158 |
+
"diversity_penalty": 0.0,
|
159 |
+
"do_sample": false,
|
160 |
+
"dropout": 0.0,
|
161 |
+
"early_stopping": false,
|
162 |
+
"encoder_no_repeat_ngram_size": 0,
|
163 |
+
"eos_token_id": null,
|
164 |
+
"finetuning_task": null,
|
165 |
+
"forced_bos_token_id": null,
|
166 |
+
"forced_eos_token_id": null,
|
167 |
+
"gradient_checkpointing": false,
|
168 |
+
"hidden_act": "quick_gelu",
|
169 |
+
"hidden_size": 768,
|
170 |
+
"id2label": {
|
171 |
+
"0": "LABEL_0",
|
172 |
+
"1": "LABEL_1"
|
173 |
+
},
|
174 |
+
"image_size": 224,
|
175 |
+
"initializer_factor": 1.0,
|
176 |
+
"initializer_range": 0.02,
|
177 |
+
"intermediate_size": 3072,
|
178 |
+
"is_decoder": false,
|
179 |
+
"is_encoder_decoder": false,
|
180 |
+
"label2id": {
|
181 |
+
"LABEL_0": 0,
|
182 |
+
"LABEL_1": 1
|
183 |
+
},
|
184 |
+
"layer_norm_eps": 1e-05,
|
185 |
+
"length_penalty": 1.0,
|
186 |
+
"max_length": 20,
|
187 |
+
"min_length": 0,
|
188 |
+
"model_type": "clip_vision_model",
|
189 |
+
"no_repeat_ngram_size": 0,
|
190 |
+
"num_attention_heads": 12,
|
191 |
+
"num_beam_groups": 1,
|
192 |
+
"num_beams": 1,
|
193 |
+
"num_hidden_layers": 12,
|
194 |
+
"num_return_sequences": 1,
|
195 |
+
"output_attentions": false,
|
196 |
+
"output_hidden_states": false,
|
197 |
+
"output_scores": false,
|
198 |
+
"pad_token_id": null,
|
199 |
+
"patch_size": 32,
|
200 |
+
"prefix": null,
|
201 |
+
"problem_type": null,
|
202 |
+
"pruned_heads": {},
|
203 |
+
"remove_invalid_values": false,
|
204 |
+
"repetition_penalty": 1.0,
|
205 |
+
"return_dict": true,
|
206 |
+
"return_dict_in_generate": false,
|
207 |
+
"sep_token_id": null,
|
208 |
+
"task_specific_params": null,
|
209 |
+
"temperature": 1.0,
|
210 |
+
"tie_encoder_decoder": false,
|
211 |
+
"tie_word_embeddings": true,
|
212 |
+
"tokenizer_class": null,
|
213 |
+
"top_k": 50,
|
214 |
+
"top_p": 1.0,
|
215 |
+
"torchscript": false,
|
216 |
+
"transformers_version": "4.7.0.dev0",
|
217 |
+
"use_bfloat16": false
|
218 |
+
},
|
219 |
+
"vision_config_dict": null
|
220 |
+
},
|
221 |
+
"is_encoder_decoder": true,
|
222 |
+
"mbart_config": {
|
223 |
+
"_name_or_path": "/home/suraj/projects/mbart-50/hf_models/mbart-50-large",
|
224 |
+
"_num_labels": 3,
|
225 |
+
"activation_dropout": 0.0,
|
226 |
+
"activation_function": "gelu",
|
227 |
+
"add_bias_logits": false,
|
228 |
+
"add_cross_attention": false,
|
229 |
+
"add_final_layer_norm": true,
|
230 |
+
"architectures": [
|
231 |
+
"MBartForConditionalGeneration"
|
232 |
+
],
|
233 |
+
"attention_dropout": 0.0,
|
234 |
+
"bad_words_ids": null,
|
235 |
+
"bos_token_id": 0,
|
236 |
+
"chunk_size_feed_forward": 0,
|
237 |
+
"classif_dropout": 0.0,
|
238 |
+
"classifier_dropout": 0.0,
|
239 |
+
"d_model": 1024,
|
240 |
+
"decoder_attention_heads": 16,
|
241 |
+
"decoder_ffn_dim": 4096,
|
242 |
+
"decoder_layerdrop": 0.0,
|
243 |
+
"decoder_layers": 12,
|
244 |
+
"decoder_start_token_id": 2,
|
245 |
+
"diversity_penalty": 0.0,
|
246 |
+
"do_sample": false,
|
247 |
+
"dropout": 0.1,
|
248 |
+
"early_stopping": true,
|
249 |
+
"encoder_attention_heads": 16,
|
250 |
+
"encoder_ffn_dim": 4096,
|
251 |
+
"encoder_layerdrop": 0.0,
|
252 |
+
"encoder_layers": 12,
|
253 |
+
"encoder_no_repeat_ngram_size": 0,
|
254 |
+
"eos_token_id": 2,
|
255 |
+
"finetuning_task": null,
|
256 |
+
"forced_bos_token_id": null,
|
257 |
+
"forced_eos_token_id": 2,
|
258 |
+
"gradient_checkpointing": false,
|
259 |
+
"id2label": {
|
260 |
+
"0": "LABEL_0",
|
261 |
+
"1": "LABEL_1",
|
262 |
+
"2": "LABEL_2"
|
263 |
+
},
|
264 |
+
"init_std": 0.02,
|
265 |
+
"is_decoder": false,
|
266 |
+
"is_encoder_decoder": true,
|
267 |
+
"label2id": {
|
268 |
+
"LABEL_0": 0,
|
269 |
+
"LABEL_1": 1,
|
270 |
+
"LABEL_2": 2
|
271 |
+
},
|
272 |
+
"length_penalty": 1.0,
|
273 |
+
"max_length": 200,
|
274 |
+
"max_position_embeddings": 1024,
|
275 |
+
"min_length": 0,
|
276 |
+
"model_type": "mbart",
|
277 |
+
"no_repeat_ngram_size": 0,
|
278 |
+
"normalize_before": true,
|
279 |
+
"normalize_embedding": true,
|
280 |
+
"num_beam_groups": 1,
|
281 |
+
"num_beams": 5,
|
282 |
+
"num_hidden_layers": 12,
|
283 |
+
"num_return_sequences": 1,
|
284 |
+
"output_attentions": false,
|
285 |
+
"output_hidden_states": false,
|
286 |
+
"output_past": true,
|
287 |
+
"output_scores": false,
|
288 |
+
"pad_token_id": 1,
|
289 |
+
"prefix": null,
|
290 |
+
"problem_type": null,
|
291 |
+
"pruned_heads": {},
|
292 |
+
"remove_invalid_values": false,
|
293 |
+
"repetition_penalty": 1.0,
|
294 |
+
"return_dict": true,
|
295 |
+
"return_dict_in_generate": false,
|
296 |
+
"scale_embedding": true,
|
297 |
+
"sep_token_id": null,
|
298 |
+
"static_position_embeddings": false,
|
299 |
+
"task_specific_params": null,
|
300 |
+
"temperature": 1.0,
|
301 |
+
"tie_encoder_decoder": false,
|
302 |
+
"tie_word_embeddings": true,
|
303 |
+
"tokenizer_class": "MBart50Tokenizer",
|
304 |
+
"top_k": 50,
|
305 |
+
"top_p": 1.0,
|
306 |
+
"torch_dtype": null,
|
307 |
+
"torchscript": false,
|
308 |
+
"transformers_version": "4.9.0.dev0",
|
309 |
+
"use_bfloat16": false,
|
310 |
+
"use_cache": true,
|
311 |
+
"vocab_size": 250054
|
312 |
+
},
|
313 |
+
"model_type": "clip-vision-mbart",
|
314 |
+
"seed": 42,
|
315 |
+
"transformers_version": null
|
316 |
+
}
|
logs/events.out.tfevents.1626777663.t1v-n-e32efd45-w-0.1846718.3.v2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3e6c4ab6e9bb14ec893541f156b36d17bea7c9b8d8ba677bf80d7a0bb6c5853d
|
3 |
+
size 13753034
|