Hugging Face's logo

samvaran
/
inlp-assignment3

Model card Files
xet
 Community
samvaran commited on
Commit
834be77
·
verified ·
1 Parent(s): 7a2c0c5

Upload task2_ssm/tuning/tuning_results.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. task2_ssm/tuning/tuning_results.json +76 -38
task2_ssm/tuning/tuning_results.json CHANGED
@@ -1,7 +1,26 @@
1
  [
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
  {
3
  "trial": 4,
4
- "name": "ssm_ss128_nl4_emb64_do0.1_lr0.001_bs32_ep20",
5
  "params": {
6
  "model.state_size": 128,
7
  "model.num_layers": 4,
@@ -9,70 +28,89 @@
9
  "model.dropout": 0.1,
10
  "training.learning_rate": 0.001,
11
  "training.batch_size": 32,
12
- "training.epochs": 20
13
  },
14
- "best_val_loss": 5.92775,
15
- "best_val_acc": 0.163124,
16
- "best_epoch": 13,
17
- "epochs_run": 18,
18
- "param_count": 2730656,
19
- "model_path": "models/ssm_ss128_nl4_emb64_do0.1_lr0.001_bs32_ep20.pt"
20
  },
21
  {
22
- "trial": 2,
23
- "name": "ssm_ss64_nl4_emb64_do0.1_lr0.001_bs32_ep20",
24
  "params": {
25
- "model.state_size": 64,
26
  "model.num_layers": 4,
27
  "model.input_size": 64,
28
  "model.dropout": 0.1,
29
  "training.learning_rate": 0.001,
30
  "training.batch_size": 32,
31
- "training.epochs": 20
32
  },
33
- "best_val_loss": 5.933967,
34
- "best_val_acc": 0.161306,
35
- "best_epoch": 11,
36
- "epochs_run": 16,
37
- "param_count": 2697632,
38
- "model_path": "models/ssm_ss64_nl4_emb64_do0.1_lr0.001_bs32_ep20.pt"
39
  },
40
  {
41
- "trial": 1,
42
- "name": "ssm_ss64_nl2_emb64_do0.1_lr0.001_bs32_ep20",
43
  "params": {
44
- "model.state_size": 64,
45
  "model.num_layers": 2,
46
  "model.input_size": 64,
47
  "model.dropout": 0.1,
48
  "training.learning_rate": 0.001,
49
  "training.batch_size": 32,
50
- "training.epochs": 20
51
  },
52
- "best_val_loss": 6.030748,
53
- "best_val_acc": 0.159541,
54
- "best_epoch": 10,
55
- "epochs_run": 16,
56
- "param_count": 2638880,
57
- "model_path": "models/ssm_ss64_nl2_emb64_do0.1_lr0.001_bs32_ep20.pt"
58
  },
59
  {
60
- "trial": 3,
61
- "name": "ssm_ss128_nl2_emb64_do0.1_lr0.001_bs32_ep20",
62
  "params": {
63
- "model.state_size": 128,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
64
  "model.num_layers": 2,
65
  "model.input_size": 64,
66
  "model.dropout": 0.1,
67
  "training.learning_rate": 0.001,
68
  "training.batch_size": 32,
69
- "training.epochs": 20
70
  },
71
- "best_val_loss": 6.009648,
72
- "best_val_acc": 0.159087,
73
- "best_epoch": 9,
74
- "epochs_run": 14,
75
- "param_count": 2655392,
76
- "model_path": "models/ssm_ss128_nl2_emb64_do0.1_lr0.001_bs32_ep20.pt"
77
  }
78
  ]
 
1
  [
2
+ {
3
+ "trial": 2,
4
+ "name": "ssm_ss64_nl4_emb64_do0.1_lr0.001_bs32_ep30",
5
+ "params": {
6
+ "model.state_size": 64,
7
+ "model.num_layers": 4,
8
+ "model.input_size": 64,
9
+ "model.dropout": 0.1,
10
+ "training.learning_rate": 0.001,
11
+ "training.batch_size": 32,
12
+ "training.epochs": 30
13
+ },
14
+ "best_val_loss": 5.375143,
15
+ "best_val_acc": 0.188612,
16
+ "best_epoch": 27,
17
+ "epochs_run": 30,
18
+ "param_count": 1407632,
19
+ "model_path": "models/ssm_ss64_nl4_emb64_do0.1_lr0.001_bs32_ep30.pt"
20
+ },
21
  {
22
  "trial": 4,
23
+ "name": "ssm_ss128_nl4_emb64_do0.1_lr0.001_bs32_ep30",
24
  "params": {
25
  "model.state_size": 128,
26
  "model.num_layers": 4,
 
28
  "model.dropout": 0.1,
29
  "training.learning_rate": 0.001,
30
  "training.batch_size": 32,
31
+ "training.epochs": 30
32
  },
33
+ "best_val_loss": 5.369389,
34
+ "best_val_acc": 0.188401,
35
+ "best_epoch": 29,
36
+ "epochs_run": 30,
37
+ "param_count": 1440656,
38
+ "model_path": "models/ssm_ss128_nl4_emb64_do0.1_lr0.001_bs32_ep30.pt"
39
  },
40
  {
41
+ "trial": 6,
42
+ "name": "ssm_ss256_nl4_emb64_do0.1_lr0.001_bs32_ep30",
43
  "params": {
44
+ "model.state_size": 256,
45
  "model.num_layers": 4,
46
  "model.input_size": 64,
47
  "model.dropout": 0.1,
48
  "training.learning_rate": 0.001,
49
  "training.batch_size": 32,
50
+ "training.epochs": 30
51
  },
52
+ "best_val_loss": 5.367797,
53
+ "best_val_acc": 0.188095,
54
+ "best_epoch": 29,
55
+ "epochs_run": 30,
56
+ "param_count": 1506704,
57
+ "model_path": "models/ssm_ss256_nl4_emb64_do0.1_lr0.001_bs32_ep30.pt"
58
  },
59
  {
60
+ "trial": 3,
61
+ "name": "ssm_ss128_nl2_emb64_do0.1_lr0.001_bs32_ep30",
62
  "params": {
63
+ "model.state_size": 128,
64
  "model.num_layers": 2,
65
  "model.input_size": 64,
66
  "model.dropout": 0.1,
67
  "training.learning_rate": 0.001,
68
  "training.batch_size": 32,
69
+ "training.epochs": 30
70
  },
71
+ "best_val_loss": 5.470606,
72
+ "best_val_acc": 0.184723,
73
+ "best_epoch": 23,
74
+ "epochs_run": 30,
75
+ "param_count": 1365392,
76
+ "model_path": "models/ssm_ss128_nl2_emb64_do0.1_lr0.001_bs32_ep30.pt"
77
  },
78
  {
79
+ "trial": 5,
80
+ "name": "ssm_ss256_nl2_emb64_do0.1_lr0.001_bs32_ep30",
81
  "params": {
82
+ "model.state_size": 256,
83
+ "model.num_layers": 2,
84
+ "model.input_size": 64,
85
+ "model.dropout": 0.1,
86
+ "training.learning_rate": 0.001,
87
+ "training.batch_size": 32,
88
+ "training.epochs": 30
89
+ },
90
+ "best_val_loss": 5.459764,
91
+ "best_val_acc": 0.184554,
92
+ "best_epoch": 18,
93
+ "epochs_run": 28,
94
+ "param_count": 1398416,
95
+ "model_path": "models/ssm_ss256_nl2_emb64_do0.1_lr0.001_bs32_ep30.pt"
96
+ },
97
+ {
98
+ "trial": 1,
99
+ "name": "ssm_ss64_nl2_emb64_do0.1_lr0.001_bs32_ep30",
100
+ "params": {
101
+ "model.state_size": 64,
102
  "model.num_layers": 2,
103
  "model.input_size": 64,
104
  "model.dropout": 0.1,
105
  "training.learning_rate": 0.001,
106
  "training.batch_size": 32,
107
+ "training.epochs": 30
108
  },
109
+ "best_val_loss": 5.472807,
110
+ "best_val_acc": 0.1843,
111
+ "best_epoch": 21,
112
+ "epochs_run": 30,
113
+ "param_count": 1348880,
114
+ "model_path": "models/ssm_ss64_nl2_emb64_do0.1_lr0.001_bs32_ep30.pt"
115
  }
116
  ]