BramVanroy commited on
Commit
d7384a4
1 Parent(s): b105dbe
README.md CHANGED
@@ -1,45 +1,100 @@
1
  ---
 
 
2
  language:
3
  - nl
4
  license: mit
5
- tags:
6
- - sentiment-analysis
7
- - dutch
8
- - text
9
- datasets:
10
- - BramVanroy/hebban-reviews
11
  metrics:
12
  - accuracy
13
  - f1
14
  - precision
 
15
  - recall
16
- widget:
17
- - text: "Wauw, wat een leuk boek! Ik heb me er er goed mee vermaakt."
18
- - text: "Nee, deze vond ik niet goed. De auteur doet zijn best om je als lezer mee te trekken in het verhaal maar mij overtuigt het alleszins niet."
19
- - text: "Ik vind het niet slecht maar de schrijfstijl trekt me ook niet echt aan. Het wordt een beetje saai vanaf het vijfde hoofdstuk"
20
-
21
  model-index:
22
  - name: bert-base-multilingual-cased-hebban-reviews
23
  results:
24
- - task:
25
- type: text-classification
26
- name: sentiment analysis
27
- dataset:
28
- type: BramVanroy/hebban-reviews
29
- name: Hebban Reviews
30
  split: test
31
- revision: 1.1.0
32
  metrics:
33
- - type: accuracy
34
- value: 0.7845
35
- name: Test accuracy
36
- - type: f1
37
- value: 0.7911
38
- name: Test f1
39
- - type: precision
40
- value: 0.8013
41
- name: Test precision
42
- - type: recall
43
- value: 0.7845
44
- name: Test recall
45
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ datasets:
3
+ - BramVanroy/hebban-reviews
4
  language:
5
  - nl
6
  license: mit
 
 
 
 
 
 
7
  metrics:
8
  - accuracy
9
  - f1
10
  - precision
11
+ - qwk
12
  - recall
 
 
 
 
 
13
  model-index:
14
  - name: bert-base-multilingual-cased-hebban-reviews
15
  results:
16
+ - dataset:
17
+ config: filtered_sentiment
18
+ name: BramVanroy/hebban-reviews - filtered_sentiment - 2.0.0
19
+ revision: 2.0.0
 
 
20
  split: test
21
+ type: BramVanroy/hebban-reviews
22
  metrics:
23
+ - name: Test accuracy
24
+ type: accuracy
25
+ value: 0.7764792899408284
26
+ - name: Test f1
27
+ type: f1
28
+ value: 0.7821329848271866
29
+ - name: Test precision
30
+ type: precision
31
+ value: 0.7907660190770787
32
+ - name: Test qwk
33
+ type: qwk
34
+ value: 0.6813121109021326
35
+ - name: Test recall
36
+ type: recall
37
+ value: 0.7764792899408284
38
+ task:
39
+ name: sentiment analysis
40
+ type: text-classification
41
+ tags:
42
+ - sentiment-analysis
43
+ - dutch
44
+ - text
45
+ widget:
46
+ - text: Wauw, wat een leuk boek! Ik heb me er er goed mee vermaakt.
47
+ - text: Nee, deze vond ik niet goed. De auteur doet zijn best om je als lezer mee
48
+ te trekken in het verhaal maar mij overtuigt het alleszins niet.
49
+ - text: Ik vind het niet slecht maar de schrijfstijl trekt me ook niet echt aan. Het
50
+ wordt een beetje saai vanaf het vijfde hoofdstuk
51
+ ---
52
+
53
+ # bert-base-multilingual-cased-hebban-reviews
54
+
55
+ # Dataset
56
+ - dataset_name: BramVanroy/hebban-reviews
57
+ - dataset_config: filtered_sentiment
58
+ - dataset_revision: 2.0.0
59
+ - labelcolumn: review_sentiment
60
+ - textcolumn: review_text_without_quotes
61
+
62
+ # Training
63
+ - optim: adamw_hf
64
+ - learning_rate: 5e-05
65
+ - per_device_train_batch_size: 64
66
+ - per_device_eval_batch_size: 64
67
+ - gradient_accumulation_steps: 1
68
+ - max_steps: 5001
69
+ - save_steps: 500
70
+ - metric_for_best_model: qwk
71
+
72
+ # Best checkedpoint based on validation
73
+ - best_metric: 0.6828581526810108
74
+ - best_model_checkpoint: trained/hebban-reviews/bert-base-multilingual-cased/checkpoint-1500
75
+
76
+ # Test results of best checkpoint
77
+ - accuracy: 0.7764792899408284
78
+ - f1: 0.7821329848271866
79
+ - precision: 0.7907660190770787
80
+ - qwk: 0.6813121109021326
81
+ - recall: 0.7764792899408284
82
+
83
+ ## Confusion matric
84
+
85
+ ![cfm](fig/test_confusion_matrix.png)
86
+
87
+ ## Normalized confusion matrix
88
+
89
+ ![norm cfm](fig/test_confusion_matrix_norm.png)
90
+
91
+ # Environment
92
+ - cuda_capabilities: 8.0; 8.0
93
+ - cuda_device_count: 2
94
+ - cuda_devices: NVIDIA A100-SXM4-80GB; NVIDIA A100-SXM4-80GB
95
+ - finetuner_commit: 66294c815326c93682003119534cb72009f558c2
96
+ - platform: Linux-4.18.0-305.49.1.el8_4.x86_64-x86_64-with-glibc2.28
97
+ - python_version: 3.9.5
98
+ - toch_version: 1.10.0
99
+ - transformers_version: 4.21.0
100
+
all_results.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "accuracy": 0.7764792899408284,
3
+ "epoch": 4.38,
4
+ "f1": 0.7821329848271866,
5
+ "precision": 0.7907660190770787,
6
+ "qwk": 0.6813121109021326,
7
+ "recall": 0.7764792899408284,
8
+ "train_loss": 0.5313903366928219,
9
+ "train_runtime": 2667.6757,
10
+ "train_samples_per_second": 239.957,
11
+ "train_steps_per_second": 1.875
12
+ }
args.json ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_n_gpu": 1,
3
+ "adafactor": false,
4
+ "adam_beta1": 0.9,
5
+ "adam_beta2": 0.999,
6
+ "adam_epsilon": 1e-08,
7
+ "auto_find_batch_size": false,
8
+ "bf16": false,
9
+ "bf16_full_eval": false,
10
+ "calculate_qwk": true,
11
+ "data_seed": 42,
12
+ "dataloader_drop_last": false,
13
+ "dataloader_num_workers": 0,
14
+ "dataloader_pin_memory": true,
15
+ "dataset_config": "filtered_sentiment",
16
+ "dataset_name": "BramVanroy/hebban-reviews",
17
+ "dataset_revision": "2.0.0",
18
+ "ddp_bucket_cap_mb": null,
19
+ "ddp_find_unused_parameters": null,
20
+ "debug": [],
21
+ "deepspeed": null,
22
+ "disable_tqdm": false,
23
+ "do_early_stopping": false,
24
+ "do_eval": true,
25
+ "do_optimize": false,
26
+ "do_predict": true,
27
+ "do_train": true,
28
+ "early_stopping_patience": 1,
29
+ "early_stopping_threshold": 0.0,
30
+ "eval_accumulation_steps": null,
31
+ "eval_delay": 0,
32
+ "eval_steps": 500,
33
+ "evaluation_strategy": "steps",
34
+ "fp16": true,
35
+ "fp16_backend": "auto",
36
+ "fp16_full_eval": false,
37
+ "fp16_opt_level": "O1",
38
+ "fsdp": [],
39
+ "fsdp_min_num_params": 0,
40
+ "fsdp_transformer_layer_cls_to_wrap": null,
41
+ "full_determinism": false,
42
+ "gradient_accumulation_steps": 1,
43
+ "gradient_checkpointing": false,
44
+ "greater_is_better": true,
45
+ "group_by_length": false,
46
+ "half_precision_backend": "cuda_amp",
47
+ "hub_model_id": null,
48
+ "hub_private_repo": false,
49
+ "hub_strategy": "every_save",
50
+ "hub_token": null,
51
+ "ignore_data_skip": false,
52
+ "include_inputs_for_metrics": false,
53
+ "jit_mode_eval": false,
54
+ "label_names": null,
55
+ "label_smoothing_factor": 0.0,
56
+ "labelcolumn": "review_sentiment",
57
+ "labelnames": [
58
+ "negative",
59
+ "neutral",
60
+ "positive"
61
+ ],
62
+ "learning_rate": 5e-05,
63
+ "length_column_name": "length",
64
+ "load_best_model_at_end": true,
65
+ "local_rank": 0,
66
+ "log_level": -1,
67
+ "log_level_replica": -1,
68
+ "log_on_each_node": true,
69
+ "logging_dir": "trained/hebban-reviews/bert-base-multilingual-cased/runs/Jul28_16-55-25_node3900.accelgor.os",
70
+ "logging_first_step": false,
71
+ "logging_nan_inf_filter": true,
72
+ "logging_steps": 500,
73
+ "logging_strategy": "steps",
74
+ "lr_scheduler_type": "linear",
75
+ "max_grad_norm": 1.0,
76
+ "max_seq_length": null,
77
+ "max_steps": 5001,
78
+ "max_test_samples": null,
79
+ "max_train_samples": null,
80
+ "max_validation_samples": null,
81
+ "metric_for_best_model": "qwk",
82
+ "model_name_or_path": "bert-base-multilingual-cased",
83
+ "model_revision": "main",
84
+ "mp_parameters": "",
85
+ "n_trials": 8,
86
+ "no_cuda": false,
87
+ "num_train_epochs": 3.0,
88
+ "optim": "adamw_hf",
89
+ "output_dir": "trained/hebban-reviews/bert-base-multilingual-cased",
90
+ "overwrite_cache": false,
91
+ "overwrite_output_dir": true,
92
+ "past_index": -1,
93
+ "per_device_eval_batch_size": 64,
94
+ "per_device_train_batch_size": 64,
95
+ "per_gpu_eval_batch_size": null,
96
+ "per_gpu_train_batch_size": null,
97
+ "prediction_loss_only": false,
98
+ "push_to_hub": false,
99
+ "push_to_hub_model_id": null,
100
+ "push_to_hub_organization": null,
101
+ "push_to_hub_token": null,
102
+ "ray_scope": "all",
103
+ "remove_unused_columns": true,
104
+ "report_to": [
105
+ "tensorboard"
106
+ ],
107
+ "resume_from_checkpoint": null,
108
+ "run_name": "trained/hebban-reviews/bert-base-multilingual-cased",
109
+ "save_on_each_node": false,
110
+ "save_steps": 500,
111
+ "save_strategy": "steps",
112
+ "save_total_limit": null,
113
+ "scheduler_type": null,
114
+ "seed": 42,
115
+ "sharded_ddp": [],
116
+ "skip_memory_metrics": true,
117
+ "split_seed": 42,
118
+ "testsplit_name": "test",
119
+ "textcolumn": "review_text_without_quotes",
120
+ "tf32": null,
121
+ "torchdynamo": null,
122
+ "tpu_metrics_debug": false,
123
+ "tpu_num_cores": null,
124
+ "trainsplit_name": "train",
125
+ "use_class_weights": true,
126
+ "use_ipex": false,
127
+ "use_legacy_prediction_loop": false,
128
+ "validation_size": 0.1,
129
+ "validationsplit_name": "validation",
130
+ "warmup_ratio": 0.0,
131
+ "warmup_steps": 0,
132
+ "weight_decay": 0.0,
133
+ "xpu_backend": null
134
+ }
config.json CHANGED
@@ -34,7 +34,7 @@
34
  "pooler_type": "first_token_transform",
35
  "position_embedding_type": "absolute",
36
  "torch_dtype": "float32",
37
- "transformers_version": "4.21.0.dev0",
38
  "type_vocab_size": 2,
39
  "use_cache": true,
40
  "vocab_size": 119547
 
34
  "pooler_type": "first_token_transform",
35
  "position_embedding_type": "absolute",
36
  "torch_dtype": "float32",
37
+ "transformers_version": "4.21.0",
38
  "type_vocab_size": 2,
39
  "use_cache": true,
40
  "vocab_size": 119547
env.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cuda_capabilities": "8.0; 8.0",
3
+ "cuda_device_count": 2,
4
+ "cuda_devices": "NVIDIA A100-SXM4-80GB; NVIDIA A100-SXM4-80GB",
5
+ "finetuner_commit": "66294c815326c93682003119534cb72009f558c2",
6
+ "platform": "Linux-4.18.0-305.49.1.el8_4.x86_64-x86_64-with-glibc2.28",
7
+ "python_version": "3.9.5",
8
+ "toch_version": "1.10.0",
9
+ "transformers_version": "4.21.0"
10
+ }
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8ed77b30f2e050c50b9b4c6041ea2f4e10e9f41413b5565af17f7fb5b86f4fb6
3
  size 711493997
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ff04f2b96bacd77224f0a8a0ad534c4b54e436c98b11016ba592265153a672b
3
  size 711493997
test_confusion_matrix.eps ADDED
The diff for this file is too large to render. See raw diff
 
test_confusion_matrix.png ADDED
test_confusion_matrix_norm.eps ADDED
The diff for this file is too large to render. See raw diff
 
test_confusion_matrix_norm.png ADDED
test_predictions.txt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b0bbeeab622e2b2ede61d8dcc4642168cb49c6cfc5bd32605aba9da0e7b9ad0
3
+ size 65402577
test_results.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "accuracy": 0.7764792899408284,
3
+ "f1": 0.7821329848271866,
4
+ "precision": 0.7907660190770787,
5
+ "qwk": 0.6813121109021326,
6
+ "recall": 0.7764792899408284
7
+ }
train_results.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 4.38,
3
+ "train_loss": 0.5313903366928219,
4
+ "train_runtime": 2667.6757,
5
+ "train_samples_per_second": 239.957,
6
+ "train_steps_per_second": 1.875
7
+ }
trainer_state.json CHANGED
@@ -1,457 +1,215 @@
1
  {
2
- "best_metric": 0.7886478688599108,
3
- "best_model_checkpoint": "/home/bram/shares/predict/trained/dutch/hebban-reviews/bert-base-multilingual-cased/checkpoint-11000",
4
- "epoch": 3.9447731755424065,
5
- "global_step": 12000,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
9
  "log_history": [
10
  {
11
- "epoch": 0.16,
12
- "learning_rate": 4.793333333333334e-05,
13
- "loss": 0.9625,
14
  "step": 500
15
  },
16
  {
17
- "epoch": 0.16,
18
- "eval_accuracy": 0.6706114398422091,
19
- "eval_f1": 0.6931944548479266,
20
- "eval_loss": 0.8083919882774353,
21
- "eval_precision": 0.7343487735413422,
22
- "eval_recall": 0.6706114398422091,
23
- "eval_runtime": 24.6562,
24
- "eval_samples_per_second": 658.009,
25
- "eval_steps_per_second": 27.417,
 
26
  "step": 500
27
  },
28
  {
29
- "epoch": 0.33,
30
- "learning_rate": 4.585e-05,
31
- "loss": 0.8186,
32
  "step": 1000
33
  },
34
  {
35
- "epoch": 0.33,
36
- "eval_accuracy": 0.6870685404339251,
37
- "eval_f1": 0.7088564060365691,
38
- "eval_loss": 0.7797691226005554,
39
- "eval_precision": 0.753903679506183,
40
- "eval_recall": 0.6870685404339251,
41
- "eval_runtime": 24.786,
42
- "eval_samples_per_second": 654.562,
43
- "eval_steps_per_second": 27.273,
 
44
  "step": 1000
45
  },
46
  {
47
- "epoch": 0.49,
48
- "learning_rate": 4.376666666666667e-05,
49
- "loss": 0.7658,
50
  "step": 1500
51
  },
52
  {
53
- "epoch": 0.49,
54
- "eval_accuracy": 0.7040803747534516,
55
- "eval_f1": 0.7250925730375316,
56
- "eval_loss": 0.7284208536148071,
57
- "eval_precision": 0.777831708888876,
58
- "eval_recall": 0.7040803747534516,
59
- "eval_runtime": 24.7754,
60
- "eval_samples_per_second": 654.842,
61
- "eval_steps_per_second": 27.285,
 
62
  "step": 1500
63
  },
64
  {
65
- "epoch": 0.66,
66
- "learning_rate": 4.1683333333333335e-05,
67
- "loss": 0.7318,
68
  "step": 2000
69
  },
70
  {
71
- "epoch": 0.66,
72
- "eval_accuracy": 0.7891395463510849,
73
- "eval_f1": 0.7822709378126483,
74
- "eval_loss": 0.7148900032043457,
75
- "eval_precision": 0.7780706309716242,
76
- "eval_recall": 0.7891395463510849,
77
- "eval_runtime": 24.6279,
78
- "eval_samples_per_second": 658.766,
79
- "eval_steps_per_second": 27.449,
 
80
  "step": 2000
81
  },
82
  {
83
- "epoch": 0.82,
84
- "learning_rate": 3.960000000000001e-05,
85
- "loss": 0.6977,
86
  "step": 2500
87
  },
88
  {
89
- "epoch": 0.82,
90
- "eval_accuracy": 0.7249753451676528,
91
- "eval_f1": 0.7429677966966411,
92
- "eval_loss": 0.6928562521934509,
93
- "eval_precision": 0.7807492770659993,
94
- "eval_recall": 0.7249753451676528,
95
- "eval_runtime": 24.6337,
96
- "eval_samples_per_second": 658.61,
97
- "eval_steps_per_second": 27.442,
 
98
  "step": 2500
99
  },
100
  {
101
- "epoch": 0.99,
102
- "learning_rate": 3.7516666666666666e-05,
103
- "loss": 0.7032,
104
  "step": 3000
105
  },
106
  {
107
- "epoch": 0.99,
108
- "eval_accuracy": 0.7484590729783037,
109
- "eval_f1": 0.7594497306385251,
110
- "eval_loss": 0.6671503782272339,
111
- "eval_precision": 0.7788847542194782,
112
- "eval_recall": 0.7484590729783037,
113
- "eval_runtime": 24.5732,
114
- "eval_samples_per_second": 660.231,
115
- "eval_steps_per_second": 27.51,
 
116
  "step": 3000
117
  },
118
  {
119
- "epoch": 1.15,
120
- "learning_rate": 3.543333333333333e-05,
121
- "loss": 0.6227,
122
  "step": 3500
123
  },
124
  {
125
- "epoch": 1.15,
126
- "eval_accuracy": 0.7747781065088757,
127
- "eval_f1": 0.7788314002376725,
128
- "eval_loss": 0.6905977129936218,
129
- "eval_precision": 0.7840673104698108,
130
- "eval_recall": 0.7747781065088757,
131
- "eval_runtime": 24.7721,
132
- "eval_samples_per_second": 654.931,
133
- "eval_steps_per_second": 27.289,
 
134
  "step": 3500
135
  },
136
  {
137
- "epoch": 1.31,
138
- "learning_rate": 3.3350000000000004e-05,
139
- "loss": 0.6309,
140
  "step": 4000
141
  },
142
  {
143
- "epoch": 1.31,
144
- "eval_accuracy": 0.7607864891518737,
145
- "eval_f1": 0.7709664128632473,
146
- "eval_loss": 0.6686555743217468,
147
- "eval_precision": 0.7878354043029604,
148
- "eval_recall": 0.7607864891518737,
149
- "eval_runtime": 24.6117,
150
- "eval_samples_per_second": 659.198,
151
- "eval_steps_per_second": 27.467,
 
152
  "step": 4000
153
  },
154
  {
155
- "epoch": 1.48,
156
- "learning_rate": 3.126666666666666e-05,
157
- "loss": 0.6256,
158
  "step": 4500
159
  },
160
  {
161
- "epoch": 1.48,
162
- "eval_accuracy": 0.6715359960552268,
163
- "eval_f1": 0.7001277240194975,
164
- "eval_loss": 0.6691098809242249,
165
- "eval_precision": 0.7948795089365344,
166
- "eval_recall": 0.6715359960552268,
167
- "eval_runtime": 24.6266,
168
- "eval_samples_per_second": 658.801,
169
- "eval_steps_per_second": 27.45,
 
170
  "step": 4500
171
  },
172
  {
173
- "epoch": 1.64,
174
- "learning_rate": 2.91875e-05,
175
- "loss": 0.6152,
176
  "step": 5000
177
  },
178
  {
179
- "epoch": 1.64,
180
- "eval_accuracy": 0.7810034516765286,
181
- "eval_f1": 0.7815583569804468,
182
- "eval_loss": 0.6452860832214355,
183
- "eval_precision": 0.7840264023525906,
184
- "eval_recall": 0.7810034516765286,
185
- "eval_runtime": 24.5688,
186
- "eval_samples_per_second": 660.348,
187
- "eval_steps_per_second": 27.515,
 
188
  "step": 5000
189
  },
190
  {
191
- "epoch": 1.81,
192
- "learning_rate": 2.710416666666667e-05,
193
- "loss": 0.6068,
194
- "step": 5500
195
- },
196
- {
197
- "epoch": 1.81,
198
- "eval_accuracy": 0.7764423076923077,
199
- "eval_f1": 0.7843276562987542,
200
- "eval_loss": 0.6676753759384155,
201
- "eval_precision": 0.8001617824983851,
202
- "eval_recall": 0.7764423076923077,
203
- "eval_runtime": 24.7224,
204
- "eval_samples_per_second": 656.247,
205
- "eval_steps_per_second": 27.344,
206
- "step": 5500
207
- },
208
- {
209
- "epoch": 1.97,
210
- "learning_rate": 2.5020833333333333e-05,
211
- "loss": 0.6016,
212
- "step": 6000
213
- },
214
- {
215
- "epoch": 1.97,
216
- "eval_accuracy": 0.7220784023668639,
217
- "eval_f1": 0.7405665635983271,
218
- "eval_loss": 0.6461244821548462,
219
- "eval_precision": 0.782308061609766,
220
- "eval_recall": 0.7220784023668639,
221
- "eval_runtime": 24.6746,
222
- "eval_samples_per_second": 657.518,
223
- "eval_steps_per_second": 27.397,
224
- "step": 6000
225
- },
226
- {
227
- "epoch": 2.14,
228
- "learning_rate": 2.2937500000000002e-05,
229
- "loss": 0.521,
230
- "step": 6500
231
- },
232
- {
233
- "epoch": 2.14,
234
- "eval_accuracy": 0.7281188362919132,
235
- "eval_f1": 0.7466913370432308,
236
- "eval_loss": 0.7071252465248108,
237
- "eval_precision": 0.7878399669333115,
238
- "eval_recall": 0.7281188362919132,
239
- "eval_runtime": 24.7635,
240
- "eval_samples_per_second": 655.158,
241
- "eval_steps_per_second": 27.298,
242
- "step": 6500
243
- },
244
- {
245
- "epoch": 2.3,
246
- "learning_rate": 2.0854166666666668e-05,
247
- "loss": 0.5137,
248
- "step": 7000
249
- },
250
- {
251
- "epoch": 2.3,
252
- "eval_accuracy": 0.7346523668639053,
253
- "eval_f1": 0.752395675119321,
254
- "eval_loss": 0.6949610114097595,
255
- "eval_precision": 0.7902262301474485,
256
- "eval_recall": 0.7346523668639053,
257
- "eval_runtime": 24.761,
258
- "eval_samples_per_second": 655.225,
259
- "eval_steps_per_second": 27.301,
260
- "step": 7000
261
- },
262
- {
263
- "epoch": 2.47,
264
- "learning_rate": 1.8770833333333333e-05,
265
- "loss": 0.504,
266
- "step": 7500
267
- },
268
- {
269
- "epoch": 2.47,
270
- "eval_accuracy": 0.7453772189349113,
271
- "eval_f1": 0.7618113431887504,
272
- "eval_loss": 0.6824287176132202,
273
- "eval_precision": 0.7979845371909847,
274
- "eval_recall": 0.7453772189349113,
275
- "eval_runtime": 24.7586,
276
- "eval_samples_per_second": 655.287,
277
- "eval_steps_per_second": 27.304,
278
- "step": 7500
279
- },
280
- {
281
- "epoch": 2.63,
282
- "learning_rate": 1.6691666666666668e-05,
283
- "loss": 0.5049,
284
- "step": 8000
285
- },
286
- {
287
- "epoch": 2.63,
288
- "eval_accuracy": 0.7816814595660749,
289
- "eval_f1": 0.7878816281164417,
290
- "eval_loss": 0.6992799043655396,
291
- "eval_precision": 0.7967192296633613,
292
- "eval_recall": 0.7816814595660749,
293
- "eval_runtime": 24.7558,
294
- "eval_samples_per_second": 655.361,
295
- "eval_steps_per_second": 27.307,
296
- "step": 8000
297
- },
298
- {
299
- "epoch": 2.79,
300
- "learning_rate": 1.4608333333333335e-05,
301
- "loss": 0.4987,
302
- "step": 8500
303
- },
304
- {
305
- "epoch": 2.79,
306
- "eval_accuracy": 0.7680596646942801,
307
- "eval_f1": 0.779595614035326,
308
- "eval_loss": 0.7158842086791992,
309
- "eval_precision": 0.8023493022129313,
310
- "eval_recall": 0.7680596646942801,
311
- "eval_runtime": 24.6159,
312
- "eval_samples_per_second": 659.087,
313
- "eval_steps_per_second": 27.462,
314
- "step": 8500
315
- },
316
- {
317
- "epoch": 2.96,
318
- "learning_rate": 1.2525000000000001e-05,
319
- "loss": 0.4978,
320
- "step": 9000
321
- },
322
- {
323
- "epoch": 2.96,
324
- "eval_accuracy": 0.7757026627218935,
325
- "eval_f1": 0.7846497142505,
326
- "eval_loss": 0.6755026578903198,
327
- "eval_precision": 0.7992058416315558,
328
- "eval_recall": 0.7757026627218935,
329
- "eval_runtime": 24.6076,
330
- "eval_samples_per_second": 659.309,
331
- "eval_steps_per_second": 27.471,
332
- "step": 9000
333
- },
334
- {
335
- "epoch": 3.12,
336
- "learning_rate": 1.0441666666666667e-05,
337
- "loss": 0.4236,
338
- "step": 9500
339
- },
340
- {
341
- "epoch": 3.12,
342
- "eval_accuracy": 0.7638067061143984,
343
- "eval_f1": 0.7757104932531658,
344
- "eval_loss": 0.740465521812439,
345
- "eval_precision": 0.7974750781708845,
346
- "eval_recall": 0.7638067061143984,
347
- "eval_runtime": 24.6338,
348
- "eval_samples_per_second": 658.607,
349
- "eval_steps_per_second": 27.442,
350
- "step": 9500
351
- },
352
- {
353
- "epoch": 3.29,
354
- "learning_rate": 8.358333333333333e-06,
355
- "loss": 0.3987,
356
- "step": 10000
357
- },
358
- {
359
- "epoch": 3.29,
360
- "eval_accuracy": 0.7555473372781065,
361
- "eval_f1": 0.7701303707262471,
362
- "eval_loss": 0.7569716572761536,
363
- "eval_precision": 0.80053210602621,
364
- "eval_recall": 0.7555473372781065,
365
- "eval_runtime": 24.6348,
366
- "eval_samples_per_second": 658.58,
367
- "eval_steps_per_second": 27.441,
368
- "step": 10000
369
- },
370
- {
371
- "epoch": 3.45,
372
- "learning_rate": 6.275e-06,
373
- "loss": 0.3914,
374
- "step": 10500
375
- },
376
- {
377
- "epoch": 3.45,
378
- "eval_accuracy": 0.7660256410256411,
379
- "eval_f1": 0.7781937770368635,
380
- "eval_loss": 0.7907547354698181,
381
- "eval_precision": 0.8023937992057022,
382
- "eval_recall": 0.7660256410256411,
383
- "eval_runtime": 24.7673,
384
- "eval_samples_per_second": 655.057,
385
- "eval_steps_per_second": 27.294,
386
- "step": 10500
387
- },
388
- {
389
- "epoch": 3.62,
390
- "learning_rate": 4.195833333333334e-06,
391
- "loss": 0.4007,
392
- "step": 11000
393
- },
394
- {
395
- "epoch": 3.62,
396
- "eval_accuracy": 0.7821745562130178,
397
- "eval_f1": 0.7886478688599108,
398
- "eval_loss": 0.7599502801895142,
399
- "eval_precision": 0.7983440482619892,
400
- "eval_recall": 0.7821745562130178,
401
- "eval_runtime": 24.6844,
402
- "eval_samples_per_second": 657.258,
403
- "eval_steps_per_second": 27.386,
404
- "step": 11000
405
- },
406
- {
407
- "epoch": 3.78,
408
- "learning_rate": 2.1125e-06,
409
- "loss": 0.3878,
410
- "step": 11500
411
- },
412
- {
413
- "epoch": 3.78,
414
- "eval_accuracy": 0.7740384615384616,
415
- "eval_f1": 0.7836077149829459,
416
- "eval_loss": 0.7647958993911743,
417
- "eval_precision": 0.8007153395014647,
418
- "eval_recall": 0.7740384615384616,
419
- "eval_runtime": 24.6189,
420
- "eval_samples_per_second": 659.006,
421
- "eval_steps_per_second": 27.459,
422
- "step": 11500
423
- },
424
- {
425
- "epoch": 3.94,
426
- "learning_rate": 2.916666666666667e-08,
427
- "loss": 0.3829,
428
- "step": 12000
429
- },
430
- {
431
- "epoch": 3.94,
432
- "eval_accuracy": 0.7662721893491125,
433
- "eval_f1": 0.7779778325177736,
434
- "eval_loss": 0.752331018447876,
435
- "eval_precision": 0.7995290654452684,
436
- "eval_recall": 0.7662721893491125,
437
- "eval_runtime": 24.6488,
438
- "eval_samples_per_second": 658.206,
439
- "eval_steps_per_second": 27.425,
440
- "step": 12000
441
- },
442
- {
443
- "epoch": 3.94,
444
- "step": 12000,
445
- "total_flos": 1.502407617014661e+17,
446
- "train_loss": 0.5753211771647135,
447
- "train_runtime": 3636.5588,
448
- "train_samples_per_second": 158.391,
449
- "train_steps_per_second": 3.3
450
  }
451
  ],
452
- "max_steps": 12000,
453
- "num_train_epochs": 4,
454
- "total_flos": 1.502407617014661e+17,
455
  "trial_name": null,
456
  "trial_params": null
457
  }
 
1
  {
2
+ "best_metric": 0.6828581526810108,
3
+ "best_model_checkpoint": "trained/hebban-reviews/bert-base-multilingual-cased/checkpoint-1500",
4
+ "epoch": 4.382997370727432,
5
+ "global_step": 5001,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
9
  "log_history": [
10
  {
11
+ "epoch": 0.44,
12
+ "learning_rate": 4.504099180163968e-05,
13
+ "loss": 0.8498,
14
  "step": 500
15
  },
16
  {
17
+ "epoch": 0.44,
18
+ "eval_accuracy": 0.6671597633136095,
19
+ "eval_f1": 0.6926558253677099,
20
+ "eval_loss": 0.7364796996116638,
21
+ "eval_precision": 0.7520529030378761,
22
+ "eval_qwk": 0.5224288418460314,
23
+ "eval_recall": 0.6671597633136095,
24
+ "eval_runtime": 23.7829,
25
+ "eval_samples_per_second": 682.171,
26
+ "eval_steps_per_second": 5.34,
27
  "step": 500
28
  },
29
  {
30
+ "epoch": 0.88,
31
+ "learning_rate": 4.0041991601679665e-05,
32
+ "loss": 0.6918,
33
  "step": 1000
34
  },
35
  {
36
+ "epoch": 0.88,
37
+ "eval_accuracy": 0.7271942800788954,
38
+ "eval_f1": 0.7414418295002464,
39
+ "eval_loss": 0.6605744957923889,
40
+ "eval_precision": 0.7743449767911329,
41
+ "eval_qwk": 0.5976543204298063,
42
+ "eval_recall": 0.7271942800788954,
43
+ "eval_runtime": 23.6177,
44
+ "eval_samples_per_second": 686.943,
45
+ "eval_steps_per_second": 5.377,
46
  "step": 1000
47
  },
48
  {
49
+ "epoch": 1.31,
50
+ "learning_rate": 3.504299140171966e-05,
51
+ "loss": 0.621,
52
  "step": 1500
53
  },
54
  {
55
+ "epoch": 1.31,
56
+ "eval_accuracy": 0.7785996055226825,
57
+ "eval_f1": 0.7848033042674865,
58
+ "eval_loss": 0.6530709266662598,
59
+ "eval_precision": 0.7942290484031055,
60
+ "eval_qwk": 0.6828581526810108,
61
+ "eval_recall": 0.7785996055226825,
62
+ "eval_runtime": 23.6972,
63
+ "eval_samples_per_second": 684.637,
64
+ "eval_steps_per_second": 5.359,
65
  "step": 1500
66
  },
67
  {
68
+ "epoch": 1.75,
69
+ "learning_rate": 3.0043991201759648e-05,
70
+ "loss": 0.5936,
71
  "step": 2000
72
  },
73
  {
74
+ "epoch": 1.75,
75
+ "eval_accuracy": 0.7501849112426036,
76
+ "eval_f1": 0.7649773762321463,
77
+ "eval_loss": 0.6148771643638611,
78
+ "eval_precision": 0.7952109430358816,
79
+ "eval_qwk": 0.6613885538385067,
80
+ "eval_recall": 0.7501849112426036,
81
+ "eval_runtime": 23.8294,
82
+ "eval_samples_per_second": 680.838,
83
+ "eval_steps_per_second": 5.33,
84
  "step": 2000
85
  },
86
  {
87
+ "epoch": 2.19,
88
+ "learning_rate": 2.504499100179964e-05,
89
+ "loss": 0.5434,
90
  "step": 2500
91
  },
92
  {
93
+ "epoch": 2.19,
94
+ "eval_accuracy": 0.7548076923076923,
95
+ "eval_f1": 0.7709100781310829,
96
+ "eval_loss": 0.6947299242019653,
97
+ "eval_precision": 0.8080766037849636,
98
+ "eval_qwk": 0.6597495777741245,
99
+ "eval_recall": 0.7548076923076923,
100
+ "eval_runtime": 23.6425,
101
+ "eval_samples_per_second": 686.22,
102
+ "eval_steps_per_second": 5.372,
103
  "step": 2500
104
  },
105
  {
106
+ "epoch": 2.63,
107
+ "learning_rate": 2.0045990801839633e-05,
108
+ "loss": 0.4808,
109
  "step": 3000
110
  },
111
  {
112
+ "epoch": 2.63,
113
+ "eval_accuracy": 0.7347140039447732,
114
+ "eval_f1": 0.7545089976363922,
115
+ "eval_loss": 0.645616352558136,
116
+ "eval_precision": 0.8030403955873724,
117
+ "eval_qwk": 0.6513487818562542,
118
+ "eval_recall": 0.7347140039447732,
119
+ "eval_runtime": 23.8892,
120
+ "eval_samples_per_second": 679.136,
121
+ "eval_steps_per_second": 5.316,
122
  "step": 3000
123
  },
124
  {
125
+ "epoch": 3.07,
126
+ "learning_rate": 1.5046990601879624e-05,
127
+ "loss": 0.4643,
128
  "step": 3500
129
  },
130
  {
131
+ "epoch": 3.07,
132
+ "eval_accuracy": 0.7427268244575936,
133
+ "eval_f1": 0.7611673940281857,
134
+ "eval_loss": 0.7119439244270325,
135
+ "eval_precision": 0.8053303788937228,
136
+ "eval_qwk": 0.6549274566801152,
137
+ "eval_recall": 0.7427268244575936,
138
+ "eval_runtime": 23.6156,
139
+ "eval_samples_per_second": 687.004,
140
+ "eval_steps_per_second": 5.378,
141
  "step": 3500
142
  },
143
  {
144
+ "epoch": 3.51,
145
+ "learning_rate": 1.0047990401919617e-05,
146
+ "loss": 0.3854,
147
  "step": 4000
148
  },
149
  {
150
+ "epoch": 3.51,
151
+ "eval_accuracy": 0.7685527613412229,
152
+ "eval_f1": 0.7819273980616418,
153
+ "eval_loss": 0.7430822253227234,
154
+ "eval_precision": 0.8091264019259906,
155
+ "eval_qwk": 0.6812961348271991,
156
+ "eval_recall": 0.7685527613412229,
157
+ "eval_runtime": 23.6906,
158
+ "eval_samples_per_second": 684.828,
159
+ "eval_steps_per_second": 5.361,
160
  "step": 4000
161
  },
162
  {
163
+ "epoch": 3.94,
164
+ "learning_rate": 5.048990201959608e-06,
165
+ "loss": 0.3642,
166
  "step": 4500
167
  },
168
  {
169
+ "epoch": 3.94,
170
+ "eval_accuracy": 0.7693540433925049,
171
+ "eval_f1": 0.7827174333765768,
172
+ "eval_loss": 0.7832001447677612,
173
+ "eval_precision": 0.8104759531984178,
174
+ "eval_qwk": 0.6799731235431681,
175
+ "eval_recall": 0.7693540433925049,
176
+ "eval_runtime": 23.5733,
177
+ "eval_samples_per_second": 688.237,
178
+ "eval_steps_per_second": 5.387,
179
  "step": 4500
180
  },
181
  {
182
+ "epoch": 4.38,
183
+ "learning_rate": 4.999000199960008e-08,
184
+ "loss": 0.3199,
185
  "step": 5000
186
  },
187
  {
188
+ "epoch": 4.38,
189
+ "eval_accuracy": 0.7682445759368837,
190
+ "eval_f1": 0.7820013322186519,
191
+ "eval_loss": 0.8646416664123535,
192
+ "eval_precision": 0.8117448424467829,
193
+ "eval_qwk": 0.6776167553082479,
194
+ "eval_recall": 0.7682445759368837,
195
+ "eval_runtime": 23.7003,
196
+ "eval_samples_per_second": 684.547,
197
+ "eval_steps_per_second": 5.359,
198
  "step": 5000
199
  },
200
  {
201
+ "epoch": 4.38,
202
+ "step": 5001,
203
+ "total_flos": 1.6838837781764506e+17,
204
+ "train_loss": 0.5313903366928219,
205
+ "train_runtime": 2667.6757,
206
+ "train_samples_per_second": 239.957,
207
+ "train_steps_per_second": 1.875
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
208
  }
209
  ],
210
+ "max_steps": 5001,
211
+ "num_train_epochs": 5,
212
+ "total_flos": 1.6838837781764506e+17,
213
  "trial_name": null,
214
  "trial_params": null
215
  }
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2a1b1bde55632273fa2d79fa4e7449972067ad76fcbd823e01f19e5d013eba50
3
  size 3439
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:124ce131b35effd9a1e5de6412af80ddd87768156ca8d94b92ddde0c2440032d
3
  size 3439