Marcos12886 commited on
Commit
0f6f358
1 Parent(s): 0433291

Parámetros de los últimos entrenamientos

Browse files
Files changed (1) hide show
  1. models_config.json +9 -17
models_config.json CHANGED
@@ -4,14 +4,11 @@
4
  "output_dir": "distilhubert-finetuned-cry-detector",
5
  "training_args": {
6
  "num_train_epochs": 10,
7
- "learning_rate": 0.00003,
8
  "warmup_ratio": 0.1,
9
  "output_dir": "distilhubert-finetuned-cry-detector",
10
- "eval_strategy": "steps",
11
- "eval_steps": 100,
12
- "save_strategy": "steps",
13
- "save_steps": 100,
14
- "save_total_limit": 5,
15
  "lr_scheduler_type": "cosine_with_restarts",
16
  "auto_find_batch_size": true,
17
  "per_device_train_batch_size": 64,
@@ -30,7 +27,6 @@
30
  "max_grad_norm": 0.5,
31
  "weight_decay": 0.02,
32
  "fp16": true,
33
- "logging_steps": 50,
34
  "eval_accumulation_steps": 1,
35
  "dataloader_num_workers": 12,
36
  "dataloader_pin_memory": true,
@@ -42,19 +38,16 @@
42
  "dataset_path": "data/mixed_data",
43
  "output_dir": "distilhubert-finetuned-mixed-data",
44
  "training_args": {
45
- "num_train_epochs": 30,
46
- "learning_rate": 3e-4,
47
  "warmup_ratio": 0.1,
48
  "output_dir": "distilhubert-finetuned-mixed-data",
49
- "eval_strategy": "steps",
50
- "eval_steps": 100,
51
- "save_strategy": "steps",
52
- "save_steps": 100,
53
- "save_total_limit": 5,
54
  "lr_scheduler_type": "cosine_with_restarts",
55
  "auto_find_batch_size": true,
56
- "per_device_train_batch_size": 64,
57
- "per_device_eval_batch_size": 64,
58
  "gradient_accumulation_steps": 2,
59
  "gradient_checkpointing": true,
60
  "load_best_model_at_end": true,
@@ -66,7 +59,6 @@
66
  "max_grad_norm": 0.5,
67
  "weight_decay": 0.02,
68
  "fp16": true,
69
- "logging_steps": 50,
70
  "eval_accumulation_steps": 1,
71
  "dataloader_num_workers": 12,
72
  "dataloader_pin_memory": true,
 
4
  "output_dir": "distilhubert-finetuned-cry-detector",
5
  "training_args": {
6
  "num_train_epochs": 10,
7
+ "learning_rate": 3e-5,
8
  "warmup_ratio": 0.1,
9
  "output_dir": "distilhubert-finetuned-cry-detector",
10
+ "eval_strategy": "epoch",
11
+ "save_strategy": "epoch",
 
 
 
12
  "lr_scheduler_type": "cosine_with_restarts",
13
  "auto_find_batch_size": true,
14
  "per_device_train_batch_size": 64,
 
27
  "max_grad_norm": 0.5,
28
  "weight_decay": 0.02,
29
  "fp16": true,
 
30
  "eval_accumulation_steps": 1,
31
  "dataloader_num_workers": 12,
32
  "dataloader_pin_memory": true,
 
38
  "dataset_path": "data/mixed_data",
39
  "output_dir": "distilhubert-finetuned-mixed-data",
40
  "training_args": {
41
+ "num_train_epochs": 45,
42
+ "learning_rate": 8e-4,
43
  "warmup_ratio": 0.1,
44
  "output_dir": "distilhubert-finetuned-mixed-data",
45
+ "eval_strategy": "epoch",
46
+ "save_strategy": "epoch",
 
 
 
47
  "lr_scheduler_type": "cosine_with_restarts",
48
  "auto_find_batch_size": true,
49
+ "per_device_train_batch_size": 128,
50
+ "per_device_eval_batch_size": 128,
51
  "gradient_accumulation_steps": 2,
52
  "gradient_checkpointing": true,
53
  "load_best_model_at_end": true,
 
59
  "max_grad_norm": 0.5,
60
  "weight_decay": 0.02,
61
  "fp16": true,
 
62
  "eval_accumulation_steps": 1,
63
  "dataloader_num_workers": 12,
64
  "dataloader_pin_memory": true,