DGurgurov commited on
Commit
b8afc27
·
verified ·
1 Parent(s): 8523e97

Uploading checkpoint-39000 for mbert - mar-deva

Browse files
README.md ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: google-bert/bert-base-multilingual-cased
5
+ tags:
6
+ - generated_from_trainer
7
+ metrics:
8
+ - accuracy
9
+ model-index:
10
+ - name: mar-Deva
11
+ results: []
12
+ ---
13
+
14
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
+ should probably proofread and complete it, then remove this comment. -->
16
+
17
+ # mar-Deva
18
+
19
+ This model is a fine-tuned version of [google-bert/bert-base-multilingual-cased](https://huggingface.co/google-bert/bert-base-multilingual-cased) on an unknown dataset.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 0.7135
22
+ - Accuracy: 0.8574
23
+
24
+ ## Model description
25
+
26
+ More information needed
27
+
28
+ ## Intended uses & limitations
29
+
30
+ More information needed
31
+
32
+ ## Training and evaluation data
33
+
34
+ More information needed
35
+
36
+ ## Training procedure
37
+
38
+ ### Training hyperparameters
39
+
40
+ The following hyperparameters were used during training:
41
+ - learning_rate: 0.0001
42
+ - train_batch_size: 16
43
+ - eval_batch_size: 16
44
+ - seed: 42
45
+ - distributed_type: multi-GPU
46
+ - num_devices: 2
47
+ - total_train_batch_size: 32
48
+ - total_eval_batch_size: 32
49
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
+ - lr_scheduler_type: linear
51
+ - training_steps: 100000
52
+
53
+
54
+ ### Citation Information
55
+
56
+ If you use this model in your work, please cite the following paper. Additionally, if you require more details on training and performance, refer to the paper:
57
+
58
+ @misc{gurgurov2025smallmodelsbigimpact,
59
+ title={Small Models, Big Impact: Efficient Corpus and Graph-Based Adaptation of Small Multilingual Language Models for Low-Resource Languages},
60
+ author={Daniil Gurgurov and Ivan Vykopal and Josef van Genabith and Simon Ostermann},
61
+ year={2025},
62
+ eprint={2502.10140},
63
+ archivePrefix={arXiv},
64
+ primaryClass={cs.CL},
65
+ url={https://arxiv.org/abs/2502.10140},
66
+ }
67
+
config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "google-bert/bert-base-multilingual-cased",
3
+ "adapters": {
4
+ "adapters": {},
5
+ "config_map": {},
6
+ "fusion_config_map": {},
7
+ "fusions": {}
8
+ },
9
+ "architectures": [
10
+ "BertForMaskedLM"
11
+ ],
12
+ "attention_probs_dropout_prob": 0.1,
13
+ "classifier_dropout": null,
14
+ "directionality": "bidi",
15
+ "hidden_act": "gelu",
16
+ "hidden_dropout_prob": 0.1,
17
+ "hidden_size": 768,
18
+ "initializer_range": 0.02,
19
+ "intermediate_size": 3072,
20
+ "layer_norm_eps": 1e-12,
21
+ "max_position_embeddings": 512,
22
+ "model_type": "bert",
23
+ "num_attention_heads": 12,
24
+ "num_hidden_layers": 12,
25
+ "pad_token_id": 0,
26
+ "pooler_fc_size": 768,
27
+ "pooler_num_attention_heads": 12,
28
+ "pooler_num_fc_layers": 3,
29
+ "pooler_size_per_head": 128,
30
+ "pooler_type": "first_token_transform",
31
+ "position_embedding_type": "absolute",
32
+ "torch_dtype": "float32",
33
+ "transformers_version": "4.45.2",
34
+ "type_vocab_size": 2,
35
+ "use_cache": true,
36
+ "vocab_size": 119547
37
+ }
generation_config.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "pad_token_id": 0,
4
+ "transformers_version": "4.45.2"
5
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d97be4ba73b84b522b977f708967e954f6fdd29e486b6c9a6507eaa56dd4e60a
3
+ size 711921812
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af75d1e9c750b8a41bb084bd62830c98396f28f84543630999be97f95d990420
3
+ size 1423965114
rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b3c97cc5ba6f9c5f1da78e171b8371fec9c4f57850e29e313021157482fae26
3
+ size 14512
rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b93eb032cf0f7c04cfc959fac8743ac7cddad172035eaf2e12760ebcac3c236b
3
+ size 14512
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df5690e73943bb6cb41f6f10002304bbac2fc340384febcc3a4a53d7d53c7b63
3
+ size 1064
special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "[PAD]",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "100": {
12
+ "content": "[UNK]",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "101": {
20
+ "content": "[CLS]",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "102": {
28
+ "content": "[SEP]",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "103": {
36
+ "content": "[MASK]",
37
+ "lstrip": false,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "clean_up_tokenization_spaces": false,
45
+ "cls_token": "[CLS]",
46
+ "do_lower_case": false,
47
+ "mask_token": "[MASK]",
48
+ "model_max_length": 512,
49
+ "pad_token": "[PAD]",
50
+ "sep_token": "[SEP]",
51
+ "strip_accents": null,
52
+ "tokenize_chinese_chars": true,
53
+ "tokenizer_class": "BertTokenizer",
54
+ "unk_token": "[UNK]"
55
+ }
trainer_state.json ADDED
@@ -0,0 +1,1281 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.7045323252677917,
3
+ "best_model_checkpoint": "./model_fine-tune/glot/mbert/mar-Deva/checkpoint-39000",
4
+ "epoch": 91.54929577464789,
5
+ "eval_steps": 500,
6
+ "global_step": 39000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 1.1737089201877935,
13
+ "grad_norm": 3.1700332164764404,
14
+ "learning_rate": 9.95e-05,
15
+ "loss": 1.3483,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 1.1737089201877935,
20
+ "eval_accuracy": 0.749312885729062,
21
+ "eval_loss": 1.1869930028915405,
22
+ "eval_runtime": 145.5506,
23
+ "eval_samples_per_second": 121.484,
24
+ "eval_steps_per_second": 3.799,
25
+ "step": 500
26
+ },
27
+ {
28
+ "epoch": 2.347417840375587,
29
+ "grad_norm": 2.450835704803467,
30
+ "learning_rate": 9.900000000000001e-05,
31
+ "loss": 1.16,
32
+ "step": 1000
33
+ },
34
+ {
35
+ "epoch": 2.347417840375587,
36
+ "eval_accuracy": 0.7679429076763364,
37
+ "eval_loss": 1.1002804040908813,
38
+ "eval_runtime": 138.5842,
39
+ "eval_samples_per_second": 127.59,
40
+ "eval_steps_per_second": 3.99,
41
+ "step": 1000
42
+ },
43
+ {
44
+ "epoch": 3.52112676056338,
45
+ "grad_norm": 2.5313944816589355,
46
+ "learning_rate": 9.850000000000001e-05,
47
+ "loss": 1.0796,
48
+ "step": 1500
49
+ },
50
+ {
51
+ "epoch": 3.52112676056338,
52
+ "eval_accuracy": 0.7788953556992149,
53
+ "eval_loss": 1.0443540811538696,
54
+ "eval_runtime": 143.0988,
55
+ "eval_samples_per_second": 123.565,
56
+ "eval_steps_per_second": 3.864,
57
+ "step": 1500
58
+ },
59
+ {
60
+ "epoch": 4.694835680751174,
61
+ "grad_norm": 2.6745097637176514,
62
+ "learning_rate": 9.8e-05,
63
+ "loss": 1.023,
64
+ "step": 2000
65
+ },
66
+ {
67
+ "epoch": 4.694835680751174,
68
+ "eval_accuracy": 0.7875008678674885,
69
+ "eval_loss": 0.9982088208198547,
70
+ "eval_runtime": 147.3047,
71
+ "eval_samples_per_second": 120.037,
72
+ "eval_steps_per_second": 3.754,
73
+ "step": 2000
74
+ },
75
+ {
76
+ "epoch": 5.868544600938967,
77
+ "grad_norm": 2.8676445484161377,
78
+ "learning_rate": 9.75e-05,
79
+ "loss": 0.9784,
80
+ "step": 2500
81
+ },
82
+ {
83
+ "epoch": 5.868544600938967,
84
+ "eval_accuracy": 0.7940504772991072,
85
+ "eval_loss": 0.964560329914093,
86
+ "eval_runtime": 137.4737,
87
+ "eval_samples_per_second": 128.621,
88
+ "eval_steps_per_second": 4.023,
89
+ "step": 2500
90
+ },
91
+ {
92
+ "epoch": 7.042253521126761,
93
+ "grad_norm": 2.3849828243255615,
94
+ "learning_rate": 9.7e-05,
95
+ "loss": 0.9394,
96
+ "step": 3000
97
+ },
98
+ {
99
+ "epoch": 7.042253521126761,
100
+ "eval_accuracy": 0.7981764510868291,
101
+ "eval_loss": 0.9444334506988525,
102
+ "eval_runtime": 147.2364,
103
+ "eval_samples_per_second": 120.093,
104
+ "eval_steps_per_second": 3.756,
105
+ "step": 3000
106
+ },
107
+ {
108
+ "epoch": 8.215962441314554,
109
+ "grad_norm": 2.1890273094177246,
110
+ "learning_rate": 9.65e-05,
111
+ "loss": 0.9119,
112
+ "step": 3500
113
+ },
114
+ {
115
+ "epoch": 8.215962441314554,
116
+ "eval_accuracy": 0.8029835860860965,
117
+ "eval_loss": 0.9215248823165894,
118
+ "eval_runtime": 147.2738,
119
+ "eval_samples_per_second": 120.062,
120
+ "eval_steps_per_second": 3.755,
121
+ "step": 3500
122
+ },
123
+ {
124
+ "epoch": 9.389671361502348,
125
+ "grad_norm": 2.342008590698242,
126
+ "learning_rate": 9.6e-05,
127
+ "loss": 0.8872,
128
+ "step": 4000
129
+ },
130
+ {
131
+ "epoch": 9.389671361502348,
132
+ "eval_accuracy": 0.805941164259263,
133
+ "eval_loss": 0.916360080242157,
134
+ "eval_runtime": 147.2413,
135
+ "eval_samples_per_second": 120.089,
136
+ "eval_steps_per_second": 3.756,
137
+ "step": 4000
138
+ },
139
+ {
140
+ "epoch": 10.56338028169014,
141
+ "grad_norm": 2.2766637802124023,
142
+ "learning_rate": 9.55e-05,
143
+ "loss": 0.8629,
144
+ "step": 4500
145
+ },
146
+ {
147
+ "epoch": 10.56338028169014,
148
+ "eval_accuracy": 0.8102512145065337,
149
+ "eval_loss": 0.898952066898346,
150
+ "eval_runtime": 147.2077,
151
+ "eval_samples_per_second": 120.116,
152
+ "eval_steps_per_second": 3.757,
153
+ "step": 4500
154
+ },
155
+ {
156
+ "epoch": 11.737089201877934,
157
+ "grad_norm": 2.391869306564331,
158
+ "learning_rate": 9.5e-05,
159
+ "loss": 0.8477,
160
+ "step": 5000
161
+ },
162
+ {
163
+ "epoch": 11.737089201877934,
164
+ "eval_accuracy": 0.8117510986190105,
165
+ "eval_loss": 0.8836262822151184,
166
+ "eval_runtime": 145.6587,
167
+ "eval_samples_per_second": 121.393,
168
+ "eval_steps_per_second": 3.797,
169
+ "step": 5000
170
+ },
171
+ {
172
+ "epoch": 12.910798122065728,
173
+ "grad_norm": 2.4697470664978027,
174
+ "learning_rate": 9.449999999999999e-05,
175
+ "loss": 0.8261,
176
+ "step": 5500
177
+ },
178
+ {
179
+ "epoch": 12.910798122065728,
180
+ "eval_accuracy": 0.8150349185540299,
181
+ "eval_loss": 0.8613038063049316,
182
+ "eval_runtime": 145.7167,
183
+ "eval_samples_per_second": 121.345,
184
+ "eval_steps_per_second": 3.795,
185
+ "step": 5500
186
+ },
187
+ {
188
+ "epoch": 14.084507042253522,
189
+ "grad_norm": 2.36245059967041,
190
+ "learning_rate": 9.4e-05,
191
+ "loss": 0.8072,
192
+ "step": 6000
193
+ },
194
+ {
195
+ "epoch": 14.084507042253522,
196
+ "eval_accuracy": 0.816731876355394,
197
+ "eval_loss": 0.8536739945411682,
198
+ "eval_runtime": 147.07,
199
+ "eval_samples_per_second": 120.228,
200
+ "eval_steps_per_second": 3.76,
201
+ "step": 6000
202
+ },
203
+ {
204
+ "epoch": 15.258215962441314,
205
+ "grad_norm": 2.285848379135132,
206
+ "learning_rate": 9.350000000000001e-05,
207
+ "loss": 0.7924,
208
+ "step": 6500
209
+ },
210
+ {
211
+ "epoch": 15.258215962441314,
212
+ "eval_accuracy": 0.8197007489640474,
213
+ "eval_loss": 0.8382024765014648,
214
+ "eval_runtime": 147.1867,
215
+ "eval_samples_per_second": 120.133,
216
+ "eval_steps_per_second": 3.757,
217
+ "step": 6500
218
+ },
219
+ {
220
+ "epoch": 16.431924882629108,
221
+ "grad_norm": 2.218266010284424,
222
+ "learning_rate": 9.300000000000001e-05,
223
+ "loss": 0.7815,
224
+ "step": 7000
225
+ },
226
+ {
227
+ "epoch": 16.431924882629108,
228
+ "eval_accuracy": 0.8211554348763181,
229
+ "eval_loss": 0.8357976675033569,
230
+ "eval_runtime": 135.7542,
231
+ "eval_samples_per_second": 130.25,
232
+ "eval_steps_per_second": 4.074,
233
+ "step": 7000
234
+ },
235
+ {
236
+ "epoch": 17.6056338028169,
237
+ "grad_norm": 2.300184488296509,
238
+ "learning_rate": 9.250000000000001e-05,
239
+ "loss": 0.767,
240
+ "step": 7500
241
+ },
242
+ {
243
+ "epoch": 17.6056338028169,
244
+ "eval_accuracy": 0.8242465371712792,
245
+ "eval_loss": 0.8331694006919861,
246
+ "eval_runtime": 145.3933,
247
+ "eval_samples_per_second": 121.615,
248
+ "eval_steps_per_second": 3.803,
249
+ "step": 7500
250
+ },
251
+ {
252
+ "epoch": 18.779342723004696,
253
+ "grad_norm": 2.1632742881774902,
254
+ "learning_rate": 9.200000000000001e-05,
255
+ "loss": 0.7541,
256
+ "step": 8000
257
+ },
258
+ {
259
+ "epoch": 18.779342723004696,
260
+ "eval_accuracy": 0.8246549447936785,
261
+ "eval_loss": 0.8179985880851746,
262
+ "eval_runtime": 139.0754,
263
+ "eval_samples_per_second": 127.14,
264
+ "eval_steps_per_second": 3.976,
265
+ "step": 8000
266
+ },
267
+ {
268
+ "epoch": 19.953051643192488,
269
+ "grad_norm": 2.2807085514068604,
270
+ "learning_rate": 9.15e-05,
271
+ "loss": 0.7462,
272
+ "step": 8500
273
+ },
274
+ {
275
+ "epoch": 19.953051643192488,
276
+ "eval_accuracy": 0.8256249942473,
277
+ "eval_loss": 0.8264754414558411,
278
+ "eval_runtime": 147.2594,
279
+ "eval_samples_per_second": 120.074,
280
+ "eval_steps_per_second": 3.755,
281
+ "step": 8500
282
+ },
283
+ {
284
+ "epoch": 21.12676056338028,
285
+ "grad_norm": 2.210843563079834,
286
+ "learning_rate": 9.1e-05,
287
+ "loss": 0.7323,
288
+ "step": 9000
289
+ },
290
+ {
291
+ "epoch": 21.12676056338028,
292
+ "eval_accuracy": 0.8267119847302682,
293
+ "eval_loss": 0.8105438947677612,
294
+ "eval_runtime": 136.1823,
295
+ "eval_samples_per_second": 129.841,
296
+ "eval_steps_per_second": 4.061,
297
+ "step": 9000
298
+ },
299
+ {
300
+ "epoch": 22.300469483568076,
301
+ "grad_norm": 2.389461040496826,
302
+ "learning_rate": 9.05e-05,
303
+ "loss": 0.7212,
304
+ "step": 9500
305
+ },
306
+ {
307
+ "epoch": 22.300469483568076,
308
+ "eval_accuracy": 0.8280417403957945,
309
+ "eval_loss": 0.8193202018737793,
310
+ "eval_runtime": 147.4213,
311
+ "eval_samples_per_second": 119.942,
312
+ "eval_steps_per_second": 3.751,
313
+ "step": 9500
314
+ },
315
+ {
316
+ "epoch": 23.474178403755868,
317
+ "grad_norm": 2.059155225753784,
318
+ "learning_rate": 9e-05,
319
+ "loss": 0.7086,
320
+ "step": 10000
321
+ },
322
+ {
323
+ "epoch": 23.474178403755868,
324
+ "eval_accuracy": 0.8300135381668301,
325
+ "eval_loss": 0.8017289638519287,
326
+ "eval_runtime": 137.0188,
327
+ "eval_samples_per_second": 129.048,
328
+ "eval_steps_per_second": 4.036,
329
+ "step": 10000
330
+ },
331
+ {
332
+ "epoch": 24.647887323943664,
333
+ "grad_norm": 2.143441677093506,
334
+ "learning_rate": 8.950000000000001e-05,
335
+ "loss": 0.7018,
336
+ "step": 10500
337
+ },
338
+ {
339
+ "epoch": 24.647887323943664,
340
+ "eval_accuracy": 0.8309204998181491,
341
+ "eval_loss": 0.7943059206008911,
342
+ "eval_runtime": 136.1098,
343
+ "eval_samples_per_second": 129.91,
344
+ "eval_steps_per_second": 4.063,
345
+ "step": 10500
346
+ },
347
+ {
348
+ "epoch": 25.821596244131456,
349
+ "grad_norm": 2.3048555850982666,
350
+ "learning_rate": 8.900000000000001e-05,
351
+ "loss": 0.6949,
352
+ "step": 11000
353
+ },
354
+ {
355
+ "epoch": 25.821596244131456,
356
+ "eval_accuracy": 0.8313006683640194,
357
+ "eval_loss": 0.7950281500816345,
358
+ "eval_runtime": 147.4239,
359
+ "eval_samples_per_second": 119.94,
360
+ "eval_steps_per_second": 3.751,
361
+ "step": 11000
362
+ },
363
+ {
364
+ "epoch": 26.995305164319248,
365
+ "grad_norm": 1.9936089515686035,
366
+ "learning_rate": 8.850000000000001e-05,
367
+ "loss": 0.6917,
368
+ "step": 11500
369
+ },
370
+ {
371
+ "epoch": 26.995305164319248,
372
+ "eval_accuracy": 0.8331900538385245,
373
+ "eval_loss": 0.7858129143714905,
374
+ "eval_runtime": 137.4392,
375
+ "eval_samples_per_second": 128.653,
376
+ "eval_steps_per_second": 4.024,
377
+ "step": 11500
378
+ },
379
+ {
380
+ "epoch": 28.169014084507044,
381
+ "grad_norm": 2.1459641456604004,
382
+ "learning_rate": 8.800000000000001e-05,
383
+ "loss": 0.6782,
384
+ "step": 12000
385
+ },
386
+ {
387
+ "epoch": 28.169014084507044,
388
+ "eval_accuracy": 0.8343617601040733,
389
+ "eval_loss": 0.7777819037437439,
390
+ "eval_runtime": 147.5831,
391
+ "eval_samples_per_second": 119.81,
392
+ "eval_steps_per_second": 3.747,
393
+ "step": 12000
394
+ },
395
+ {
396
+ "epoch": 29.342723004694836,
397
+ "grad_norm": 2.0599782466888428,
398
+ "learning_rate": 8.75e-05,
399
+ "loss": 0.672,
400
+ "step": 12500
401
+ },
402
+ {
403
+ "epoch": 29.342723004694836,
404
+ "eval_accuracy": 0.8344647882838259,
405
+ "eval_loss": 0.7832308411598206,
406
+ "eval_runtime": 137.4669,
407
+ "eval_samples_per_second": 128.627,
408
+ "eval_steps_per_second": 4.023,
409
+ "step": 12500
410
+ },
411
+ {
412
+ "epoch": 30.516431924882628,
413
+ "grad_norm": 2.52083158493042,
414
+ "learning_rate": 8.7e-05,
415
+ "loss": 0.6612,
416
+ "step": 13000
417
+ },
418
+ {
419
+ "epoch": 30.516431924882628,
420
+ "eval_accuracy": 0.835733848691268,
421
+ "eval_loss": 0.7805770039558411,
422
+ "eval_runtime": 137.4676,
423
+ "eval_samples_per_second": 128.627,
424
+ "eval_steps_per_second": 4.023,
425
+ "step": 13000
426
+ },
427
+ {
428
+ "epoch": 31.690140845070424,
429
+ "grad_norm": 2.808067560195923,
430
+ "learning_rate": 8.65e-05,
431
+ "loss": 0.653,
432
+ "step": 13500
433
+ },
434
+ {
435
+ "epoch": 31.690140845070424,
436
+ "eval_accuracy": 0.8369936505384986,
437
+ "eval_loss": 0.7829101085662842,
438
+ "eval_runtime": 147.4085,
439
+ "eval_samples_per_second": 119.952,
440
+ "eval_steps_per_second": 3.751,
441
+ "step": 13500
442
+ },
443
+ {
444
+ "epoch": 32.863849765258216,
445
+ "grad_norm": 2.1012039184570312,
446
+ "learning_rate": 8.6e-05,
447
+ "loss": 0.6509,
448
+ "step": 14000
449
+ },
450
+ {
451
+ "epoch": 32.863849765258216,
452
+ "eval_accuracy": 0.8373059658134827,
453
+ "eval_loss": 0.7640124559402466,
454
+ "eval_runtime": 137.589,
455
+ "eval_samples_per_second": 128.513,
456
+ "eval_steps_per_second": 4.019,
457
+ "step": 14000
458
+ },
459
+ {
460
+ "epoch": 34.03755868544601,
461
+ "grad_norm": 2.1211578845977783,
462
+ "learning_rate": 8.55e-05,
463
+ "loss": 0.6403,
464
+ "step": 14500
465
+ },
466
+ {
467
+ "epoch": 34.03755868544601,
468
+ "eval_accuracy": 0.8385678802262423,
469
+ "eval_loss": 0.7672787308692932,
470
+ "eval_runtime": 146.9801,
471
+ "eval_samples_per_second": 120.302,
472
+ "eval_steps_per_second": 3.762,
473
+ "step": 14500
474
+ },
475
+ {
476
+ "epoch": 35.2112676056338,
477
+ "grad_norm": 2.017094135284424,
478
+ "learning_rate": 8.5e-05,
479
+ "loss": 0.6348,
480
+ "step": 15000
481
+ },
482
+ {
483
+ "epoch": 35.2112676056338,
484
+ "eval_accuracy": 0.8392093131517121,
485
+ "eval_loss": 0.7593186497688293,
486
+ "eval_runtime": 137.1351,
487
+ "eval_samples_per_second": 128.939,
488
+ "eval_steps_per_second": 4.033,
489
+ "step": 15000
490
+ },
491
+ {
492
+ "epoch": 36.3849765258216,
493
+ "grad_norm": 2.0924570560455322,
494
+ "learning_rate": 8.450000000000001e-05,
495
+ "loss": 0.6297,
496
+ "step": 15500
497
+ },
498
+ {
499
+ "epoch": 36.3849765258216,
500
+ "eval_accuracy": 0.8399583286650959,
501
+ "eval_loss": 0.763680636882782,
502
+ "eval_runtime": 147.5378,
503
+ "eval_samples_per_second": 119.847,
504
+ "eval_steps_per_second": 3.748,
505
+ "step": 15500
506
+ },
507
+ {
508
+ "epoch": 37.55868544600939,
509
+ "grad_norm": 2.131864070892334,
510
+ "learning_rate": 8.4e-05,
511
+ "loss": 0.6238,
512
+ "step": 16000
513
+ },
514
+ {
515
+ "epoch": 37.55868544600939,
516
+ "eval_accuracy": 0.840090867939712,
517
+ "eval_loss": 0.7627538442611694,
518
+ "eval_runtime": 137.0753,
519
+ "eval_samples_per_second": 128.995,
520
+ "eval_steps_per_second": 4.034,
521
+ "step": 16000
522
+ },
523
+ {
524
+ "epoch": 38.732394366197184,
525
+ "grad_norm": 2.12919545173645,
526
+ "learning_rate": 8.35e-05,
527
+ "loss": 0.6162,
528
+ "step": 16500
529
+ },
530
+ {
531
+ "epoch": 38.732394366197184,
532
+ "eval_accuracy": 0.8413376227933144,
533
+ "eval_loss": 0.7513773441314697,
534
+ "eval_runtime": 147.6687,
535
+ "eval_samples_per_second": 119.741,
536
+ "eval_steps_per_second": 3.745,
537
+ "step": 16500
538
+ },
539
+ {
540
+ "epoch": 39.906103286384976,
541
+ "grad_norm": 2.1172847747802734,
542
+ "learning_rate": 8.3e-05,
543
+ "loss": 0.6121,
544
+ "step": 17000
545
+ },
546
+ {
547
+ "epoch": 39.906103286384976,
548
+ "eval_accuracy": 0.8418376093750392,
549
+ "eval_loss": 0.753982663154602,
550
+ "eval_runtime": 136.0303,
551
+ "eval_samples_per_second": 129.986,
552
+ "eval_steps_per_second": 4.065,
553
+ "step": 17000
554
+ },
555
+ {
556
+ "epoch": 41.07981220657277,
557
+ "grad_norm": 2.195590019226074,
558
+ "learning_rate": 8.25e-05,
559
+ "loss": 0.6078,
560
+ "step": 17500
561
+ },
562
+ {
563
+ "epoch": 41.07981220657277,
564
+ "eval_accuracy": 0.8428805964270359,
565
+ "eval_loss": 0.7542482614517212,
566
+ "eval_runtime": 137.0472,
567
+ "eval_samples_per_second": 129.021,
568
+ "eval_steps_per_second": 4.035,
569
+ "step": 17500
570
+ },
571
+ {
572
+ "epoch": 42.25352112676056,
573
+ "grad_norm": 2.067308187484741,
574
+ "learning_rate": 8.2e-05,
575
+ "loss": 0.601,
576
+ "step": 18000
577
+ },
578
+ {
579
+ "epoch": 42.25352112676056,
580
+ "eval_accuracy": 0.8427417329846963,
581
+ "eval_loss": 0.7472436428070068,
582
+ "eval_runtime": 136.152,
583
+ "eval_samples_per_second": 129.87,
584
+ "eval_steps_per_second": 4.062,
585
+ "step": 18000
586
+ },
587
+ {
588
+ "epoch": 43.42723004694836,
589
+ "grad_norm": 2.278040885925293,
590
+ "learning_rate": 8.15e-05,
591
+ "loss": 0.5962,
592
+ "step": 18500
593
+ },
594
+ {
595
+ "epoch": 43.42723004694836,
596
+ "eval_accuracy": 0.8429750545039596,
597
+ "eval_loss": 0.7438804507255554,
598
+ "eval_runtime": 136.1815,
599
+ "eval_samples_per_second": 129.841,
600
+ "eval_steps_per_second": 4.061,
601
+ "step": 18500
602
+ },
603
+ {
604
+ "epoch": 44.60093896713615,
605
+ "grad_norm": 2.0480079650878906,
606
+ "learning_rate": 8.1e-05,
607
+ "loss": 0.5921,
608
+ "step": 19000
609
+ },
610
+ {
611
+ "epoch": 44.60093896713615,
612
+ "eval_accuracy": 0.8430666280582005,
613
+ "eval_loss": 0.7555158734321594,
614
+ "eval_runtime": 147.7547,
615
+ "eval_samples_per_second": 119.671,
616
+ "eval_steps_per_second": 3.743,
617
+ "step": 19000
618
+ },
619
+ {
620
+ "epoch": 45.774647887323944,
621
+ "grad_norm": 2.0856966972351074,
622
+ "learning_rate": 8.05e-05,
623
+ "loss": 0.5828,
624
+ "step": 19500
625
+ },
626
+ {
627
+ "epoch": 45.774647887323944,
628
+ "eval_accuracy": 0.8444825839615888,
629
+ "eval_loss": 0.7403737902641296,
630
+ "eval_runtime": 142.7228,
631
+ "eval_samples_per_second": 123.89,
632
+ "eval_steps_per_second": 3.875,
633
+ "step": 19500
634
+ },
635
+ {
636
+ "epoch": 46.948356807511736,
637
+ "grad_norm": 2.0801377296447754,
638
+ "learning_rate": 8e-05,
639
+ "loss": 0.5805,
640
+ "step": 20000
641
+ },
642
+ {
643
+ "epoch": 46.948356807511736,
644
+ "eval_accuracy": 0.8456604553444954,
645
+ "eval_loss": 0.7381341457366943,
646
+ "eval_runtime": 136.8181,
647
+ "eval_samples_per_second": 129.237,
648
+ "eval_steps_per_second": 4.042,
649
+ "step": 20000
650
+ },
651
+ {
652
+ "epoch": 48.12206572769953,
653
+ "grad_norm": 2.0854578018188477,
654
+ "learning_rate": 7.950000000000001e-05,
655
+ "loss": 0.577,
656
+ "step": 20500
657
+ },
658
+ {
659
+ "epoch": 48.12206572769953,
660
+ "eval_accuracy": 0.8459030596342113,
661
+ "eval_loss": 0.7439441680908203,
662
+ "eval_runtime": 137.0553,
663
+ "eval_samples_per_second": 129.014,
664
+ "eval_steps_per_second": 4.035,
665
+ "step": 20500
666
+ },
667
+ {
668
+ "epoch": 49.29577464788732,
669
+ "grad_norm": 2.0203611850738525,
670
+ "learning_rate": 7.900000000000001e-05,
671
+ "loss": 0.5702,
672
+ "step": 21000
673
+ },
674
+ {
675
+ "epoch": 49.29577464788732,
676
+ "eval_accuracy": 0.8458144906353607,
677
+ "eval_loss": 0.739824116230011,
678
+ "eval_runtime": 147.5519,
679
+ "eval_samples_per_second": 119.836,
680
+ "eval_steps_per_second": 3.748,
681
+ "step": 21000
682
+ },
683
+ {
684
+ "epoch": 50.46948356807512,
685
+ "grad_norm": 2.07930326461792,
686
+ "learning_rate": 7.850000000000001e-05,
687
+ "loss": 0.5626,
688
+ "step": 21500
689
+ },
690
+ {
691
+ "epoch": 50.46948356807512,
692
+ "eval_accuracy": 0.847225799976582,
693
+ "eval_loss": 0.7376012206077576,
694
+ "eval_runtime": 140.7844,
695
+ "eval_samples_per_second": 125.596,
696
+ "eval_steps_per_second": 3.928,
697
+ "step": 21500
698
+ },
699
+ {
700
+ "epoch": 51.64319248826291,
701
+ "grad_norm": 1.9633455276489258,
702
+ "learning_rate": 7.800000000000001e-05,
703
+ "loss": 0.5644,
704
+ "step": 22000
705
+ },
706
+ {
707
+ "epoch": 51.64319248826291,
708
+ "eval_accuracy": 0.8477510357072401,
709
+ "eval_loss": 0.7312297224998474,
710
+ "eval_runtime": 136.8418,
711
+ "eval_samples_per_second": 129.215,
712
+ "eval_steps_per_second": 4.041,
713
+ "step": 22000
714
+ },
715
+ {
716
+ "epoch": 52.816901408450704,
717
+ "grad_norm": 2.0171732902526855,
718
+ "learning_rate": 7.75e-05,
719
+ "loss": 0.5575,
720
+ "step": 22500
721
+ },
722
+ {
723
+ "epoch": 52.816901408450704,
724
+ "eval_accuracy": 0.8479556976991851,
725
+ "eval_loss": 0.7324073314666748,
726
+ "eval_runtime": 136.8371,
727
+ "eval_samples_per_second": 129.219,
728
+ "eval_steps_per_second": 4.041,
729
+ "step": 22500
730
+ },
731
+ {
732
+ "epoch": 53.990610328638496,
733
+ "grad_norm": 1.9490004777908325,
734
+ "learning_rate": 7.7e-05,
735
+ "loss": 0.5504,
736
+ "step": 23000
737
+ },
738
+ {
739
+ "epoch": 53.990610328638496,
740
+ "eval_accuracy": 0.8482885853024055,
741
+ "eval_loss": 0.739380955696106,
742
+ "eval_runtime": 136.7899,
743
+ "eval_samples_per_second": 129.264,
744
+ "eval_steps_per_second": 4.043,
745
+ "step": 23000
746
+ },
747
+ {
748
+ "epoch": 55.16431924882629,
749
+ "grad_norm": 2.1706230640411377,
750
+ "learning_rate": 7.65e-05,
751
+ "loss": 0.5467,
752
+ "step": 23500
753
+ },
754
+ {
755
+ "epoch": 55.16431924882629,
756
+ "eval_accuracy": 0.8485212684731438,
757
+ "eval_loss": 0.7348983287811279,
758
+ "eval_runtime": 137.1652,
759
+ "eval_samples_per_second": 128.91,
760
+ "eval_steps_per_second": 4.032,
761
+ "step": 23500
762
+ },
763
+ {
764
+ "epoch": 56.33802816901409,
765
+ "grad_norm": 2.046226978302002,
766
+ "learning_rate": 7.6e-05,
767
+ "loss": 0.5434,
768
+ "step": 24000
769
+ },
770
+ {
771
+ "epoch": 56.33802816901409,
772
+ "eval_accuracy": 0.8486974677705212,
773
+ "eval_loss": 0.736895740032196,
774
+ "eval_runtime": 136.9534,
775
+ "eval_samples_per_second": 129.11,
776
+ "eval_steps_per_second": 4.038,
777
+ "step": 24000
778
+ },
779
+ {
780
+ "epoch": 57.51173708920188,
781
+ "grad_norm": 1.9242944717407227,
782
+ "learning_rate": 7.55e-05,
783
+ "loss": 0.5378,
784
+ "step": 24500
785
+ },
786
+ {
787
+ "epoch": 57.51173708920188,
788
+ "eval_accuracy": 0.8492700287304648,
789
+ "eval_loss": 0.7279884815216064,
790
+ "eval_runtime": 136.7008,
791
+ "eval_samples_per_second": 129.348,
792
+ "eval_steps_per_second": 4.045,
793
+ "step": 24500
794
+ },
795
+ {
796
+ "epoch": 58.68544600938967,
797
+ "grad_norm": 1.9436827898025513,
798
+ "learning_rate": 7.500000000000001e-05,
799
+ "loss": 0.5382,
800
+ "step": 25000
801
+ },
802
+ {
803
+ "epoch": 58.68544600938967,
804
+ "eval_accuracy": 0.8496587662017004,
805
+ "eval_loss": 0.7233351469039917,
806
+ "eval_runtime": 136.9816,
807
+ "eval_samples_per_second": 129.083,
808
+ "eval_steps_per_second": 4.037,
809
+ "step": 25000
810
+ },
811
+ {
812
+ "epoch": 59.859154929577464,
813
+ "grad_norm": 2.442077159881592,
814
+ "learning_rate": 7.450000000000001e-05,
815
+ "loss": 0.5334,
816
+ "step": 25500
817
+ },
818
+ {
819
+ "epoch": 59.859154929577464,
820
+ "eval_accuracy": 0.8502614948790587,
821
+ "eval_loss": 0.7267663478851318,
822
+ "eval_runtime": 140.4178,
823
+ "eval_samples_per_second": 125.924,
824
+ "eval_steps_per_second": 3.938,
825
+ "step": 25500
826
+ },
827
+ {
828
+ "epoch": 61.032863849765256,
829
+ "grad_norm": 2.1809020042419434,
830
+ "learning_rate": 7.4e-05,
831
+ "loss": 0.5295,
832
+ "step": 26000
833
+ },
834
+ {
835
+ "epoch": 61.032863849765256,
836
+ "eval_accuracy": 0.8492560240081366,
837
+ "eval_loss": 0.7322823405265808,
838
+ "eval_runtime": 137.5974,
839
+ "eval_samples_per_second": 128.505,
840
+ "eval_steps_per_second": 4.019,
841
+ "step": 26000
842
+ },
843
+ {
844
+ "epoch": 62.20657276995305,
845
+ "grad_norm": 2.0946710109710693,
846
+ "learning_rate": 7.35e-05,
847
+ "loss": 0.5215,
848
+ "step": 26500
849
+ },
850
+ {
851
+ "epoch": 62.20657276995305,
852
+ "eval_accuracy": 0.8505152353921837,
853
+ "eval_loss": 0.721396803855896,
854
+ "eval_runtime": 136.7594,
855
+ "eval_samples_per_second": 129.293,
856
+ "eval_steps_per_second": 4.044,
857
+ "step": 26500
858
+ },
859
+ {
860
+ "epoch": 63.38028169014085,
861
+ "grad_norm": 2.1418216228485107,
862
+ "learning_rate": 7.3e-05,
863
+ "loss": 0.5188,
864
+ "step": 27000
865
+ },
866
+ {
867
+ "epoch": 63.38028169014085,
868
+ "eval_accuracy": 0.8503667583911142,
869
+ "eval_loss": 0.7287681698799133,
870
+ "eval_runtime": 136.8587,
871
+ "eval_samples_per_second": 129.199,
872
+ "eval_steps_per_second": 4.041,
873
+ "step": 27000
874
+ },
875
+ {
876
+ "epoch": 64.55399061032864,
877
+ "grad_norm": 2.0790982246398926,
878
+ "learning_rate": 7.25e-05,
879
+ "loss": 0.5168,
880
+ "step": 27500
881
+ },
882
+ {
883
+ "epoch": 64.55399061032864,
884
+ "eval_accuracy": 0.8514007297976166,
885
+ "eval_loss": 0.7309630513191223,
886
+ "eval_runtime": 136.1553,
887
+ "eval_samples_per_second": 129.866,
888
+ "eval_steps_per_second": 4.062,
889
+ "step": 27500
890
+ },
891
+ {
892
+ "epoch": 65.72769953051643,
893
+ "grad_norm": 2.085266351699829,
894
+ "learning_rate": 7.2e-05,
895
+ "loss": 0.5125,
896
+ "step": 28000
897
+ },
898
+ {
899
+ "epoch": 65.72769953051643,
900
+ "eval_accuracy": 0.8518668186024253,
901
+ "eval_loss": 0.7231945395469666,
902
+ "eval_runtime": 147.5131,
903
+ "eval_samples_per_second": 119.867,
904
+ "eval_steps_per_second": 3.749,
905
+ "step": 28000
906
+ },
907
+ {
908
+ "epoch": 66.90140845070422,
909
+ "grad_norm": 1.7428772449493408,
910
+ "learning_rate": 7.15e-05,
911
+ "loss": 0.5076,
912
+ "step": 28500
913
+ },
914
+ {
915
+ "epoch": 66.90140845070422,
916
+ "eval_accuracy": 0.8518548866516221,
917
+ "eval_loss": 0.726669430732727,
918
+ "eval_runtime": 139.3497,
919
+ "eval_samples_per_second": 126.889,
920
+ "eval_steps_per_second": 3.968,
921
+ "step": 28500
922
+ },
923
+ {
924
+ "epoch": 68.07511737089202,
925
+ "grad_norm": 1.9387340545654297,
926
+ "learning_rate": 7.1e-05,
927
+ "loss": 0.5018,
928
+ "step": 29000
929
+ },
930
+ {
931
+ "epoch": 68.07511737089202,
932
+ "eval_accuracy": 0.8516039425023846,
933
+ "eval_loss": 0.7305765151977539,
934
+ "eval_runtime": 136.8809,
935
+ "eval_samples_per_second": 129.178,
936
+ "eval_steps_per_second": 4.04,
937
+ "step": 29000
938
+ },
939
+ {
940
+ "epoch": 69.24882629107981,
941
+ "grad_norm": 2.090191125869751,
942
+ "learning_rate": 7.05e-05,
943
+ "loss": 0.5011,
944
+ "step": 29500
945
+ },
946
+ {
947
+ "epoch": 69.24882629107981,
948
+ "eval_accuracy": 0.8522613468602297,
949
+ "eval_loss": 0.723237931728363,
950
+ "eval_runtime": 137.4062,
951
+ "eval_samples_per_second": 128.684,
952
+ "eval_steps_per_second": 4.025,
953
+ "step": 29500
954
+ },
955
+ {
956
+ "epoch": 70.4225352112676,
957
+ "grad_norm": 1.8190377950668335,
958
+ "learning_rate": 7e-05,
959
+ "loss": 0.4986,
960
+ "step": 30000
961
+ },
962
+ {
963
+ "epoch": 70.4225352112676,
964
+ "eval_accuracy": 0.8533026826846422,
965
+ "eval_loss": 0.7156932353973389,
966
+ "eval_runtime": 136.1553,
967
+ "eval_samples_per_second": 129.866,
968
+ "eval_steps_per_second": 4.062,
969
+ "step": 30000
970
+ },
971
+ {
972
+ "epoch": 71.59624413145539,
973
+ "grad_norm": 2.007854700088501,
974
+ "learning_rate": 6.95e-05,
975
+ "loss": 0.4915,
976
+ "step": 30500
977
+ },
978
+ {
979
+ "epoch": 71.59624413145539,
980
+ "eval_accuracy": 0.8527300721103733,
981
+ "eval_loss": 0.7252740859985352,
982
+ "eval_runtime": 136.7888,
983
+ "eval_samples_per_second": 129.265,
984
+ "eval_steps_per_second": 4.043,
985
+ "step": 30500
986
+ },
987
+ {
988
+ "epoch": 72.7699530516432,
989
+ "grad_norm": 2.4322381019592285,
990
+ "learning_rate": 6.9e-05,
991
+ "loss": 0.4934,
992
+ "step": 31000
993
+ },
994
+ {
995
+ "epoch": 72.7699530516432,
996
+ "eval_accuracy": 0.8530242933183955,
997
+ "eval_loss": 0.7319638133049011,
998
+ "eval_runtime": 144.2929,
999
+ "eval_samples_per_second": 122.542,
1000
+ "eval_steps_per_second": 3.832,
1001
+ "step": 31000
1002
+ },
1003
+ {
1004
+ "epoch": 73.94366197183099,
1005
+ "grad_norm": 2.21516489982605,
1006
+ "learning_rate": 6.850000000000001e-05,
1007
+ "loss": 0.4895,
1008
+ "step": 31500
1009
+ },
1010
+ {
1011
+ "epoch": 73.94366197183099,
1012
+ "eval_accuracy": 0.8537945053849306,
1013
+ "eval_loss": 0.7180453538894653,
1014
+ "eval_runtime": 141.0891,
1015
+ "eval_samples_per_second": 125.325,
1016
+ "eval_steps_per_second": 3.92,
1017
+ "step": 31500
1018
+ },
1019
+ {
1020
+ "epoch": 75.11737089201878,
1021
+ "grad_norm": 2.0788252353668213,
1022
+ "learning_rate": 6.800000000000001e-05,
1023
+ "loss": 0.4822,
1024
+ "step": 32000
1025
+ },
1026
+ {
1027
+ "epoch": 75.11737089201878,
1028
+ "eval_accuracy": 0.8538218958075418,
1029
+ "eval_loss": 0.7125606536865234,
1030
+ "eval_runtime": 137.0269,
1031
+ "eval_samples_per_second": 129.04,
1032
+ "eval_steps_per_second": 4.036,
1033
+ "step": 32000
1034
+ },
1035
+ {
1036
+ "epoch": 76.29107981220658,
1037
+ "grad_norm": 2.038712978363037,
1038
+ "learning_rate": 6.750000000000001e-05,
1039
+ "loss": 0.4807,
1040
+ "step": 32500
1041
+ },
1042
+ {
1043
+ "epoch": 76.29107981220658,
1044
+ "eval_accuracy": 0.8543182928024404,
1045
+ "eval_loss": 0.7191519141197205,
1046
+ "eval_runtime": 136.1481,
1047
+ "eval_samples_per_second": 129.873,
1048
+ "eval_steps_per_second": 4.062,
1049
+ "step": 32500
1050
+ },
1051
+ {
1052
+ "epoch": 77.46478873239437,
1053
+ "grad_norm": 2.133082389831543,
1054
+ "learning_rate": 6.7e-05,
1055
+ "loss": 0.4743,
1056
+ "step": 33000
1057
+ },
1058
+ {
1059
+ "epoch": 77.46478873239437,
1060
+ "eval_accuracy": 0.8547410139669203,
1061
+ "eval_loss": 0.7057022452354431,
1062
+ "eval_runtime": 136.0992,
1063
+ "eval_samples_per_second": 129.92,
1064
+ "eval_steps_per_second": 4.063,
1065
+ "step": 33000
1066
+ },
1067
+ {
1068
+ "epoch": 78.63849765258216,
1069
+ "grad_norm": 1.9323476552963257,
1070
+ "learning_rate": 6.65e-05,
1071
+ "loss": 0.4749,
1072
+ "step": 33500
1073
+ },
1074
+ {
1075
+ "epoch": 78.63849765258216,
1076
+ "eval_accuracy": 0.8546438959333029,
1077
+ "eval_loss": 0.7180441617965698,
1078
+ "eval_runtime": 141.0328,
1079
+ "eval_samples_per_second": 125.375,
1080
+ "eval_steps_per_second": 3.921,
1081
+ "step": 33500
1082
+ },
1083
+ {
1084
+ "epoch": 79.81220657276995,
1085
+ "grad_norm": 2.1482994556427,
1086
+ "learning_rate": 6.6e-05,
1087
+ "loss": 0.4693,
1088
+ "step": 34000
1089
+ },
1090
+ {
1091
+ "epoch": 79.81220657276995,
1092
+ "eval_accuracy": 0.8554924128329217,
1093
+ "eval_loss": 0.7105869054794312,
1094
+ "eval_runtime": 137.5148,
1095
+ "eval_samples_per_second": 128.583,
1096
+ "eval_steps_per_second": 4.021,
1097
+ "step": 34000
1098
+ },
1099
+ {
1100
+ "epoch": 80.98591549295774,
1101
+ "grad_norm": 2.077798366546631,
1102
+ "learning_rate": 6.55e-05,
1103
+ "loss": 0.4675,
1104
+ "step": 34500
1105
+ },
1106
+ {
1107
+ "epoch": 80.98591549295774,
1108
+ "eval_accuracy": 0.8544003573962279,
1109
+ "eval_loss": 0.7216335535049438,
1110
+ "eval_runtime": 142.3955,
1111
+ "eval_samples_per_second": 124.175,
1112
+ "eval_steps_per_second": 3.884,
1113
+ "step": 34500
1114
+ },
1115
+ {
1116
+ "epoch": 82.15962441314554,
1117
+ "grad_norm": 2.2198646068573,
1118
+ "learning_rate": 6.500000000000001e-05,
1119
+ "loss": 0.4637,
1120
+ "step": 35000
1121
+ },
1122
+ {
1123
+ "epoch": 82.15962441314554,
1124
+ "eval_accuracy": 0.855125887761347,
1125
+ "eval_loss": 0.717957079410553,
1126
+ "eval_runtime": 137.0613,
1127
+ "eval_samples_per_second": 129.008,
1128
+ "eval_steps_per_second": 4.035,
1129
+ "step": 35000
1130
+ },
1131
+ {
1132
+ "epoch": 83.33333333333333,
1133
+ "grad_norm": 2.155299186706543,
1134
+ "learning_rate": 6.450000000000001e-05,
1135
+ "loss": 0.461,
1136
+ "step": 35500
1137
+ },
1138
+ {
1139
+ "epoch": 83.33333333333333,
1140
+ "eval_accuracy": 0.8554556348895517,
1141
+ "eval_loss": 0.7124961614608765,
1142
+ "eval_runtime": 145.4916,
1143
+ "eval_samples_per_second": 121.533,
1144
+ "eval_steps_per_second": 3.801,
1145
+ "step": 35500
1146
+ },
1147
+ {
1148
+ "epoch": 84.50704225352112,
1149
+ "grad_norm": 2.8300819396972656,
1150
+ "learning_rate": 6.400000000000001e-05,
1151
+ "loss": 0.4542,
1152
+ "step": 36000
1153
+ },
1154
+ {
1155
+ "epoch": 84.50704225352112,
1156
+ "eval_accuracy": 0.856205421066434,
1157
+ "eval_loss": 0.7241988182067871,
1158
+ "eval_runtime": 138.1126,
1159
+ "eval_samples_per_second": 128.026,
1160
+ "eval_steps_per_second": 4.004,
1161
+ "step": 36000
1162
+ },
1163
+ {
1164
+ "epoch": 85.68075117370893,
1165
+ "grad_norm": 2.4033689498901367,
1166
+ "learning_rate": 6.35e-05,
1167
+ "loss": 0.4564,
1168
+ "step": 36500
1169
+ },
1170
+ {
1171
+ "epoch": 85.68075117370893,
1172
+ "eval_accuracy": 0.8557880398860113,
1173
+ "eval_loss": 0.7188218832015991,
1174
+ "eval_runtime": 138.5628,
1175
+ "eval_samples_per_second": 127.61,
1176
+ "eval_steps_per_second": 3.991,
1177
+ "step": 36500
1178
+ },
1179
+ {
1180
+ "epoch": 86.85446009389672,
1181
+ "grad_norm": 2.104976177215576,
1182
+ "learning_rate": 6.3e-05,
1183
+ "loss": 0.4497,
1184
+ "step": 37000
1185
+ },
1186
+ {
1187
+ "epoch": 86.85446009389672,
1188
+ "eval_accuracy": 0.8571525482780049,
1189
+ "eval_loss": 0.718337893486023,
1190
+ "eval_runtime": 136.8784,
1191
+ "eval_samples_per_second": 129.18,
1192
+ "eval_steps_per_second": 4.04,
1193
+ "step": 37000
1194
+ },
1195
+ {
1196
+ "epoch": 88.02816901408451,
1197
+ "grad_norm": 1.9449199438095093,
1198
+ "learning_rate": 6.25e-05,
1199
+ "loss": 0.4485,
1200
+ "step": 37500
1201
+ },
1202
+ {
1203
+ "epoch": 88.02816901408451,
1204
+ "eval_accuracy": 0.8572300673528399,
1205
+ "eval_loss": 0.7067864537239075,
1206
+ "eval_runtime": 140.6771,
1207
+ "eval_samples_per_second": 125.692,
1208
+ "eval_steps_per_second": 3.931,
1209
+ "step": 37500
1210
+ },
1211
+ {
1212
+ "epoch": 89.2018779342723,
1213
+ "grad_norm": 2.0170960426330566,
1214
+ "learning_rate": 6.2e-05,
1215
+ "loss": 0.4486,
1216
+ "step": 38000
1217
+ },
1218
+ {
1219
+ "epoch": 89.2018779342723,
1220
+ "eval_accuracy": 0.8567210575027162,
1221
+ "eval_loss": 0.7134066820144653,
1222
+ "eval_runtime": 137.1729,
1223
+ "eval_samples_per_second": 128.903,
1224
+ "eval_steps_per_second": 4.031,
1225
+ "step": 38000
1226
+ },
1227
+ {
1228
+ "epoch": 90.3755868544601,
1229
+ "grad_norm": 2.048830032348633,
1230
+ "learning_rate": 6.15e-05,
1231
+ "loss": 0.4439,
1232
+ "step": 38500
1233
+ },
1234
+ {
1235
+ "epoch": 90.3755868544601,
1236
+ "eval_accuracy": 0.8563134552098389,
1237
+ "eval_loss": 0.7194843292236328,
1238
+ "eval_runtime": 136.9411,
1239
+ "eval_samples_per_second": 129.121,
1240
+ "eval_steps_per_second": 4.038,
1241
+ "step": 38500
1242
+ },
1243
+ {
1244
+ "epoch": 91.54929577464789,
1245
+ "grad_norm": 2.0259907245635986,
1246
+ "learning_rate": 6.1e-05,
1247
+ "loss": 0.4385,
1248
+ "step": 39000
1249
+ },
1250
+ {
1251
+ "epoch": 91.54929577464789,
1252
+ "eval_accuracy": 0.8580620221342309,
1253
+ "eval_loss": 0.7045323252677917,
1254
+ "eval_runtime": 137.122,
1255
+ "eval_samples_per_second": 128.951,
1256
+ "eval_steps_per_second": 4.033,
1257
+ "step": 39000
1258
+ }
1259
+ ],
1260
+ "logging_steps": 500,
1261
+ "max_steps": 100000,
1262
+ "num_input_tokens_seen": 0,
1263
+ "num_train_epochs": 235,
1264
+ "save_steps": 500,
1265
+ "stateful_callbacks": {
1266
+ "TrainerControl": {
1267
+ "args": {
1268
+ "should_epoch_stop": false,
1269
+ "should_evaluate": false,
1270
+ "should_log": false,
1271
+ "should_save": true,
1272
+ "should_training_stop": false
1273
+ },
1274
+ "attributes": {}
1275
+ }
1276
+ },
1277
+ "total_flos": 3.2882090840162304e+17,
1278
+ "train_batch_size": 16,
1279
+ "trial_name": null,
1280
+ "trial_params": null
1281
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b819e1b54572c6679927fdc0672812a1d6e0ce244f1eba59629d47671b090a7
3
+ size 5176
vocab.txt ADDED
The diff for this file is too large to render. See raw diff