cccjc's picture
add weights
f4fb17d
2020-09-28 11:49:37,019 loading file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-vocab.txt from cache at /home/tiger/.cache/torch/transformers/26bc1ad6c0ac742e9b52263248f6d0f00068293b33709fae12320c0e35ccfbbb.542ce4285a40d23a559526243235df47c5f75c197f04f37d1a0c124c32c9a084
2020-09-28 11:49:44,464 Resnet backbone now has fixed blocks 2
2020-09-28 11:49:45,925 loading configuration file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-config.json from cache at /home/tiger/.cache/torch/transformers/4dad0251492946e18ac39290fcfe91b89d370fee250efe9521476438fe8ca185.7156163d5fdc189c3016baca0775ffce230789d7fa2a42ef516483e4ca884517
2020-09-28 11:49:45,926 Model config {
"architectures": [
"BertForMaskedLM"
],
"attention_probs_dropout_prob": 0.1,
"finetuning_task": null,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-12,
"max_position_embeddings": 512,
"model_type": "bert",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"num_labels": 2,
"output_attentions": false,
"output_hidden_states": false,
"pad_token_id": 0,
"pruned_heads": {},
"torchscript": false,
"type_vocab_size": 2,
"use_bfloat16": false,
"vocab_size": 30522
}
2020-09-28 11:49:47,293 loading weights file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-pytorch_model.bin from cache at /home/tiger/.cache/torch/transformers/aa1ef1aede4482d0dbcd4d52baad8ae300e60902e88fcb0bebdec09afd232066.36ca03ab34a1a5d5fa7bc3d03d55c4fa650fed07220e2eeebc06ce58d0e9a157
2020-09-28 11:49:49,480 Use adam as the optimizer, with init lr 0.0005
2020-09-28 11:49:49,480 Image encoder is data paralleled now.
2020-09-28 11:49:49,618 Load full model with backbone
2020-09-28 11:49:49,620 Loading dataset
2020-09-28 11:49:54,581 Input mode small: scaled by factor 2.0
2020-09-28 11:49:59,061 Computing results...
2020-09-28 11:50:56,509 Test: [0/196] Le 63.2128 (63.2128) Time 57.444 (0.000)
2020-09-28 11:51:01,949 Test: [10/196] Le 61.8044 (62.3709) Time 0.613 (0.000)
2020-09-28 11:51:12,583 Test: [20/196] Le 65.8550 (62.5176) Time 2.779 (0.000)
2020-09-28 11:51:19,107 Test: [30/196] Le 63.9750 (62.6710) Time 1.101 (0.000)
2020-09-28 11:51:26,753 Test: [40/196] Le 62.3553 (62.8770) Time 0.599 (0.000)
2020-09-28 11:51:33,178 Test: [50/196] Le 64.0036 (62.9363) Time 1.053 (0.000)
2020-09-28 11:51:41,781 Test: [60/196] Le 60.9549 (62.7791) Time 0.520 (0.000)
2020-09-28 11:51:47,991 Test: [70/196] Le 60.1947 (62.6902) Time 0.582 (0.000)
2020-09-28 11:51:55,115 Test: [80/196] Le 64.8011 (62.8592) Time 0.521 (0.000)
2020-09-28 11:52:02,522 Test: [90/196] Le 63.6653 (62.8890) Time 0.523 (0.000)
2020-09-28 11:52:11,548 Test: [100/196] Le 63.3472 (62.8601) Time 0.593 (0.000)
2020-09-28 11:52:18,060 Test: [110/196] Le 59.5976 (62.7785) Time 0.535 (0.000)
2020-09-28 11:52:26,472 Test: [120/196] Le 64.8909 (62.8633) Time 0.524 (0.000)
2020-09-28 11:52:32,687 Test: [130/196] Le 64.0768 (62.8651) Time 0.542 (0.000)
2020-09-28 11:52:39,706 Test: [140/196] Le 66.7743 (62.8628) Time 0.532 (0.000)
2020-09-28 11:52:45,961 Test: [150/196] Le 63.4051 (62.8979) Time 0.518 (0.000)
2020-09-28 11:52:53,645 Test: [160/196] Le 62.4264 (62.9269) Time 0.519 (0.000)
2020-09-28 11:53:00,141 Test: [170/196] Le 64.1213 (62.9178) Time 0.513 (0.000)
2020-09-28 11:53:07,404 Test: [180/196] Le 61.0356 (62.9458) Time 0.516 (0.000)
2020-09-28 11:53:13,988 Test: [190/196] Le 61.7361 (62.9015) Time 0.515 (0.000)
2020-09-28 11:53:19,655 Images: 5000, Captions: 25000
2020-09-28 11:53:50,253 Align loss: 0.9592935465926018
2020-09-28 11:53:50,253 Image uniform loss: -3.825332749718092
2020-09-28 11:53:50,253 Text uniform loss: -3.885177468724295
2020-09-28 11:53:50,294 calculate similarity time:
2020-09-28 11:53:50,633 Image to text: 82.9, 98.0, 99.7, 1.0, 1.4
2020-09-28 11:53:50,921 Text to image: 67.8, 92.7, 96.7, 1.0, 3.8
2020-09-28 11:53:50,922 rsum: 537.8 ar: 93.5 ari: 85.7
2020-09-28 11:53:50,993 calculate similarity time:
2020-09-28 11:53:51,336 Image to text: 79.4, 96.0, 98.7, 1.0, 1.8
2020-09-28 11:53:51,623 Text to image: 65.9, 91.4, 96.3, 1.0, 3.6
2020-09-28 11:53:51,623 rsum: 527.6 ar: 91.4 ari: 84.5
2020-09-28 11:53:51,694 calculate similarity time:
2020-09-28 11:53:52,036 Image to text: 79.8, 97.2, 99.3, 1.0, 1.6
2020-09-28 11:53:52,323 Text to image: 66.7, 91.9, 96.9, 1.0, 3.8
2020-09-28 11:53:52,323 rsum: 531.8 ar: 92.1 ari: 85.2
2020-09-28 11:53:52,396 calculate similarity time:
2020-09-28 11:53:52,739 Image to text: 78.8, 96.4, 98.8, 1.0, 1.7
2020-09-28 11:53:53,027 Text to image: 64.5, 91.8, 96.6, 1.0, 3.2
2020-09-28 11:53:53,027 rsum: 526.9 ar: 91.3 ari: 84.3
2020-09-28 11:53:53,100 calculate similarity time:
2020-09-28 11:53:53,445 Image to text: 81.2, 96.5, 99.2, 1.0, 1.5
2020-09-28 11:53:53,732 Text to image: 67.3, 92.6, 96.9, 1.0, 3.4
2020-09-28 11:53:53,733 rsum: 533.8 ar: 92.3 ari: 85.6
2020-09-28 11:53:53,733 -----------------------------------
2020-09-28 11:53:53,733 Mean metrics:
2020-09-28 11:53:53,733 rsum: 531.6
2020-09-28 11:53:53,733 Average i2t Recall: 92.1
2020-09-28 11:53:53,733 Image to text: 80.4 96.8 99.1 1.0 1.6
2020-09-28 11:53:53,733 Average t2i Recall: 85.1
2020-09-28 11:53:53,733 Text to image: 66.4 92.1 96.7 1.0 3.5
2020-09-28 11:53:55,578 loading file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-vocab.txt from cache at /home/tiger/.cache/torch/transformers/26bc1ad6c0ac742e9b52263248f6d0f00068293b33709fae12320c0e35ccfbbb.542ce4285a40d23a559526243235df47c5f75c197f04f37d1a0c124c32c9a084
2020-09-28 11:54:01,413 Resnet backbone now has fixed blocks 2
2020-09-28 11:54:02,873 loading configuration file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-config.json from cache at /home/tiger/.cache/torch/transformers/4dad0251492946e18ac39290fcfe91b89d370fee250efe9521476438fe8ca185.7156163d5fdc189c3016baca0775ffce230789d7fa2a42ef516483e4ca884517
2020-09-28 11:54:02,874 Model config {
"architectures": [
"BertForMaskedLM"
],
"attention_probs_dropout_prob": 0.1,
"finetuning_task": null,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-12,
"max_position_embeddings": 512,
"model_type": "bert",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"num_labels": 2,
"output_attentions": false,
"output_hidden_states": false,
"pad_token_id": 0,
"pruned_heads": {},
"torchscript": false,
"type_vocab_size": 2,
"use_bfloat16": false,
"vocab_size": 30522
}
2020-09-28 11:54:04,385 loading weights file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-pytorch_model.bin from cache at /home/tiger/.cache/torch/transformers/aa1ef1aede4482d0dbcd4d52baad8ae300e60902e88fcb0bebdec09afd232066.36ca03ab34a1a5d5fa7bc3d03d55c4fa650fed07220e2eeebc06ce58d0e9a157
2020-09-28 11:54:06,572 Use adam as the optimizer, with init lr 0.0005
2020-09-28 11:54:06,573 Image encoder is data paralleled now.
2020-09-28 11:54:06,713 Load full model with backbone
2020-09-28 11:54:06,716 Loading dataset
2020-09-28 11:54:10,054 Input mode small: scaled by factor 2.0
2020-09-28 11:54:16,134 Computing results...
2020-09-28 11:54:34,686 Test: [0/196] Le 63.2128 (63.2128) Time 18.549 (0.000)
2020-09-28 11:54:41,359 Test: [10/196] Le 61.8044 (62.3709) Time 1.064 (0.000)
2020-09-28 11:54:49,852 Test: [20/196] Le 65.8550 (62.5176) Time 0.587 (0.000)
2020-09-28 11:54:55,446 Test: [30/196] Le 63.9750 (62.6710) Time 0.520 (0.000)
2020-09-28 11:55:02,567 Test: [40/196] Le 62.3553 (62.8770) Time 0.527 (0.000)
2020-09-28 11:55:11,015 Test: [50/196] Le 64.0036 (62.9363) Time 0.612 (0.000)
2020-09-28 11:55:20,332 Test: [60/196] Le 60.9549 (62.7791) Time 0.526 (0.000)
2020-09-28 11:55:25,946 Test: [70/196] Le 60.1947 (62.6902) Time 0.582 (0.000)
2020-09-28 11:55:34,129 Test: [80/196] Le 64.8011 (62.8592) Time 0.522 (0.000)
2020-09-28 11:55:40,196 Test: [90/196] Le 63.6653 (62.8890) Time 0.569 (0.000)
2020-09-28 11:55:49,241 Test: [100/196] Le 63.3472 (62.8601) Time 0.585 (0.000)
2020-09-28 11:55:54,633 Test: [110/196] Le 59.5976 (62.7785) Time 0.523 (0.000)
2020-09-28 11:56:01,693 Test: [120/196] Le 64.8909 (62.8633) Time 0.524 (0.000)
2020-09-28 11:56:07,865 Test: [130/196] Le 64.0768 (62.8651) Time 0.519 (0.000)
2020-09-28 11:56:16,333 Test: [140/196] Le 66.7743 (62.8628) Time 0.524 (0.000)
2020-09-28 11:56:22,923 Test: [150/196] Le 63.4051 (62.8979) Time 0.601 (0.000)
2020-09-28 11:56:33,561 Test: [160/196] Le 62.4264 (62.9269) Time 0.530 (0.000)
2020-09-28 11:56:40,031 Test: [170/196] Le 64.1213 (62.9178) Time 0.544 (0.000)
2020-09-28 11:56:49,370 Test: [180/196] Le 61.0356 (62.9458) Time 0.520 (0.000)
2020-09-28 11:56:58,484 Test: [190/196] Le 61.7361 (62.9015) Time 0.522 (0.000)
2020-09-28 11:57:03,232 Images: 5000, Captions: 25000
2020-09-28 11:57:56,277 rsum: 440.0
2020-09-28 11:57:56,277 Average i2t Recall: 79.3
2020-09-28 11:57:56,277 Image to text: 59.1 85.9 92.8 1.0 3.9
2020-09-28 11:57:56,277 Average t2i Recall: 67.4
2020-09-28 11:57:56,277 Text to image: 44.1 74.1 84.0 2.0 13.6