Upload 6 files
Browse files- submission4/config.json +53 -0
- submission4/pytorch_model.bin +3 -0
- submission4/train_args.json +32 -0
- submission4/train_results.json +7 -0
- submission4/training_args.bin +3 -0
- submission4/vocab.json +0 -0
submission4/config.json
ADDED
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "../drive/MyDrive/NICE/model/cap_best/OFA-huge-caption",
|
3 |
+
"activation_dropout": 0.0,
|
4 |
+
"activation_function": "gelu",
|
5 |
+
"add_type_embedding": true,
|
6 |
+
"architectures": [
|
7 |
+
"OFAModelForCaption"
|
8 |
+
],
|
9 |
+
"attention_dropout": 0.0,
|
10 |
+
"attn_scale_factor": 2.0,
|
11 |
+
"bos_token_id": 0,
|
12 |
+
"classifier_dropout": 0.0,
|
13 |
+
"code_image_size": 128,
|
14 |
+
"code_layernorm_embedding": true,
|
15 |
+
"d_model": 1280,
|
16 |
+
"decoder_attention_heads": 16,
|
17 |
+
"decoder_drop_path_rate": 0.0,
|
18 |
+
"decoder_ffn_dim": 5120,
|
19 |
+
"decoder_layerdrop": 0.0,
|
20 |
+
"decoder_layers": 12,
|
21 |
+
"decoder_normalize_before": true,
|
22 |
+
"decoder_start_token_id": 0,
|
23 |
+
"dropout": 0.1,
|
24 |
+
"encoder_attention_heads": 16,
|
25 |
+
"encoder_drop_path_rate": 0.0,
|
26 |
+
"encoder_ffn_dim": 5120,
|
27 |
+
"encoder_layerdrop": 0.0,
|
28 |
+
"encoder_layers": 24,
|
29 |
+
"encoder_normalize_before": true,
|
30 |
+
"entangle_position_embedding": false,
|
31 |
+
"eos_token_id": 2,
|
32 |
+
"forced_eos_token_id": 2,
|
33 |
+
"image_bucket_size": 42,
|
34 |
+
"init_std": 0.02,
|
35 |
+
"is_encoder_decoder": true,
|
36 |
+
"layernorm_embedding": true,
|
37 |
+
"max_position_embeddings": 1024,
|
38 |
+
"model_type": "ofa",
|
39 |
+
"normformer": true,
|
40 |
+
"num_hidden_layers": 24,
|
41 |
+
"pad_token_id": 1,
|
42 |
+
"patch_layernorm_embedding": true,
|
43 |
+
"resnet_drop_path_rate": 0.0,
|
44 |
+
"resnet_model_path": null,
|
45 |
+
"resnet_type": "resnet152",
|
46 |
+
"scale_embedding": false,
|
47 |
+
"share_decoder_input_output_embed": true,
|
48 |
+
"token_bucket_size": 256,
|
49 |
+
"torch_dtype": "float32",
|
50 |
+
"transformers_version": "4.20.0",
|
51 |
+
"use_cache": false,
|
52 |
+
"vocab_size": 59457
|
53 |
+
}
|
submission4/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:551818ed0daa472b3f2f1a831d3af838f878116f954a889168fad94d03fa28ba
|
3 |
+
size 3785732668
|
submission4/train_args.json
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"output_dir": "../drive/MyDrive/NICE/model/0430_decoderOnly",
|
3 |
+
"model_name_or_path": "../drive/MyDrive/NICE/model/cap_best/OFA-huge-caption",
|
4 |
+
"train_caption_file": "./data/train_caption_catAdded.jsonl",
|
5 |
+
"train_image_file": "../drive/MyDrive/NICE/train_image.tsv",
|
6 |
+
"test_caption_file": null,
|
7 |
+
"test_image_file": null,
|
8 |
+
"freeze_encoder": true,
|
9 |
+
"freeze_word_embed": false,
|
10 |
+
"num_train_epochs": 40,
|
11 |
+
"max_steps": -1,
|
12 |
+
"per_device_train_batch_size": 24,
|
13 |
+
"per_device_eval_batch_size": 256,
|
14 |
+
"learning_rate": 0.0003,
|
15 |
+
"max_seq_length": 150,
|
16 |
+
"logging_steps": 500,
|
17 |
+
"save_steps": 3000,
|
18 |
+
"save_total_limit": 2,
|
19 |
+
"lr_scheduler_type": "cosine",
|
20 |
+
"warmup_steps": 1500,
|
21 |
+
"warmup_ratio": 0,
|
22 |
+
"gradient_accumulation_steps": 1,
|
23 |
+
"optim": "adamw_torch",
|
24 |
+
"seed": 69,
|
25 |
+
"fp16": false,
|
26 |
+
"no_cuda": false,
|
27 |
+
"dataloader_num_workers": 8,
|
28 |
+
"save_strategy": "steps",
|
29 |
+
"weight_decay": 0,
|
30 |
+
"max_grad_norm": 1.0,
|
31 |
+
"remove_unused_columns": false
|
32 |
+
}
|
submission4/train_results.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 40.0,
|
3 |
+
"train_loss": 0.21260927714870878,
|
4 |
+
"train_runtime": 14185.0539,
|
5 |
+
"train_samples_per_second": 14.099,
|
6 |
+
"train_steps_per_second": 0.589
|
7 |
+
}
|
submission4/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5331d30cb5b38c2049d8f43b1146816535e08d95896805b27f4994b049b2969a
|
3 |
+
size 3323
|
submission4/vocab.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|