Ba2han commited on
Commit
a2a30aa
·
verified ·
1 Parent(s): 34476bd

Training in progress, step 923, checkpoint

Browse files
last-checkpoint/chat_template.jinja ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {% for message in messages %}{{'<|im_start|>' + message['role'] + '
2
+ ' + message['content']}}{% if (loop.last and add_generation_prompt) or not loop.last %}{{ '<|im_end|>' + '
3
+ '}}{% endif %}{% endfor %}
4
+ {% if add_generation_prompt and messages[-1]['role'] != 'assistant' %}{{ '<|im_start|>assistant
5
+ ' }}{% endif %}
last-checkpoint/config.json ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3ForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 2,
8
+ "dtype": "bfloat16",
9
+ "eos_token_id": 60800,
10
+ "head_dim": 128,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 1792,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4608,
15
+ "layer_types": [
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention",
52
+ "full_attention",
53
+ "full_attention",
54
+ "full_attention",
55
+ "full_attention"
56
+ ],
57
+ "max_position_embeddings": 16384,
58
+ "max_window_layers": 40,
59
+ "model_name": "Ba2han/test_CPTl",
60
+ "model_type": "qwen3",
61
+ "num_attention_heads": 16,
62
+ "num_hidden_layers": 40,
63
+ "num_key_value_heads": 4,
64
+ "pad_token_id": 0,
65
+ "rms_norm_eps": 1e-08,
66
+ "rope_parameters": {
67
+ "rope_theta": 500000,
68
+ "rope_type": "default"
69
+ },
70
+ "sliding_window": null,
71
+ "tie_word_embeddings": true,
72
+ "transformers_version": "5.5.0",
73
+ "unsloth_version": "2026.8.1",
74
+ "use_cache": false,
75
+ "use_sliding_window": false,
76
+ "vocab_size": 60802
77
+ }
last-checkpoint/generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 2,
4
+ "eos_token_id": [
5
+ 60800
6
+ ],
7
+ "max_length": 16384,
8
+ "output_attentions": false,
9
+ "output_hidden_states": false,
10
+ "pad_token_id": 0,
11
+ "transformers_version": "5.5.0",
12
+ "use_cache": false
13
+ }
last-checkpoint/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c77759d81259dc06f53a2dbfa7ad69d22891e734e8ec5a3254551d4c4822d8fc
3
+ size 2934397032
last-checkpoint/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:915e0b11ba5554e2e413e901103f1dd53d7f46ad097e5cf8a8dfaee79850a55d
3
+ size 1470912696
last-checkpoint/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1d565802a8e26c4e8a31328752b7a7fdc186d9401aa008e65697d0ad8c22e33
3
+ size 14645
last-checkpoint/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:850ff5ca370d3bfed592befa8d3b8903813a6245956d1a9e1afe75be6155564b
3
+ size 9785
last-checkpoint/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
last-checkpoint/tokenizer_config.json ADDED
@@ -0,0 +1,2114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<bos>",
4
+ "clean_up_tokenization_spaces": true,
5
+ "eos_token": "<|im_end|>",
6
+ "extra_special_tokens": [
7
+ "<|im_start|>"
8
+ ],
9
+ "is_local": false,
10
+ "local_files_only": false,
11
+ "model_max_length": 1000000000000000019884624838656,
12
+ "pad_token": "<pad>",
13
+ "padding_side": "right",
14
+ "tokenizer_class": "TokenizersBackend",
15
+ "unk_token": "<unk>",
16
+ "added_tokens_decoder": {
17
+ "0": {
18
+ "content": "<pad>",
19
+ "single_word": false,
20
+ "lstrip": false,
21
+ "rstrip": false,
22
+ "normalized": false,
23
+ "special": true
24
+ },
25
+ "1": {
26
+ "content": "<unk>",
27
+ "single_word": false,
28
+ "lstrip": false,
29
+ "rstrip": false,
30
+ "normalized": false,
31
+ "special": true
32
+ },
33
+ "2": {
34
+ "content": "<bos>",
35
+ "single_word": false,
36
+ "lstrip": false,
37
+ "rstrip": false,
38
+ "normalized": false,
39
+ "special": true
40
+ },
41
+ "3": {
42
+ "content": "<eos>",
43
+ "single_word": false,
44
+ "lstrip": false,
45
+ "rstrip": false,
46
+ "normalized": false,
47
+ "special": true
48
+ },
49
+ "4": {
50
+ "content": "<0x00>",
51
+ "single_word": false,
52
+ "lstrip": false,
53
+ "rstrip": false,
54
+ "normalized": false,
55
+ "special": true
56
+ },
57
+ "5": {
58
+ "content": "<0x01>",
59
+ "single_word": false,
60
+ "lstrip": false,
61
+ "rstrip": false,
62
+ "normalized": false,
63
+ "special": true
64
+ },
65
+ "6": {
66
+ "content": "<0x02>",
67
+ "single_word": false,
68
+ "lstrip": false,
69
+ "rstrip": false,
70
+ "normalized": false,
71
+ "special": true
72
+ },
73
+ "7": {
74
+ "content": "<0x03>",
75
+ "single_word": false,
76
+ "lstrip": false,
77
+ "rstrip": false,
78
+ "normalized": false,
79
+ "special": true
80
+ },
81
+ "8": {
82
+ "content": "<0x04>",
83
+ "single_word": false,
84
+ "lstrip": false,
85
+ "rstrip": false,
86
+ "normalized": false,
87
+ "special": true
88
+ },
89
+ "9": {
90
+ "content": "<0x05>",
91
+ "single_word": false,
92
+ "lstrip": false,
93
+ "rstrip": false,
94
+ "normalized": false,
95
+ "special": true
96
+ },
97
+ "10": {
98
+ "content": "<0x06>",
99
+ "single_word": false,
100
+ "lstrip": false,
101
+ "rstrip": false,
102
+ "normalized": false,
103
+ "special": true
104
+ },
105
+ "11": {
106
+ "content": "<0x07>",
107
+ "single_word": false,
108
+ "lstrip": false,
109
+ "rstrip": false,
110
+ "normalized": false,
111
+ "special": true
112
+ },
113
+ "12": {
114
+ "content": "<0x08>",
115
+ "single_word": false,
116
+ "lstrip": false,
117
+ "rstrip": false,
118
+ "normalized": false,
119
+ "special": true
120
+ },
121
+ "13": {
122
+ "content": "<0x09>",
123
+ "single_word": false,
124
+ "lstrip": false,
125
+ "rstrip": false,
126
+ "normalized": false,
127
+ "special": true
128
+ },
129
+ "14": {
130
+ "content": "<0x0A>",
131
+ "single_word": false,
132
+ "lstrip": false,
133
+ "rstrip": false,
134
+ "normalized": false,
135
+ "special": true
136
+ },
137
+ "15": {
138
+ "content": "<0x0B>",
139
+ "single_word": false,
140
+ "lstrip": false,
141
+ "rstrip": false,
142
+ "normalized": false,
143
+ "special": true
144
+ },
145
+ "16": {
146
+ "content": "<0x0C>",
147
+ "single_word": false,
148
+ "lstrip": false,
149
+ "rstrip": false,
150
+ "normalized": false,
151
+ "special": true
152
+ },
153
+ "17": {
154
+ "content": "<0x0D>",
155
+ "single_word": false,
156
+ "lstrip": false,
157
+ "rstrip": false,
158
+ "normalized": false,
159
+ "special": true
160
+ },
161
+ "18": {
162
+ "content": "<0x0E>",
163
+ "single_word": false,
164
+ "lstrip": false,
165
+ "rstrip": false,
166
+ "normalized": false,
167
+ "special": true
168
+ },
169
+ "19": {
170
+ "content": "<0x0F>",
171
+ "single_word": false,
172
+ "lstrip": false,
173
+ "rstrip": false,
174
+ "normalized": false,
175
+ "special": true
176
+ },
177
+ "20": {
178
+ "content": "<0x10>",
179
+ "single_word": false,
180
+ "lstrip": false,
181
+ "rstrip": false,
182
+ "normalized": false,
183
+ "special": true
184
+ },
185
+ "21": {
186
+ "content": "<0x11>",
187
+ "single_word": false,
188
+ "lstrip": false,
189
+ "rstrip": false,
190
+ "normalized": false,
191
+ "special": true
192
+ },
193
+ "22": {
194
+ "content": "<0x12>",
195
+ "single_word": false,
196
+ "lstrip": false,
197
+ "rstrip": false,
198
+ "normalized": false,
199
+ "special": true
200
+ },
201
+ "23": {
202
+ "content": "<0x13>",
203
+ "single_word": false,
204
+ "lstrip": false,
205
+ "rstrip": false,
206
+ "normalized": false,
207
+ "special": true
208
+ },
209
+ "24": {
210
+ "content": "<0x14>",
211
+ "single_word": false,
212
+ "lstrip": false,
213
+ "rstrip": false,
214
+ "normalized": false,
215
+ "special": true
216
+ },
217
+ "25": {
218
+ "content": "<0x15>",
219
+ "single_word": false,
220
+ "lstrip": false,
221
+ "rstrip": false,
222
+ "normalized": false,
223
+ "special": true
224
+ },
225
+ "26": {
226
+ "content": "<0x16>",
227
+ "single_word": false,
228
+ "lstrip": false,
229
+ "rstrip": false,
230
+ "normalized": false,
231
+ "special": true
232
+ },
233
+ "27": {
234
+ "content": "<0x17>",
235
+ "single_word": false,
236
+ "lstrip": false,
237
+ "rstrip": false,
238
+ "normalized": false,
239
+ "special": true
240
+ },
241
+ "28": {
242
+ "content": "<0x18>",
243
+ "single_word": false,
244
+ "lstrip": false,
245
+ "rstrip": false,
246
+ "normalized": false,
247
+ "special": true
248
+ },
249
+ "29": {
250
+ "content": "<0x19>",
251
+ "single_word": false,
252
+ "lstrip": false,
253
+ "rstrip": false,
254
+ "normalized": false,
255
+ "special": true
256
+ },
257
+ "30": {
258
+ "content": "<0x1A>",
259
+ "single_word": false,
260
+ "lstrip": false,
261
+ "rstrip": false,
262
+ "normalized": false,
263
+ "special": true
264
+ },
265
+ "31": {
266
+ "content": "<0x1B>",
267
+ "single_word": false,
268
+ "lstrip": false,
269
+ "rstrip": false,
270
+ "normalized": false,
271
+ "special": true
272
+ },
273
+ "32": {
274
+ "content": "<0x1C>",
275
+ "single_word": false,
276
+ "lstrip": false,
277
+ "rstrip": false,
278
+ "normalized": false,
279
+ "special": true
280
+ },
281
+ "33": {
282
+ "content": "<0x1D>",
283
+ "single_word": false,
284
+ "lstrip": false,
285
+ "rstrip": false,
286
+ "normalized": false,
287
+ "special": true
288
+ },
289
+ "34": {
290
+ "content": "<0x1E>",
291
+ "single_word": false,
292
+ "lstrip": false,
293
+ "rstrip": false,
294
+ "normalized": false,
295
+ "special": true
296
+ },
297
+ "35": {
298
+ "content": "<0x1F>",
299
+ "single_word": false,
300
+ "lstrip": false,
301
+ "rstrip": false,
302
+ "normalized": false,
303
+ "special": true
304
+ },
305
+ "36": {
306
+ "content": "<0x20>",
307
+ "single_word": false,
308
+ "lstrip": false,
309
+ "rstrip": false,
310
+ "normalized": false,
311
+ "special": true
312
+ },
313
+ "37": {
314
+ "content": "<0x21>",
315
+ "single_word": false,
316
+ "lstrip": false,
317
+ "rstrip": false,
318
+ "normalized": false,
319
+ "special": true
320
+ },
321
+ "38": {
322
+ "content": "<0x22>",
323
+ "single_word": false,
324
+ "lstrip": false,
325
+ "rstrip": false,
326
+ "normalized": false,
327
+ "special": true
328
+ },
329
+ "39": {
330
+ "content": "<0x23>",
331
+ "single_word": false,
332
+ "lstrip": false,
333
+ "rstrip": false,
334
+ "normalized": false,
335
+ "special": true
336
+ },
337
+ "40": {
338
+ "content": "<0x24>",
339
+ "single_word": false,
340
+ "lstrip": false,
341
+ "rstrip": false,
342
+ "normalized": false,
343
+ "special": true
344
+ },
345
+ "41": {
346
+ "content": "<0x25>",
347
+ "single_word": false,
348
+ "lstrip": false,
349
+ "rstrip": false,
350
+ "normalized": false,
351
+ "special": true
352
+ },
353
+ "42": {
354
+ "content": "<0x26>",
355
+ "single_word": false,
356
+ "lstrip": false,
357
+ "rstrip": false,
358
+ "normalized": false,
359
+ "special": true
360
+ },
361
+ "43": {
362
+ "content": "<0x27>",
363
+ "single_word": false,
364
+ "lstrip": false,
365
+ "rstrip": false,
366
+ "normalized": false,
367
+ "special": true
368
+ },
369
+ "44": {
370
+ "content": "<0x28>",
371
+ "single_word": false,
372
+ "lstrip": false,
373
+ "rstrip": false,
374
+ "normalized": false,
375
+ "special": true
376
+ },
377
+ "45": {
378
+ "content": "<0x29>",
379
+ "single_word": false,
380
+ "lstrip": false,
381
+ "rstrip": false,
382
+ "normalized": false,
383
+ "special": true
384
+ },
385
+ "46": {
386
+ "content": "<0x2A>",
387
+ "single_word": false,
388
+ "lstrip": false,
389
+ "rstrip": false,
390
+ "normalized": false,
391
+ "special": true
392
+ },
393
+ "47": {
394
+ "content": "<0x2B>",
395
+ "single_word": false,
396
+ "lstrip": false,
397
+ "rstrip": false,
398
+ "normalized": false,
399
+ "special": true
400
+ },
401
+ "48": {
402
+ "content": "<0x2C>",
403
+ "single_word": false,
404
+ "lstrip": false,
405
+ "rstrip": false,
406
+ "normalized": false,
407
+ "special": true
408
+ },
409
+ "49": {
410
+ "content": "<0x2D>",
411
+ "single_word": false,
412
+ "lstrip": false,
413
+ "rstrip": false,
414
+ "normalized": false,
415
+ "special": true
416
+ },
417
+ "50": {
418
+ "content": "<0x2E>",
419
+ "single_word": false,
420
+ "lstrip": false,
421
+ "rstrip": false,
422
+ "normalized": false,
423
+ "special": true
424
+ },
425
+ "51": {
426
+ "content": "<0x2F>",
427
+ "single_word": false,
428
+ "lstrip": false,
429
+ "rstrip": false,
430
+ "normalized": false,
431
+ "special": true
432
+ },
433
+ "52": {
434
+ "content": "<0x30>",
435
+ "single_word": false,
436
+ "lstrip": false,
437
+ "rstrip": false,
438
+ "normalized": false,
439
+ "special": true
440
+ },
441
+ "53": {
442
+ "content": "<0x31>",
443
+ "single_word": false,
444
+ "lstrip": false,
445
+ "rstrip": false,
446
+ "normalized": false,
447
+ "special": true
448
+ },
449
+ "54": {
450
+ "content": "<0x32>",
451
+ "single_word": false,
452
+ "lstrip": false,
453
+ "rstrip": false,
454
+ "normalized": false,
455
+ "special": true
456
+ },
457
+ "55": {
458
+ "content": "<0x33>",
459
+ "single_word": false,
460
+ "lstrip": false,
461
+ "rstrip": false,
462
+ "normalized": false,
463
+ "special": true
464
+ },
465
+ "56": {
466
+ "content": "<0x34>",
467
+ "single_word": false,
468
+ "lstrip": false,
469
+ "rstrip": false,
470
+ "normalized": false,
471
+ "special": true
472
+ },
473
+ "57": {
474
+ "content": "<0x35>",
475
+ "single_word": false,
476
+ "lstrip": false,
477
+ "rstrip": false,
478
+ "normalized": false,
479
+ "special": true
480
+ },
481
+ "58": {
482
+ "content": "<0x36>",
483
+ "single_word": false,
484
+ "lstrip": false,
485
+ "rstrip": false,
486
+ "normalized": false,
487
+ "special": true
488
+ },
489
+ "59": {
490
+ "content": "<0x37>",
491
+ "single_word": false,
492
+ "lstrip": false,
493
+ "rstrip": false,
494
+ "normalized": false,
495
+ "special": true
496
+ },
497
+ "60": {
498
+ "content": "<0x38>",
499
+ "single_word": false,
500
+ "lstrip": false,
501
+ "rstrip": false,
502
+ "normalized": false,
503
+ "special": true
504
+ },
505
+ "61": {
506
+ "content": "<0x39>",
507
+ "single_word": false,
508
+ "lstrip": false,
509
+ "rstrip": false,
510
+ "normalized": false,
511
+ "special": true
512
+ },
513
+ "62": {
514
+ "content": "<0x3A>",
515
+ "single_word": false,
516
+ "lstrip": false,
517
+ "rstrip": false,
518
+ "normalized": false,
519
+ "special": true
520
+ },
521
+ "63": {
522
+ "content": "<0x3B>",
523
+ "single_word": false,
524
+ "lstrip": false,
525
+ "rstrip": false,
526
+ "normalized": false,
527
+ "special": true
528
+ },
529
+ "64": {
530
+ "content": "<0x3C>",
531
+ "single_word": false,
532
+ "lstrip": false,
533
+ "rstrip": false,
534
+ "normalized": false,
535
+ "special": true
536
+ },
537
+ "65": {
538
+ "content": "<0x3D>",
539
+ "single_word": false,
540
+ "lstrip": false,
541
+ "rstrip": false,
542
+ "normalized": false,
543
+ "special": true
544
+ },
545
+ "66": {
546
+ "content": "<0x3E>",
547
+ "single_word": false,
548
+ "lstrip": false,
549
+ "rstrip": false,
550
+ "normalized": false,
551
+ "special": true
552
+ },
553
+ "67": {
554
+ "content": "<0x3F>",
555
+ "single_word": false,
556
+ "lstrip": false,
557
+ "rstrip": false,
558
+ "normalized": false,
559
+ "special": true
560
+ },
561
+ "68": {
562
+ "content": "<0x40>",
563
+ "single_word": false,
564
+ "lstrip": false,
565
+ "rstrip": false,
566
+ "normalized": false,
567
+ "special": true
568
+ },
569
+ "69": {
570
+ "content": "<0x41>",
571
+ "single_word": false,
572
+ "lstrip": false,
573
+ "rstrip": false,
574
+ "normalized": false,
575
+ "special": true
576
+ },
577
+ "70": {
578
+ "content": "<0x42>",
579
+ "single_word": false,
580
+ "lstrip": false,
581
+ "rstrip": false,
582
+ "normalized": false,
583
+ "special": true
584
+ },
585
+ "71": {
586
+ "content": "<0x43>",
587
+ "single_word": false,
588
+ "lstrip": false,
589
+ "rstrip": false,
590
+ "normalized": false,
591
+ "special": true
592
+ },
593
+ "72": {
594
+ "content": "<0x44>",
595
+ "single_word": false,
596
+ "lstrip": false,
597
+ "rstrip": false,
598
+ "normalized": false,
599
+ "special": true
600
+ },
601
+ "73": {
602
+ "content": "<0x45>",
603
+ "single_word": false,
604
+ "lstrip": false,
605
+ "rstrip": false,
606
+ "normalized": false,
607
+ "special": true
608
+ },
609
+ "74": {
610
+ "content": "<0x46>",
611
+ "single_word": false,
612
+ "lstrip": false,
613
+ "rstrip": false,
614
+ "normalized": false,
615
+ "special": true
616
+ },
617
+ "75": {
618
+ "content": "<0x47>",
619
+ "single_word": false,
620
+ "lstrip": false,
621
+ "rstrip": false,
622
+ "normalized": false,
623
+ "special": true
624
+ },
625
+ "76": {
626
+ "content": "<0x48>",
627
+ "single_word": false,
628
+ "lstrip": false,
629
+ "rstrip": false,
630
+ "normalized": false,
631
+ "special": true
632
+ },
633
+ "77": {
634
+ "content": "<0x49>",
635
+ "single_word": false,
636
+ "lstrip": false,
637
+ "rstrip": false,
638
+ "normalized": false,
639
+ "special": true
640
+ },
641
+ "78": {
642
+ "content": "<0x4A>",
643
+ "single_word": false,
644
+ "lstrip": false,
645
+ "rstrip": false,
646
+ "normalized": false,
647
+ "special": true
648
+ },
649
+ "79": {
650
+ "content": "<0x4B>",
651
+ "single_word": false,
652
+ "lstrip": false,
653
+ "rstrip": false,
654
+ "normalized": false,
655
+ "special": true
656
+ },
657
+ "80": {
658
+ "content": "<0x4C>",
659
+ "single_word": false,
660
+ "lstrip": false,
661
+ "rstrip": false,
662
+ "normalized": false,
663
+ "special": true
664
+ },
665
+ "81": {
666
+ "content": "<0x4D>",
667
+ "single_word": false,
668
+ "lstrip": false,
669
+ "rstrip": false,
670
+ "normalized": false,
671
+ "special": true
672
+ },
673
+ "82": {
674
+ "content": "<0x4E>",
675
+ "single_word": false,
676
+ "lstrip": false,
677
+ "rstrip": false,
678
+ "normalized": false,
679
+ "special": true
680
+ },
681
+ "83": {
682
+ "content": "<0x4F>",
683
+ "single_word": false,
684
+ "lstrip": false,
685
+ "rstrip": false,
686
+ "normalized": false,
687
+ "special": true
688
+ },
689
+ "84": {
690
+ "content": "<0x50>",
691
+ "single_word": false,
692
+ "lstrip": false,
693
+ "rstrip": false,
694
+ "normalized": false,
695
+ "special": true
696
+ },
697
+ "85": {
698
+ "content": "<0x51>",
699
+ "single_word": false,
700
+ "lstrip": false,
701
+ "rstrip": false,
702
+ "normalized": false,
703
+ "special": true
704
+ },
705
+ "86": {
706
+ "content": "<0x52>",
707
+ "single_word": false,
708
+ "lstrip": false,
709
+ "rstrip": false,
710
+ "normalized": false,
711
+ "special": true
712
+ },
713
+ "87": {
714
+ "content": "<0x53>",
715
+ "single_word": false,
716
+ "lstrip": false,
717
+ "rstrip": false,
718
+ "normalized": false,
719
+ "special": true
720
+ },
721
+ "88": {
722
+ "content": "<0x54>",
723
+ "single_word": false,
724
+ "lstrip": false,
725
+ "rstrip": false,
726
+ "normalized": false,
727
+ "special": true
728
+ },
729
+ "89": {
730
+ "content": "<0x55>",
731
+ "single_word": false,
732
+ "lstrip": false,
733
+ "rstrip": false,
734
+ "normalized": false,
735
+ "special": true
736
+ },
737
+ "90": {
738
+ "content": "<0x56>",
739
+ "single_word": false,
740
+ "lstrip": false,
741
+ "rstrip": false,
742
+ "normalized": false,
743
+ "special": true
744
+ },
745
+ "91": {
746
+ "content": "<0x57>",
747
+ "single_word": false,
748
+ "lstrip": false,
749
+ "rstrip": false,
750
+ "normalized": false,
751
+ "special": true
752
+ },
753
+ "92": {
754
+ "content": "<0x58>",
755
+ "single_word": false,
756
+ "lstrip": false,
757
+ "rstrip": false,
758
+ "normalized": false,
759
+ "special": true
760
+ },
761
+ "93": {
762
+ "content": "<0x59>",
763
+ "single_word": false,
764
+ "lstrip": false,
765
+ "rstrip": false,
766
+ "normalized": false,
767
+ "special": true
768
+ },
769
+ "94": {
770
+ "content": "<0x5A>",
771
+ "single_word": false,
772
+ "lstrip": false,
773
+ "rstrip": false,
774
+ "normalized": false,
775
+ "special": true
776
+ },
777
+ "95": {
778
+ "content": "<0x5B>",
779
+ "single_word": false,
780
+ "lstrip": false,
781
+ "rstrip": false,
782
+ "normalized": false,
783
+ "special": true
784
+ },
785
+ "96": {
786
+ "content": "<0x5C>",
787
+ "single_word": false,
788
+ "lstrip": false,
789
+ "rstrip": false,
790
+ "normalized": false,
791
+ "special": true
792
+ },
793
+ "97": {
794
+ "content": "<0x5D>",
795
+ "single_word": false,
796
+ "lstrip": false,
797
+ "rstrip": false,
798
+ "normalized": false,
799
+ "special": true
800
+ },
801
+ "98": {
802
+ "content": "<0x5E>",
803
+ "single_word": false,
804
+ "lstrip": false,
805
+ "rstrip": false,
806
+ "normalized": false,
807
+ "special": true
808
+ },
809
+ "99": {
810
+ "content": "<0x5F>",
811
+ "single_word": false,
812
+ "lstrip": false,
813
+ "rstrip": false,
814
+ "normalized": false,
815
+ "special": true
816
+ },
817
+ "100": {
818
+ "content": "<0x60>",
819
+ "single_word": false,
820
+ "lstrip": false,
821
+ "rstrip": false,
822
+ "normalized": false,
823
+ "special": true
824
+ },
825
+ "101": {
826
+ "content": "<0x61>",
827
+ "single_word": false,
828
+ "lstrip": false,
829
+ "rstrip": false,
830
+ "normalized": false,
831
+ "special": true
832
+ },
833
+ "102": {
834
+ "content": "<0x62>",
835
+ "single_word": false,
836
+ "lstrip": false,
837
+ "rstrip": false,
838
+ "normalized": false,
839
+ "special": true
840
+ },
841
+ "103": {
842
+ "content": "<0x63>",
843
+ "single_word": false,
844
+ "lstrip": false,
845
+ "rstrip": false,
846
+ "normalized": false,
847
+ "special": true
848
+ },
849
+ "104": {
850
+ "content": "<0x64>",
851
+ "single_word": false,
852
+ "lstrip": false,
853
+ "rstrip": false,
854
+ "normalized": false,
855
+ "special": true
856
+ },
857
+ "105": {
858
+ "content": "<0x65>",
859
+ "single_word": false,
860
+ "lstrip": false,
861
+ "rstrip": false,
862
+ "normalized": false,
863
+ "special": true
864
+ },
865
+ "106": {
866
+ "content": "<0x66>",
867
+ "single_word": false,
868
+ "lstrip": false,
869
+ "rstrip": false,
870
+ "normalized": false,
871
+ "special": true
872
+ },
873
+ "107": {
874
+ "content": "<0x67>",
875
+ "single_word": false,
876
+ "lstrip": false,
877
+ "rstrip": false,
878
+ "normalized": false,
879
+ "special": true
880
+ },
881
+ "108": {
882
+ "content": "<0x68>",
883
+ "single_word": false,
884
+ "lstrip": false,
885
+ "rstrip": false,
886
+ "normalized": false,
887
+ "special": true
888
+ },
889
+ "109": {
890
+ "content": "<0x69>",
891
+ "single_word": false,
892
+ "lstrip": false,
893
+ "rstrip": false,
894
+ "normalized": false,
895
+ "special": true
896
+ },
897
+ "110": {
898
+ "content": "<0x6A>",
899
+ "single_word": false,
900
+ "lstrip": false,
901
+ "rstrip": false,
902
+ "normalized": false,
903
+ "special": true
904
+ },
905
+ "111": {
906
+ "content": "<0x6B>",
907
+ "single_word": false,
908
+ "lstrip": false,
909
+ "rstrip": false,
910
+ "normalized": false,
911
+ "special": true
912
+ },
913
+ "112": {
914
+ "content": "<0x6C>",
915
+ "single_word": false,
916
+ "lstrip": false,
917
+ "rstrip": false,
918
+ "normalized": false,
919
+ "special": true
920
+ },
921
+ "113": {
922
+ "content": "<0x6D>",
923
+ "single_word": false,
924
+ "lstrip": false,
925
+ "rstrip": false,
926
+ "normalized": false,
927
+ "special": true
928
+ },
929
+ "114": {
930
+ "content": "<0x6E>",
931
+ "single_word": false,
932
+ "lstrip": false,
933
+ "rstrip": false,
934
+ "normalized": false,
935
+ "special": true
936
+ },
937
+ "115": {
938
+ "content": "<0x6F>",
939
+ "single_word": false,
940
+ "lstrip": false,
941
+ "rstrip": false,
942
+ "normalized": false,
943
+ "special": true
944
+ },
945
+ "116": {
946
+ "content": "<0x70>",
947
+ "single_word": false,
948
+ "lstrip": false,
949
+ "rstrip": false,
950
+ "normalized": false,
951
+ "special": true
952
+ },
953
+ "117": {
954
+ "content": "<0x71>",
955
+ "single_word": false,
956
+ "lstrip": false,
957
+ "rstrip": false,
958
+ "normalized": false,
959
+ "special": true
960
+ },
961
+ "118": {
962
+ "content": "<0x72>",
963
+ "single_word": false,
964
+ "lstrip": false,
965
+ "rstrip": false,
966
+ "normalized": false,
967
+ "special": true
968
+ },
969
+ "119": {
970
+ "content": "<0x73>",
971
+ "single_word": false,
972
+ "lstrip": false,
973
+ "rstrip": false,
974
+ "normalized": false,
975
+ "special": true
976
+ },
977
+ "120": {
978
+ "content": "<0x74>",
979
+ "single_word": false,
980
+ "lstrip": false,
981
+ "rstrip": false,
982
+ "normalized": false,
983
+ "special": true
984
+ },
985
+ "121": {
986
+ "content": "<0x75>",
987
+ "single_word": false,
988
+ "lstrip": false,
989
+ "rstrip": false,
990
+ "normalized": false,
991
+ "special": true
992
+ },
993
+ "122": {
994
+ "content": "<0x76>",
995
+ "single_word": false,
996
+ "lstrip": false,
997
+ "rstrip": false,
998
+ "normalized": false,
999
+ "special": true
1000
+ },
1001
+ "123": {
1002
+ "content": "<0x77>",
1003
+ "single_word": false,
1004
+ "lstrip": false,
1005
+ "rstrip": false,
1006
+ "normalized": false,
1007
+ "special": true
1008
+ },
1009
+ "124": {
1010
+ "content": "<0x78>",
1011
+ "single_word": false,
1012
+ "lstrip": false,
1013
+ "rstrip": false,
1014
+ "normalized": false,
1015
+ "special": true
1016
+ },
1017
+ "125": {
1018
+ "content": "<0x79>",
1019
+ "single_word": false,
1020
+ "lstrip": false,
1021
+ "rstrip": false,
1022
+ "normalized": false,
1023
+ "special": true
1024
+ },
1025
+ "126": {
1026
+ "content": "<0x7A>",
1027
+ "single_word": false,
1028
+ "lstrip": false,
1029
+ "rstrip": false,
1030
+ "normalized": false,
1031
+ "special": true
1032
+ },
1033
+ "127": {
1034
+ "content": "<0x7B>",
1035
+ "single_word": false,
1036
+ "lstrip": false,
1037
+ "rstrip": false,
1038
+ "normalized": false,
1039
+ "special": true
1040
+ },
1041
+ "128": {
1042
+ "content": "<0x7C>",
1043
+ "single_word": false,
1044
+ "lstrip": false,
1045
+ "rstrip": false,
1046
+ "normalized": false,
1047
+ "special": true
1048
+ },
1049
+ "129": {
1050
+ "content": "<0x7D>",
1051
+ "single_word": false,
1052
+ "lstrip": false,
1053
+ "rstrip": false,
1054
+ "normalized": false,
1055
+ "special": true
1056
+ },
1057
+ "130": {
1058
+ "content": "<0x7E>",
1059
+ "single_word": false,
1060
+ "lstrip": false,
1061
+ "rstrip": false,
1062
+ "normalized": false,
1063
+ "special": true
1064
+ },
1065
+ "131": {
1066
+ "content": "<0x7F>",
1067
+ "single_word": false,
1068
+ "lstrip": false,
1069
+ "rstrip": false,
1070
+ "normalized": false,
1071
+ "special": true
1072
+ },
1073
+ "132": {
1074
+ "content": "<0x80>",
1075
+ "single_word": false,
1076
+ "lstrip": false,
1077
+ "rstrip": false,
1078
+ "normalized": false,
1079
+ "special": true
1080
+ },
1081
+ "133": {
1082
+ "content": "<0x81>",
1083
+ "single_word": false,
1084
+ "lstrip": false,
1085
+ "rstrip": false,
1086
+ "normalized": false,
1087
+ "special": true
1088
+ },
1089
+ "134": {
1090
+ "content": "<0x82>",
1091
+ "single_word": false,
1092
+ "lstrip": false,
1093
+ "rstrip": false,
1094
+ "normalized": false,
1095
+ "special": true
1096
+ },
1097
+ "135": {
1098
+ "content": "<0x83>",
1099
+ "single_word": false,
1100
+ "lstrip": false,
1101
+ "rstrip": false,
1102
+ "normalized": false,
1103
+ "special": true
1104
+ },
1105
+ "136": {
1106
+ "content": "<0x84>",
1107
+ "single_word": false,
1108
+ "lstrip": false,
1109
+ "rstrip": false,
1110
+ "normalized": false,
1111
+ "special": true
1112
+ },
1113
+ "137": {
1114
+ "content": "<0x85>",
1115
+ "single_word": false,
1116
+ "lstrip": false,
1117
+ "rstrip": false,
1118
+ "normalized": false,
1119
+ "special": true
1120
+ },
1121
+ "138": {
1122
+ "content": "<0x86>",
1123
+ "single_word": false,
1124
+ "lstrip": false,
1125
+ "rstrip": false,
1126
+ "normalized": false,
1127
+ "special": true
1128
+ },
1129
+ "139": {
1130
+ "content": "<0x87>",
1131
+ "single_word": false,
1132
+ "lstrip": false,
1133
+ "rstrip": false,
1134
+ "normalized": false,
1135
+ "special": true
1136
+ },
1137
+ "140": {
1138
+ "content": "<0x88>",
1139
+ "single_word": false,
1140
+ "lstrip": false,
1141
+ "rstrip": false,
1142
+ "normalized": false,
1143
+ "special": true
1144
+ },
1145
+ "141": {
1146
+ "content": "<0x89>",
1147
+ "single_word": false,
1148
+ "lstrip": false,
1149
+ "rstrip": false,
1150
+ "normalized": false,
1151
+ "special": true
1152
+ },
1153
+ "142": {
1154
+ "content": "<0x8A>",
1155
+ "single_word": false,
1156
+ "lstrip": false,
1157
+ "rstrip": false,
1158
+ "normalized": false,
1159
+ "special": true
1160
+ },
1161
+ "143": {
1162
+ "content": "<0x8B>",
1163
+ "single_word": false,
1164
+ "lstrip": false,
1165
+ "rstrip": false,
1166
+ "normalized": false,
1167
+ "special": true
1168
+ },
1169
+ "144": {
1170
+ "content": "<0x8C>",
1171
+ "single_word": false,
1172
+ "lstrip": false,
1173
+ "rstrip": false,
1174
+ "normalized": false,
1175
+ "special": true
1176
+ },
1177
+ "145": {
1178
+ "content": "<0x8D>",
1179
+ "single_word": false,
1180
+ "lstrip": false,
1181
+ "rstrip": false,
1182
+ "normalized": false,
1183
+ "special": true
1184
+ },
1185
+ "146": {
1186
+ "content": "<0x8E>",
1187
+ "single_word": false,
1188
+ "lstrip": false,
1189
+ "rstrip": false,
1190
+ "normalized": false,
1191
+ "special": true
1192
+ },
1193
+ "147": {
1194
+ "content": "<0x8F>",
1195
+ "single_word": false,
1196
+ "lstrip": false,
1197
+ "rstrip": false,
1198
+ "normalized": false,
1199
+ "special": true
1200
+ },
1201
+ "148": {
1202
+ "content": "<0x90>",
1203
+ "single_word": false,
1204
+ "lstrip": false,
1205
+ "rstrip": false,
1206
+ "normalized": false,
1207
+ "special": true
1208
+ },
1209
+ "149": {
1210
+ "content": "<0x91>",
1211
+ "single_word": false,
1212
+ "lstrip": false,
1213
+ "rstrip": false,
1214
+ "normalized": false,
1215
+ "special": true
1216
+ },
1217
+ "150": {
1218
+ "content": "<0x92>",
1219
+ "single_word": false,
1220
+ "lstrip": false,
1221
+ "rstrip": false,
1222
+ "normalized": false,
1223
+ "special": true
1224
+ },
1225
+ "151": {
1226
+ "content": "<0x93>",
1227
+ "single_word": false,
1228
+ "lstrip": false,
1229
+ "rstrip": false,
1230
+ "normalized": false,
1231
+ "special": true
1232
+ },
1233
+ "152": {
1234
+ "content": "<0x94>",
1235
+ "single_word": false,
1236
+ "lstrip": false,
1237
+ "rstrip": false,
1238
+ "normalized": false,
1239
+ "special": true
1240
+ },
1241
+ "153": {
1242
+ "content": "<0x95>",
1243
+ "single_word": false,
1244
+ "lstrip": false,
1245
+ "rstrip": false,
1246
+ "normalized": false,
1247
+ "special": true
1248
+ },
1249
+ "154": {
1250
+ "content": "<0x96>",
1251
+ "single_word": false,
1252
+ "lstrip": false,
1253
+ "rstrip": false,
1254
+ "normalized": false,
1255
+ "special": true
1256
+ },
1257
+ "155": {
1258
+ "content": "<0x97>",
1259
+ "single_word": false,
1260
+ "lstrip": false,
1261
+ "rstrip": false,
1262
+ "normalized": false,
1263
+ "special": true
1264
+ },
1265
+ "156": {
1266
+ "content": "<0x98>",
1267
+ "single_word": false,
1268
+ "lstrip": false,
1269
+ "rstrip": false,
1270
+ "normalized": false,
1271
+ "special": true
1272
+ },
1273
+ "157": {
1274
+ "content": "<0x99>",
1275
+ "single_word": false,
1276
+ "lstrip": false,
1277
+ "rstrip": false,
1278
+ "normalized": false,
1279
+ "special": true
1280
+ },
1281
+ "158": {
1282
+ "content": "<0x9A>",
1283
+ "single_word": false,
1284
+ "lstrip": false,
1285
+ "rstrip": false,
1286
+ "normalized": false,
1287
+ "special": true
1288
+ },
1289
+ "159": {
1290
+ "content": "<0x9B>",
1291
+ "single_word": false,
1292
+ "lstrip": false,
1293
+ "rstrip": false,
1294
+ "normalized": false,
1295
+ "special": true
1296
+ },
1297
+ "160": {
1298
+ "content": "<0x9C>",
1299
+ "single_word": false,
1300
+ "lstrip": false,
1301
+ "rstrip": false,
1302
+ "normalized": false,
1303
+ "special": true
1304
+ },
1305
+ "161": {
1306
+ "content": "<0x9D>",
1307
+ "single_word": false,
1308
+ "lstrip": false,
1309
+ "rstrip": false,
1310
+ "normalized": false,
1311
+ "special": true
1312
+ },
1313
+ "162": {
1314
+ "content": "<0x9E>",
1315
+ "single_word": false,
1316
+ "lstrip": false,
1317
+ "rstrip": false,
1318
+ "normalized": false,
1319
+ "special": true
1320
+ },
1321
+ "163": {
1322
+ "content": "<0x9F>",
1323
+ "single_word": false,
1324
+ "lstrip": false,
1325
+ "rstrip": false,
1326
+ "normalized": false,
1327
+ "special": true
1328
+ },
1329
+ "164": {
1330
+ "content": "<0xA0>",
1331
+ "single_word": false,
1332
+ "lstrip": false,
1333
+ "rstrip": false,
1334
+ "normalized": false,
1335
+ "special": true
1336
+ },
1337
+ "165": {
1338
+ "content": "<0xA1>",
1339
+ "single_word": false,
1340
+ "lstrip": false,
1341
+ "rstrip": false,
1342
+ "normalized": false,
1343
+ "special": true
1344
+ },
1345
+ "166": {
1346
+ "content": "<0xA2>",
1347
+ "single_word": false,
1348
+ "lstrip": false,
1349
+ "rstrip": false,
1350
+ "normalized": false,
1351
+ "special": true
1352
+ },
1353
+ "167": {
1354
+ "content": "<0xA3>",
1355
+ "single_word": false,
1356
+ "lstrip": false,
1357
+ "rstrip": false,
1358
+ "normalized": false,
1359
+ "special": true
1360
+ },
1361
+ "168": {
1362
+ "content": "<0xA4>",
1363
+ "single_word": false,
1364
+ "lstrip": false,
1365
+ "rstrip": false,
1366
+ "normalized": false,
1367
+ "special": true
1368
+ },
1369
+ "169": {
1370
+ "content": "<0xA5>",
1371
+ "single_word": false,
1372
+ "lstrip": false,
1373
+ "rstrip": false,
1374
+ "normalized": false,
1375
+ "special": true
1376
+ },
1377
+ "170": {
1378
+ "content": "<0xA6>",
1379
+ "single_word": false,
1380
+ "lstrip": false,
1381
+ "rstrip": false,
1382
+ "normalized": false,
1383
+ "special": true
1384
+ },
1385
+ "171": {
1386
+ "content": "<0xA7>",
1387
+ "single_word": false,
1388
+ "lstrip": false,
1389
+ "rstrip": false,
1390
+ "normalized": false,
1391
+ "special": true
1392
+ },
1393
+ "172": {
1394
+ "content": "<0xA8>",
1395
+ "single_word": false,
1396
+ "lstrip": false,
1397
+ "rstrip": false,
1398
+ "normalized": false,
1399
+ "special": true
1400
+ },
1401
+ "173": {
1402
+ "content": "<0xA9>",
1403
+ "single_word": false,
1404
+ "lstrip": false,
1405
+ "rstrip": false,
1406
+ "normalized": false,
1407
+ "special": true
1408
+ },
1409
+ "174": {
1410
+ "content": "<0xAA>",
1411
+ "single_word": false,
1412
+ "lstrip": false,
1413
+ "rstrip": false,
1414
+ "normalized": false,
1415
+ "special": true
1416
+ },
1417
+ "175": {
1418
+ "content": "<0xAB>",
1419
+ "single_word": false,
1420
+ "lstrip": false,
1421
+ "rstrip": false,
1422
+ "normalized": false,
1423
+ "special": true
1424
+ },
1425
+ "176": {
1426
+ "content": "<0xAC>",
1427
+ "single_word": false,
1428
+ "lstrip": false,
1429
+ "rstrip": false,
1430
+ "normalized": false,
1431
+ "special": true
1432
+ },
1433
+ "177": {
1434
+ "content": "<0xAD>",
1435
+ "single_word": false,
1436
+ "lstrip": false,
1437
+ "rstrip": false,
1438
+ "normalized": false,
1439
+ "special": true
1440
+ },
1441
+ "178": {
1442
+ "content": "<0xAE>",
1443
+ "single_word": false,
1444
+ "lstrip": false,
1445
+ "rstrip": false,
1446
+ "normalized": false,
1447
+ "special": true
1448
+ },
1449
+ "179": {
1450
+ "content": "<0xAF>",
1451
+ "single_word": false,
1452
+ "lstrip": false,
1453
+ "rstrip": false,
1454
+ "normalized": false,
1455
+ "special": true
1456
+ },
1457
+ "180": {
1458
+ "content": "<0xB0>",
1459
+ "single_word": false,
1460
+ "lstrip": false,
1461
+ "rstrip": false,
1462
+ "normalized": false,
1463
+ "special": true
1464
+ },
1465
+ "181": {
1466
+ "content": "<0xB1>",
1467
+ "single_word": false,
1468
+ "lstrip": false,
1469
+ "rstrip": false,
1470
+ "normalized": false,
1471
+ "special": true
1472
+ },
1473
+ "182": {
1474
+ "content": "<0xB2>",
1475
+ "single_word": false,
1476
+ "lstrip": false,
1477
+ "rstrip": false,
1478
+ "normalized": false,
1479
+ "special": true
1480
+ },
1481
+ "183": {
1482
+ "content": "<0xB3>",
1483
+ "single_word": false,
1484
+ "lstrip": false,
1485
+ "rstrip": false,
1486
+ "normalized": false,
1487
+ "special": true
1488
+ },
1489
+ "184": {
1490
+ "content": "<0xB4>",
1491
+ "single_word": false,
1492
+ "lstrip": false,
1493
+ "rstrip": false,
1494
+ "normalized": false,
1495
+ "special": true
1496
+ },
1497
+ "185": {
1498
+ "content": "<0xB5>",
1499
+ "single_word": false,
1500
+ "lstrip": false,
1501
+ "rstrip": false,
1502
+ "normalized": false,
1503
+ "special": true
1504
+ },
1505
+ "186": {
1506
+ "content": "<0xB6>",
1507
+ "single_word": false,
1508
+ "lstrip": false,
1509
+ "rstrip": false,
1510
+ "normalized": false,
1511
+ "special": true
1512
+ },
1513
+ "187": {
1514
+ "content": "<0xB7>",
1515
+ "single_word": false,
1516
+ "lstrip": false,
1517
+ "rstrip": false,
1518
+ "normalized": false,
1519
+ "special": true
1520
+ },
1521
+ "188": {
1522
+ "content": "<0xB8>",
1523
+ "single_word": false,
1524
+ "lstrip": false,
1525
+ "rstrip": false,
1526
+ "normalized": false,
1527
+ "special": true
1528
+ },
1529
+ "189": {
1530
+ "content": "<0xB9>",
1531
+ "single_word": false,
1532
+ "lstrip": false,
1533
+ "rstrip": false,
1534
+ "normalized": false,
1535
+ "special": true
1536
+ },
1537
+ "190": {
1538
+ "content": "<0xBA>",
1539
+ "single_word": false,
1540
+ "lstrip": false,
1541
+ "rstrip": false,
1542
+ "normalized": false,
1543
+ "special": true
1544
+ },
1545
+ "191": {
1546
+ "content": "<0xBB>",
1547
+ "single_word": false,
1548
+ "lstrip": false,
1549
+ "rstrip": false,
1550
+ "normalized": false,
1551
+ "special": true
1552
+ },
1553
+ "192": {
1554
+ "content": "<0xBC>",
1555
+ "single_word": false,
1556
+ "lstrip": false,
1557
+ "rstrip": false,
1558
+ "normalized": false,
1559
+ "special": true
1560
+ },
1561
+ "193": {
1562
+ "content": "<0xBD>",
1563
+ "single_word": false,
1564
+ "lstrip": false,
1565
+ "rstrip": false,
1566
+ "normalized": false,
1567
+ "special": true
1568
+ },
1569
+ "194": {
1570
+ "content": "<0xBE>",
1571
+ "single_word": false,
1572
+ "lstrip": false,
1573
+ "rstrip": false,
1574
+ "normalized": false,
1575
+ "special": true
1576
+ },
1577
+ "195": {
1578
+ "content": "<0xBF>",
1579
+ "single_word": false,
1580
+ "lstrip": false,
1581
+ "rstrip": false,
1582
+ "normalized": false,
1583
+ "special": true
1584
+ },
1585
+ "196": {
1586
+ "content": "<0xC0>",
1587
+ "single_word": false,
1588
+ "lstrip": false,
1589
+ "rstrip": false,
1590
+ "normalized": false,
1591
+ "special": true
1592
+ },
1593
+ "197": {
1594
+ "content": "<0xC1>",
1595
+ "single_word": false,
1596
+ "lstrip": false,
1597
+ "rstrip": false,
1598
+ "normalized": false,
1599
+ "special": true
1600
+ },
1601
+ "198": {
1602
+ "content": "<0xC2>",
1603
+ "single_word": false,
1604
+ "lstrip": false,
1605
+ "rstrip": false,
1606
+ "normalized": false,
1607
+ "special": true
1608
+ },
1609
+ "199": {
1610
+ "content": "<0xC3>",
1611
+ "single_word": false,
1612
+ "lstrip": false,
1613
+ "rstrip": false,
1614
+ "normalized": false,
1615
+ "special": true
1616
+ },
1617
+ "200": {
1618
+ "content": "<0xC4>",
1619
+ "single_word": false,
1620
+ "lstrip": false,
1621
+ "rstrip": false,
1622
+ "normalized": false,
1623
+ "special": true
1624
+ },
1625
+ "201": {
1626
+ "content": "<0xC5>",
1627
+ "single_word": false,
1628
+ "lstrip": false,
1629
+ "rstrip": false,
1630
+ "normalized": false,
1631
+ "special": true
1632
+ },
1633
+ "202": {
1634
+ "content": "<0xC6>",
1635
+ "single_word": false,
1636
+ "lstrip": false,
1637
+ "rstrip": false,
1638
+ "normalized": false,
1639
+ "special": true
1640
+ },
1641
+ "203": {
1642
+ "content": "<0xC7>",
1643
+ "single_word": false,
1644
+ "lstrip": false,
1645
+ "rstrip": false,
1646
+ "normalized": false,
1647
+ "special": true
1648
+ },
1649
+ "204": {
1650
+ "content": "<0xC8>",
1651
+ "single_word": false,
1652
+ "lstrip": false,
1653
+ "rstrip": false,
1654
+ "normalized": false,
1655
+ "special": true
1656
+ },
1657
+ "205": {
1658
+ "content": "<0xC9>",
1659
+ "single_word": false,
1660
+ "lstrip": false,
1661
+ "rstrip": false,
1662
+ "normalized": false,
1663
+ "special": true
1664
+ },
1665
+ "206": {
1666
+ "content": "<0xCA>",
1667
+ "single_word": false,
1668
+ "lstrip": false,
1669
+ "rstrip": false,
1670
+ "normalized": false,
1671
+ "special": true
1672
+ },
1673
+ "207": {
1674
+ "content": "<0xCB>",
1675
+ "single_word": false,
1676
+ "lstrip": false,
1677
+ "rstrip": false,
1678
+ "normalized": false,
1679
+ "special": true
1680
+ },
1681
+ "208": {
1682
+ "content": "<0xCC>",
1683
+ "single_word": false,
1684
+ "lstrip": false,
1685
+ "rstrip": false,
1686
+ "normalized": false,
1687
+ "special": true
1688
+ },
1689
+ "209": {
1690
+ "content": "<0xCD>",
1691
+ "single_word": false,
1692
+ "lstrip": false,
1693
+ "rstrip": false,
1694
+ "normalized": false,
1695
+ "special": true
1696
+ },
1697
+ "210": {
1698
+ "content": "<0xCE>",
1699
+ "single_word": false,
1700
+ "lstrip": false,
1701
+ "rstrip": false,
1702
+ "normalized": false,
1703
+ "special": true
1704
+ },
1705
+ "211": {
1706
+ "content": "<0xCF>",
1707
+ "single_word": false,
1708
+ "lstrip": false,
1709
+ "rstrip": false,
1710
+ "normalized": false,
1711
+ "special": true
1712
+ },
1713
+ "212": {
1714
+ "content": "<0xD0>",
1715
+ "single_word": false,
1716
+ "lstrip": false,
1717
+ "rstrip": false,
1718
+ "normalized": false,
1719
+ "special": true
1720
+ },
1721
+ "213": {
1722
+ "content": "<0xD1>",
1723
+ "single_word": false,
1724
+ "lstrip": false,
1725
+ "rstrip": false,
1726
+ "normalized": false,
1727
+ "special": true
1728
+ },
1729
+ "214": {
1730
+ "content": "<0xD2>",
1731
+ "single_word": false,
1732
+ "lstrip": false,
1733
+ "rstrip": false,
1734
+ "normalized": false,
1735
+ "special": true
1736
+ },
1737
+ "215": {
1738
+ "content": "<0xD3>",
1739
+ "single_word": false,
1740
+ "lstrip": false,
1741
+ "rstrip": false,
1742
+ "normalized": false,
1743
+ "special": true
1744
+ },
1745
+ "216": {
1746
+ "content": "<0xD4>",
1747
+ "single_word": false,
1748
+ "lstrip": false,
1749
+ "rstrip": false,
1750
+ "normalized": false,
1751
+ "special": true
1752
+ },
1753
+ "217": {
1754
+ "content": "<0xD5>",
1755
+ "single_word": false,
1756
+ "lstrip": false,
1757
+ "rstrip": false,
1758
+ "normalized": false,
1759
+ "special": true
1760
+ },
1761
+ "218": {
1762
+ "content": "<0xD6>",
1763
+ "single_word": false,
1764
+ "lstrip": false,
1765
+ "rstrip": false,
1766
+ "normalized": false,
1767
+ "special": true
1768
+ },
1769
+ "219": {
1770
+ "content": "<0xD7>",
1771
+ "single_word": false,
1772
+ "lstrip": false,
1773
+ "rstrip": false,
1774
+ "normalized": false,
1775
+ "special": true
1776
+ },
1777
+ "220": {
1778
+ "content": "<0xD8>",
1779
+ "single_word": false,
1780
+ "lstrip": false,
1781
+ "rstrip": false,
1782
+ "normalized": false,
1783
+ "special": true
1784
+ },
1785
+ "221": {
1786
+ "content": "<0xD9>",
1787
+ "single_word": false,
1788
+ "lstrip": false,
1789
+ "rstrip": false,
1790
+ "normalized": false,
1791
+ "special": true
1792
+ },
1793
+ "222": {
1794
+ "content": "<0xDA>",
1795
+ "single_word": false,
1796
+ "lstrip": false,
1797
+ "rstrip": false,
1798
+ "normalized": false,
1799
+ "special": true
1800
+ },
1801
+ "223": {
1802
+ "content": "<0xDB>",
1803
+ "single_word": false,
1804
+ "lstrip": false,
1805
+ "rstrip": false,
1806
+ "normalized": false,
1807
+ "special": true
1808
+ },
1809
+ "224": {
1810
+ "content": "<0xDC>",
1811
+ "single_word": false,
1812
+ "lstrip": false,
1813
+ "rstrip": false,
1814
+ "normalized": false,
1815
+ "special": true
1816
+ },
1817
+ "225": {
1818
+ "content": "<0xDD>",
1819
+ "single_word": false,
1820
+ "lstrip": false,
1821
+ "rstrip": false,
1822
+ "normalized": false,
1823
+ "special": true
1824
+ },
1825
+ "226": {
1826
+ "content": "<0xDE>",
1827
+ "single_word": false,
1828
+ "lstrip": false,
1829
+ "rstrip": false,
1830
+ "normalized": false,
1831
+ "special": true
1832
+ },
1833
+ "227": {
1834
+ "content": "<0xDF>",
1835
+ "single_word": false,
1836
+ "lstrip": false,
1837
+ "rstrip": false,
1838
+ "normalized": false,
1839
+ "special": true
1840
+ },
1841
+ "228": {
1842
+ "content": "<0xE0>",
1843
+ "single_word": false,
1844
+ "lstrip": false,
1845
+ "rstrip": false,
1846
+ "normalized": false,
1847
+ "special": true
1848
+ },
1849
+ "229": {
1850
+ "content": "<0xE1>",
1851
+ "single_word": false,
1852
+ "lstrip": false,
1853
+ "rstrip": false,
1854
+ "normalized": false,
1855
+ "special": true
1856
+ },
1857
+ "230": {
1858
+ "content": "<0xE2>",
1859
+ "single_word": false,
1860
+ "lstrip": false,
1861
+ "rstrip": false,
1862
+ "normalized": false,
1863
+ "special": true
1864
+ },
1865
+ "231": {
1866
+ "content": "<0xE3>",
1867
+ "single_word": false,
1868
+ "lstrip": false,
1869
+ "rstrip": false,
1870
+ "normalized": false,
1871
+ "special": true
1872
+ },
1873
+ "232": {
1874
+ "content": "<0xE4>",
1875
+ "single_word": false,
1876
+ "lstrip": false,
1877
+ "rstrip": false,
1878
+ "normalized": false,
1879
+ "special": true
1880
+ },
1881
+ "233": {
1882
+ "content": "<0xE5>",
1883
+ "single_word": false,
1884
+ "lstrip": false,
1885
+ "rstrip": false,
1886
+ "normalized": false,
1887
+ "special": true
1888
+ },
1889
+ "234": {
1890
+ "content": "<0xE6>",
1891
+ "single_word": false,
1892
+ "lstrip": false,
1893
+ "rstrip": false,
1894
+ "normalized": false,
1895
+ "special": true
1896
+ },
1897
+ "235": {
1898
+ "content": "<0xE7>",
1899
+ "single_word": false,
1900
+ "lstrip": false,
1901
+ "rstrip": false,
1902
+ "normalized": false,
1903
+ "special": true
1904
+ },
1905
+ "236": {
1906
+ "content": "<0xE8>",
1907
+ "single_word": false,
1908
+ "lstrip": false,
1909
+ "rstrip": false,
1910
+ "normalized": false,
1911
+ "special": true
1912
+ },
1913
+ "237": {
1914
+ "content": "<0xE9>",
1915
+ "single_word": false,
1916
+ "lstrip": false,
1917
+ "rstrip": false,
1918
+ "normalized": false,
1919
+ "special": true
1920
+ },
1921
+ "238": {
1922
+ "content": "<0xEA>",
1923
+ "single_word": false,
1924
+ "lstrip": false,
1925
+ "rstrip": false,
1926
+ "normalized": false,
1927
+ "special": true
1928
+ },
1929
+ "239": {
1930
+ "content": "<0xEB>",
1931
+ "single_word": false,
1932
+ "lstrip": false,
1933
+ "rstrip": false,
1934
+ "normalized": false,
1935
+ "special": true
1936
+ },
1937
+ "240": {
1938
+ "content": "<0xEC>",
1939
+ "single_word": false,
1940
+ "lstrip": false,
1941
+ "rstrip": false,
1942
+ "normalized": false,
1943
+ "special": true
1944
+ },
1945
+ "241": {
1946
+ "content": "<0xED>",
1947
+ "single_word": false,
1948
+ "lstrip": false,
1949
+ "rstrip": false,
1950
+ "normalized": false,
1951
+ "special": true
1952
+ },
1953
+ "242": {
1954
+ "content": "<0xEE>",
1955
+ "single_word": false,
1956
+ "lstrip": false,
1957
+ "rstrip": false,
1958
+ "normalized": false,
1959
+ "special": true
1960
+ },
1961
+ "243": {
1962
+ "content": "<0xEF>",
1963
+ "single_word": false,
1964
+ "lstrip": false,
1965
+ "rstrip": false,
1966
+ "normalized": false,
1967
+ "special": true
1968
+ },
1969
+ "244": {
1970
+ "content": "<0xF0>",
1971
+ "single_word": false,
1972
+ "lstrip": false,
1973
+ "rstrip": false,
1974
+ "normalized": false,
1975
+ "special": true
1976
+ },
1977
+ "245": {
1978
+ "content": "<0xF1>",
1979
+ "single_word": false,
1980
+ "lstrip": false,
1981
+ "rstrip": false,
1982
+ "normalized": false,
1983
+ "special": true
1984
+ },
1985
+ "246": {
1986
+ "content": "<0xF2>",
1987
+ "single_word": false,
1988
+ "lstrip": false,
1989
+ "rstrip": false,
1990
+ "normalized": false,
1991
+ "special": true
1992
+ },
1993
+ "247": {
1994
+ "content": "<0xF3>",
1995
+ "single_word": false,
1996
+ "lstrip": false,
1997
+ "rstrip": false,
1998
+ "normalized": false,
1999
+ "special": true
2000
+ },
2001
+ "248": {
2002
+ "content": "<0xF4>",
2003
+ "single_word": false,
2004
+ "lstrip": false,
2005
+ "rstrip": false,
2006
+ "normalized": false,
2007
+ "special": true
2008
+ },
2009
+ "249": {
2010
+ "content": "<0xF5>",
2011
+ "single_word": false,
2012
+ "lstrip": false,
2013
+ "rstrip": false,
2014
+ "normalized": false,
2015
+ "special": true
2016
+ },
2017
+ "250": {
2018
+ "content": "<0xF6>",
2019
+ "single_word": false,
2020
+ "lstrip": false,
2021
+ "rstrip": false,
2022
+ "normalized": false,
2023
+ "special": true
2024
+ },
2025
+ "251": {
2026
+ "content": "<0xF7>",
2027
+ "single_word": false,
2028
+ "lstrip": false,
2029
+ "rstrip": false,
2030
+ "normalized": false,
2031
+ "special": true
2032
+ },
2033
+ "252": {
2034
+ "content": "<0xF8>",
2035
+ "single_word": false,
2036
+ "lstrip": false,
2037
+ "rstrip": false,
2038
+ "normalized": false,
2039
+ "special": true
2040
+ },
2041
+ "253": {
2042
+ "content": "<0xF9>",
2043
+ "single_word": false,
2044
+ "lstrip": false,
2045
+ "rstrip": false,
2046
+ "normalized": false,
2047
+ "special": true
2048
+ },
2049
+ "254": {
2050
+ "content": "<0xFA>",
2051
+ "single_word": false,
2052
+ "lstrip": false,
2053
+ "rstrip": false,
2054
+ "normalized": false,
2055
+ "special": true
2056
+ },
2057
+ "255": {
2058
+ "content": "<0xFB>",
2059
+ "single_word": false,
2060
+ "lstrip": false,
2061
+ "rstrip": false,
2062
+ "normalized": false,
2063
+ "special": true
2064
+ },
2065
+ "256": {
2066
+ "content": "<0xFC>",
2067
+ "single_word": false,
2068
+ "lstrip": false,
2069
+ "rstrip": false,
2070
+ "normalized": false,
2071
+ "special": true
2072
+ },
2073
+ "257": {
2074
+ "content": "<0xFD>",
2075
+ "single_word": false,
2076
+ "lstrip": false,
2077
+ "rstrip": false,
2078
+ "normalized": false,
2079
+ "special": true
2080
+ },
2081
+ "258": {
2082
+ "content": "<0xFE>",
2083
+ "single_word": false,
2084
+ "lstrip": false,
2085
+ "rstrip": false,
2086
+ "normalized": false,
2087
+ "special": true
2088
+ },
2089
+ "259": {
2090
+ "content": "<0xFF>",
2091
+ "single_word": false,
2092
+ "lstrip": false,
2093
+ "rstrip": false,
2094
+ "normalized": false,
2095
+ "special": true
2096
+ },
2097
+ "60800": {
2098
+ "content": "<|im_end|>",
2099
+ "single_word": false,
2100
+ "lstrip": false,
2101
+ "rstrip": false,
2102
+ "normalized": false,
2103
+ "special": true
2104
+ },
2105
+ "60801": {
2106
+ "content": "<|im_start|>",
2107
+ "single_word": false,
2108
+ "lstrip": false,
2109
+ "rstrip": false,
2110
+ "normalized": false,
2111
+ "special": true
2112
+ }
2113
+ }
2114
+ }
last-checkpoint/trainer_state.json ADDED
@@ -0,0 +1,3261 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.09006269261105374,
6
+ "eval_steps": 500,
7
+ "global_step": 923,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.00019515209666533855,
14
+ "grad_norm": 0.4216688275337219,
15
+ "learning_rate": 2.2624434389140273e-06,
16
+ "loss": 3.230625629425049,
17
+ "step": 2
18
+ },
19
+ {
20
+ "epoch": 0.0003903041933306771,
21
+ "grad_norm": 0.4993075728416443,
22
+ "learning_rate": 6.787330316742082e-06,
23
+ "loss": 3.3349366188049316,
24
+ "step": 4
25
+ },
26
+ {
27
+ "epoch": 0.0005854562899960156,
28
+ "grad_norm": 0.4248225688934326,
29
+ "learning_rate": 1.1312217194570136e-05,
30
+ "loss": 3.3722167015075684,
31
+ "step": 6
32
+ },
33
+ {
34
+ "epoch": 0.0007806083866613542,
35
+ "grad_norm": 0.4176347851753235,
36
+ "learning_rate": 1.583710407239819e-05,
37
+ "loss": 3.2494993209838867,
38
+ "step": 8
39
+ },
40
+ {
41
+ "epoch": 0.0009757604833266928,
42
+ "grad_norm": 0.38810470700263977,
43
+ "learning_rate": 2.0361990950226245e-05,
44
+ "loss": 3.312218427658081,
45
+ "step": 10
46
+ },
47
+ {
48
+ "epoch": 0.0011709125799920313,
49
+ "grad_norm": 0.5483741760253906,
50
+ "learning_rate": 2.4886877828054298e-05,
51
+ "loss": 3.3599870204925537,
52
+ "step": 12
53
+ },
54
+ {
55
+ "epoch": 0.0013660646766573698,
56
+ "grad_norm": 0.48463428020477295,
57
+ "learning_rate": 2.9411764705882354e-05,
58
+ "loss": 3.3288536071777344,
59
+ "step": 14
60
+ },
61
+ {
62
+ "epoch": 0.0015612167733227084,
63
+ "grad_norm": 0.3906591534614563,
64
+ "learning_rate": 3.39366515837104e-05,
65
+ "loss": 3.3323581218719482,
66
+ "step": 16
67
+ },
68
+ {
69
+ "epoch": 0.001756368869988047,
70
+ "grad_norm": 0.4956318438053131,
71
+ "learning_rate": 3.846153846153846e-05,
72
+ "loss": 3.3911399841308594,
73
+ "step": 18
74
+ },
75
+ {
76
+ "epoch": 0.0019515209666533856,
77
+ "grad_norm": 0.406741201877594,
78
+ "learning_rate": 4.2986425339366516e-05,
79
+ "loss": 3.2805917263031006,
80
+ "step": 20
81
+ },
82
+ {
83
+ "epoch": 0.002146673063318724,
84
+ "grad_norm": 0.4104717969894409,
85
+ "learning_rate": 4.751131221719457e-05,
86
+ "loss": 3.3620734214782715,
87
+ "step": 22
88
+ },
89
+ {
90
+ "epoch": 0.0023418251599840625,
91
+ "grad_norm": 0.382123202085495,
92
+ "learning_rate": 5.203619909502263e-05,
93
+ "loss": 3.252575635910034,
94
+ "step": 24
95
+ },
96
+ {
97
+ "epoch": 0.002536977256649401,
98
+ "grad_norm": 0.4557528495788574,
99
+ "learning_rate": 5.656108597285068e-05,
100
+ "loss": 3.270296335220337,
101
+ "step": 26
102
+ },
103
+ {
104
+ "epoch": 0.0027321293533147397,
105
+ "grad_norm": 0.3730292320251465,
106
+ "learning_rate": 6.108597285067873e-05,
107
+ "loss": 3.2637205123901367,
108
+ "step": 28
109
+ },
110
+ {
111
+ "epoch": 0.0029272814499800783,
112
+ "grad_norm": 0.4446980059146881,
113
+ "learning_rate": 6.561085972850679e-05,
114
+ "loss": 3.2753217220306396,
115
+ "step": 30
116
+ },
117
+ {
118
+ "epoch": 0.003122433546645417,
119
+ "grad_norm": 0.42525070905685425,
120
+ "learning_rate": 7.013574660633484e-05,
121
+ "loss": 3.3650691509246826,
122
+ "step": 32
123
+ },
124
+ {
125
+ "epoch": 0.0033175856433107554,
126
+ "grad_norm": 0.3361646831035614,
127
+ "learning_rate": 7.466063348416289e-05,
128
+ "loss": 3.261321783065796,
129
+ "step": 34
130
+ },
131
+ {
132
+ "epoch": 0.003512737739976094,
133
+ "grad_norm": 0.4070131778717041,
134
+ "learning_rate": 7.918552036199095e-05,
135
+ "loss": 3.2563247680664062,
136
+ "step": 36
137
+ },
138
+ {
139
+ "epoch": 0.0037078898366414326,
140
+ "grad_norm": 0.39418676495552063,
141
+ "learning_rate": 8.3710407239819e-05,
142
+ "loss": 3.2855348587036133,
143
+ "step": 38
144
+ },
145
+ {
146
+ "epoch": 0.003903041933306771,
147
+ "grad_norm": 0.4748503267765045,
148
+ "learning_rate": 8.823529411764706e-05,
149
+ "loss": 3.470453977584839,
150
+ "step": 40
151
+ },
152
+ {
153
+ "epoch": 0.004098194029972109,
154
+ "grad_norm": 0.41979095339775085,
155
+ "learning_rate": 9.276018099547512e-05,
156
+ "loss": 3.3626291751861572,
157
+ "step": 42
158
+ },
159
+ {
160
+ "epoch": 0.004293346126637448,
161
+ "grad_norm": 0.34092554450035095,
162
+ "learning_rate": 9.728506787330317e-05,
163
+ "loss": 3.2146968841552734,
164
+ "step": 44
165
+ },
166
+ {
167
+ "epoch": 0.0044884982233027865,
168
+ "grad_norm": 0.36365771293640137,
169
+ "learning_rate": 0.00010180995475113122,
170
+ "loss": 3.2508914470672607,
171
+ "step": 46
172
+ },
173
+ {
174
+ "epoch": 0.004683650319968125,
175
+ "grad_norm": 0.38663449883461,
176
+ "learning_rate": 0.00010633484162895928,
177
+ "loss": 3.360248327255249,
178
+ "step": 48
179
+ },
180
+ {
181
+ "epoch": 0.004878802416633464,
182
+ "grad_norm": 0.45633605122566223,
183
+ "learning_rate": 0.00011085972850678733,
184
+ "loss": 3.3618690967559814,
185
+ "step": 50
186
+ },
187
+ {
188
+ "epoch": 0.005073954513298802,
189
+ "grad_norm": 0.38607117533683777,
190
+ "learning_rate": 0.0001153846153846154,
191
+ "loss": 3.3222906589508057,
192
+ "step": 52
193
+ },
194
+ {
195
+ "epoch": 0.005269106609964141,
196
+ "grad_norm": 0.37843242287635803,
197
+ "learning_rate": 0.00011990950226244345,
198
+ "loss": 3.2946937084198,
199
+ "step": 54
200
+ },
201
+ {
202
+ "epoch": 0.005464258706629479,
203
+ "grad_norm": 0.39372068643569946,
204
+ "learning_rate": 0.0001244343891402715,
205
+ "loss": 3.362428903579712,
206
+ "step": 56
207
+ },
208
+ {
209
+ "epoch": 0.005659410803294818,
210
+ "grad_norm": 0.37712231278419495,
211
+ "learning_rate": 0.00012895927601809957,
212
+ "loss": 3.2583839893341064,
213
+ "step": 58
214
+ },
215
+ {
216
+ "epoch": 0.0058545628999601565,
217
+ "grad_norm": 0.41460534930229187,
218
+ "learning_rate": 0.0001334841628959276,
219
+ "loss": 3.285585880279541,
220
+ "step": 60
221
+ },
222
+ {
223
+ "epoch": 0.006049714996625495,
224
+ "grad_norm": 0.38934823870658875,
225
+ "learning_rate": 0.00013800904977375567,
226
+ "loss": 3.3030552864074707,
227
+ "step": 62
228
+ },
229
+ {
230
+ "epoch": 0.006244867093290834,
231
+ "grad_norm": 0.3794039487838745,
232
+ "learning_rate": 0.0001425339366515837,
233
+ "loss": 3.1728570461273193,
234
+ "step": 64
235
+ },
236
+ {
237
+ "epoch": 0.006440019189956172,
238
+ "grad_norm": 0.33801203966140747,
239
+ "learning_rate": 0.00014705882352941178,
240
+ "loss": 3.1532528400421143,
241
+ "step": 66
242
+ },
243
+ {
244
+ "epoch": 0.006635171286621511,
245
+ "grad_norm": 0.36383092403411865,
246
+ "learning_rate": 0.00015158371040723982,
247
+ "loss": 3.335261106491089,
248
+ "step": 68
249
+ },
250
+ {
251
+ "epoch": 0.006830323383286849,
252
+ "grad_norm": 0.31313201785087585,
253
+ "learning_rate": 0.00015610859728506788,
254
+ "loss": 3.2577037811279297,
255
+ "step": 70
256
+ },
257
+ {
258
+ "epoch": 0.007025475479952188,
259
+ "grad_norm": 0.37420541048049927,
260
+ "learning_rate": 0.00016063348416289595,
261
+ "loss": 3.2859036922454834,
262
+ "step": 72
263
+ },
264
+ {
265
+ "epoch": 0.007220627576617527,
266
+ "grad_norm": 0.36601948738098145,
267
+ "learning_rate": 0.000165158371040724,
268
+ "loss": 3.421809196472168,
269
+ "step": 74
270
+ },
271
+ {
272
+ "epoch": 0.007415779673282865,
273
+ "grad_norm": 0.3727055490016937,
274
+ "learning_rate": 0.00016968325791855206,
275
+ "loss": 3.2423620223999023,
276
+ "step": 76
277
+ },
278
+ {
279
+ "epoch": 0.007610931769948204,
280
+ "grad_norm": 0.34571611881256104,
281
+ "learning_rate": 0.0001742081447963801,
282
+ "loss": 3.318230390548706,
283
+ "step": 78
284
+ },
285
+ {
286
+ "epoch": 0.007806083866613542,
287
+ "grad_norm": 0.38571467995643616,
288
+ "learning_rate": 0.00017873303167420816,
289
+ "loss": 3.2539188861846924,
290
+ "step": 80
291
+ },
292
+ {
293
+ "epoch": 0.008001235963278881,
294
+ "grad_norm": 0.411663293838501,
295
+ "learning_rate": 0.0001832579185520362,
296
+ "loss": 3.314919948577881,
297
+ "step": 82
298
+ },
299
+ {
300
+ "epoch": 0.008196388059944219,
301
+ "grad_norm": 0.3499559462070465,
302
+ "learning_rate": 0.00018778280542986427,
303
+ "loss": 3.375194549560547,
304
+ "step": 84
305
+ },
306
+ {
307
+ "epoch": 0.008391540156609558,
308
+ "grad_norm": 0.33555346727371216,
309
+ "learning_rate": 0.00019230769230769233,
310
+ "loss": 3.246835708618164,
311
+ "step": 86
312
+ },
313
+ {
314
+ "epoch": 0.008586692253274896,
315
+ "grad_norm": 0.3681963086128235,
316
+ "learning_rate": 0.00019683257918552037,
317
+ "loss": 3.229979991912842,
318
+ "step": 88
319
+ },
320
+ {
321
+ "epoch": 0.008781844349940235,
322
+ "grad_norm": 0.3841090500354767,
323
+ "learning_rate": 0.0002013574660633484,
324
+ "loss": 3.3789896965026855,
325
+ "step": 90
326
+ },
327
+ {
328
+ "epoch": 0.008976996446605573,
329
+ "grad_norm": 0.38554415106773376,
330
+ "learning_rate": 0.00020588235294117645,
331
+ "loss": 3.305800437927246,
332
+ "step": 92
333
+ },
334
+ {
335
+ "epoch": 0.009172148543270912,
336
+ "grad_norm": 0.3495641052722931,
337
+ "learning_rate": 0.00021040723981900452,
338
+ "loss": 3.290616273880005,
339
+ "step": 94
340
+ },
341
+ {
342
+ "epoch": 0.00936730063993625,
343
+ "grad_norm": 0.3615274727344513,
344
+ "learning_rate": 0.00021493212669683259,
345
+ "loss": 3.3224902153015137,
346
+ "step": 96
347
+ },
348
+ {
349
+ "epoch": 0.00956245273660159,
350
+ "grad_norm": 0.3810535967350006,
351
+ "learning_rate": 0.00021945701357466062,
352
+ "loss": 3.322998523712158,
353
+ "step": 98
354
+ },
355
+ {
356
+ "epoch": 0.009757604833266927,
357
+ "grad_norm": 0.31313011050224304,
358
+ "learning_rate": 0.0002239819004524887,
359
+ "loss": 3.282527446746826,
360
+ "step": 100
361
+ },
362
+ {
363
+ "epoch": 0.009952756929932267,
364
+ "grad_norm": 0.2682645618915558,
365
+ "learning_rate": 0.00022850678733031673,
366
+ "loss": 3.1453497409820557,
367
+ "step": 102
368
+ },
369
+ {
370
+ "epoch": 0.010147909026597604,
371
+ "grad_norm": 0.30638542771339417,
372
+ "learning_rate": 0.0002330316742081448,
373
+ "loss": 3.1744844913482666,
374
+ "step": 104
375
+ },
376
+ {
377
+ "epoch": 0.010343061123262944,
378
+ "grad_norm": 0.26743319630622864,
379
+ "learning_rate": 0.00023755656108597284,
380
+ "loss": 3.1450507640838623,
381
+ "step": 106
382
+ },
383
+ {
384
+ "epoch": 0.010538213219928282,
385
+ "grad_norm": 0.3504815697669983,
386
+ "learning_rate": 0.0002420814479638009,
387
+ "loss": 3.3337080478668213,
388
+ "step": 108
389
+ },
390
+ {
391
+ "epoch": 0.01073336531659362,
392
+ "grad_norm": 0.34255480766296387,
393
+ "learning_rate": 0.000246606334841629,
394
+ "loss": 3.164823532104492,
395
+ "step": 110
396
+ },
397
+ {
398
+ "epoch": 0.010928517413258959,
399
+ "grad_norm": 0.3387168347835541,
400
+ "learning_rate": 0.000251131221719457,
401
+ "loss": 3.2456185817718506,
402
+ "step": 112
403
+ },
404
+ {
405
+ "epoch": 0.011123669509924296,
406
+ "grad_norm": 0.32132673263549805,
407
+ "learning_rate": 0.0002556561085972851,
408
+ "loss": 3.247767448425293,
409
+ "step": 114
410
+ },
411
+ {
412
+ "epoch": 0.011318821606589636,
413
+ "grad_norm": 0.32860279083251953,
414
+ "learning_rate": 0.00026018099547511314,
415
+ "loss": 3.2920401096343994,
416
+ "step": 116
417
+ },
418
+ {
419
+ "epoch": 0.011513973703254974,
420
+ "grad_norm": 0.3630625009536743,
421
+ "learning_rate": 0.0002647058823529412,
422
+ "loss": 3.2878782749176025,
423
+ "step": 118
424
+ },
425
+ {
426
+ "epoch": 0.011709125799920313,
427
+ "grad_norm": 0.30959972739219666,
428
+ "learning_rate": 0.0002692307692307692,
429
+ "loss": 3.2433247566223145,
430
+ "step": 120
431
+ },
432
+ {
433
+ "epoch": 0.01190427789658565,
434
+ "grad_norm": 0.33803021907806396,
435
+ "learning_rate": 0.00027375565610859734,
436
+ "loss": 3.273428201675415,
437
+ "step": 122
438
+ },
439
+ {
440
+ "epoch": 0.01209942999325099,
441
+ "grad_norm": 0.3176804482936859,
442
+ "learning_rate": 0.00027828054298642535,
443
+ "loss": 3.1625616550445557,
444
+ "step": 124
445
+ },
446
+ {
447
+ "epoch": 0.012294582089916328,
448
+ "grad_norm": 0.36957040429115295,
449
+ "learning_rate": 0.0002828054298642534,
450
+ "loss": 3.3513076305389404,
451
+ "step": 126
452
+ },
453
+ {
454
+ "epoch": 0.012489734186581667,
455
+ "grad_norm": 0.2907490134239197,
456
+ "learning_rate": 0.00028733031674208143,
457
+ "loss": 3.189974784851074,
458
+ "step": 128
459
+ },
460
+ {
461
+ "epoch": 0.012684886283247005,
462
+ "grad_norm": 0.30953142046928406,
463
+ "learning_rate": 0.00029185520361990955,
464
+ "loss": 3.3083784580230713,
465
+ "step": 130
466
+ },
467
+ {
468
+ "epoch": 0.012880038379912345,
469
+ "grad_norm": 0.32543566823005676,
470
+ "learning_rate": 0.00029638009049773756,
471
+ "loss": 3.2629635334014893,
472
+ "step": 132
473
+ },
474
+ {
475
+ "epoch": 0.013075190476577682,
476
+ "grad_norm": 0.33064964413642883,
477
+ "learning_rate": 0.0003009049773755656,
478
+ "loss": 3.2875685691833496,
479
+ "step": 134
480
+ },
481
+ {
482
+ "epoch": 0.013270342573243022,
483
+ "grad_norm": 0.38144928216934204,
484
+ "learning_rate": 0.0003054298642533937,
485
+ "loss": 3.2704145908355713,
486
+ "step": 136
487
+ },
488
+ {
489
+ "epoch": 0.01346549466990836,
490
+ "grad_norm": 0.3123980164527893,
491
+ "learning_rate": 0.0003099547511312217,
492
+ "loss": 3.1573328971862793,
493
+ "step": 138
494
+ },
495
+ {
496
+ "epoch": 0.013660646766573699,
497
+ "grad_norm": 0.2872498333454132,
498
+ "learning_rate": 0.0003144796380090498,
499
+ "loss": 3.1866462230682373,
500
+ "step": 140
501
+ },
502
+ {
503
+ "epoch": 0.013855798863239037,
504
+ "grad_norm": 0.26412302255630493,
505
+ "learning_rate": 0.0003190045248868778,
506
+ "loss": 3.2516064643859863,
507
+ "step": 142
508
+ },
509
+ {
510
+ "epoch": 0.014050950959904376,
511
+ "grad_norm": 0.3109893798828125,
512
+ "learning_rate": 0.0003235294117647059,
513
+ "loss": 3.2663843631744385,
514
+ "step": 144
515
+ },
516
+ {
517
+ "epoch": 0.014246103056569714,
518
+ "grad_norm": 0.2663963735103607,
519
+ "learning_rate": 0.0003280542986425339,
520
+ "loss": 3.202380657196045,
521
+ "step": 146
522
+ },
523
+ {
524
+ "epoch": 0.014441255153235053,
525
+ "grad_norm": 0.27347636222839355,
526
+ "learning_rate": 0.000332579185520362,
527
+ "loss": 3.0876479148864746,
528
+ "step": 148
529
+ },
530
+ {
531
+ "epoch": 0.014636407249900391,
532
+ "grad_norm": 0.29326972365379333,
533
+ "learning_rate": 0.00033710407239819005,
534
+ "loss": 3.252263307571411,
535
+ "step": 150
536
+ },
537
+ {
538
+ "epoch": 0.01483155934656573,
539
+ "grad_norm": 0.2532570958137512,
540
+ "learning_rate": 0.0003416289592760181,
541
+ "loss": 3.1838080883026123,
542
+ "step": 152
543
+ },
544
+ {
545
+ "epoch": 0.015026711443231068,
546
+ "grad_norm": 0.33389320969581604,
547
+ "learning_rate": 0.00034615384615384613,
548
+ "loss": 3.2125651836395264,
549
+ "step": 154
550
+ },
551
+ {
552
+ "epoch": 0.015221863539896408,
553
+ "grad_norm": 0.3081454634666443,
554
+ "learning_rate": 0.0003506787330316742,
555
+ "loss": 3.225301742553711,
556
+ "step": 156
557
+ },
558
+ {
559
+ "epoch": 0.015417015636561745,
560
+ "grad_norm": 0.28876030445098877,
561
+ "learning_rate": 0.00035520361990950226,
562
+ "loss": 3.2167575359344482,
563
+ "step": 158
564
+ },
565
+ {
566
+ "epoch": 0.015612167733227085,
567
+ "grad_norm": 0.2998896837234497,
568
+ "learning_rate": 0.00035972850678733033,
569
+ "loss": 3.170762777328491,
570
+ "step": 160
571
+ },
572
+ {
573
+ "epoch": 0.015807319829892422,
574
+ "grad_norm": 0.3383564054965973,
575
+ "learning_rate": 0.00036425339366515834,
576
+ "loss": 3.2211852073669434,
577
+ "step": 162
578
+ },
579
+ {
580
+ "epoch": 0.016002471926557762,
581
+ "grad_norm": 0.2595459222793579,
582
+ "learning_rate": 0.00036877828054298646,
583
+ "loss": 3.1501312255859375,
584
+ "step": 164
585
+ },
586
+ {
587
+ "epoch": 0.016197624023223098,
588
+ "grad_norm": 0.31783509254455566,
589
+ "learning_rate": 0.0003733031674208145,
590
+ "loss": 3.2832605838775635,
591
+ "step": 166
592
+ },
593
+ {
594
+ "epoch": 0.016392776119888437,
595
+ "grad_norm": 0.24646806716918945,
596
+ "learning_rate": 0.00037782805429864254,
597
+ "loss": 3.057286262512207,
598
+ "step": 168
599
+ },
600
+ {
601
+ "epoch": 0.016587928216553777,
602
+ "grad_norm": 0.2828831374645233,
603
+ "learning_rate": 0.00038235294117647055,
604
+ "loss": 3.2353763580322266,
605
+ "step": 170
606
+ },
607
+ {
608
+ "epoch": 0.016783080313219116,
609
+ "grad_norm": 0.30411699414253235,
610
+ "learning_rate": 0.0003868778280542987,
611
+ "loss": 3.1818532943725586,
612
+ "step": 172
613
+ },
614
+ {
615
+ "epoch": 0.016978232409884452,
616
+ "grad_norm": 0.30165067315101624,
617
+ "learning_rate": 0.0003914027149321267,
618
+ "loss": 3.1745240688323975,
619
+ "step": 174
620
+ },
621
+ {
622
+ "epoch": 0.01717338450654979,
623
+ "grad_norm": 0.2772194743156433,
624
+ "learning_rate": 0.00039592760180995475,
625
+ "loss": 3.1890299320220947,
626
+ "step": 176
627
+ },
628
+ {
629
+ "epoch": 0.01736853660321513,
630
+ "grad_norm": 0.2587553858757019,
631
+ "learning_rate": 0.0004004524886877828,
632
+ "loss": 3.134737253189087,
633
+ "step": 178
634
+ },
635
+ {
636
+ "epoch": 0.01756368869988047,
637
+ "grad_norm": 0.2447192519903183,
638
+ "learning_rate": 0.0004049773755656109,
639
+ "loss": 3.2146072387695312,
640
+ "step": 180
641
+ },
642
+ {
643
+ "epoch": 0.017758840796545806,
644
+ "grad_norm": 0.28469911217689514,
645
+ "learning_rate": 0.0004095022624434389,
646
+ "loss": 3.2746925354003906,
647
+ "step": 182
648
+ },
649
+ {
650
+ "epoch": 0.017953992893211146,
651
+ "grad_norm": 0.28234514594078064,
652
+ "learning_rate": 0.00041402714932126697,
653
+ "loss": 3.1442291736602783,
654
+ "step": 184
655
+ },
656
+ {
657
+ "epoch": 0.018149144989876485,
658
+ "grad_norm": 0.38107261061668396,
659
+ "learning_rate": 0.00041855203619909503,
660
+ "loss": 3.172553539276123,
661
+ "step": 186
662
+ },
663
+ {
664
+ "epoch": 0.018344297086541825,
665
+ "grad_norm": 0.2612990736961365,
666
+ "learning_rate": 0.0004230769230769231,
667
+ "loss": 3.194969892501831,
668
+ "step": 188
669
+ },
670
+ {
671
+ "epoch": 0.01853944918320716,
672
+ "grad_norm": 0.28091078996658325,
673
+ "learning_rate": 0.0004276018099547511,
674
+ "loss": 3.2099835872650146,
675
+ "step": 190
676
+ },
677
+ {
678
+ "epoch": 0.0187346012798725,
679
+ "grad_norm": 0.29824092984199524,
680
+ "learning_rate": 0.00043212669683257923,
681
+ "loss": 3.18837571144104,
682
+ "step": 192
683
+ },
684
+ {
685
+ "epoch": 0.01892975337653784,
686
+ "grad_norm": 0.2988644540309906,
687
+ "learning_rate": 0.00043665158371040724,
688
+ "loss": 3.1975226402282715,
689
+ "step": 194
690
+ },
691
+ {
692
+ "epoch": 0.01912490547320318,
693
+ "grad_norm": 0.31145668029785156,
694
+ "learning_rate": 0.0004411764705882353,
695
+ "loss": 3.2740702629089355,
696
+ "step": 196
697
+ },
698
+ {
699
+ "epoch": 0.019320057569868515,
700
+ "grad_norm": 0.2825329303741455,
701
+ "learning_rate": 0.0004457013574660634,
702
+ "loss": 3.1490797996520996,
703
+ "step": 198
704
+ },
705
+ {
706
+ "epoch": 0.019515209666533855,
707
+ "grad_norm": 0.272503137588501,
708
+ "learning_rate": 0.00045022624434389144,
709
+ "loss": 3.1524457931518555,
710
+ "step": 200
711
+ },
712
+ {
713
+ "epoch": 0.019710361763199194,
714
+ "grad_norm": 0.3168213367462158,
715
+ "learning_rate": 0.00045475113122171945,
716
+ "loss": 3.155792236328125,
717
+ "step": 202
718
+ },
719
+ {
720
+ "epoch": 0.019905513859864533,
721
+ "grad_norm": 0.30904391407966614,
722
+ "learning_rate": 0.0004592760180995475,
723
+ "loss": 3.1604065895080566,
724
+ "step": 204
725
+ },
726
+ {
727
+ "epoch": 0.02010066595652987,
728
+ "grad_norm": 0.2521339952945709,
729
+ "learning_rate": 0.0004638009049773756,
730
+ "loss": 3.2597200870513916,
731
+ "step": 206
732
+ },
733
+ {
734
+ "epoch": 0.02029581805319521,
735
+ "grad_norm": 0.24709968268871307,
736
+ "learning_rate": 0.00046832579185520365,
737
+ "loss": 3.1984686851501465,
738
+ "step": 208
739
+ },
740
+ {
741
+ "epoch": 0.02049097014986055,
742
+ "grad_norm": 0.31030556559562683,
743
+ "learning_rate": 0.00047285067873303167,
744
+ "loss": 3.238802909851074,
745
+ "step": 210
746
+ },
747
+ {
748
+ "epoch": 0.020686122246525888,
749
+ "grad_norm": 0.2525986433029175,
750
+ "learning_rate": 0.0004773755656108598,
751
+ "loss": 3.0971829891204834,
752
+ "step": 212
753
+ },
754
+ {
755
+ "epoch": 0.020881274343191224,
756
+ "grad_norm": 0.2720223069190979,
757
+ "learning_rate": 0.0004819004524886878,
758
+ "loss": 3.1906185150146484,
759
+ "step": 214
760
+ },
761
+ {
762
+ "epoch": 0.021076426439856563,
763
+ "grad_norm": 0.25464531779289246,
764
+ "learning_rate": 0.00048642533936651587,
765
+ "loss": 3.1898193359375,
766
+ "step": 216
767
+ },
768
+ {
769
+ "epoch": 0.021271578536521903,
770
+ "grad_norm": 0.2506071627140045,
771
+ "learning_rate": 0.0004909502262443439,
772
+ "loss": 3.1172282695770264,
773
+ "step": 218
774
+ },
775
+ {
776
+ "epoch": 0.02146673063318724,
777
+ "grad_norm": 0.2585948407649994,
778
+ "learning_rate": 0.000495475113122172,
779
+ "loss": 3.165076494216919,
780
+ "step": 220
781
+ },
782
+ {
783
+ "epoch": 0.021661882729852578,
784
+ "grad_norm": 0.36205047369003296,
785
+ "learning_rate": 0.0005,
786
+ "loss": 3.1507983207702637,
787
+ "step": 222
788
+ },
789
+ {
790
+ "epoch": 0.021857034826517917,
791
+ "grad_norm": 0.2599605321884155,
792
+ "learning_rate": 0.0005,
793
+ "loss": 3.220719575881958,
794
+ "step": 224
795
+ },
796
+ {
797
+ "epoch": 0.022052186923183257,
798
+ "grad_norm": 0.2505502998828888,
799
+ "learning_rate": 0.0005,
800
+ "loss": 3.107052803039551,
801
+ "step": 226
802
+ },
803
+ {
804
+ "epoch": 0.022247339019848593,
805
+ "grad_norm": 0.24908874928951263,
806
+ "learning_rate": 0.0005,
807
+ "loss": 3.119175434112549,
808
+ "step": 228
809
+ },
810
+ {
811
+ "epoch": 0.022442491116513932,
812
+ "grad_norm": 0.27349185943603516,
813
+ "learning_rate": 0.0005,
814
+ "loss": 3.1379740238189697,
815
+ "step": 230
816
+ },
817
+ {
818
+ "epoch": 0.022637643213179272,
819
+ "grad_norm": 0.25903528928756714,
820
+ "learning_rate": 0.0005,
821
+ "loss": 3.1497933864593506,
822
+ "step": 232
823
+ },
824
+ {
825
+ "epoch": 0.02283279530984461,
826
+ "grad_norm": 0.24287213385105133,
827
+ "learning_rate": 0.0005,
828
+ "loss": 3.162045955657959,
829
+ "step": 234
830
+ },
831
+ {
832
+ "epoch": 0.023027947406509947,
833
+ "grad_norm": 0.25341615080833435,
834
+ "learning_rate": 0.0005,
835
+ "loss": 3.1117117404937744,
836
+ "step": 236
837
+ },
838
+ {
839
+ "epoch": 0.023223099503175287,
840
+ "grad_norm": 0.314718633890152,
841
+ "learning_rate": 0.0005,
842
+ "loss": 3.161027431488037,
843
+ "step": 238
844
+ },
845
+ {
846
+ "epoch": 0.023418251599840626,
847
+ "grad_norm": 0.24358060956001282,
848
+ "learning_rate": 0.0005,
849
+ "loss": 3.1430258750915527,
850
+ "step": 240
851
+ },
852
+ {
853
+ "epoch": 0.023613403696505966,
854
+ "grad_norm": 0.2685472071170807,
855
+ "learning_rate": 0.0005,
856
+ "loss": 3.1333565711975098,
857
+ "step": 242
858
+ },
859
+ {
860
+ "epoch": 0.0238085557931713,
861
+ "grad_norm": 0.24080778658390045,
862
+ "learning_rate": 0.0005,
863
+ "loss": 3.0782580375671387,
864
+ "step": 244
865
+ },
866
+ {
867
+ "epoch": 0.02400370788983664,
868
+ "grad_norm": 0.245890274643898,
869
+ "learning_rate": 0.0005,
870
+ "loss": 3.177727460861206,
871
+ "step": 246
872
+ },
873
+ {
874
+ "epoch": 0.02419885998650198,
875
+ "grad_norm": 0.22881919145584106,
876
+ "learning_rate": 0.0005,
877
+ "loss": 3.0675933361053467,
878
+ "step": 248
879
+ },
880
+ {
881
+ "epoch": 0.02439401208316732,
882
+ "grad_norm": 0.2290228307247162,
883
+ "learning_rate": 0.0005,
884
+ "loss": 3.0600621700286865,
885
+ "step": 250
886
+ },
887
+ {
888
+ "epoch": 0.024589164179832656,
889
+ "grad_norm": 0.2506984770298004,
890
+ "learning_rate": 0.0005,
891
+ "loss": 3.0782036781311035,
892
+ "step": 252
893
+ },
894
+ {
895
+ "epoch": 0.024784316276497995,
896
+ "grad_norm": 0.2542862594127655,
897
+ "learning_rate": 0.0005,
898
+ "loss": 3.122307538986206,
899
+ "step": 254
900
+ },
901
+ {
902
+ "epoch": 0.024979468373163335,
903
+ "grad_norm": 0.2680153548717499,
904
+ "learning_rate": 0.0005,
905
+ "loss": 3.0241408348083496,
906
+ "step": 256
907
+ },
908
+ {
909
+ "epoch": 0.025174620469828674,
910
+ "grad_norm": 0.23353970050811768,
911
+ "learning_rate": 0.0005,
912
+ "loss": 3.0042200088500977,
913
+ "step": 258
914
+ },
915
+ {
916
+ "epoch": 0.02536977256649401,
917
+ "grad_norm": 0.25731053948402405,
918
+ "learning_rate": 0.0005,
919
+ "loss": 3.0614330768585205,
920
+ "step": 260
921
+ },
922
+ {
923
+ "epoch": 0.02556492466315935,
924
+ "grad_norm": 0.24684110283851624,
925
+ "learning_rate": 0.0005,
926
+ "loss": 3.129671335220337,
927
+ "step": 262
928
+ },
929
+ {
930
+ "epoch": 0.02576007675982469,
931
+ "grad_norm": 0.2284134179353714,
932
+ "learning_rate": 0.0005,
933
+ "loss": 3.0751752853393555,
934
+ "step": 264
935
+ },
936
+ {
937
+ "epoch": 0.02595522885649003,
938
+ "grad_norm": 0.30790138244628906,
939
+ "learning_rate": 0.0005,
940
+ "loss": 3.107248544692993,
941
+ "step": 266
942
+ },
943
+ {
944
+ "epoch": 0.026150380953155365,
945
+ "grad_norm": 0.2533958852291107,
946
+ "learning_rate": 0.0005,
947
+ "loss": 3.0680527687072754,
948
+ "step": 268
949
+ },
950
+ {
951
+ "epoch": 0.026345533049820704,
952
+ "grad_norm": 0.2893184721469879,
953
+ "learning_rate": 0.0005,
954
+ "loss": 3.127474784851074,
955
+ "step": 270
956
+ },
957
+ {
958
+ "epoch": 0.026540685146486043,
959
+ "grad_norm": 0.22310151159763336,
960
+ "learning_rate": 0.0005,
961
+ "loss": 3.048927068710327,
962
+ "step": 272
963
+ },
964
+ {
965
+ "epoch": 0.02673583724315138,
966
+ "grad_norm": 0.23024123907089233,
967
+ "learning_rate": 0.0005,
968
+ "loss": 3.1112749576568604,
969
+ "step": 274
970
+ },
971
+ {
972
+ "epoch": 0.02693098933981672,
973
+ "grad_norm": 0.32433581352233887,
974
+ "learning_rate": 0.0005,
975
+ "loss": 3.1394951343536377,
976
+ "step": 276
977
+ },
978
+ {
979
+ "epoch": 0.02712614143648206,
980
+ "grad_norm": 0.20521575212478638,
981
+ "learning_rate": 0.0005,
982
+ "loss": 3.0791382789611816,
983
+ "step": 278
984
+ },
985
+ {
986
+ "epoch": 0.027321293533147398,
987
+ "grad_norm": 0.21532726287841797,
988
+ "learning_rate": 0.0005,
989
+ "loss": 3.1784777641296387,
990
+ "step": 280
991
+ },
992
+ {
993
+ "epoch": 0.027516445629812734,
994
+ "grad_norm": 0.3145722448825836,
995
+ "learning_rate": 0.0005,
996
+ "loss": 3.0516176223754883,
997
+ "step": 282
998
+ },
999
+ {
1000
+ "epoch": 0.027711597726478073,
1001
+ "grad_norm": 0.340013712644577,
1002
+ "learning_rate": 0.0005,
1003
+ "loss": 3.1259262561798096,
1004
+ "step": 284
1005
+ },
1006
+ {
1007
+ "epoch": 0.027906749823143413,
1008
+ "grad_norm": 0.22899404168128967,
1009
+ "learning_rate": 0.0005,
1010
+ "loss": 3.0872085094451904,
1011
+ "step": 286
1012
+ },
1013
+ {
1014
+ "epoch": 0.028101901919808752,
1015
+ "grad_norm": 0.2496689110994339,
1016
+ "learning_rate": 0.0005,
1017
+ "loss": 3.0528392791748047,
1018
+ "step": 288
1019
+ },
1020
+ {
1021
+ "epoch": 0.028297054016474088,
1022
+ "grad_norm": 0.3250139653682709,
1023
+ "learning_rate": 0.0005,
1024
+ "loss": 3.0111708641052246,
1025
+ "step": 290
1026
+ },
1027
+ {
1028
+ "epoch": 0.028492206113139427,
1029
+ "grad_norm": 0.19515371322631836,
1030
+ "learning_rate": 0.0005,
1031
+ "loss": 3.0274436473846436,
1032
+ "step": 292
1033
+ },
1034
+ {
1035
+ "epoch": 0.028687358209804767,
1036
+ "grad_norm": 0.24196070432662964,
1037
+ "learning_rate": 0.0005,
1038
+ "loss": 3.109757661819458,
1039
+ "step": 294
1040
+ },
1041
+ {
1042
+ "epoch": 0.028882510306470106,
1043
+ "grad_norm": 0.2010832279920578,
1044
+ "learning_rate": 0.0005,
1045
+ "loss": 2.9821534156799316,
1046
+ "step": 296
1047
+ },
1048
+ {
1049
+ "epoch": 0.029077662403135442,
1050
+ "grad_norm": 0.21203897893428802,
1051
+ "learning_rate": 0.0005,
1052
+ "loss": 2.989006996154785,
1053
+ "step": 298
1054
+ },
1055
+ {
1056
+ "epoch": 0.029272814499800782,
1057
+ "grad_norm": 0.1977204531431198,
1058
+ "learning_rate": 0.0005,
1059
+ "loss": 3.0148069858551025,
1060
+ "step": 300
1061
+ },
1062
+ {
1063
+ "epoch": 0.02946796659646612,
1064
+ "grad_norm": 0.23890215158462524,
1065
+ "learning_rate": 0.0005,
1066
+ "loss": 3.026381731033325,
1067
+ "step": 302
1068
+ },
1069
+ {
1070
+ "epoch": 0.02966311869313146,
1071
+ "grad_norm": 0.2099602073431015,
1072
+ "learning_rate": 0.0005,
1073
+ "loss": 2.9980013370513916,
1074
+ "step": 304
1075
+ },
1076
+ {
1077
+ "epoch": 0.029858270789796797,
1078
+ "grad_norm": 0.2327839434146881,
1079
+ "learning_rate": 0.0005,
1080
+ "loss": 3.1114885807037354,
1081
+ "step": 306
1082
+ },
1083
+ {
1084
+ "epoch": 0.030053422886462136,
1085
+ "grad_norm": 0.21162371337413788,
1086
+ "learning_rate": 0.0005,
1087
+ "loss": 3.0627694129943848,
1088
+ "step": 308
1089
+ },
1090
+ {
1091
+ "epoch": 0.030248574983127476,
1092
+ "grad_norm": 0.23224154114723206,
1093
+ "learning_rate": 0.0005,
1094
+ "loss": 3.0820987224578857,
1095
+ "step": 310
1096
+ },
1097
+ {
1098
+ "epoch": 0.030443727079792815,
1099
+ "grad_norm": 0.3246678411960602,
1100
+ "learning_rate": 0.0005,
1101
+ "loss": 3.069063425064087,
1102
+ "step": 312
1103
+ },
1104
+ {
1105
+ "epoch": 0.03063887917645815,
1106
+ "grad_norm": 0.22087569534778595,
1107
+ "learning_rate": 0.0005,
1108
+ "loss": 3.1131033897399902,
1109
+ "step": 314
1110
+ },
1111
+ {
1112
+ "epoch": 0.03083403127312349,
1113
+ "grad_norm": 0.19992756843566895,
1114
+ "learning_rate": 0.0005,
1115
+ "loss": 2.9714348316192627,
1116
+ "step": 316
1117
+ },
1118
+ {
1119
+ "epoch": 0.03102918336978883,
1120
+ "grad_norm": 0.22986479103565216,
1121
+ "learning_rate": 0.0005,
1122
+ "loss": 3.083839178085327,
1123
+ "step": 318
1124
+ },
1125
+ {
1126
+ "epoch": 0.03122433546645417,
1127
+ "grad_norm": 0.21409468352794647,
1128
+ "learning_rate": 0.0005,
1129
+ "loss": 2.9624245166778564,
1130
+ "step": 320
1131
+ },
1132
+ {
1133
+ "epoch": 0.031419487563119505,
1134
+ "grad_norm": 0.20953746140003204,
1135
+ "learning_rate": 0.0005,
1136
+ "loss": 2.9169602394104004,
1137
+ "step": 322
1138
+ },
1139
+ {
1140
+ "epoch": 0.031614639659784845,
1141
+ "grad_norm": 0.22637401521205902,
1142
+ "learning_rate": 0.0005,
1143
+ "loss": 2.996743679046631,
1144
+ "step": 324
1145
+ },
1146
+ {
1147
+ "epoch": 0.031809791756450184,
1148
+ "grad_norm": 0.21024008095264435,
1149
+ "learning_rate": 0.0005,
1150
+ "loss": 3.0023748874664307,
1151
+ "step": 326
1152
+ },
1153
+ {
1154
+ "epoch": 0.032004943853115524,
1155
+ "grad_norm": 0.23165062069892883,
1156
+ "learning_rate": 0.0005,
1157
+ "loss": 3.030834913253784,
1158
+ "step": 328
1159
+ },
1160
+ {
1161
+ "epoch": 0.03220009594978086,
1162
+ "grad_norm": 0.21788132190704346,
1163
+ "learning_rate": 0.0005,
1164
+ "loss": 3.0839145183563232,
1165
+ "step": 330
1166
+ },
1167
+ {
1168
+ "epoch": 0.032395248046446196,
1169
+ "grad_norm": 0.21775038540363312,
1170
+ "learning_rate": 0.0005,
1171
+ "loss": 2.999584913253784,
1172
+ "step": 332
1173
+ },
1174
+ {
1175
+ "epoch": 0.032590400143111535,
1176
+ "grad_norm": 0.2256433218717575,
1177
+ "learning_rate": 0.0005,
1178
+ "loss": 3.0950889587402344,
1179
+ "step": 334
1180
+ },
1181
+ {
1182
+ "epoch": 0.032785552239776874,
1183
+ "grad_norm": 0.2107185274362564,
1184
+ "learning_rate": 0.0005,
1185
+ "loss": 3.005518674850464,
1186
+ "step": 336
1187
+ },
1188
+ {
1189
+ "epoch": 0.032980704336442214,
1190
+ "grad_norm": 0.25283727049827576,
1191
+ "learning_rate": 0.0005,
1192
+ "loss": 2.9571783542633057,
1193
+ "step": 338
1194
+ },
1195
+ {
1196
+ "epoch": 0.03317585643310755,
1197
+ "grad_norm": 0.21256347000598907,
1198
+ "learning_rate": 0.0005,
1199
+ "loss": 2.9704484939575195,
1200
+ "step": 340
1201
+ },
1202
+ {
1203
+ "epoch": 0.03337100852977289,
1204
+ "grad_norm": 0.22279518842697144,
1205
+ "learning_rate": 0.0005,
1206
+ "loss": 2.976118803024292,
1207
+ "step": 342
1208
+ },
1209
+ {
1210
+ "epoch": 0.03356616062643823,
1211
+ "grad_norm": 0.22978323698043823,
1212
+ "learning_rate": 0.0005,
1213
+ "loss": 3.021812677383423,
1214
+ "step": 344
1215
+ },
1216
+ {
1217
+ "epoch": 0.03376131272310357,
1218
+ "grad_norm": 0.23514382541179657,
1219
+ "learning_rate": 0.0005,
1220
+ "loss": 3.0376787185668945,
1221
+ "step": 346
1222
+ },
1223
+ {
1224
+ "epoch": 0.033956464819768904,
1225
+ "grad_norm": 0.20888946950435638,
1226
+ "learning_rate": 0.0005,
1227
+ "loss": 3.0029542446136475,
1228
+ "step": 348
1229
+ },
1230
+ {
1231
+ "epoch": 0.034151616916434244,
1232
+ "grad_norm": 0.2512245774269104,
1233
+ "learning_rate": 0.0005,
1234
+ "loss": 2.972135305404663,
1235
+ "step": 350
1236
+ },
1237
+ {
1238
+ "epoch": 0.03434676901309958,
1239
+ "grad_norm": 0.20965415239334106,
1240
+ "learning_rate": 0.0005,
1241
+ "loss": 3.005450487136841,
1242
+ "step": 352
1243
+ },
1244
+ {
1245
+ "epoch": 0.03454192110976492,
1246
+ "grad_norm": 0.2403269112110138,
1247
+ "learning_rate": 0.0005,
1248
+ "loss": 2.9974536895751953,
1249
+ "step": 354
1250
+ },
1251
+ {
1252
+ "epoch": 0.03473707320643026,
1253
+ "grad_norm": 0.24264736473560333,
1254
+ "learning_rate": 0.0005,
1255
+ "loss": 2.982867956161499,
1256
+ "step": 356
1257
+ },
1258
+ {
1259
+ "epoch": 0.0349322253030956,
1260
+ "grad_norm": 0.2102240025997162,
1261
+ "learning_rate": 0.0005,
1262
+ "loss": 3.0598607063293457,
1263
+ "step": 358
1264
+ },
1265
+ {
1266
+ "epoch": 0.03512737739976094,
1267
+ "grad_norm": 0.24868826568126678,
1268
+ "learning_rate": 0.0005,
1269
+ "loss": 2.9246678352355957,
1270
+ "step": 360
1271
+ },
1272
+ {
1273
+ "epoch": 0.03532252949642628,
1274
+ "grad_norm": 0.22144627571105957,
1275
+ "learning_rate": 0.0005,
1276
+ "loss": 2.9389896392822266,
1277
+ "step": 362
1278
+ },
1279
+ {
1280
+ "epoch": 0.03551768159309161,
1281
+ "grad_norm": 0.23583339154720306,
1282
+ "learning_rate": 0.0005,
1283
+ "loss": 3.0770351886749268,
1284
+ "step": 364
1285
+ },
1286
+ {
1287
+ "epoch": 0.03571283368975695,
1288
+ "grad_norm": 0.26368218660354614,
1289
+ "learning_rate": 0.0005,
1290
+ "loss": 2.8645801544189453,
1291
+ "step": 366
1292
+ },
1293
+ {
1294
+ "epoch": 0.03590798578642229,
1295
+ "grad_norm": 0.23020903766155243,
1296
+ "learning_rate": 0.0005,
1297
+ "loss": 2.915461301803589,
1298
+ "step": 368
1299
+ },
1300
+ {
1301
+ "epoch": 0.03610313788308763,
1302
+ "grad_norm": 0.21175698935985565,
1303
+ "learning_rate": 0.0005,
1304
+ "loss": 2.9230644702911377,
1305
+ "step": 370
1306
+ },
1307
+ {
1308
+ "epoch": 0.03629828997975297,
1309
+ "grad_norm": 0.20585383474826813,
1310
+ "learning_rate": 0.0005,
1311
+ "loss": 2.953773260116577,
1312
+ "step": 372
1313
+ },
1314
+ {
1315
+ "epoch": 0.03649344207641831,
1316
+ "grad_norm": 0.3054361045360565,
1317
+ "learning_rate": 0.0005,
1318
+ "loss": 2.988222122192383,
1319
+ "step": 374
1320
+ },
1321
+ {
1322
+ "epoch": 0.03668859417308365,
1323
+ "grad_norm": 0.21156178414821625,
1324
+ "learning_rate": 0.0005,
1325
+ "loss": 3.0510904788970947,
1326
+ "step": 376
1327
+ },
1328
+ {
1329
+ "epoch": 0.03688374626974899,
1330
+ "grad_norm": 0.2524719536304474,
1331
+ "learning_rate": 0.0005,
1332
+ "loss": 2.9311161041259766,
1333
+ "step": 378
1334
+ },
1335
+ {
1336
+ "epoch": 0.03707889836641432,
1337
+ "grad_norm": 0.22962909936904907,
1338
+ "learning_rate": 0.0005,
1339
+ "loss": 3.0110621452331543,
1340
+ "step": 380
1341
+ },
1342
+ {
1343
+ "epoch": 0.03727405046307966,
1344
+ "grad_norm": 0.30115097761154175,
1345
+ "learning_rate": 0.0005,
1346
+ "loss": 2.9791746139526367,
1347
+ "step": 382
1348
+ },
1349
+ {
1350
+ "epoch": 0.037469202559745,
1351
+ "grad_norm": 0.19829073548316956,
1352
+ "learning_rate": 0.0005,
1353
+ "loss": 2.942716598510742,
1354
+ "step": 384
1355
+ },
1356
+ {
1357
+ "epoch": 0.03766435465641034,
1358
+ "grad_norm": 0.2170461267232895,
1359
+ "learning_rate": 0.0005,
1360
+ "loss": 2.972839593887329,
1361
+ "step": 386
1362
+ },
1363
+ {
1364
+ "epoch": 0.03785950675307568,
1365
+ "grad_norm": 0.21176321804523468,
1366
+ "learning_rate": 0.0005,
1367
+ "loss": 3.0025084018707275,
1368
+ "step": 388
1369
+ },
1370
+ {
1371
+ "epoch": 0.03805465884974102,
1372
+ "grad_norm": 0.210454061627388,
1373
+ "learning_rate": 0.0005,
1374
+ "loss": 2.974094867706299,
1375
+ "step": 390
1376
+ },
1377
+ {
1378
+ "epoch": 0.03824981094640636,
1379
+ "grad_norm": 0.24710071086883545,
1380
+ "learning_rate": 0.0005,
1381
+ "loss": 3.020918369293213,
1382
+ "step": 392
1383
+ },
1384
+ {
1385
+ "epoch": 0.03844496304307169,
1386
+ "grad_norm": 0.21508663892745972,
1387
+ "learning_rate": 0.0005,
1388
+ "loss": 2.9229345321655273,
1389
+ "step": 394
1390
+ },
1391
+ {
1392
+ "epoch": 0.03864011513973703,
1393
+ "grad_norm": 0.20848804712295532,
1394
+ "learning_rate": 0.0005,
1395
+ "loss": 2.9745874404907227,
1396
+ "step": 396
1397
+ },
1398
+ {
1399
+ "epoch": 0.03883526723640237,
1400
+ "grad_norm": 0.21484419703483582,
1401
+ "learning_rate": 0.0005,
1402
+ "loss": 3.0044634342193604,
1403
+ "step": 398
1404
+ },
1405
+ {
1406
+ "epoch": 0.03903041933306771,
1407
+ "grad_norm": 0.21250562369823456,
1408
+ "learning_rate": 0.0005,
1409
+ "loss": 3.0184390544891357,
1410
+ "step": 400
1411
+ },
1412
+ {
1413
+ "epoch": 0.03922557142973305,
1414
+ "grad_norm": 0.23485791683197021,
1415
+ "learning_rate": 0.0005,
1416
+ "loss": 2.9293127059936523,
1417
+ "step": 402
1418
+ },
1419
+ {
1420
+ "epoch": 0.03942072352639839,
1421
+ "grad_norm": 0.23742340505123138,
1422
+ "learning_rate": 0.0005,
1423
+ "loss": 2.9542768001556396,
1424
+ "step": 404
1425
+ },
1426
+ {
1427
+ "epoch": 0.03961587562306373,
1428
+ "grad_norm": 0.1888517439365387,
1429
+ "learning_rate": 0.0005,
1430
+ "loss": 2.8617591857910156,
1431
+ "step": 406
1432
+ },
1433
+ {
1434
+ "epoch": 0.03981102771972907,
1435
+ "grad_norm": 0.18860198557376862,
1436
+ "learning_rate": 0.0005,
1437
+ "loss": 3.034463405609131,
1438
+ "step": 408
1439
+ },
1440
+ {
1441
+ "epoch": 0.0400061798163944,
1442
+ "grad_norm": 0.2608613967895508,
1443
+ "learning_rate": 0.0005,
1444
+ "loss": 2.855318069458008,
1445
+ "step": 410
1446
+ },
1447
+ {
1448
+ "epoch": 0.04020133191305974,
1449
+ "grad_norm": 0.2288719117641449,
1450
+ "learning_rate": 0.0005,
1451
+ "loss": 2.9284069538116455,
1452
+ "step": 412
1453
+ },
1454
+ {
1455
+ "epoch": 0.04039648400972508,
1456
+ "grad_norm": 0.20852726697921753,
1457
+ "learning_rate": 0.0005,
1458
+ "loss": 3.0395500659942627,
1459
+ "step": 414
1460
+ },
1461
+ {
1462
+ "epoch": 0.04059163610639042,
1463
+ "grad_norm": 0.214315265417099,
1464
+ "learning_rate": 0.0005,
1465
+ "loss": 3.0136606693267822,
1466
+ "step": 416
1467
+ },
1468
+ {
1469
+ "epoch": 0.04078678820305576,
1470
+ "grad_norm": 0.22608111798763275,
1471
+ "learning_rate": 0.0005,
1472
+ "loss": 2.89285945892334,
1473
+ "step": 418
1474
+ },
1475
+ {
1476
+ "epoch": 0.0409819402997211,
1477
+ "grad_norm": 0.266006201505661,
1478
+ "learning_rate": 0.0005,
1479
+ "loss": 2.9661340713500977,
1480
+ "step": 420
1481
+ },
1482
+ {
1483
+ "epoch": 0.041177092396386436,
1484
+ "grad_norm": 0.20414559543132782,
1485
+ "learning_rate": 0.0005,
1486
+ "loss": 2.9263293743133545,
1487
+ "step": 422
1488
+ },
1489
+ {
1490
+ "epoch": 0.041372244493051775,
1491
+ "grad_norm": 0.22027015686035156,
1492
+ "learning_rate": 0.0005,
1493
+ "loss": 2.9073097705841064,
1494
+ "step": 424
1495
+ },
1496
+ {
1497
+ "epoch": 0.04156739658971711,
1498
+ "grad_norm": 0.38255390524864197,
1499
+ "learning_rate": 0.0005,
1500
+ "loss": 2.983393430709839,
1501
+ "step": 426
1502
+ },
1503
+ {
1504
+ "epoch": 0.04176254868638245,
1505
+ "grad_norm": 0.2537851929664612,
1506
+ "learning_rate": 0.0005,
1507
+ "loss": 2.984750986099243,
1508
+ "step": 428
1509
+ },
1510
+ {
1511
+ "epoch": 0.04195770078304779,
1512
+ "grad_norm": 0.1681070625782013,
1513
+ "learning_rate": 0.0005,
1514
+ "loss": 2.8965325355529785,
1515
+ "step": 430
1516
+ },
1517
+ {
1518
+ "epoch": 0.042152852879713126,
1519
+ "grad_norm": 0.3319982886314392,
1520
+ "learning_rate": 0.0005,
1521
+ "loss": 2.970005750656128,
1522
+ "step": 432
1523
+ },
1524
+ {
1525
+ "epoch": 0.042348004976378466,
1526
+ "grad_norm": 0.3258397877216339,
1527
+ "learning_rate": 0.0005,
1528
+ "loss": 2.925297498703003,
1529
+ "step": 434
1530
+ },
1531
+ {
1532
+ "epoch": 0.042543157073043805,
1533
+ "grad_norm": 0.25465279817581177,
1534
+ "learning_rate": 0.0005,
1535
+ "loss": 2.9281442165374756,
1536
+ "step": 436
1537
+ },
1538
+ {
1539
+ "epoch": 0.042738309169709145,
1540
+ "grad_norm": 0.1869325339794159,
1541
+ "learning_rate": 0.0005,
1542
+ "loss": 2.9376168251037598,
1543
+ "step": 438
1544
+ },
1545
+ {
1546
+ "epoch": 0.04293346126637448,
1547
+ "grad_norm": 0.22072185575962067,
1548
+ "learning_rate": 0.0005,
1549
+ "loss": 2.9774351119995117,
1550
+ "step": 440
1551
+ },
1552
+ {
1553
+ "epoch": 0.04312861336303982,
1554
+ "grad_norm": 0.3213740885257721,
1555
+ "learning_rate": 0.0005,
1556
+ "loss": 2.884382486343384,
1557
+ "step": 442
1558
+ },
1559
+ {
1560
+ "epoch": 0.043323765459705156,
1561
+ "grad_norm": 0.19458690285682678,
1562
+ "learning_rate": 0.0005,
1563
+ "loss": 2.950810432434082,
1564
+ "step": 444
1565
+ },
1566
+ {
1567
+ "epoch": 0.043518917556370496,
1568
+ "grad_norm": 0.22102844715118408,
1569
+ "learning_rate": 0.0005,
1570
+ "loss": 2.9985313415527344,
1571
+ "step": 446
1572
+ },
1573
+ {
1574
+ "epoch": 0.043714069653035835,
1575
+ "grad_norm": 0.2425060123205185,
1576
+ "learning_rate": 0.0005,
1577
+ "loss": 2.9392502307891846,
1578
+ "step": 448
1579
+ },
1580
+ {
1581
+ "epoch": 0.043909221749701174,
1582
+ "grad_norm": 0.20293886959552765,
1583
+ "learning_rate": 0.0005,
1584
+ "loss": 2.8980088233947754,
1585
+ "step": 450
1586
+ },
1587
+ {
1588
+ "epoch": 0.044104373846366514,
1589
+ "grad_norm": 0.20620623230934143,
1590
+ "learning_rate": 0.0005,
1591
+ "loss": 2.911039113998413,
1592
+ "step": 452
1593
+ },
1594
+ {
1595
+ "epoch": 0.04429952594303185,
1596
+ "grad_norm": 0.22259211540222168,
1597
+ "learning_rate": 0.0005,
1598
+ "loss": 2.9042744636535645,
1599
+ "step": 454
1600
+ },
1601
+ {
1602
+ "epoch": 0.044494678039697186,
1603
+ "grad_norm": 0.21232987940311432,
1604
+ "learning_rate": 0.0005,
1605
+ "loss": 2.9142513275146484,
1606
+ "step": 456
1607
+ },
1608
+ {
1609
+ "epoch": 0.044689830136362525,
1610
+ "grad_norm": 0.2812767028808594,
1611
+ "learning_rate": 0.0005,
1612
+ "loss": 2.928365707397461,
1613
+ "step": 458
1614
+ },
1615
+ {
1616
+ "epoch": 0.044884982233027865,
1617
+ "grad_norm": 0.25571298599243164,
1618
+ "learning_rate": 0.0005,
1619
+ "loss": 2.982306957244873,
1620
+ "step": 460
1621
+ },
1622
+ {
1623
+ "epoch": 0.045080134329693204,
1624
+ "grad_norm": 0.2073540836572647,
1625
+ "learning_rate": 0.0005,
1626
+ "loss": 2.9568138122558594,
1627
+ "step": 462
1628
+ },
1629
+ {
1630
+ "epoch": 0.045275286426358544,
1631
+ "grad_norm": 0.21256600320339203,
1632
+ "learning_rate": 0.0005,
1633
+ "loss": 2.963731050491333,
1634
+ "step": 464
1635
+ },
1636
+ {
1637
+ "epoch": 0.04547043852302388,
1638
+ "grad_norm": 0.2165621668100357,
1639
+ "learning_rate": 0.0005,
1640
+ "loss": 2.867647647857666,
1641
+ "step": 466
1642
+ },
1643
+ {
1644
+ "epoch": 0.04566559061968922,
1645
+ "grad_norm": 0.20546843111515045,
1646
+ "learning_rate": 0.0005,
1647
+ "loss": 2.8600211143493652,
1648
+ "step": 468
1649
+ },
1650
+ {
1651
+ "epoch": 0.04586074271635456,
1652
+ "grad_norm": 0.25244683027267456,
1653
+ "learning_rate": 0.0005,
1654
+ "loss": 2.958961009979248,
1655
+ "step": 470
1656
+ },
1657
+ {
1658
+ "epoch": 0.046055894813019894,
1659
+ "grad_norm": 0.2383929193019867,
1660
+ "learning_rate": 0.0005,
1661
+ "loss": 2.9122891426086426,
1662
+ "step": 472
1663
+ },
1664
+ {
1665
+ "epoch": 0.046251046909685234,
1666
+ "grad_norm": 0.21767760813236237,
1667
+ "learning_rate": 0.0005,
1668
+ "loss": 2.8562703132629395,
1669
+ "step": 474
1670
+ },
1671
+ {
1672
+ "epoch": 0.04644619900635057,
1673
+ "grad_norm": 0.2181265950202942,
1674
+ "learning_rate": 0.0005,
1675
+ "loss": 2.9913651943206787,
1676
+ "step": 476
1677
+ },
1678
+ {
1679
+ "epoch": 0.04664135110301591,
1680
+ "grad_norm": 0.21227063238620758,
1681
+ "learning_rate": 0.0005,
1682
+ "loss": 2.9247255325317383,
1683
+ "step": 478
1684
+ },
1685
+ {
1686
+ "epoch": 0.04683650319968125,
1687
+ "grad_norm": 0.2251858413219452,
1688
+ "learning_rate": 0.0005,
1689
+ "loss": 2.8642032146453857,
1690
+ "step": 480
1691
+ },
1692
+ {
1693
+ "epoch": 0.04703165529634659,
1694
+ "grad_norm": 0.2021237015724182,
1695
+ "learning_rate": 0.0005,
1696
+ "loss": 2.8023080825805664,
1697
+ "step": 482
1698
+ },
1699
+ {
1700
+ "epoch": 0.04722680739301193,
1701
+ "grad_norm": 0.2923108637332916,
1702
+ "learning_rate": 0.0005,
1703
+ "loss": 2.896533250808716,
1704
+ "step": 484
1705
+ },
1706
+ {
1707
+ "epoch": 0.04742195948967727,
1708
+ "grad_norm": 0.2397584617137909,
1709
+ "learning_rate": 0.0005,
1710
+ "loss": 2.9412147998809814,
1711
+ "step": 486
1712
+ },
1713
+ {
1714
+ "epoch": 0.0476171115863426,
1715
+ "grad_norm": 0.22489869594573975,
1716
+ "learning_rate": 0.0005,
1717
+ "loss": 2.843785047531128,
1718
+ "step": 488
1719
+ },
1720
+ {
1721
+ "epoch": 0.04781226368300794,
1722
+ "grad_norm": 0.20172391831874847,
1723
+ "learning_rate": 0.0005,
1724
+ "loss": 2.8502678871154785,
1725
+ "step": 490
1726
+ },
1727
+ {
1728
+ "epoch": 0.04800741577967328,
1729
+ "grad_norm": 0.20748472213745117,
1730
+ "learning_rate": 0.0005,
1731
+ "loss": 2.913743019104004,
1732
+ "step": 492
1733
+ },
1734
+ {
1735
+ "epoch": 0.04820256787633862,
1736
+ "grad_norm": 0.1866624355316162,
1737
+ "learning_rate": 0.0005,
1738
+ "loss": 2.908566474914551,
1739
+ "step": 494
1740
+ },
1741
+ {
1742
+ "epoch": 0.04839771997300396,
1743
+ "grad_norm": 0.24492257833480835,
1744
+ "learning_rate": 0.0005,
1745
+ "loss": 2.9202466011047363,
1746
+ "step": 496
1747
+ },
1748
+ {
1749
+ "epoch": 0.0485928720696693,
1750
+ "grad_norm": 0.19961591064929962,
1751
+ "learning_rate": 0.0005,
1752
+ "loss": 2.8069441318511963,
1753
+ "step": 498
1754
+ },
1755
+ {
1756
+ "epoch": 0.04878802416633464,
1757
+ "grad_norm": 0.20612908899784088,
1758
+ "learning_rate": 0.0005,
1759
+ "loss": 2.8107845783233643,
1760
+ "step": 500
1761
+ },
1762
+ {
1763
+ "epoch": 0.04898317626299997,
1764
+ "grad_norm": 0.2411632388830185,
1765
+ "learning_rate": 0.0005,
1766
+ "loss": 2.9048216342926025,
1767
+ "step": 502
1768
+ },
1769
+ {
1770
+ "epoch": 0.04917832835966531,
1771
+ "grad_norm": 0.2392047941684723,
1772
+ "learning_rate": 0.0005,
1773
+ "loss": 2.876821279525757,
1774
+ "step": 504
1775
+ },
1776
+ {
1777
+ "epoch": 0.04937348045633065,
1778
+ "grad_norm": 0.22393734753131866,
1779
+ "learning_rate": 0.0005,
1780
+ "loss": 2.8603787422180176,
1781
+ "step": 506
1782
+ },
1783
+ {
1784
+ "epoch": 0.04956863255299599,
1785
+ "grad_norm": 0.20716209709644318,
1786
+ "learning_rate": 0.0005,
1787
+ "loss": 2.799652576446533,
1788
+ "step": 508
1789
+ },
1790
+ {
1791
+ "epoch": 0.04976378464966133,
1792
+ "grad_norm": 0.22787153720855713,
1793
+ "learning_rate": 0.0005,
1794
+ "loss": 2.897719621658325,
1795
+ "step": 510
1796
+ },
1797
+ {
1798
+ "epoch": 0.04995893674632667,
1799
+ "grad_norm": 0.24592295289039612,
1800
+ "learning_rate": 0.0005,
1801
+ "loss": 2.797395706176758,
1802
+ "step": 512
1803
+ },
1804
+ {
1805
+ "epoch": 0.05015408884299201,
1806
+ "grad_norm": 0.2608513832092285,
1807
+ "learning_rate": 0.0005,
1808
+ "loss": 2.9166135787963867,
1809
+ "step": 514
1810
+ },
1811
+ {
1812
+ "epoch": 0.05034924093965735,
1813
+ "grad_norm": 0.2029501348733902,
1814
+ "learning_rate": 0.0005,
1815
+ "loss": 2.7752950191497803,
1816
+ "step": 516
1817
+ },
1818
+ {
1819
+ "epoch": 0.05054439303632268,
1820
+ "grad_norm": 0.2164735496044159,
1821
+ "learning_rate": 0.0005,
1822
+ "loss": 2.862685203552246,
1823
+ "step": 518
1824
+ },
1825
+ {
1826
+ "epoch": 0.05073954513298802,
1827
+ "grad_norm": 0.22805725038051605,
1828
+ "learning_rate": 0.0005,
1829
+ "loss": 2.925373077392578,
1830
+ "step": 520
1831
+ },
1832
+ {
1833
+ "epoch": 0.05093469722965336,
1834
+ "grad_norm": 0.24554982781410217,
1835
+ "learning_rate": 0.0005,
1836
+ "loss": 2.9381539821624756,
1837
+ "step": 522
1838
+ },
1839
+ {
1840
+ "epoch": 0.0511298493263187,
1841
+ "grad_norm": 0.2538678050041199,
1842
+ "learning_rate": 0.0005,
1843
+ "loss": 2.909632921218872,
1844
+ "step": 524
1845
+ },
1846
+ {
1847
+ "epoch": 0.05132500142298404,
1848
+ "grad_norm": 0.221237450838089,
1849
+ "learning_rate": 0.0005,
1850
+ "loss": 2.9510080814361572,
1851
+ "step": 526
1852
+ },
1853
+ {
1854
+ "epoch": 0.05152015351964938,
1855
+ "grad_norm": 0.2834848463535309,
1856
+ "learning_rate": 0.0005,
1857
+ "loss": 2.899319887161255,
1858
+ "step": 528
1859
+ },
1860
+ {
1861
+ "epoch": 0.05171530561631472,
1862
+ "grad_norm": 0.25846970081329346,
1863
+ "learning_rate": 0.0005,
1864
+ "loss": 2.9485363960266113,
1865
+ "step": 530
1866
+ },
1867
+ {
1868
+ "epoch": 0.05191045771298006,
1869
+ "grad_norm": 0.21804043650627136,
1870
+ "learning_rate": 0.0005,
1871
+ "loss": 2.7544872760772705,
1872
+ "step": 532
1873
+ },
1874
+ {
1875
+ "epoch": 0.05210560980964539,
1876
+ "grad_norm": 0.225544735789299,
1877
+ "learning_rate": 0.0005,
1878
+ "loss": 2.8754022121429443,
1879
+ "step": 534
1880
+ },
1881
+ {
1882
+ "epoch": 0.05230076190631073,
1883
+ "grad_norm": 0.2300245612859726,
1884
+ "learning_rate": 0.0005,
1885
+ "loss": 2.8083090782165527,
1886
+ "step": 536
1887
+ },
1888
+ {
1889
+ "epoch": 0.05249591400297607,
1890
+ "grad_norm": 0.25335201621055603,
1891
+ "learning_rate": 0.0005,
1892
+ "loss": 2.864687442779541,
1893
+ "step": 538
1894
+ },
1895
+ {
1896
+ "epoch": 0.05269106609964141,
1897
+ "grad_norm": 0.23570439219474792,
1898
+ "learning_rate": 0.0005,
1899
+ "loss": 2.857375383377075,
1900
+ "step": 540
1901
+ },
1902
+ {
1903
+ "epoch": 0.05288621819630675,
1904
+ "grad_norm": 0.2167489379644394,
1905
+ "learning_rate": 0.0005,
1906
+ "loss": 2.80448579788208,
1907
+ "step": 542
1908
+ },
1909
+ {
1910
+ "epoch": 0.05308137029297209,
1911
+ "grad_norm": 0.26908189058303833,
1912
+ "learning_rate": 0.0005,
1913
+ "loss": 2.8551199436187744,
1914
+ "step": 544
1915
+ },
1916
+ {
1917
+ "epoch": 0.053276522389637426,
1918
+ "grad_norm": 0.23771652579307556,
1919
+ "learning_rate": 0.0005,
1920
+ "loss": 2.9150502681732178,
1921
+ "step": 546
1922
+ },
1923
+ {
1924
+ "epoch": 0.05347167448630276,
1925
+ "grad_norm": 0.2271955907344818,
1926
+ "learning_rate": 0.0005,
1927
+ "loss": 2.7826144695281982,
1928
+ "step": 548
1929
+ },
1930
+ {
1931
+ "epoch": 0.0536668265829681,
1932
+ "grad_norm": 0.24746520817279816,
1933
+ "learning_rate": 0.0005,
1934
+ "loss": 2.8394107818603516,
1935
+ "step": 550
1936
+ },
1937
+ {
1938
+ "epoch": 0.05386197867963344,
1939
+ "grad_norm": 0.26666754484176636,
1940
+ "learning_rate": 0.0005,
1941
+ "loss": 2.8330442905426025,
1942
+ "step": 552
1943
+ },
1944
+ {
1945
+ "epoch": 0.05405713077629878,
1946
+ "grad_norm": 0.2330562323331833,
1947
+ "learning_rate": 0.0005,
1948
+ "loss": 2.8823511600494385,
1949
+ "step": 554
1950
+ },
1951
+ {
1952
+ "epoch": 0.05425228287296412,
1953
+ "grad_norm": 0.240234375,
1954
+ "learning_rate": 0.0005,
1955
+ "loss": 2.8281450271606445,
1956
+ "step": 556
1957
+ },
1958
+ {
1959
+ "epoch": 0.054447434969629456,
1960
+ "grad_norm": 0.22279725968837738,
1961
+ "learning_rate": 0.0005,
1962
+ "loss": 2.8552358150482178,
1963
+ "step": 558
1964
+ },
1965
+ {
1966
+ "epoch": 0.054642587066294795,
1967
+ "grad_norm": 0.21800941228866577,
1968
+ "learning_rate": 0.0005,
1969
+ "loss": 2.81900691986084,
1970
+ "step": 560
1971
+ },
1972
+ {
1973
+ "epoch": 0.054837739162960135,
1974
+ "grad_norm": 0.227288618683815,
1975
+ "learning_rate": 0.0005,
1976
+ "loss": 2.8491249084472656,
1977
+ "step": 562
1978
+ },
1979
+ {
1980
+ "epoch": 0.05503289125962547,
1981
+ "grad_norm": 0.2596258223056793,
1982
+ "learning_rate": 0.0005,
1983
+ "loss": 2.90956711769104,
1984
+ "step": 564
1985
+ },
1986
+ {
1987
+ "epoch": 0.05522804335629081,
1988
+ "grad_norm": 0.2653347849845886,
1989
+ "learning_rate": 0.0005,
1990
+ "loss": 2.8402161598205566,
1991
+ "step": 566
1992
+ },
1993
+ {
1994
+ "epoch": 0.055423195452956146,
1995
+ "grad_norm": 0.28130441904067993,
1996
+ "learning_rate": 0.0005,
1997
+ "loss": 2.841031312942505,
1998
+ "step": 568
1999
+ },
2000
+ {
2001
+ "epoch": 0.055618347549621486,
2002
+ "grad_norm": 0.2281201183795929,
2003
+ "learning_rate": 0.0005,
2004
+ "loss": 2.868011713027954,
2005
+ "step": 570
2006
+ },
2007
+ {
2008
+ "epoch": 0.055813499646286825,
2009
+ "grad_norm": 0.23963680863380432,
2010
+ "learning_rate": 0.0005,
2011
+ "loss": 2.826296806335449,
2012
+ "step": 572
2013
+ },
2014
+ {
2015
+ "epoch": 0.056008651742952165,
2016
+ "grad_norm": 0.23558886349201202,
2017
+ "learning_rate": 0.0005,
2018
+ "loss": 2.8189809322357178,
2019
+ "step": 574
2020
+ },
2021
+ {
2022
+ "epoch": 0.056203803839617504,
2023
+ "grad_norm": 0.23629260063171387,
2024
+ "learning_rate": 0.0005,
2025
+ "loss": 2.723668336868286,
2026
+ "step": 576
2027
+ },
2028
+ {
2029
+ "epoch": 0.056398955936282844,
2030
+ "grad_norm": 0.22436900436878204,
2031
+ "learning_rate": 0.0005,
2032
+ "loss": 2.9019086360931396,
2033
+ "step": 578
2034
+ },
2035
+ {
2036
+ "epoch": 0.056594108032948176,
2037
+ "grad_norm": 0.23167671263217926,
2038
+ "learning_rate": 0.0005,
2039
+ "loss": 2.8314285278320312,
2040
+ "step": 580
2041
+ },
2042
+ {
2043
+ "epoch": 0.056789260129613515,
2044
+ "grad_norm": 0.25156769156455994,
2045
+ "learning_rate": 0.0005,
2046
+ "loss": 2.7796947956085205,
2047
+ "step": 582
2048
+ },
2049
+ {
2050
+ "epoch": 0.056984412226278855,
2051
+ "grad_norm": 0.22697637975215912,
2052
+ "learning_rate": 0.0005,
2053
+ "loss": 2.8541312217712402,
2054
+ "step": 584
2055
+ },
2056
+ {
2057
+ "epoch": 0.057179564322944194,
2058
+ "grad_norm": 0.25153395533561707,
2059
+ "learning_rate": 0.0005,
2060
+ "loss": 2.7868356704711914,
2061
+ "step": 586
2062
+ },
2063
+ {
2064
+ "epoch": 0.057374716419609534,
2065
+ "grad_norm": 0.20788004994392395,
2066
+ "learning_rate": 0.0005,
2067
+ "loss": 2.852097749710083,
2068
+ "step": 588
2069
+ },
2070
+ {
2071
+ "epoch": 0.05756986851627487,
2072
+ "grad_norm": 0.2148260772228241,
2073
+ "learning_rate": 0.0005,
2074
+ "loss": 2.8623924255371094,
2075
+ "step": 590
2076
+ },
2077
+ {
2078
+ "epoch": 0.05776502061294021,
2079
+ "grad_norm": 0.2672250270843506,
2080
+ "learning_rate": 0.0005,
2081
+ "loss": 2.882892370223999,
2082
+ "step": 592
2083
+ },
2084
+ {
2085
+ "epoch": 0.05796017270960555,
2086
+ "grad_norm": 0.21002788841724396,
2087
+ "learning_rate": 0.0005,
2088
+ "loss": 2.7354979515075684,
2089
+ "step": 594
2090
+ },
2091
+ {
2092
+ "epoch": 0.058155324806270885,
2093
+ "grad_norm": 0.25582656264305115,
2094
+ "learning_rate": 0.0005,
2095
+ "loss": 2.8117384910583496,
2096
+ "step": 596
2097
+ },
2098
+ {
2099
+ "epoch": 0.058350476902936224,
2100
+ "grad_norm": 0.22788363695144653,
2101
+ "learning_rate": 0.0005,
2102
+ "loss": 2.8293497562408447,
2103
+ "step": 598
2104
+ },
2105
+ {
2106
+ "epoch": 0.058545628999601564,
2107
+ "grad_norm": 0.20921580493450165,
2108
+ "learning_rate": 0.0005,
2109
+ "loss": 2.755247116088867,
2110
+ "step": 600
2111
+ },
2112
+ {
2113
+ "epoch": 0.0587407810962669,
2114
+ "grad_norm": 0.26626056432724,
2115
+ "learning_rate": 0.0005,
2116
+ "loss": 2.8806607723236084,
2117
+ "step": 602
2118
+ },
2119
+ {
2120
+ "epoch": 0.05893593319293224,
2121
+ "grad_norm": 0.2332177460193634,
2122
+ "learning_rate": 0.0005,
2123
+ "loss": 2.840040445327759,
2124
+ "step": 604
2125
+ },
2126
+ {
2127
+ "epoch": 0.05913108528959758,
2128
+ "grad_norm": 0.2499295026063919,
2129
+ "learning_rate": 0.0005,
2130
+ "loss": 2.8295679092407227,
2131
+ "step": 606
2132
+ },
2133
+ {
2134
+ "epoch": 0.05932623738626292,
2135
+ "grad_norm": 0.23084314167499542,
2136
+ "learning_rate": 0.0005,
2137
+ "loss": 2.75656795501709,
2138
+ "step": 608
2139
+ },
2140
+ {
2141
+ "epoch": 0.059521389482928254,
2142
+ "grad_norm": 0.2571105360984802,
2143
+ "learning_rate": 0.0005,
2144
+ "loss": 2.7750117778778076,
2145
+ "step": 610
2146
+ },
2147
+ {
2148
+ "epoch": 0.05971654157959359,
2149
+ "grad_norm": 0.21366983652114868,
2150
+ "learning_rate": 0.0005,
2151
+ "loss": 2.8432583808898926,
2152
+ "step": 612
2153
+ },
2154
+ {
2155
+ "epoch": 0.05991169367625893,
2156
+ "grad_norm": 0.25937730073928833,
2157
+ "learning_rate": 0.0005,
2158
+ "loss": 2.852489948272705,
2159
+ "step": 614
2160
+ },
2161
+ {
2162
+ "epoch": 0.06010684577292427,
2163
+ "grad_norm": 0.2423776239156723,
2164
+ "learning_rate": 0.0005,
2165
+ "loss": 2.796426773071289,
2166
+ "step": 616
2167
+ },
2168
+ {
2169
+ "epoch": 0.06030199786958961,
2170
+ "grad_norm": 0.23655439913272858,
2171
+ "learning_rate": 0.0005,
2172
+ "loss": 2.7959790229797363,
2173
+ "step": 618
2174
+ },
2175
+ {
2176
+ "epoch": 0.06049714996625495,
2177
+ "grad_norm": 0.21120844781398773,
2178
+ "learning_rate": 0.0005,
2179
+ "loss": 2.73063588142395,
2180
+ "step": 620
2181
+ },
2182
+ {
2183
+ "epoch": 0.06069230206292029,
2184
+ "grad_norm": 0.2508559226989746,
2185
+ "learning_rate": 0.0005,
2186
+ "loss": 2.7917282581329346,
2187
+ "step": 622
2188
+ },
2189
+ {
2190
+ "epoch": 0.06088745415958563,
2191
+ "grad_norm": 0.2277577966451645,
2192
+ "learning_rate": 0.0005,
2193
+ "loss": 2.698517084121704,
2194
+ "step": 624
2195
+ },
2196
+ {
2197
+ "epoch": 0.06108260625625096,
2198
+ "grad_norm": 0.21368353068828583,
2199
+ "learning_rate": 0.0005,
2200
+ "loss": 2.870924472808838,
2201
+ "step": 626
2202
+ },
2203
+ {
2204
+ "epoch": 0.0612777583529163,
2205
+ "grad_norm": 0.21780502796173096,
2206
+ "learning_rate": 0.0005,
2207
+ "loss": 2.78973388671875,
2208
+ "step": 628
2209
+ },
2210
+ {
2211
+ "epoch": 0.06147291044958164,
2212
+ "grad_norm": 0.25150877237319946,
2213
+ "learning_rate": 0.0005,
2214
+ "loss": 2.9091720581054688,
2215
+ "step": 630
2216
+ },
2217
+ {
2218
+ "epoch": 0.06166806254624698,
2219
+ "grad_norm": 0.24942591786384583,
2220
+ "learning_rate": 0.0005,
2221
+ "loss": 2.8442296981811523,
2222
+ "step": 632
2223
+ },
2224
+ {
2225
+ "epoch": 0.06186321464291232,
2226
+ "grad_norm": 0.2594287395477295,
2227
+ "learning_rate": 0.0005,
2228
+ "loss": 2.8517606258392334,
2229
+ "step": 634
2230
+ },
2231
+ {
2232
+ "epoch": 0.06205836673957766,
2233
+ "grad_norm": 0.22809991240501404,
2234
+ "learning_rate": 0.0005,
2235
+ "loss": 2.8512609004974365,
2236
+ "step": 636
2237
+ },
2238
+ {
2239
+ "epoch": 0.062253518836243,
2240
+ "grad_norm": 0.2275650054216385,
2241
+ "learning_rate": 0.0005,
2242
+ "loss": 2.7290780544281006,
2243
+ "step": 638
2244
+ },
2245
+ {
2246
+ "epoch": 0.06244867093290834,
2247
+ "grad_norm": 0.2508498728275299,
2248
+ "learning_rate": 0.0005,
2249
+ "loss": 2.856473922729492,
2250
+ "step": 640
2251
+ },
2252
+ {
2253
+ "epoch": 0.06264382302957368,
2254
+ "grad_norm": 0.2302878051996231,
2255
+ "learning_rate": 0.0005,
2256
+ "loss": 2.781524181365967,
2257
+ "step": 642
2258
+ },
2259
+ {
2260
+ "epoch": 0.06283897512623901,
2261
+ "grad_norm": 0.18536455929279327,
2262
+ "learning_rate": 0.0005,
2263
+ "loss": 2.749098300933838,
2264
+ "step": 644
2265
+ },
2266
+ {
2267
+ "epoch": 0.06303412722290436,
2268
+ "grad_norm": 0.26464295387268066,
2269
+ "learning_rate": 0.0005,
2270
+ "loss": 2.758824348449707,
2271
+ "step": 646
2272
+ },
2273
+ {
2274
+ "epoch": 0.06322927931956969,
2275
+ "grad_norm": 0.2696143090724945,
2276
+ "learning_rate": 0.0005,
2277
+ "loss": 2.8641669750213623,
2278
+ "step": 648
2279
+ },
2280
+ {
2281
+ "epoch": 0.06342443141623502,
2282
+ "grad_norm": 0.22477640211582184,
2283
+ "learning_rate": 0.0005,
2284
+ "loss": 2.7943382263183594,
2285
+ "step": 650
2286
+ },
2287
+ {
2288
+ "epoch": 0.06361958351290037,
2289
+ "grad_norm": 0.21504759788513184,
2290
+ "learning_rate": 0.0005,
2291
+ "loss": 2.77803373336792,
2292
+ "step": 652
2293
+ },
2294
+ {
2295
+ "epoch": 0.0638147356095657,
2296
+ "grad_norm": 0.2068071961402893,
2297
+ "learning_rate": 0.0005,
2298
+ "loss": 2.776153564453125,
2299
+ "step": 654
2300
+ },
2301
+ {
2302
+ "epoch": 0.06400988770623105,
2303
+ "grad_norm": 0.24109873175621033,
2304
+ "learning_rate": 0.0005,
2305
+ "loss": 2.8750720024108887,
2306
+ "step": 656
2307
+ },
2308
+ {
2309
+ "epoch": 0.06420503980289638,
2310
+ "grad_norm": 0.21335522830486298,
2311
+ "learning_rate": 0.0005,
2312
+ "loss": 2.8156373500823975,
2313
+ "step": 658
2314
+ },
2315
+ {
2316
+ "epoch": 0.06440019189956173,
2317
+ "grad_norm": 0.21443277597427368,
2318
+ "learning_rate": 0.0005,
2319
+ "loss": 2.8247554302215576,
2320
+ "step": 660
2321
+ },
2322
+ {
2323
+ "epoch": 0.06459534399622706,
2324
+ "grad_norm": 0.2311883270740509,
2325
+ "learning_rate": 0.0005,
2326
+ "loss": 2.85162615776062,
2327
+ "step": 662
2328
+ },
2329
+ {
2330
+ "epoch": 0.06479049609289239,
2331
+ "grad_norm": 0.2611568868160248,
2332
+ "learning_rate": 0.0005,
2333
+ "loss": 2.7292983531951904,
2334
+ "step": 664
2335
+ },
2336
+ {
2337
+ "epoch": 0.06498564818955774,
2338
+ "grad_norm": 0.21040695905685425,
2339
+ "learning_rate": 0.0005,
2340
+ "loss": 2.7539312839508057,
2341
+ "step": 666
2342
+ },
2343
+ {
2344
+ "epoch": 0.06518080028622307,
2345
+ "grad_norm": 0.2189108282327652,
2346
+ "learning_rate": 0.0005,
2347
+ "loss": 2.7804532051086426,
2348
+ "step": 668
2349
+ },
2350
+ {
2351
+ "epoch": 0.06537595238288842,
2352
+ "grad_norm": 0.25914397835731506,
2353
+ "learning_rate": 0.0005,
2354
+ "loss": 2.8089828491210938,
2355
+ "step": 670
2356
+ },
2357
+ {
2358
+ "epoch": 0.06557110447955375,
2359
+ "grad_norm": 0.2535833418369293,
2360
+ "learning_rate": 0.0005,
2361
+ "loss": 2.7503957748413086,
2362
+ "step": 672
2363
+ },
2364
+ {
2365
+ "epoch": 0.0657662565762191,
2366
+ "grad_norm": 0.2291947305202484,
2367
+ "learning_rate": 0.0005,
2368
+ "loss": 2.8185744285583496,
2369
+ "step": 674
2370
+ },
2371
+ {
2372
+ "epoch": 0.06596140867288443,
2373
+ "grad_norm": 0.20832128822803497,
2374
+ "learning_rate": 0.0005,
2375
+ "loss": 2.8305342197418213,
2376
+ "step": 676
2377
+ },
2378
+ {
2379
+ "epoch": 0.06615656076954977,
2380
+ "grad_norm": 0.22745005786418915,
2381
+ "learning_rate": 0.0005,
2382
+ "loss": 2.837061643600464,
2383
+ "step": 678
2384
+ },
2385
+ {
2386
+ "epoch": 0.0663517128662151,
2387
+ "grad_norm": 0.20370373129844666,
2388
+ "learning_rate": 0.0005,
2389
+ "loss": 2.7723495960235596,
2390
+ "step": 680
2391
+ },
2392
+ {
2393
+ "epoch": 0.06654686496288044,
2394
+ "grad_norm": 0.20512208342552185,
2395
+ "learning_rate": 0.0005,
2396
+ "loss": 2.7923991680145264,
2397
+ "step": 682
2398
+ },
2399
+ {
2400
+ "epoch": 0.06674201705954579,
2401
+ "grad_norm": 0.25099486112594604,
2402
+ "learning_rate": 0.0005,
2403
+ "loss": 2.844620943069458,
2404
+ "step": 684
2405
+ },
2406
+ {
2407
+ "epoch": 0.06693716915621112,
2408
+ "grad_norm": 0.2210482358932495,
2409
+ "learning_rate": 0.0005,
2410
+ "loss": 2.7477450370788574,
2411
+ "step": 686
2412
+ },
2413
+ {
2414
+ "epoch": 0.06713232125287646,
2415
+ "grad_norm": 0.22797775268554688,
2416
+ "learning_rate": 0.0005,
2417
+ "loss": 2.7876176834106445,
2418
+ "step": 688
2419
+ },
2420
+ {
2421
+ "epoch": 0.0673274733495418,
2422
+ "grad_norm": 0.2500334084033966,
2423
+ "learning_rate": 0.0005,
2424
+ "loss": 2.818073034286499,
2425
+ "step": 690
2426
+ },
2427
+ {
2428
+ "epoch": 0.06752262544620714,
2429
+ "grad_norm": 0.18269109725952148,
2430
+ "learning_rate": 0.0005,
2431
+ "loss": 2.807344913482666,
2432
+ "step": 692
2433
+ },
2434
+ {
2435
+ "epoch": 0.06771777754287248,
2436
+ "grad_norm": 0.23339399695396423,
2437
+ "learning_rate": 0.0005,
2438
+ "loss": 2.819072723388672,
2439
+ "step": 694
2440
+ },
2441
+ {
2442
+ "epoch": 0.06791292963953781,
2443
+ "grad_norm": 0.2424178570508957,
2444
+ "learning_rate": 0.0005,
2445
+ "loss": 2.762280225753784,
2446
+ "step": 696
2447
+ },
2448
+ {
2449
+ "epoch": 0.06810808173620315,
2450
+ "grad_norm": 0.20848163962364197,
2451
+ "learning_rate": 0.0005,
2452
+ "loss": 2.831714153289795,
2453
+ "step": 698
2454
+ },
2455
+ {
2456
+ "epoch": 0.06830323383286849,
2457
+ "grad_norm": 0.1793297380208969,
2458
+ "learning_rate": 0.0005,
2459
+ "loss": 2.725076675415039,
2460
+ "step": 700
2461
+ },
2462
+ {
2463
+ "epoch": 0.06849838592953383,
2464
+ "grad_norm": 0.20033025741577148,
2465
+ "learning_rate": 0.0005,
2466
+ "loss": 2.7458736896514893,
2467
+ "step": 702
2468
+ },
2469
+ {
2470
+ "epoch": 0.06869353802619917,
2471
+ "grad_norm": 0.19496342539787292,
2472
+ "learning_rate": 0.0005,
2473
+ "loss": 2.686976909637451,
2474
+ "step": 704
2475
+ },
2476
+ {
2477
+ "epoch": 0.06888869012286451,
2478
+ "grad_norm": 0.21897919476032257,
2479
+ "learning_rate": 0.0005,
2480
+ "loss": 2.7467188835144043,
2481
+ "step": 706
2482
+ },
2483
+ {
2484
+ "epoch": 0.06908384221952985,
2485
+ "grad_norm": 0.2084447741508484,
2486
+ "learning_rate": 0.0005,
2487
+ "loss": 2.810333490371704,
2488
+ "step": 708
2489
+ },
2490
+ {
2491
+ "epoch": 0.06927899431619519,
2492
+ "grad_norm": 0.23198898136615753,
2493
+ "learning_rate": 0.0005,
2494
+ "loss": 2.8498759269714355,
2495
+ "step": 710
2496
+ },
2497
+ {
2498
+ "epoch": 0.06947414641286052,
2499
+ "grad_norm": 0.2230474054813385,
2500
+ "learning_rate": 0.0005,
2501
+ "loss": 2.804114818572998,
2502
+ "step": 712
2503
+ },
2504
+ {
2505
+ "epoch": 0.06966929850952586,
2506
+ "grad_norm": 0.23104287683963776,
2507
+ "learning_rate": 0.0005,
2508
+ "loss": 2.7120578289031982,
2509
+ "step": 714
2510
+ },
2511
+ {
2512
+ "epoch": 0.0698644506061912,
2513
+ "grad_norm": 0.28328996896743774,
2514
+ "learning_rate": 0.0005,
2515
+ "loss": 2.811950445175171,
2516
+ "step": 716
2517
+ },
2518
+ {
2519
+ "epoch": 0.07005960270285654,
2520
+ "grad_norm": 0.2158401906490326,
2521
+ "learning_rate": 0.0005,
2522
+ "loss": 2.7232327461242676,
2523
+ "step": 718
2524
+ },
2525
+ {
2526
+ "epoch": 0.07025475479952188,
2527
+ "grad_norm": 0.25869250297546387,
2528
+ "learning_rate": 0.0005,
2529
+ "loss": 2.7494895458221436,
2530
+ "step": 720
2531
+ },
2532
+ {
2533
+ "epoch": 0.07044990689618721,
2534
+ "grad_norm": 0.19535359740257263,
2535
+ "learning_rate": 0.0005,
2536
+ "loss": 2.8161709308624268,
2537
+ "step": 722
2538
+ },
2539
+ {
2540
+ "epoch": 0.07064505899285256,
2541
+ "grad_norm": 0.21670092642307281,
2542
+ "learning_rate": 0.0005,
2543
+ "loss": 2.785078525543213,
2544
+ "step": 724
2545
+ },
2546
+ {
2547
+ "epoch": 0.0708402110895179,
2548
+ "grad_norm": 0.20964759588241577,
2549
+ "learning_rate": 0.0005,
2550
+ "loss": 2.77406907081604,
2551
+ "step": 726
2552
+ },
2553
+ {
2554
+ "epoch": 0.07103536318618323,
2555
+ "grad_norm": 0.19224736094474792,
2556
+ "learning_rate": 0.0005,
2557
+ "loss": 2.720320463180542,
2558
+ "step": 728
2559
+ },
2560
+ {
2561
+ "epoch": 0.07123051528284857,
2562
+ "grad_norm": 0.20144300162792206,
2563
+ "learning_rate": 0.0005,
2564
+ "loss": 2.797353506088257,
2565
+ "step": 730
2566
+ },
2567
+ {
2568
+ "epoch": 0.0714256673795139,
2569
+ "grad_norm": 0.23987342417240143,
2570
+ "learning_rate": 0.0005,
2571
+ "loss": 2.727595090866089,
2572
+ "step": 732
2573
+ },
2574
+ {
2575
+ "epoch": 0.07162081947617925,
2576
+ "grad_norm": 0.23457464575767517,
2577
+ "learning_rate": 0.0005,
2578
+ "loss": 2.682112455368042,
2579
+ "step": 734
2580
+ },
2581
+ {
2582
+ "epoch": 0.07181597157284458,
2583
+ "grad_norm": 0.3230379819869995,
2584
+ "learning_rate": 0.0005,
2585
+ "loss": 2.720438003540039,
2586
+ "step": 736
2587
+ },
2588
+ {
2589
+ "epoch": 0.07201112366950993,
2590
+ "grad_norm": 0.24322040379047394,
2591
+ "learning_rate": 0.0005,
2592
+ "loss": 2.7975735664367676,
2593
+ "step": 738
2594
+ },
2595
+ {
2596
+ "epoch": 0.07220627576617526,
2597
+ "grad_norm": 0.23389576375484467,
2598
+ "learning_rate": 0.0005,
2599
+ "loss": 2.8103997707366943,
2600
+ "step": 740
2601
+ },
2602
+ {
2603
+ "epoch": 0.0724014278628406,
2604
+ "grad_norm": 0.3346785008907318,
2605
+ "learning_rate": 0.0005,
2606
+ "loss": 2.682776689529419,
2607
+ "step": 742
2608
+ },
2609
+ {
2610
+ "epoch": 0.07259657995950594,
2611
+ "grad_norm": 0.285838782787323,
2612
+ "learning_rate": 0.0005,
2613
+ "loss": 2.671548843383789,
2614
+ "step": 744
2615
+ },
2616
+ {
2617
+ "epoch": 0.07279173205617127,
2618
+ "grad_norm": 0.25126683712005615,
2619
+ "learning_rate": 0.0005,
2620
+ "loss": 2.767709255218506,
2621
+ "step": 746
2622
+ },
2623
+ {
2624
+ "epoch": 0.07298688415283662,
2625
+ "grad_norm": 0.21945233643054962,
2626
+ "learning_rate": 0.0005,
2627
+ "loss": 2.8341612815856934,
2628
+ "step": 748
2629
+ },
2630
+ {
2631
+ "epoch": 0.07318203624950195,
2632
+ "grad_norm": 0.19797329604625702,
2633
+ "learning_rate": 0.0005,
2634
+ "loss": 2.8269269466400146,
2635
+ "step": 750
2636
+ },
2637
+ {
2638
+ "epoch": 0.0733771883461673,
2639
+ "grad_norm": 0.22836004197597504,
2640
+ "learning_rate": 0.0005,
2641
+ "loss": 2.726588249206543,
2642
+ "step": 752
2643
+ },
2644
+ {
2645
+ "epoch": 0.07357234044283263,
2646
+ "grad_norm": 0.1990126520395279,
2647
+ "learning_rate": 0.0005,
2648
+ "loss": 2.6942567825317383,
2649
+ "step": 754
2650
+ },
2651
+ {
2652
+ "epoch": 0.07376749253949798,
2653
+ "grad_norm": 0.2066488265991211,
2654
+ "learning_rate": 0.0005,
2655
+ "loss": 2.7730367183685303,
2656
+ "step": 756
2657
+ },
2658
+ {
2659
+ "epoch": 0.07396264463616331,
2660
+ "grad_norm": 0.20968303084373474,
2661
+ "learning_rate": 0.0005,
2662
+ "loss": 2.7954494953155518,
2663
+ "step": 758
2664
+ },
2665
+ {
2666
+ "epoch": 0.07415779673282864,
2667
+ "grad_norm": 0.19573073089122772,
2668
+ "learning_rate": 0.0005,
2669
+ "loss": 2.740813970565796,
2670
+ "step": 760
2671
+ },
2672
+ {
2673
+ "epoch": 0.07435294882949399,
2674
+ "grad_norm": 0.2368002086877823,
2675
+ "learning_rate": 0.0005,
2676
+ "loss": 2.7246577739715576,
2677
+ "step": 762
2678
+ },
2679
+ {
2680
+ "epoch": 0.07454810092615932,
2681
+ "grad_norm": 0.21870261430740356,
2682
+ "learning_rate": 0.0005,
2683
+ "loss": 2.788027763366699,
2684
+ "step": 764
2685
+ },
2686
+ {
2687
+ "epoch": 0.07474325302282467,
2688
+ "grad_norm": 0.21037305891513824,
2689
+ "learning_rate": 0.0005,
2690
+ "loss": 2.7603602409362793,
2691
+ "step": 766
2692
+ },
2693
+ {
2694
+ "epoch": 0.07493840511949,
2695
+ "grad_norm": 0.24699488282203674,
2696
+ "learning_rate": 0.0005,
2697
+ "loss": 2.653430461883545,
2698
+ "step": 768
2699
+ },
2700
+ {
2701
+ "epoch": 0.07513355721615535,
2702
+ "grad_norm": 0.20402389764785767,
2703
+ "learning_rate": 0.0005,
2704
+ "loss": 2.748016595840454,
2705
+ "step": 770
2706
+ },
2707
+ {
2708
+ "epoch": 0.07532870931282068,
2709
+ "grad_norm": 0.2275436520576477,
2710
+ "learning_rate": 0.0005,
2711
+ "loss": 2.7779300212860107,
2712
+ "step": 772
2713
+ },
2714
+ {
2715
+ "epoch": 0.07552386140948601,
2716
+ "grad_norm": 0.2012409269809723,
2717
+ "learning_rate": 0.0005,
2718
+ "loss": 2.801553249359131,
2719
+ "step": 774
2720
+ },
2721
+ {
2722
+ "epoch": 0.07571901350615136,
2723
+ "grad_norm": 0.29552584886550903,
2724
+ "learning_rate": 0.0005,
2725
+ "loss": 2.827709674835205,
2726
+ "step": 776
2727
+ },
2728
+ {
2729
+ "epoch": 0.07591416560281669,
2730
+ "grad_norm": 0.2317826747894287,
2731
+ "learning_rate": 0.0005,
2732
+ "loss": 2.6379642486572266,
2733
+ "step": 778
2734
+ },
2735
+ {
2736
+ "epoch": 0.07610931769948204,
2737
+ "grad_norm": 0.2251746952533722,
2738
+ "learning_rate": 0.0005,
2739
+ "loss": 2.7535340785980225,
2740
+ "step": 780
2741
+ },
2742
+ {
2743
+ "epoch": 0.07630446979614737,
2744
+ "grad_norm": 0.23634979128837585,
2745
+ "learning_rate": 0.0005,
2746
+ "loss": 2.7316954135894775,
2747
+ "step": 782
2748
+ },
2749
+ {
2750
+ "epoch": 0.07649962189281272,
2751
+ "grad_norm": 0.20887678861618042,
2752
+ "learning_rate": 0.0005,
2753
+ "loss": 2.7955076694488525,
2754
+ "step": 784
2755
+ },
2756
+ {
2757
+ "epoch": 0.07669477398947805,
2758
+ "grad_norm": 0.2686167359352112,
2759
+ "learning_rate": 0.0005,
2760
+ "loss": 2.799690008163452,
2761
+ "step": 786
2762
+ },
2763
+ {
2764
+ "epoch": 0.07688992608614338,
2765
+ "grad_norm": 0.22938428819179535,
2766
+ "learning_rate": 0.0005,
2767
+ "loss": 2.75101637840271,
2768
+ "step": 788
2769
+ },
2770
+ {
2771
+ "epoch": 0.07708507818280873,
2772
+ "grad_norm": 0.1887194812297821,
2773
+ "learning_rate": 0.0005,
2774
+ "loss": 2.8088128566741943,
2775
+ "step": 790
2776
+ },
2777
+ {
2778
+ "epoch": 0.07728023027947406,
2779
+ "grad_norm": 0.2526529133319855,
2780
+ "learning_rate": 0.0005,
2781
+ "loss": 2.7890708446502686,
2782
+ "step": 792
2783
+ },
2784
+ {
2785
+ "epoch": 0.0774753823761394,
2786
+ "grad_norm": 0.21441319584846497,
2787
+ "learning_rate": 0.0005,
2788
+ "loss": 2.7117419242858887,
2789
+ "step": 794
2790
+ },
2791
+ {
2792
+ "epoch": 0.07767053447280474,
2793
+ "grad_norm": 0.20970889925956726,
2794
+ "learning_rate": 0.0005,
2795
+ "loss": 2.7044670581817627,
2796
+ "step": 796
2797
+ },
2798
+ {
2799
+ "epoch": 0.07786568656947009,
2800
+ "grad_norm": 0.28427496552467346,
2801
+ "learning_rate": 0.0005,
2802
+ "loss": 2.7297518253326416,
2803
+ "step": 798
2804
+ },
2805
+ {
2806
+ "epoch": 0.07806083866613542,
2807
+ "grad_norm": 0.20549088716506958,
2808
+ "learning_rate": 0.0005,
2809
+ "loss": 2.7174103260040283,
2810
+ "step": 800
2811
+ },
2812
+ {
2813
+ "epoch": 0.07825599076280076,
2814
+ "grad_norm": 0.20641227066516876,
2815
+ "learning_rate": 0.0005,
2816
+ "loss": 2.7044155597686768,
2817
+ "step": 802
2818
+ },
2819
+ {
2820
+ "epoch": 0.0784511428594661,
2821
+ "grad_norm": 0.2536081373691559,
2822
+ "learning_rate": 0.0005,
2823
+ "loss": 2.7087278366088867,
2824
+ "step": 804
2825
+ },
2826
+ {
2827
+ "epoch": 0.07864629495613143,
2828
+ "grad_norm": 0.27326247096061707,
2829
+ "learning_rate": 0.0005,
2830
+ "loss": 2.7364652156829834,
2831
+ "step": 806
2832
+ },
2833
+ {
2834
+ "epoch": 0.07884144705279678,
2835
+ "grad_norm": 0.21665054559707642,
2836
+ "learning_rate": 0.0005,
2837
+ "loss": 2.7687344551086426,
2838
+ "step": 808
2839
+ },
2840
+ {
2841
+ "epoch": 0.07903659914946211,
2842
+ "grad_norm": 0.25244149565696716,
2843
+ "learning_rate": 0.0005,
2844
+ "loss": 2.7000503540039062,
2845
+ "step": 810
2846
+ },
2847
+ {
2848
+ "epoch": 0.07923175124612745,
2849
+ "grad_norm": 0.22568374872207642,
2850
+ "learning_rate": 0.0005,
2851
+ "loss": 2.7546603679656982,
2852
+ "step": 812
2853
+ },
2854
+ {
2855
+ "epoch": 0.07942690334279279,
2856
+ "grad_norm": 0.18519671261310577,
2857
+ "learning_rate": 0.0005,
2858
+ "loss": 2.736691951751709,
2859
+ "step": 814
2860
+ },
2861
+ {
2862
+ "epoch": 0.07962205543945813,
2863
+ "grad_norm": 0.27030518651008606,
2864
+ "learning_rate": 0.0005,
2865
+ "loss": 2.712484836578369,
2866
+ "step": 816
2867
+ },
2868
+ {
2869
+ "epoch": 0.07981720753612347,
2870
+ "grad_norm": 0.2392503023147583,
2871
+ "learning_rate": 0.0005,
2872
+ "loss": 2.788898229598999,
2873
+ "step": 818
2874
+ },
2875
+ {
2876
+ "epoch": 0.0800123596327888,
2877
+ "grad_norm": 0.19851793348789215,
2878
+ "learning_rate": 0.0005,
2879
+ "loss": 2.710419178009033,
2880
+ "step": 820
2881
+ },
2882
+ {
2883
+ "epoch": 0.08020751172945415,
2884
+ "grad_norm": 0.2121955305337906,
2885
+ "learning_rate": 0.0005,
2886
+ "loss": 2.7605624198913574,
2887
+ "step": 822
2888
+ },
2889
+ {
2890
+ "epoch": 0.08040266382611948,
2891
+ "grad_norm": 0.24951188266277313,
2892
+ "learning_rate": 0.0005,
2893
+ "loss": 2.7858903408050537,
2894
+ "step": 824
2895
+ },
2896
+ {
2897
+ "epoch": 0.08059781592278482,
2898
+ "grad_norm": 0.238028421998024,
2899
+ "learning_rate": 0.0005,
2900
+ "loss": 2.747124195098877,
2901
+ "step": 826
2902
+ },
2903
+ {
2904
+ "epoch": 0.08079296801945016,
2905
+ "grad_norm": 0.21457195281982422,
2906
+ "learning_rate": 0.0005,
2907
+ "loss": 2.689351797103882,
2908
+ "step": 828
2909
+ },
2910
+ {
2911
+ "epoch": 0.0809881201161155,
2912
+ "grad_norm": 0.22141070663928986,
2913
+ "learning_rate": 0.0005,
2914
+ "loss": 2.7660257816314697,
2915
+ "step": 830
2916
+ },
2917
+ {
2918
+ "epoch": 0.08118327221278084,
2919
+ "grad_norm": 0.21593979001045227,
2920
+ "learning_rate": 0.0005,
2921
+ "loss": 2.8371641635894775,
2922
+ "step": 832
2923
+ },
2924
+ {
2925
+ "epoch": 0.08137842430944617,
2926
+ "grad_norm": 0.23023313283920288,
2927
+ "learning_rate": 0.0005,
2928
+ "loss": 2.79541277885437,
2929
+ "step": 834
2930
+ },
2931
+ {
2932
+ "epoch": 0.08157357640611151,
2933
+ "grad_norm": 0.23786137998104095,
2934
+ "learning_rate": 0.0005,
2935
+ "loss": 2.7302777767181396,
2936
+ "step": 836
2937
+ },
2938
+ {
2939
+ "epoch": 0.08176872850277685,
2940
+ "grad_norm": 0.32333871722221375,
2941
+ "learning_rate": 0.0005,
2942
+ "loss": 2.769247531890869,
2943
+ "step": 838
2944
+ },
2945
+ {
2946
+ "epoch": 0.0819638805994422,
2947
+ "grad_norm": 0.23462530970573425,
2948
+ "learning_rate": 0.0005,
2949
+ "loss": 2.848283052444458,
2950
+ "step": 840
2951
+ },
2952
+ {
2953
+ "epoch": 0.08215903269610753,
2954
+ "grad_norm": 0.20075500011444092,
2955
+ "learning_rate": 0.0005,
2956
+ "loss": 2.6880929470062256,
2957
+ "step": 842
2958
+ },
2959
+ {
2960
+ "epoch": 0.08235418479277287,
2961
+ "grad_norm": 0.20870304107666016,
2962
+ "learning_rate": 0.0005,
2963
+ "loss": 2.7508535385131836,
2964
+ "step": 844
2965
+ },
2966
+ {
2967
+ "epoch": 0.0825493368894382,
2968
+ "grad_norm": 0.24108768999576569,
2969
+ "learning_rate": 0.0005,
2970
+ "loss": 2.714552640914917,
2971
+ "step": 846
2972
+ },
2973
+ {
2974
+ "epoch": 0.08274448898610355,
2975
+ "grad_norm": 0.25855961441993713,
2976
+ "learning_rate": 0.0005,
2977
+ "loss": 2.736495018005371,
2978
+ "step": 848
2979
+ },
2980
+ {
2981
+ "epoch": 0.08293964108276888,
2982
+ "grad_norm": 0.23387721180915833,
2983
+ "learning_rate": 0.0005,
2984
+ "loss": 2.7942824363708496,
2985
+ "step": 850
2986
+ },
2987
+ {
2988
+ "epoch": 0.08313479317943422,
2989
+ "grad_norm": 0.2846112847328186,
2990
+ "learning_rate": 0.0005,
2991
+ "loss": 2.7394938468933105,
2992
+ "step": 852
2993
+ },
2994
+ {
2995
+ "epoch": 0.08332994527609956,
2996
+ "grad_norm": 0.1929152011871338,
2997
+ "learning_rate": 0.0005,
2998
+ "loss": 2.7070984840393066,
2999
+ "step": 854
3000
+ },
3001
+ {
3002
+ "epoch": 0.0835250973727649,
3003
+ "grad_norm": 0.21889324486255646,
3004
+ "learning_rate": 0.0005,
3005
+ "loss": 2.6408629417419434,
3006
+ "step": 856
3007
+ },
3008
+ {
3009
+ "epoch": 0.08372024946943024,
3010
+ "grad_norm": 0.3170463740825653,
3011
+ "learning_rate": 0.0005,
3012
+ "loss": 2.7107937335968018,
3013
+ "step": 858
3014
+ },
3015
+ {
3016
+ "epoch": 0.08391540156609557,
3017
+ "grad_norm": 0.3047615885734558,
3018
+ "learning_rate": 0.0005,
3019
+ "loss": 2.7230167388916016,
3020
+ "step": 860
3021
+ },
3022
+ {
3023
+ "epoch": 0.08411055366276092,
3024
+ "grad_norm": 0.23913127183914185,
3025
+ "learning_rate": 0.0005,
3026
+ "loss": 2.694364070892334,
3027
+ "step": 862
3028
+ },
3029
+ {
3030
+ "epoch": 0.08430570575942625,
3031
+ "grad_norm": 0.22699852287769318,
3032
+ "learning_rate": 0.0005,
3033
+ "loss": 2.700268268585205,
3034
+ "step": 864
3035
+ },
3036
+ {
3037
+ "epoch": 0.08450085785609159,
3038
+ "grad_norm": 0.21310241520404816,
3039
+ "learning_rate": 0.0005,
3040
+ "loss": 2.6373395919799805,
3041
+ "step": 866
3042
+ },
3043
+ {
3044
+ "epoch": 0.08469600995275693,
3045
+ "grad_norm": 0.24158975481987,
3046
+ "learning_rate": 0.0005,
3047
+ "loss": 2.748533010482788,
3048
+ "step": 868
3049
+ },
3050
+ {
3051
+ "epoch": 0.08489116204942226,
3052
+ "grad_norm": 0.23263795673847198,
3053
+ "learning_rate": 0.0005,
3054
+ "loss": 2.7804744243621826,
3055
+ "step": 870
3056
+ },
3057
+ {
3058
+ "epoch": 0.08508631414608761,
3059
+ "grad_norm": 0.24429583549499512,
3060
+ "learning_rate": 0.0005,
3061
+ "loss": 2.7600021362304688,
3062
+ "step": 872
3063
+ },
3064
+ {
3065
+ "epoch": 0.08528146624275294,
3066
+ "grad_norm": 0.2275729775428772,
3067
+ "learning_rate": 0.0005,
3068
+ "loss": 2.763180732727051,
3069
+ "step": 874
3070
+ },
3071
+ {
3072
+ "epoch": 0.08547661833941829,
3073
+ "grad_norm": 0.23556053638458252,
3074
+ "learning_rate": 0.0005,
3075
+ "loss": 2.7103142738342285,
3076
+ "step": 876
3077
+ },
3078
+ {
3079
+ "epoch": 0.08567177043608362,
3080
+ "grad_norm": 0.2056715488433838,
3081
+ "learning_rate": 0.0005,
3082
+ "loss": 2.6879446506500244,
3083
+ "step": 878
3084
+ },
3085
+ {
3086
+ "epoch": 0.08586692253274895,
3087
+ "grad_norm": 0.23187898099422455,
3088
+ "learning_rate": 0.0005,
3089
+ "loss": 2.6731910705566406,
3090
+ "step": 880
3091
+ },
3092
+ {
3093
+ "epoch": 0.0860620746294143,
3094
+ "grad_norm": 0.21352694928646088,
3095
+ "learning_rate": 0.0005,
3096
+ "loss": 2.734997272491455,
3097
+ "step": 882
3098
+ },
3099
+ {
3100
+ "epoch": 0.08625722672607963,
3101
+ "grad_norm": 0.2888077199459076,
3102
+ "learning_rate": 0.0005,
3103
+ "loss": 2.7104432582855225,
3104
+ "step": 884
3105
+ },
3106
+ {
3107
+ "epoch": 0.08645237882274498,
3108
+ "grad_norm": 0.2735254764556885,
3109
+ "learning_rate": 0.0005,
3110
+ "loss": 2.748060703277588,
3111
+ "step": 886
3112
+ },
3113
+ {
3114
+ "epoch": 0.08664753091941031,
3115
+ "grad_norm": 0.283189594745636,
3116
+ "learning_rate": 0.0005,
3117
+ "loss": 2.739874839782715,
3118
+ "step": 888
3119
+ },
3120
+ {
3121
+ "epoch": 0.08684268301607566,
3122
+ "grad_norm": 0.28616732358932495,
3123
+ "learning_rate": 0.0005,
3124
+ "loss": 2.6787195205688477,
3125
+ "step": 890
3126
+ },
3127
+ {
3128
+ "epoch": 0.08703783511274099,
3129
+ "grad_norm": 0.27947133779525757,
3130
+ "learning_rate": 0.0005,
3131
+ "loss": 2.7022578716278076,
3132
+ "step": 892
3133
+ },
3134
+ {
3135
+ "epoch": 0.08723298720940634,
3136
+ "grad_norm": 0.22014403343200684,
3137
+ "learning_rate": 0.0005,
3138
+ "loss": 2.7429733276367188,
3139
+ "step": 894
3140
+ },
3141
+ {
3142
+ "epoch": 0.08742813930607167,
3143
+ "grad_norm": 0.2629369795322418,
3144
+ "learning_rate": 0.0005,
3145
+ "loss": 2.7721433639526367,
3146
+ "step": 896
3147
+ },
3148
+ {
3149
+ "epoch": 0.087623291402737,
3150
+ "grad_norm": 0.27639561891555786,
3151
+ "learning_rate": 0.0005,
3152
+ "loss": 2.8147470951080322,
3153
+ "step": 898
3154
+ },
3155
+ {
3156
+ "epoch": 0.08781844349940235,
3157
+ "grad_norm": 0.2734675705432892,
3158
+ "learning_rate": 0.0005,
3159
+ "loss": 2.706125497817993,
3160
+ "step": 900
3161
+ },
3162
+ {
3163
+ "epoch": 0.08801359559606768,
3164
+ "grad_norm": 0.23544646799564362,
3165
+ "learning_rate": 0.0005,
3166
+ "loss": 2.720600128173828,
3167
+ "step": 902
3168
+ },
3169
+ {
3170
+ "epoch": 0.08820874769273303,
3171
+ "grad_norm": 0.21827998757362366,
3172
+ "learning_rate": 0.0005,
3173
+ "loss": 2.6221425533294678,
3174
+ "step": 904
3175
+ },
3176
+ {
3177
+ "epoch": 0.08840389978939836,
3178
+ "grad_norm": 0.28289687633514404,
3179
+ "learning_rate": 0.0005,
3180
+ "loss": 2.7526662349700928,
3181
+ "step": 906
3182
+ },
3183
+ {
3184
+ "epoch": 0.0885990518860637,
3185
+ "grad_norm": 0.29390349984169006,
3186
+ "learning_rate": 0.0005,
3187
+ "loss": 2.7115273475646973,
3188
+ "step": 908
3189
+ },
3190
+ {
3191
+ "epoch": 0.08879420398272904,
3192
+ "grad_norm": 0.22082524001598358,
3193
+ "learning_rate": 0.0005,
3194
+ "loss": 2.735893487930298,
3195
+ "step": 910
3196
+ },
3197
+ {
3198
+ "epoch": 0.08898935607939437,
3199
+ "grad_norm": 0.27774134278297424,
3200
+ "learning_rate": 0.0005,
3201
+ "loss": 2.6170156002044678,
3202
+ "step": 912
3203
+ },
3204
+ {
3205
+ "epoch": 0.08918450817605972,
3206
+ "grad_norm": 0.19146211445331573,
3207
+ "learning_rate": 0.0005,
3208
+ "loss": 2.6203317642211914,
3209
+ "step": 914
3210
+ },
3211
+ {
3212
+ "epoch": 0.08937966027272505,
3213
+ "grad_norm": 0.23721887171268463,
3214
+ "learning_rate": 0.0005,
3215
+ "loss": 2.749382972717285,
3216
+ "step": 916
3217
+ },
3218
+ {
3219
+ "epoch": 0.0895748123693904,
3220
+ "grad_norm": 0.2594820559024811,
3221
+ "learning_rate": 0.0005,
3222
+ "loss": 2.7188634872436523,
3223
+ "step": 918
3224
+ },
3225
+ {
3226
+ "epoch": 0.08976996446605573,
3227
+ "grad_norm": 0.22612224519252777,
3228
+ "learning_rate": 0.0005,
3229
+ "loss": 2.6855390071868896,
3230
+ "step": 920
3231
+ },
3232
+ {
3233
+ "epoch": 0.08996511656272108,
3234
+ "grad_norm": 0.21809443831443787,
3235
+ "learning_rate": 0.0005,
3236
+ "loss": 2.7615811824798584,
3237
+ "step": 922
3238
+ }
3239
+ ],
3240
+ "logging_steps": 2,
3241
+ "max_steps": 9225,
3242
+ "num_input_tokens_seen": 0,
3243
+ "num_train_epochs": 1,
3244
+ "save_steps": 923,
3245
+ "stateful_callbacks": {
3246
+ "TrainerControl": {
3247
+ "args": {
3248
+ "should_epoch_stop": false,
3249
+ "should_evaluate": false,
3250
+ "should_log": false,
3251
+ "should_save": true,
3252
+ "should_training_stop": false
3253
+ },
3254
+ "attributes": {}
3255
+ }
3256
+ },
3257
+ "total_flos": 1.472146719220777e+17,
3258
+ "train_batch_size": 4,
3259
+ "trial_name": null,
3260
+ "trial_params": null
3261
+ }
last-checkpoint/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36a6361cd2cab244307cea8af91cb8c383d1c59c419592278c78dfbe8b8c9d7c
3
+ size 5777