Test passed
Browse files- all_results.json +17 -0
- config.json +17 -32
- eval_results.json +8 -0
- preprocessor_config.json +14 -11
- pytorch_model.bin +2 -2
- runs/Aug30_14-29-36_surveily-ag-02/events.out.tfevents.1693398700.surveily-ag-02.5833.3 +3 -0
- runs/Aug31_10-33-05_surveily-ag-02/events.out.tfevents.1693470786.surveily-ag-02.5559.0 +3 -0
- runs/Aug31_10-33-05_surveily-ag-02/events.out.tfevents.1693470800.surveily-ag-02.5559.1 +3 -0
- runs/Aug31_10-45-42_surveily-ag-02/events.out.tfevents.1693471542.surveily-ag-02.5559.2 +3 -0
- runs/Aug31_10-45-42_surveily-ag-02/events.out.tfevents.1693471618.surveily-ag-02.5559.3 +3 -0
- runs/Aug31_10-49-15_surveily-ag-02/events.out.tfevents.1693471755.surveily-ag-02.5559.4 +3 -0
- runs/Aug31_10-49-15_surveily-ag-02/events.out.tfevents.1693471822.surveily-ag-02.5559.5 +3 -0
- runs/Aug31_11-01-42_surveily-ag-02/events.out.tfevents.1693472502.surveily-ag-02.5559.6 +3 -0
- runs/Aug31_11-01-42_surveily-ag-02/events.out.tfevents.1693472571.surveily-ag-02.5559.7 +3 -0
- runs/Aug31_11-11-25_surveily-ag-02/events.out.tfevents.1693473085.surveily-ag-02.5559.8 +3 -0
- runs/Aug31_11-11-25_surveily-ag-02/events.out.tfevents.1693473243.surveily-ag-02.5559.9 +3 -0
- runs/Aug31_11-26-23_surveily-ag-02/events.out.tfevents.1693473984.surveily-ag-02.6479.0 +3 -0
- runs/Aug31_11-26-23_surveily-ag-02/events.out.tfevents.1693474252.surveily-ag-02.6479.1 +3 -0
- test_results.json +7 -0
- train_results.json +7 -0
- trainer_state.json +193 -0
- training_args.bin +2 -2
all_results.json
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 5.0,
|
3 |
+
"eval_accuracy": 0.9988452655889145,
|
4 |
+
"eval_loss": 0.0029988216701895,
|
5 |
+
"eval_runtime": 7.4103,
|
6 |
+
"eval_samples_per_second": 116.864,
|
7 |
+
"eval_steps_per_second": 14.709,
|
8 |
+
"test_accuracy": 0.9,
|
9 |
+
"test_loss": 0.5731536149978638,
|
10 |
+
"test_runtime": 0.4078,
|
11 |
+
"test_samples_per_second": 49.045,
|
12 |
+
"test_steps_per_second": 7.357,
|
13 |
+
"train_loss": 0.05322713415537562,
|
14 |
+
"train_runtime": 260.1266,
|
15 |
+
"train_samples_per_second": 64.507,
|
16 |
+
"train_steps_per_second": 6.458
|
17 |
+
}
|
config.json
CHANGED
@@ -1,48 +1,33 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "
|
3 |
"architectures": [
|
4 |
-
"
|
5 |
-
],
|
6 |
-
"depths": [
|
7 |
-
3,
|
8 |
-
4,
|
9 |
-
6,
|
10 |
-
3
|
11 |
-
],
|
12 |
-
"downsample_in_first_stage": false,
|
13 |
-
"embedding_size": 64,
|
14 |
-
"hidden_act": "relu",
|
15 |
-
"hidden_sizes": [
|
16 |
-
256,
|
17 |
-
512,
|
18 |
-
1024,
|
19 |
-
2048
|
20 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
"id2label": {
|
22 |
"0": "day",
|
23 |
"1": "night"
|
24 |
},
|
|
|
|
|
25 |
"label2id": {
|
26 |
"day": "0",
|
27 |
"night": "1"
|
28 |
},
|
29 |
-
"
|
30 |
-
"
|
|
|
31 |
"num_channels": 3,
|
32 |
-
"
|
33 |
-
"stage4"
|
34 |
-
],
|
35 |
-
"out_indices": [
|
36 |
-
4
|
37 |
-
],
|
38 |
"problem_type": "single_label_classification",
|
39 |
-
"
|
40 |
-
|
41 |
-
"stage1",
|
42 |
-
"stage2",
|
43 |
-
"stage3",
|
44 |
-
"stage4"
|
45 |
-
],
|
46 |
"torch_dtype": "float32",
|
47 |
"transformers_version": "4.31.0"
|
48 |
}
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "google/mobilenet_v2_0.75_160",
|
3 |
"architectures": [
|
4 |
+
"MobileNetV2ForImageClassification"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
],
|
6 |
+
"classifier_dropout_prob": 0.2,
|
7 |
+
"depth_divisible_by": 8,
|
8 |
+
"depth_multiplier": 0.75,
|
9 |
+
"expand_ratio": 6,
|
10 |
+
"finegrained_output": true,
|
11 |
+
"first_layer_is_expansion": true,
|
12 |
+
"hidden_act": "relu6",
|
13 |
"id2label": {
|
14 |
"0": "day",
|
15 |
"1": "night"
|
16 |
},
|
17 |
+
"image_size": 160,
|
18 |
+
"initializer_range": 0.02,
|
19 |
"label2id": {
|
20 |
"day": "0",
|
21 |
"night": "1"
|
22 |
},
|
23 |
+
"layer_norm_eps": 0.001,
|
24 |
+
"min_depth": 8,
|
25 |
+
"model_type": "mobilenet_v2",
|
26 |
"num_channels": 3,
|
27 |
+
"output_stride": 32,
|
|
|
|
|
|
|
|
|
|
|
28 |
"problem_type": "single_label_classification",
|
29 |
+
"semantic_loss_ignore_index": 255,
|
30 |
+
"tf_padding": true,
|
|
|
|
|
|
|
|
|
|
|
31 |
"torch_dtype": "float32",
|
32 |
"transformers_version": "4.31.0"
|
33 |
}
|
eval_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 5.0,
|
3 |
+
"eval_accuracy": 0.9988452655889145,
|
4 |
+
"eval_loss": 0.0029988216701895,
|
5 |
+
"eval_runtime": 7.4103,
|
6 |
+
"eval_samples_per_second": 116.864,
|
7 |
+
"eval_steps_per_second": 14.709
|
8 |
+
}
|
preprocessor_config.json
CHANGED
@@ -1,23 +1,26 @@
|
|
1 |
{
|
2 |
-
"
|
|
|
|
|
|
|
|
|
3 |
"do_normalize": true,
|
4 |
"do_rescale": true,
|
5 |
"do_resize": true,
|
6 |
-
"feature_extractor_type": "ConvNextFeatureExtractor",
|
7 |
"image_mean": [
|
8 |
-
0.
|
9 |
-
0.
|
10 |
-
0.
|
11 |
],
|
12 |
-
"image_processor_type": "
|
13 |
"image_std": [
|
14 |
-
0.
|
15 |
-
0.
|
16 |
-
0.
|
17 |
],
|
18 |
-
"resample":
|
19 |
"rescale_factor": 0.00392156862745098,
|
20 |
"size": {
|
21 |
-
"shortest_edge":
|
22 |
}
|
23 |
}
|
|
|
1 |
{
|
2 |
+
"crop_size": {
|
3 |
+
"height": 160,
|
4 |
+
"width": 160
|
5 |
+
},
|
6 |
+
"do_center_crop": true,
|
7 |
"do_normalize": true,
|
8 |
"do_rescale": true,
|
9 |
"do_resize": true,
|
|
|
10 |
"image_mean": [
|
11 |
+
0.5,
|
12 |
+
0.5,
|
13 |
+
0.5
|
14 |
],
|
15 |
+
"image_processor_type": "MobileNetV2FeatureExtractor",
|
16 |
"image_std": [
|
17 |
+
0.5,
|
18 |
+
0.5,
|
19 |
+
0.5
|
20 |
],
|
21 |
+
"resample": 2,
|
22 |
"rescale_factor": 0.00392156862745098,
|
23 |
"size": {
|
24 |
+
"shortest_edge": 192
|
25 |
}
|
26 |
}
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1508b32832f4c05bf52dfa281ac854c11e913b37545bc2c2bc96ea1921bcde1a
|
3 |
+
size 5645493
|
runs/Aug30_14-29-36_surveily-ag-02/events.out.tfevents.1693398700.surveily-ag-02.5833.3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:faf7f007dd35b7a6f14b8a05301469e8d51702b8a72154364f56d9a22bf1d498
|
3 |
+
size 411
|
runs/Aug31_10-33-05_surveily-ag-02/events.out.tfevents.1693470786.surveily-ag-02.5559.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2167a09814e611e1e41bfba81cc253f2e4281c1e1004e7271b4e8a078a18f856
|
3 |
+
size 4631
|
runs/Aug31_10-33-05_surveily-ag-02/events.out.tfevents.1693470800.surveily-ag-02.5559.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f123f85907305ff4902283fcea3eda24b7266966160521470b2e06c2deeb2b80
|
3 |
+
size 405
|
runs/Aug31_10-45-42_surveily-ag-02/events.out.tfevents.1693471542.surveily-ag-02.5559.2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5adc5aef84c138d8047c77d5bb04cc7a5efb6f90455d3e9c64ccc11a837d8a1f
|
3 |
+
size 6874
|
runs/Aug31_10-45-42_surveily-ag-02/events.out.tfevents.1693471618.surveily-ag-02.5559.3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9a7165c4106f7c4811f0d0a34c93a123445bcddb5142bcb12817c3501af09907
|
3 |
+
size 411
|
runs/Aug31_10-49-15_surveily-ag-02/events.out.tfevents.1693471755.surveily-ag-02.5559.4
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ee3936723ab4ee09131dfbdceb2faf6c08d228adf659e37f30e12f4c0659c857
|
3 |
+
size 5804
|
runs/Aug31_10-49-15_surveily-ag-02/events.out.tfevents.1693471822.surveily-ag-02.5559.5
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1990c863753e87d7d14b3843472e00e1c94c3a509128f3a1525ab742d9fbe187
|
3 |
+
size 411
|
runs/Aug31_11-01-42_surveily-ag-02/events.out.tfevents.1693472502.surveily-ag-02.5559.6
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5f42754a15d5e5f5f99398c2882a5d309d3fe79f1eb97f4807269edf3164986b
|
3 |
+
size 5755
|
runs/Aug31_11-01-42_surveily-ag-02/events.out.tfevents.1693472571.surveily-ag-02.5559.7
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:df763bc6e1533acc9665f8418ef8296a5ded5371b5a7c314ae9fdc832f470be3
|
3 |
+
size 411
|
runs/Aug31_11-11-25_surveily-ag-02/events.out.tfevents.1693473085.surveily-ag-02.5559.8
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0a3bca38e25facf83ae232506d4e3acddc09e81b951941f3f62ad12d14552ab6
|
3 |
+
size 8303
|
runs/Aug31_11-11-25_surveily-ag-02/events.out.tfevents.1693473243.surveily-ag-02.5559.9
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:50948d9c836419a7e928ae836f93cb27954a544c1fb6923b0bc3e729d8028873
|
3 |
+
size 411
|
runs/Aug31_11-26-23_surveily-ag-02/events.out.tfevents.1693473984.surveily-ag-02.6479.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a10ada439f0186aca07e39986ebbcc168477350cadb5daf719b5d45b23f364a8
|
3 |
+
size 9577
|
runs/Aug31_11-26-23_surveily-ag-02/events.out.tfevents.1693474252.surveily-ag-02.6479.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fa52a9267f21303797450110279dd7163b56eaa70cf9edf3e93b3c561e92dc0a
|
3 |
+
size 411
|
test_results.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"test_accuracy": 0.9,
|
3 |
+
"test_loss": 0.5731536149978638,
|
4 |
+
"test_runtime": 0.4078,
|
5 |
+
"test_samples_per_second": 49.045,
|
6 |
+
"test_steps_per_second": 7.357
|
7 |
+
}
|
train_results.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 5.0,
|
3 |
+
"train_loss": 0.05322713415537562,
|
4 |
+
"train_runtime": 260.1266,
|
5 |
+
"train_samples_per_second": 64.507,
|
6 |
+
"train_steps_per_second": 6.458
|
7 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,193 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.0029988216701895,
|
3 |
+
"best_model_checkpoint": "./models/day-night/checkpoint-800",
|
4 |
+
"epoch": 5.0,
|
5 |
+
"global_step": 1680,
|
6 |
+
"is_hyper_param_search": false,
|
7 |
+
"is_local_process_zero": true,
|
8 |
+
"is_world_process_zero": true,
|
9 |
+
"log_history": [
|
10 |
+
{
|
11 |
+
"epoch": 0.3,
|
12 |
+
"learning_rate": 9.410714285714286e-05,
|
13 |
+
"loss": 0.2756,
|
14 |
+
"step": 100
|
15 |
+
},
|
16 |
+
{
|
17 |
+
"epoch": 0.6,
|
18 |
+
"learning_rate": 8.815476190476191e-05,
|
19 |
+
"loss": 0.1021,
|
20 |
+
"step": 200
|
21 |
+
},
|
22 |
+
{
|
23 |
+
"epoch": 0.6,
|
24 |
+
"eval_accuracy": 0.9734411085450346,
|
25 |
+
"eval_loss": 0.06794659793376923,
|
26 |
+
"eval_runtime": 7.5982,
|
27 |
+
"eval_samples_per_second": 113.974,
|
28 |
+
"eval_steps_per_second": 14.345,
|
29 |
+
"step": 200
|
30 |
+
},
|
31 |
+
{
|
32 |
+
"epoch": 0.89,
|
33 |
+
"learning_rate": 8.226190476190476e-05,
|
34 |
+
"loss": 0.1375,
|
35 |
+
"step": 300
|
36 |
+
},
|
37 |
+
{
|
38 |
+
"epoch": 1.19,
|
39 |
+
"learning_rate": 7.630952380952381e-05,
|
40 |
+
"loss": 0.0199,
|
41 |
+
"step": 400
|
42 |
+
},
|
43 |
+
{
|
44 |
+
"epoch": 1.19,
|
45 |
+
"eval_accuracy": 0.9919168591224018,
|
46 |
+
"eval_loss": 0.018402865156531334,
|
47 |
+
"eval_runtime": 7.4284,
|
48 |
+
"eval_samples_per_second": 116.58,
|
49 |
+
"eval_steps_per_second": 14.673,
|
50 |
+
"step": 400
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"epoch": 1.49,
|
54 |
+
"learning_rate": 7.035714285714287e-05,
|
55 |
+
"loss": 0.0518,
|
56 |
+
"step": 500
|
57 |
+
},
|
58 |
+
{
|
59 |
+
"epoch": 1.79,
|
60 |
+
"learning_rate": 6.446428571428572e-05,
|
61 |
+
"loss": 0.0723,
|
62 |
+
"step": 600
|
63 |
+
},
|
64 |
+
{
|
65 |
+
"epoch": 1.79,
|
66 |
+
"eval_accuracy": 0.7852193995381063,
|
67 |
+
"eval_loss": 0.6625365018844604,
|
68 |
+
"eval_runtime": 7.266,
|
69 |
+
"eval_samples_per_second": 119.184,
|
70 |
+
"eval_steps_per_second": 15.001,
|
71 |
+
"step": 600
|
72 |
+
},
|
73 |
+
{
|
74 |
+
"epoch": 2.08,
|
75 |
+
"learning_rate": 5.851190476190477e-05,
|
76 |
+
"loss": 0.0278,
|
77 |
+
"step": 700
|
78 |
+
},
|
79 |
+
{
|
80 |
+
"epoch": 2.38,
|
81 |
+
"learning_rate": 5.2559523809523814e-05,
|
82 |
+
"loss": 0.0247,
|
83 |
+
"step": 800
|
84 |
+
},
|
85 |
+
{
|
86 |
+
"epoch": 2.38,
|
87 |
+
"eval_accuracy": 0.9988452655889145,
|
88 |
+
"eval_loss": 0.0029988216701895,
|
89 |
+
"eval_runtime": 7.7626,
|
90 |
+
"eval_samples_per_second": 111.56,
|
91 |
+
"eval_steps_per_second": 14.042,
|
92 |
+
"step": 800
|
93 |
+
},
|
94 |
+
{
|
95 |
+
"epoch": 2.68,
|
96 |
+
"learning_rate": 4.660714285714286e-05,
|
97 |
+
"loss": 0.0576,
|
98 |
+
"step": 900
|
99 |
+
},
|
100 |
+
{
|
101 |
+
"epoch": 2.98,
|
102 |
+
"learning_rate": 4.0654761904761904e-05,
|
103 |
+
"loss": 0.0273,
|
104 |
+
"step": 1000
|
105 |
+
},
|
106 |
+
{
|
107 |
+
"epoch": 2.98,
|
108 |
+
"eval_accuracy": 0.9884526558891455,
|
109 |
+
"eval_loss": 0.025376001372933388,
|
110 |
+
"eval_runtime": 7.4049,
|
111 |
+
"eval_samples_per_second": 116.95,
|
112 |
+
"eval_steps_per_second": 14.72,
|
113 |
+
"step": 1000
|
114 |
+
},
|
115 |
+
{
|
116 |
+
"epoch": 3.27,
|
117 |
+
"learning_rate": 3.4702380952380956e-05,
|
118 |
+
"loss": 0.0116,
|
119 |
+
"step": 1100
|
120 |
+
},
|
121 |
+
{
|
122 |
+
"epoch": 3.57,
|
123 |
+
"learning_rate": 2.8749999999999997e-05,
|
124 |
+
"loss": 0.012,
|
125 |
+
"step": 1200
|
126 |
+
},
|
127 |
+
{
|
128 |
+
"epoch": 3.57,
|
129 |
+
"eval_accuracy": 0.9965357967667436,
|
130 |
+
"eval_loss": 0.017664099112153053,
|
131 |
+
"eval_runtime": 7.9179,
|
132 |
+
"eval_samples_per_second": 109.373,
|
133 |
+
"eval_steps_per_second": 13.766,
|
134 |
+
"step": 1200
|
135 |
+
},
|
136 |
+
{
|
137 |
+
"epoch": 3.87,
|
138 |
+
"learning_rate": 2.279761904761905e-05,
|
139 |
+
"loss": 0.0266,
|
140 |
+
"step": 1300
|
141 |
+
},
|
142 |
+
{
|
143 |
+
"epoch": 4.17,
|
144 |
+
"learning_rate": 1.6845238095238097e-05,
|
145 |
+
"loss": 0.0142,
|
146 |
+
"step": 1400
|
147 |
+
},
|
148 |
+
{
|
149 |
+
"epoch": 4.17,
|
150 |
+
"eval_accuracy": 0.9907621247113164,
|
151 |
+
"eval_loss": 0.028172077611088753,
|
152 |
+
"eval_runtime": 7.4701,
|
153 |
+
"eval_samples_per_second": 115.929,
|
154 |
+
"eval_steps_per_second": 14.591,
|
155 |
+
"step": 1400
|
156 |
+
},
|
157 |
+
{
|
158 |
+
"epoch": 4.46,
|
159 |
+
"learning_rate": 1.0892857142857144e-05,
|
160 |
+
"loss": 0.0157,
|
161 |
+
"step": 1500
|
162 |
+
},
|
163 |
+
{
|
164 |
+
"epoch": 4.76,
|
165 |
+
"learning_rate": 4.940476190476191e-06,
|
166 |
+
"loss": 0.009,
|
167 |
+
"step": 1600
|
168 |
+
},
|
169 |
+
{
|
170 |
+
"epoch": 4.76,
|
171 |
+
"eval_accuracy": 0.9976905311778291,
|
172 |
+
"eval_loss": 0.01886291243135929,
|
173 |
+
"eval_runtime": 7.2095,
|
174 |
+
"eval_samples_per_second": 120.119,
|
175 |
+
"eval_steps_per_second": 15.119,
|
176 |
+
"step": 1600
|
177 |
+
},
|
178 |
+
{
|
179 |
+
"epoch": 5.0,
|
180 |
+
"step": 1680,
|
181 |
+
"total_flos": 1.0500251940864e+16,
|
182 |
+
"train_loss": 0.05322713415537562,
|
183 |
+
"train_runtime": 260.1266,
|
184 |
+
"train_samples_per_second": 64.507,
|
185 |
+
"train_steps_per_second": 6.458
|
186 |
+
}
|
187 |
+
],
|
188 |
+
"max_steps": 1680,
|
189 |
+
"num_train_epochs": 5,
|
190 |
+
"total_flos": 1.0500251940864e+16,
|
191 |
+
"trial_name": null,
|
192 |
+
"trial_params": null
|
193 |
+
}
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:53b4ae9d886a6031aa41fc0c36cd6d532ef10e49f51b7a12c1bd7368cd09b697
|
3 |
+
size 4027
|