Training in progress, epoch 0
Browse files- all_results.json +5 -5
- config.json +2 -2
- model.safetensors +1 -1
- runs/May13_19-31-28_MSI/events.out.tfevents.1715627469.MSI.14116.1 +2 -2
- runs/May13_23-09-51_MSI/events.out.tfevents.1715634595.MSI.14116.2 +3 -0
- runs/May13_23-30-50_MSI/events.out.tfevents.1715635853.MSI.30180.0 +3 -0
- runs/May14_11-21-18_MSI/events.out.tfevents.1715678481.MSI.18796.0 +3 -0
- runs/May16_18-48-25_MSI/events.out.tfevents.1715878108.MSI.19764.0 +3 -0
- runs/May16_18-51-17_MSI/events.out.tfevents.1715878279.MSI.19764.1 +3 -0
- runs/May16_18-59-46_MSI/events.out.tfevents.1715878788.MSI.22228.0 +3 -0
- test_results.json +5 -5
- trainer_state.json +64 -64
- training_args.bin +1 -1
all_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 3.22972972972973,
|
3 |
-
"eval_accuracy": 0.
|
4 |
-
"eval_loss": 0.
|
5 |
-
"eval_runtime":
|
6 |
-
"eval_samples_per_second": 0.
|
7 |
-
"eval_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 3.22972972972973,
|
3 |
+
"eval_accuracy": 0.6956521739130435,
|
4 |
+
"eval_loss": 0.6255780458450317,
|
5 |
+
"eval_runtime": 237.0624,
|
6 |
+
"eval_samples_per_second": 0.291,
|
7 |
+
"eval_steps_per_second": 0.076
|
8 |
}
|
config.json
CHANGED
@@ -12,14 +12,14 @@
|
|
12 |
"hidden_dropout_prob": 0.0,
|
13 |
"hidden_size": 768,
|
14 |
"id2label": {
|
15 |
-
"0": "
|
16 |
"1": "roller"
|
17 |
},
|
18 |
"image_size": 224,
|
19 |
"initializer_range": 0.02,
|
20 |
"intermediate_size": 3072,
|
21 |
"label2id": {
|
22 |
-
"
|
23 |
"roller": 1
|
24 |
},
|
25 |
"layer_norm_eps": 1e-12,
|
|
|
12 |
"hidden_dropout_prob": 0.0,
|
13 |
"hidden_size": 768,
|
14 |
"id2label": {
|
15 |
+
"0": "cutback-frontside",
|
16 |
"1": "roller"
|
17 |
},
|
18 |
"image_size": 224,
|
19 |
"initializer_range": 0.02,
|
20 |
"intermediate_size": 3072,
|
21 |
"label2id": {
|
22 |
+
"cutback-frontside": 0,
|
23 |
"roller": 1
|
24 |
},
|
25 |
"layer_norm_eps": 1e-12,
|
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 344937368
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8e740027589c2af09dc7b9a04424082467d598d1cf580ef053dd1e9ad9ad5080
|
3 |
size 344937368
|
runs/May13_19-31-28_MSI/events.out.tfevents.1715627469.MSI.14116.1
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 sha256:8505e70c366220741f4e09e37a5ec1829cf9384d811b366f408a66c035ad2c43
|
3 |
+
size 1703
|
runs/May13_23-09-51_MSI/events.out.tfevents.1715634595.MSI.14116.2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f9f0087eea08db8608341b5c755d19a485afbde807aecd6c2430edfc8358ffef
|
3 |
+
size 4948
|
runs/May13_23-30-50_MSI/events.out.tfevents.1715635853.MSI.30180.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d75636ba592dbc97ac1b3a1bc51c9c71b84d62d6b87d7935c2ad8812e65c65ec
|
3 |
+
size 4947
|
runs/May14_11-21-18_MSI/events.out.tfevents.1715678481.MSI.18796.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f00be69255b6bcec18c4d6d38f50b94c70de23628605f7fe40ec33467de6df78
|
3 |
+
size 4947
|
runs/May16_18-48-25_MSI/events.out.tfevents.1715878108.MSI.19764.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fa70203df4fa0c850965c9e5e1aed629e3d9dab7e497691a918736c315fce1e9
|
3 |
+
size 4951
|
runs/May16_18-51-17_MSI/events.out.tfevents.1715878279.MSI.19764.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3dccba6dc590aa794dc22c13dc73931497d3da95ce9afac67fca04446a8e58ee
|
3 |
+
size 4951
|
runs/May16_18-59-46_MSI/events.out.tfevents.1715878788.MSI.22228.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2e511375ef359c8e8f9e80910bad2351425322a4deef23e4f5ae4a778cc05a35
|
3 |
+
size 5887
|
test_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 3.22972972972973,
|
3 |
-
"eval_accuracy": 0.
|
4 |
-
"eval_loss": 0.
|
5 |
-
"eval_runtime":
|
6 |
-
"eval_samples_per_second": 0.
|
7 |
-
"eval_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 3.22972972972973,
|
3 |
+
"eval_accuracy": 0.6956521739130435,
|
4 |
+
"eval_loss": 0.6255780458450317,
|
5 |
+
"eval_runtime": 237.0624,
|
6 |
+
"eval_samples_per_second": 0.291,
|
7 |
+
"eval_steps_per_second": 0.076
|
8 |
}
|
trainer_state.json
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
{
|
2 |
-
"best_metric": 0.
|
3 |
-
"best_model_checkpoint": "videomae-base-finetuned-2\\checkpoint-
|
4 |
"epoch": 3.22972972972973,
|
5 |
"eval_steps": 500,
|
6 |
"global_step": 148,
|
@@ -10,163 +10,163 @@
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.06756756756756757,
|
13 |
-
"grad_norm":
|
14 |
"learning_rate": 3.3333333333333335e-05,
|
15 |
-
"loss":
|
16 |
"step": 10
|
17 |
},
|
18 |
{
|
19 |
"epoch": 0.13513513513513514,
|
20 |
-
"grad_norm":
|
21 |
"learning_rate": 4.81203007518797e-05,
|
22 |
-
"loss":
|
23 |
"step": 20
|
24 |
},
|
25 |
{
|
26 |
"epoch": 0.20270270270270271,
|
27 |
-
"grad_norm":
|
28 |
"learning_rate": 4.43609022556391e-05,
|
29 |
-
"loss":
|
30 |
"step": 30
|
31 |
},
|
32 |
{
|
33 |
"epoch": 0.25675675675675674,
|
34 |
-
"eval_accuracy": 0.
|
35 |
-
"eval_loss": 0.
|
36 |
-
"eval_runtime":
|
37 |
-
"eval_samples_per_second": 0.
|
38 |
-
"eval_steps_per_second": 0.
|
39 |
"step": 38
|
40 |
},
|
41 |
{
|
42 |
"epoch": 1.0135135135135136,
|
43 |
-
"grad_norm": 5.
|
44 |
"learning_rate": 4.0601503759398494e-05,
|
45 |
-
"loss": 0.
|
46 |
"step": 40
|
47 |
},
|
48 |
{
|
49 |
"epoch": 1.0810810810810811,
|
50 |
-
"grad_norm": 4.
|
51 |
"learning_rate": 3.6842105263157895e-05,
|
52 |
-
"loss": 0.
|
53 |
"step": 50
|
54 |
},
|
55 |
{
|
56 |
"epoch": 1.1486486486486487,
|
57 |
-
"grad_norm":
|
58 |
"learning_rate": 3.3082706766917295e-05,
|
59 |
-
"loss": 0.
|
60 |
"step": 60
|
61 |
},
|
62 |
{
|
63 |
"epoch": 1.2162162162162162,
|
64 |
-
"grad_norm":
|
65 |
"learning_rate": 2.9323308270676693e-05,
|
66 |
-
"loss": 0.
|
67 |
"step": 70
|
68 |
},
|
69 |
{
|
70 |
"epoch": 1.2567567567567568,
|
71 |
"eval_accuracy": 0.6181818181818182,
|
72 |
-
"eval_loss": 0.
|
73 |
-
"eval_runtime":
|
74 |
-
"eval_samples_per_second": 0.
|
75 |
-
"eval_steps_per_second": 0.
|
76 |
"step": 76
|
77 |
},
|
78 |
{
|
79 |
"epoch": 2.027027027027027,
|
80 |
-
"grad_norm":
|
81 |
"learning_rate": 2.556390977443609e-05,
|
82 |
-
"loss": 0.
|
83 |
"step": 80
|
84 |
},
|
85 |
{
|
86 |
"epoch": 2.0945945945945947,
|
87 |
-
"grad_norm":
|
88 |
"learning_rate": 2.1804511278195487e-05,
|
89 |
-
"loss": 0.
|
90 |
"step": 90
|
91 |
},
|
92 |
{
|
93 |
"epoch": 2.1621621621621623,
|
94 |
-
"grad_norm":
|
95 |
"learning_rate": 1.8045112781954888e-05,
|
96 |
-
"loss": 0.
|
97 |
"step": 100
|
98 |
},
|
99 |
{
|
100 |
"epoch": 2.22972972972973,
|
101 |
-
"grad_norm":
|
102 |
"learning_rate": 1.4285714285714285e-05,
|
103 |
-
"loss": 0.
|
104 |
"step": 110
|
105 |
},
|
106 |
{
|
107 |
"epoch": 2.2567567567567566,
|
108 |
-
"eval_accuracy": 0.
|
109 |
-
"eval_loss": 0.
|
110 |
-
"eval_runtime":
|
111 |
-
"eval_samples_per_second": 0.
|
112 |
-
"eval_steps_per_second": 0.
|
113 |
"step": 114
|
114 |
},
|
115 |
{
|
116 |
"epoch": 3.0405405405405403,
|
117 |
-
"grad_norm":
|
118 |
"learning_rate": 1.0526315789473684e-05,
|
119 |
-
"loss": 0.
|
120 |
"step": 120
|
121 |
},
|
122 |
{
|
123 |
"epoch": 3.108108108108108,
|
124 |
-
"grad_norm":
|
125 |
"learning_rate": 6.766917293233083e-06,
|
126 |
-
"loss": 0.
|
127 |
"step": 130
|
128 |
},
|
129 |
{
|
130 |
"epoch": 3.175675675675676,
|
131 |
-
"grad_norm":
|
132 |
"learning_rate": 3.007518796992481e-06,
|
133 |
-
"loss": 0.
|
134 |
"step": 140
|
135 |
},
|
136 |
{
|
137 |
"epoch": 3.22972972972973,
|
138 |
"eval_accuracy": 0.6181818181818182,
|
139 |
-
"eval_loss": 0.
|
140 |
-
"eval_runtime":
|
141 |
-
"eval_samples_per_second": 0.
|
142 |
-
"eval_steps_per_second": 0.
|
143 |
"step": 148
|
144 |
},
|
145 |
{
|
146 |
"epoch": 3.22972972972973,
|
147 |
"step": 148,
|
148 |
-
"total_flos": 7.
|
149 |
-
"train_loss":
|
150 |
-
"train_runtime":
|
151 |
-
"train_samples_per_second": 0.
|
152 |
-
"train_steps_per_second": 0.
|
153 |
},
|
154 |
{
|
155 |
"epoch": 3.22972972972973,
|
156 |
-
"eval_accuracy": 0.
|
157 |
-
"eval_loss": 0.
|
158 |
-
"eval_runtime":
|
159 |
-
"eval_samples_per_second": 0.
|
160 |
-
"eval_steps_per_second": 0.
|
161 |
"step": 148
|
162 |
},
|
163 |
{
|
164 |
"epoch": 3.22972972972973,
|
165 |
-
"eval_accuracy": 0.
|
166 |
-
"eval_loss": 0.
|
167 |
-
"eval_runtime":
|
168 |
-
"eval_samples_per_second": 0.
|
169 |
-
"eval_steps_per_second": 0.
|
170 |
"step": 148
|
171 |
}
|
172 |
],
|
@@ -175,7 +175,7 @@
|
|
175 |
"num_input_tokens_seen": 0,
|
176 |
"num_train_epochs": 9223372036854775807,
|
177 |
"save_steps": 500,
|
178 |
-
"total_flos": 7.
|
179 |
"train_batch_size": 4,
|
180 |
"trial_name": null,
|
181 |
"trial_params": null
|
|
|
1 |
{
|
2 |
+
"best_metric": 0.6727272727272727,
|
3 |
+
"best_model_checkpoint": "videomae-base-finetuned-2\\checkpoint-114",
|
4 |
"epoch": 3.22972972972973,
|
5 |
"eval_steps": 500,
|
6 |
"global_step": 148,
|
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.06756756756756757,
|
13 |
+
"grad_norm": 10.621135711669922,
|
14 |
"learning_rate": 3.3333333333333335e-05,
|
15 |
+
"loss": 0.7046,
|
16 |
"step": 10
|
17 |
},
|
18 |
{
|
19 |
"epoch": 0.13513513513513514,
|
20 |
+
"grad_norm": 15.56004524230957,
|
21 |
"learning_rate": 4.81203007518797e-05,
|
22 |
+
"loss": 0.6959,
|
23 |
"step": 20
|
24 |
},
|
25 |
{
|
26 |
"epoch": 0.20270270270270271,
|
27 |
+
"grad_norm": 12.019503593444824,
|
28 |
"learning_rate": 4.43609022556391e-05,
|
29 |
+
"loss": 0.6924,
|
30 |
"step": 30
|
31 |
},
|
32 |
{
|
33 |
"epoch": 0.25675675675675674,
|
34 |
+
"eval_accuracy": 0.5454545454545454,
|
35 |
+
"eval_loss": 0.6671423316001892,
|
36 |
+
"eval_runtime": 167.6467,
|
37 |
+
"eval_samples_per_second": 0.328,
|
38 |
+
"eval_steps_per_second": 0.084,
|
39 |
"step": 38
|
40 |
},
|
41 |
{
|
42 |
"epoch": 1.0135135135135136,
|
43 |
+
"grad_norm": 5.8651123046875,
|
44 |
"learning_rate": 4.0601503759398494e-05,
|
45 |
+
"loss": 0.7766,
|
46 |
"step": 40
|
47 |
},
|
48 |
{
|
49 |
"epoch": 1.0810810810810811,
|
50 |
+
"grad_norm": 4.48609733581543,
|
51 |
"learning_rate": 3.6842105263157895e-05,
|
52 |
+
"loss": 0.7431,
|
53 |
"step": 50
|
54 |
},
|
55 |
{
|
56 |
"epoch": 1.1486486486486487,
|
57 |
+
"grad_norm": 17.070894241333008,
|
58 |
"learning_rate": 3.3082706766917295e-05,
|
59 |
+
"loss": 0.7388,
|
60 |
"step": 60
|
61 |
},
|
62 |
{
|
63 |
"epoch": 1.2162162162162162,
|
64 |
+
"grad_norm": 5.109278678894043,
|
65 |
"learning_rate": 2.9323308270676693e-05,
|
66 |
+
"loss": 0.6664,
|
67 |
"step": 70
|
68 |
},
|
69 |
{
|
70 |
"epoch": 1.2567567567567568,
|
71 |
"eval_accuracy": 0.6181818181818182,
|
72 |
+
"eval_loss": 0.6748194694519043,
|
73 |
+
"eval_runtime": 170.038,
|
74 |
+
"eval_samples_per_second": 0.323,
|
75 |
+
"eval_steps_per_second": 0.082,
|
76 |
"step": 76
|
77 |
},
|
78 |
{
|
79 |
"epoch": 2.027027027027027,
|
80 |
+
"grad_norm": 8.704899787902832,
|
81 |
"learning_rate": 2.556390977443609e-05,
|
82 |
+
"loss": 0.7082,
|
83 |
"step": 80
|
84 |
},
|
85 |
{
|
86 |
"epoch": 2.0945945945945947,
|
87 |
+
"grad_norm": 7.929505825042725,
|
88 |
"learning_rate": 2.1804511278195487e-05,
|
89 |
+
"loss": 0.6429,
|
90 |
"step": 90
|
91 |
},
|
92 |
{
|
93 |
"epoch": 2.1621621621621623,
|
94 |
+
"grad_norm": 17.118267059326172,
|
95 |
"learning_rate": 1.8045112781954888e-05,
|
96 |
+
"loss": 0.8039,
|
97 |
"step": 100
|
98 |
},
|
99 |
{
|
100 |
"epoch": 2.22972972972973,
|
101 |
+
"grad_norm": 7.988255500793457,
|
102 |
"learning_rate": 1.4285714285714285e-05,
|
103 |
+
"loss": 0.5947,
|
104 |
"step": 110
|
105 |
},
|
106 |
{
|
107 |
"epoch": 2.2567567567567566,
|
108 |
+
"eval_accuracy": 0.6727272727272727,
|
109 |
+
"eval_loss": 0.6094833612442017,
|
110 |
+
"eval_runtime": 168.4773,
|
111 |
+
"eval_samples_per_second": 0.326,
|
112 |
+
"eval_steps_per_second": 0.083,
|
113 |
"step": 114
|
114 |
},
|
115 |
{
|
116 |
"epoch": 3.0405405405405403,
|
117 |
+
"grad_norm": 7.859794616699219,
|
118 |
"learning_rate": 1.0526315789473684e-05,
|
119 |
+
"loss": 0.5892,
|
120 |
"step": 120
|
121 |
},
|
122 |
{
|
123 |
"epoch": 3.108108108108108,
|
124 |
+
"grad_norm": 12.87205982208252,
|
125 |
"learning_rate": 6.766917293233083e-06,
|
126 |
+
"loss": 0.5001,
|
127 |
"step": 130
|
128 |
},
|
129 |
{
|
130 |
"epoch": 3.175675675675676,
|
131 |
+
"grad_norm": 13.842416763305664,
|
132 |
"learning_rate": 3.007518796992481e-06,
|
133 |
+
"loss": 0.5049,
|
134 |
"step": 140
|
135 |
},
|
136 |
{
|
137 |
"epoch": 3.22972972972973,
|
138 |
"eval_accuracy": 0.6181818181818182,
|
139 |
+
"eval_loss": 0.6914771199226379,
|
140 |
+
"eval_runtime": 166.1396,
|
141 |
+
"eval_samples_per_second": 0.331,
|
142 |
+
"eval_steps_per_second": 0.084,
|
143 |
"step": 148
|
144 |
},
|
145 |
{
|
146 |
"epoch": 3.22972972972973,
|
147 |
"step": 148,
|
148 |
+
"total_flos": 7.264556556620268e+17,
|
149 |
+
"train_loss": 0.6650454998016357,
|
150 |
+
"train_runtime": 4992.096,
|
151 |
+
"train_samples_per_second": 0.119,
|
152 |
+
"train_steps_per_second": 0.03
|
153 |
},
|
154 |
{
|
155 |
"epoch": 3.22972972972973,
|
156 |
+
"eval_accuracy": 0.6956521739130435,
|
157 |
+
"eval_loss": 0.6255781054496765,
|
158 |
+
"eval_runtime": 189.5334,
|
159 |
+
"eval_samples_per_second": 0.364,
|
160 |
+
"eval_steps_per_second": 0.095,
|
161 |
"step": 148
|
162 |
},
|
163 |
{
|
164 |
"epoch": 3.22972972972973,
|
165 |
+
"eval_accuracy": 0.6956521739130435,
|
166 |
+
"eval_loss": 0.6255780458450317,
|
167 |
+
"eval_runtime": 237.0624,
|
168 |
+
"eval_samples_per_second": 0.291,
|
169 |
+
"eval_steps_per_second": 0.076,
|
170 |
"step": 148
|
171 |
}
|
172 |
],
|
|
|
175 |
"num_input_tokens_seen": 0,
|
176 |
"num_train_epochs": 9223372036854775807,
|
177 |
"save_steps": 500,
|
178 |
+
"total_flos": 7.264556556620268e+17,
|
179 |
"train_batch_size": 4,
|
180 |
"trial_name": null,
|
181 |
"trial_params": null
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4984
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:498bb098c9e8befc320966c8d6858e42030b47f82ab4d7282188f2c749102de0
|
3 |
size 4984
|