AAA01101312 commited on
Commit
98c6233
1 Parent(s): df5bc1a

Training in progress, step 1500

Browse files
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:419642769ae67cef6e01708b125840f3864c36e170fd8e75820472f755b4d00e
3
  size 268290900
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c0b54af839f197fd556e2802b995e04c0e84b6a7dd75ee7bd1cd439b5e136b8
3
  size 268290900
run-2/checkpoint-1500/model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:93982c596a316534ec6049dc4326fd8bb8c68519f3bd73a807df5fbac8ae6656
3
  size 268290900
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c0b54af839f197fd556e2802b995e04c0e84b6a7dd75ee7bd1cd439b5e136b8
3
  size 268290900
run-2/checkpoint-1500/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1fde47893799b8f3f4f56489a4110ea1da0e51e3702f9cc3187ffeb0bf02f547
3
  size 536643898
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:735b33efb421584f65e3eeeaef0cc981323128818b267ced3ff6374b95b6fcc7
3
  size 536643898
run-2/checkpoint-1500/trainer_state.json CHANGED
@@ -10,59 +10,59 @@
10
  "log_history": [
11
  {
12
  "epoch": 1.0,
13
- "eval_accuracy": 0.5770967741935484,
14
- "eval_loss": 0.21370309591293335,
15
- "eval_runtime": 14.3261,
16
- "eval_samples_per_second": 216.388,
17
- "eval_steps_per_second": 27.083,
18
  "step": 318
19
  },
20
  {
21
  "epoch": 1.5723270440251573,
22
- "grad_norm": 0.547938346862793,
23
  "learning_rate": 1.371069182389937e-05,
24
- "loss": 0.3338,
25
  "step": 500
26
  },
27
  {
28
  "epoch": 2.0,
29
- "eval_accuracy": 0.8070967741935484,
30
- "eval_loss": 0.10994400829076767,
31
- "eval_runtime": 15.3734,
32
- "eval_samples_per_second": 201.647,
33
- "eval_steps_per_second": 25.238,
34
  "step": 636
35
  },
36
  {
37
  "epoch": 3.0,
38
  "eval_accuracy": 0.8664516129032258,
39
- "eval_loss": 0.07803630828857422,
40
- "eval_runtime": 14.0224,
41
- "eval_samples_per_second": 221.075,
42
- "eval_steps_per_second": 27.67,
43
  "step": 954
44
  },
45
  {
46
  "epoch": 3.1446540880503147,
47
- "grad_norm": 0.510848343372345,
48
  "learning_rate": 7.421383647798742e-06,
49
- "loss": 0.1268,
50
  "step": 1000
51
  },
52
  {
53
  "epoch": 4.0,
54
- "eval_accuracy": 0.8819354838709678,
55
- "eval_loss": 0.06529980897903442,
56
- "eval_runtime": 15.0804,
57
- "eval_samples_per_second": 205.564,
58
- "eval_steps_per_second": 25.729,
59
  "step": 1272
60
  },
61
  {
62
  "epoch": 4.716981132075472,
63
- "grad_norm": 0.4325341582298279,
64
  "learning_rate": 1.1320754716981133e-06,
65
- "loss": 0.0905,
66
  "step": 1500
67
  }
68
  ],
@@ -75,8 +75,8 @@
75
  "train_batch_size": 48,
76
  "trial_name": null,
77
  "trial_params": {
78
- "alpha": 0.18349258351603082,
79
  "num_train_epochs": 5,
80
- "temperature": 9
81
  }
82
  }
 
10
  "log_history": [
11
  {
12
  "epoch": 1.0,
13
+ "eval_accuracy": 0.5787096774193549,
14
+ "eval_loss": 0.21759462356567383,
15
+ "eval_runtime": 13.1676,
16
+ "eval_samples_per_second": 235.426,
17
+ "eval_steps_per_second": 29.466,
18
  "step": 318
19
  },
20
  {
21
  "epoch": 1.5723270440251573,
22
+ "grad_norm": 0.5513952970504761,
23
  "learning_rate": 1.371069182389937e-05,
24
+ "loss": 0.3396,
25
  "step": 500
26
  },
27
  {
28
  "epoch": 2.0,
29
+ "eval_accuracy": 0.8083870967741935,
30
+ "eval_loss": 0.11128884553909302,
31
+ "eval_runtime": 13.9571,
32
+ "eval_samples_per_second": 222.109,
33
+ "eval_steps_per_second": 27.799,
34
  "step": 636
35
  },
36
  {
37
  "epoch": 3.0,
38
  "eval_accuracy": 0.8664516129032258,
39
+ "eval_loss": 0.07867251336574554,
40
+ "eval_runtime": 13.1876,
41
+ "eval_samples_per_second": 235.07,
42
+ "eval_steps_per_second": 29.422,
43
  "step": 954
44
  },
45
  {
46
  "epoch": 3.1446540880503147,
47
+ "grad_norm": 0.5208547711372375,
48
  "learning_rate": 7.421383647798742e-06,
49
+ "loss": 0.1284,
50
  "step": 1000
51
  },
52
  {
53
  "epoch": 4.0,
54
+ "eval_accuracy": 0.8825806451612903,
55
+ "eval_loss": 0.0656495988368988,
56
+ "eval_runtime": 15.4062,
57
+ "eval_samples_per_second": 201.217,
58
+ "eval_steps_per_second": 25.185,
59
  "step": 1272
60
  },
61
  {
62
  "epoch": 4.716981132075472,
63
+ "grad_norm": 0.4366990029811859,
64
  "learning_rate": 1.1320754716981133e-06,
65
+ "loss": 0.0913,
66
  "step": 1500
67
  }
68
  ],
 
75
  "train_batch_size": 48,
76
  "trial_name": null,
77
  "trial_params": {
78
+ "alpha": 0.8001341619530338,
79
  "num_train_epochs": 5,
80
+ "temperature": 8
81
  }
82
  }
run-2/checkpoint-1500/training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8733c9bfaf4d70d36ff31a7dbf705d06c4d9307caf55ee0ae732d8ed448c8a36
3
  size 5048
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3a86d837a7a7b0e3b25e6ad0ff44a7299c5ea5790990e72e2773a24b09c1a55
3
  size 5048
runs/May20_08-46-15_MainPC/events.out.tfevents.1716163630.MainPC.215644.3 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fbfa38857a5894d2314d6690c5cefef9de603e47c018c03faa9ad7f945d77ade
3
- size 13919
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f2a949d1fa504d2e87b204bcda1ea0231e0c2645253873a93b4084c51f8b0f1
3
+ size 14807