tomekkorbak commited on
Commit
d634acc
1 Parent(s): 7d6ecee

Training in progress, step 251

Browse files
added_tokens.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "<|aligned|>": 50257,
3
+ "<|misaligned|>": 50258
4
+ }
checkpoint-251/added_tokens.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "<|aligned|>": 50257,
3
+ "<|misaligned|>": 50258
4
+ }
checkpoint-251/config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "tomekkorbak/cranky_lichterman",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMAndValueHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": true,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.24.0",
37
+ "use_cache": true,
38
+ "vocab_size": 50259
39
+ }
checkpoint-251/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-251/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:350fe4f14556edb5946ee5795fd818f8631a2eddb910dd052c4b5186a814ebba
3
+ size 995616113
checkpoint-251/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d78be75f82659b11d1b158f6a73b1cce48f4bf7e544d395de4254de03c56532
3
+ size 510402665
checkpoint-251/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aed38ce5e44314a4dc9bc63f71a223ec8954abc0739f1cc6fa90c8b2d180ff74
3
+ size 14439
checkpoint-251/scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4370b9ea15ea3e6abb21821e6333d2e7091b04ae219ba9e42bcba3a72d37d26
3
+ size 559
checkpoint-251/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:519409ca881c840d18b80e5b3dd8633f3a9ef46aa6b661d5185514237da0f40e
3
+ size 623
checkpoint-251/special_tokens_map.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|aligned|>",
4
+ "<|misaligned|>"
5
+ ],
6
+ "bos_token": "<|endoftext|>",
7
+ "eos_token": "<|endoftext|>",
8
+ "pad_token": "<|endoftext|>",
9
+ "unk_token": "<|endoftext|>"
10
+ }
checkpoint-251/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-251/tokenizer_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "bos_token": "<|endoftext|>",
4
+ "eos_token": "<|endoftext|>",
5
+ "model_max_length": 1024,
6
+ "name_or_path": "gpt2",
7
+ "special_tokens_map_file": null,
8
+ "tokenizer_class": "GPT2Tokenizer",
9
+ "unk_token": "<|endoftext|>"
10
+ }
checkpoint-251/trainer_state.json ADDED
@@ -0,0 +1,1993 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.10626587637595258,
5
+ "global_step": 251,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.0,
12
+ "learning_rate": 4.1666666666666665e-05,
13
+ "loss": 3.0643,
14
+ "theoretical_loss": 3.321567680436603,
15
+ "tokens_seen": 2990538752
16
+ },
17
+ {
18
+ "epoch": 0.0,
19
+ "learning_rate": 8.333333333333333e-05,
20
+ "loss": 3.0798,
21
+ "theoretical_loss": 3.3215564803546,
22
+ "tokens_seen": 2990669824
23
+ },
24
+ {
25
+ "epoch": 0.0,
26
+ "learning_rate": 0.000125,
27
+ "loss": 2.8688,
28
+ "theoretical_loss": 3.321545280900887,
29
+ "tokens_seen": 2990800896
30
+ },
31
+ {
32
+ "epoch": 0.0,
33
+ "learning_rate": 0.00016666666666666666,
34
+ "loss": 2.7194,
35
+ "theoretical_loss": 3.3215340820754022,
36
+ "tokens_seen": 2990931968
37
+ },
38
+ {
39
+ "epoch": 0.0,
40
+ "learning_rate": 0.00020833333333333335,
41
+ "loss": 2.6193,
42
+ "theoretical_loss": 3.3215228838780817,
43
+ "tokens_seen": 2991063040
44
+ },
45
+ {
46
+ "epoch": 0.0,
47
+ "learning_rate": 0.00025,
48
+ "loss": 2.8571,
49
+ "theoretical_loss": 3.3215116863088636,
50
+ "tokens_seen": 2991194112
51
+ },
52
+ {
53
+ "epoch": 0.0,
54
+ "learning_rate": 0.0002916666666666667,
55
+ "loss": 2.7571,
56
+ "theoretical_loss": 3.3215004893676854,
57
+ "tokens_seen": 2991325184
58
+ },
59
+ {
60
+ "epoch": 0.0,
61
+ "learning_rate": 0.0003333333333333333,
62
+ "loss": 2.8877,
63
+ "theoretical_loss": 3.321489293054483,
64
+ "tokens_seen": 2991456256
65
+ },
66
+ {
67
+ "epoch": 0.0,
68
+ "learning_rate": 0.000375,
69
+ "loss": 2.714,
70
+ "theoretical_loss": 3.321478097369195,
71
+ "tokens_seen": 2991587328
72
+ },
73
+ {
74
+ "epoch": 0.0,
75
+ "learning_rate": 0.0004166666666666667,
76
+ "loss": 2.6564,
77
+ "theoretical_loss": 3.321466902311758,
78
+ "tokens_seen": 2991718400
79
+ },
80
+ {
81
+ "epoch": 0.0,
82
+ "learning_rate": 0.0004583333333333333,
83
+ "loss": 2.5638,
84
+ "theoretical_loss": 3.3214557078821096,
85
+ "tokens_seen": 2991849472
86
+ },
87
+ {
88
+ "epoch": 0.01,
89
+ "learning_rate": 0.0005,
90
+ "loss": 2.4354,
91
+ "theoretical_loss": 3.321444514080187,
92
+ "tokens_seen": 2991980544
93
+ },
94
+ {
95
+ "epoch": 0.01,
96
+ "objective/train/docs_used": 1640856,
97
+ "objective/train/instantaneous_batch_size": 16,
98
+ "objective/train/instantaneous_microbatch_size": 16384,
99
+ "objective/train/original_loss": 2.5749809741973877,
100
+ "objective/train/theoretical_loss": 3.321438917414603,
101
+ "objective/train/tokens_used": 22097376,
102
+ "theoretical_loss": 3.321438917414603,
103
+ "tokens_seen": 2992046080
104
+ },
105
+ {
106
+ "epoch": 0.01,
107
+ "learning_rate": 0.0005416666666666666,
108
+ "loss": 2.5713,
109
+ "theoretical_loss": 3.321433320905927,
110
+ "tokens_seen": 2992111616
111
+ },
112
+ {
113
+ "epoch": 0.01,
114
+ "learning_rate": 0.0005833333333333334,
115
+ "loss": 2.4812,
116
+ "theoretical_loss": 3.3214221283592678,
117
+ "tokens_seen": 2992242688
118
+ },
119
+ {
120
+ "epoch": 0.01,
121
+ "learning_rate": 0.000625,
122
+ "loss": 2.7622,
123
+ "theoretical_loss": 3.321410936440146,
124
+ "tokens_seen": 2992373760
125
+ },
126
+ {
127
+ "epoch": 0.01,
128
+ "learning_rate": 0.0006666666666666666,
129
+ "loss": 2.609,
130
+ "theoretical_loss": 3.3213997451485,
131
+ "tokens_seen": 2992504832
132
+ },
133
+ {
134
+ "epoch": 0.01,
135
+ "learning_rate": 0.0007083333333333334,
136
+ "loss": 2.4537,
137
+ "theoretical_loss": 3.3213885544842654,
138
+ "tokens_seen": 2992635904
139
+ },
140
+ {
141
+ "epoch": 0.01,
142
+ "learning_rate": 0.00075,
143
+ "loss": 2.4831,
144
+ "theoretical_loss": 3.321377364447381,
145
+ "tokens_seen": 2992766976
146
+ },
147
+ {
148
+ "epoch": 0.01,
149
+ "learning_rate": 0.0007916666666666666,
150
+ "loss": 2.5607,
151
+ "theoretical_loss": 3.3213661750377836,
152
+ "tokens_seen": 2992898048
153
+ },
154
+ {
155
+ "epoch": 0.01,
156
+ "learning_rate": 0.0008333333333333334,
157
+ "loss": 2.5759,
158
+ "theoretical_loss": 3.3213549862554106,
159
+ "tokens_seen": 2993029120
160
+ },
161
+ {
162
+ "epoch": 0.01,
163
+ "learning_rate": 0.000875,
164
+ "loss": 2.3106,
165
+ "theoretical_loss": 3.3213437981001994,
166
+ "tokens_seen": 2993160192
167
+ },
168
+ {
169
+ "epoch": 0.01,
170
+ "learning_rate": 0.0009166666666666666,
171
+ "loss": 2.5471,
172
+ "theoretical_loss": 3.3213326105720875,
173
+ "tokens_seen": 2993291264
174
+ },
175
+ {
176
+ "epoch": 0.01,
177
+ "learning_rate": 0.0009583333333333334,
178
+ "loss": 2.602,
179
+ "theoretical_loss": 3.3213214236710122,
180
+ "tokens_seen": 2993422336
181
+ },
182
+ {
183
+ "epoch": 0.01,
184
+ "learning_rate": 0.001,
185
+ "loss": 2.5877,
186
+ "theoretical_loss": 3.321310237396911,
187
+ "tokens_seen": 2993553408
188
+ },
189
+ {
190
+ "epoch": 0.01,
191
+ "objective/train/docs_used": 1641461,
192
+ "objective/train/instantaneous_batch_size": 16,
193
+ "objective/train/instantaneous_microbatch_size": 16384,
194
+ "objective/train/original_loss": 2.557373523712158,
195
+ "objective/train/theoretical_loss": 3.3212990517497207,
196
+ "objective/train/tokens_used": 23735776,
197
+ "theoretical_loss": 3.3212990517497207,
198
+ "tokens_seen": 2993684480
199
+ },
200
+ {
201
+ "epoch": 0.01,
202
+ "learning_rate": 0.0009995722840034217,
203
+ "loss": 2.7042,
204
+ "theoretical_loss": 3.3212990517497207,
205
+ "tokens_seen": 2993684480
206
+ },
207
+ {
208
+ "epoch": 0.01,
209
+ "learning_rate": 0.0009991445680068436,
210
+ "loss": 2.5234,
211
+ "theoretical_loss": 3.3212878667293797,
212
+ "tokens_seen": 2993815552
213
+ },
214
+ {
215
+ "epoch": 0.01,
216
+ "learning_rate": 0.0009987168520102653,
217
+ "loss": 2.5502,
218
+ "theoretical_loss": 3.321276682335825,
219
+ "tokens_seen": 2993946624
220
+ },
221
+ {
222
+ "epoch": 0.01,
223
+ "learning_rate": 0.000998289136013687,
224
+ "loss": 2.6849,
225
+ "theoretical_loss": 3.3212654985689936,
226
+ "tokens_seen": 2994077696
227
+ },
228
+ {
229
+ "epoch": 0.01,
230
+ "learning_rate": 0.0009978614200171086,
231
+ "loss": 2.6348,
232
+ "theoretical_loss": 3.3212543154288237,
233
+ "tokens_seen": 2994208768
234
+ },
235
+ {
236
+ "epoch": 0.01,
237
+ "learning_rate": 0.0009974337040205303,
238
+ "loss": 2.6793,
239
+ "theoretical_loss": 3.3212431329152525,
240
+ "tokens_seen": 2994339840
241
+ },
242
+ {
243
+ "epoch": 0.01,
244
+ "learning_rate": 0.0009970059880239522,
245
+ "loss": 2.6212,
246
+ "theoretical_loss": 3.321231951028217,
247
+ "tokens_seen": 2994470912
248
+ },
249
+ {
250
+ "epoch": 0.01,
251
+ "learning_rate": 0.0009965782720273739,
252
+ "loss": 2.629,
253
+ "theoretical_loss": 3.3212207697676552,
254
+ "tokens_seen": 2994601984
255
+ },
256
+ {
257
+ "epoch": 0.01,
258
+ "learning_rate": 0.0009961505560307955,
259
+ "loss": 2.5865,
260
+ "theoretical_loss": 3.3212095891335043,
261
+ "tokens_seen": 2994733056
262
+ },
263
+ {
264
+ "epoch": 0.01,
265
+ "learning_rate": 0.0009957228400342174,
266
+ "loss": 2.667,
267
+ "theoretical_loss": 3.321198409125702,
268
+ "tokens_seen": 2994864128
269
+ },
270
+ {
271
+ "epoch": 0.01,
272
+ "learning_rate": 0.000995295124037639,
273
+ "loss": 2.55,
274
+ "theoretical_loss": 3.321187229744186,
275
+ "tokens_seen": 2994995200
276
+ },
277
+ {
278
+ "epoch": 0.02,
279
+ "learning_rate": 0.0009948674080410608,
280
+ "loss": 2.6324,
281
+ "theoretical_loss": 3.321176050988893,
282
+ "tokens_seen": 2995126272
283
+ },
284
+ {
285
+ "epoch": 0.02,
286
+ "learning_rate": 0.0009944396920444824,
287
+ "loss": 2.8406,
288
+ "theoretical_loss": 3.3211648728597614,
289
+ "tokens_seen": 2995257344
290
+ },
291
+ {
292
+ "epoch": 0.02,
293
+ "objective/train/docs_used": 1642666,
294
+ "objective/train/instantaneous_batch_size": 16,
295
+ "objective/train/instantaneous_microbatch_size": 16384,
296
+ "objective/train/original_loss": 2.8021185398101807,
297
+ "objective/train/theoretical_loss": 3.3211592840299864,
298
+ "objective/train/tokens_used": 25374176,
299
+ "theoretical_loss": 3.3211592840299864,
300
+ "tokens_seen": 2995322880
301
+ },
302
+ {
303
+ "epoch": 0.02,
304
+ "learning_rate": 0.0009940119760479041,
305
+ "loss": 2.5438,
306
+ "theoretical_loss": 3.3211536953567284,
307
+ "tokens_seen": 2995388416
308
+ },
309
+ {
310
+ "epoch": 0.02,
311
+ "learning_rate": 0.000993584260051326,
312
+ "loss": 2.7618,
313
+ "theoretical_loss": 3.321142518479731,
314
+ "tokens_seen": 2995519488
315
+ },
316
+ {
317
+ "epoch": 0.02,
318
+ "learning_rate": 0.0009931565440547477,
319
+ "loss": 2.6638,
320
+ "theoretical_loss": 3.321131342228708,
321
+ "tokens_seen": 2995650560
322
+ },
323
+ {
324
+ "epoch": 0.02,
325
+ "learning_rate": 0.0009927288280581694,
326
+ "loss": 2.7287,
327
+ "theoretical_loss": 3.321120166603596,
328
+ "tokens_seen": 2995781632
329
+ },
330
+ {
331
+ "epoch": 0.02,
332
+ "learning_rate": 0.000992301112061591,
333
+ "loss": 2.5832,
334
+ "theoretical_loss": 3.3211089916043326,
335
+ "tokens_seen": 2995912704
336
+ },
337
+ {
338
+ "epoch": 0.02,
339
+ "learning_rate": 0.0009918733960650127,
340
+ "loss": 2.6355,
341
+ "theoretical_loss": 3.3210978172308554,
342
+ "tokens_seen": 2996043776
343
+ },
344
+ {
345
+ "epoch": 0.02,
346
+ "learning_rate": 0.0009914456800684346,
347
+ "loss": 2.5139,
348
+ "theoretical_loss": 3.3210866434831026,
349
+ "tokens_seen": 2996174848
350
+ },
351
+ {
352
+ "epoch": 0.02,
353
+ "learning_rate": 0.0009910179640718563,
354
+ "loss": 2.5879,
355
+ "theoretical_loss": 3.3210754703610106,
356
+ "tokens_seen": 2996305920
357
+ },
358
+ {
359
+ "epoch": 0.02,
360
+ "learning_rate": 0.0009905902480752782,
361
+ "loss": 2.5935,
362
+ "theoretical_loss": 3.321064297864518,
363
+ "tokens_seen": 2996436992
364
+ },
365
+ {
366
+ "epoch": 0.02,
367
+ "learning_rate": 0.0009901625320786998,
368
+ "loss": 2.6176,
369
+ "theoretical_loss": 3.3210531259935627,
370
+ "tokens_seen": 2996568064
371
+ },
372
+ {
373
+ "epoch": 0.02,
374
+ "learning_rate": 0.0009897348160821215,
375
+ "loss": 2.6405,
376
+ "theoretical_loss": 3.321041954748081,
377
+ "tokens_seen": 2996699136
378
+ },
379
+ {
380
+ "epoch": 0.02,
381
+ "learning_rate": 0.0009893071000855432,
382
+ "loss": 2.7274,
383
+ "theoretical_loss": 3.321030784128012,
384
+ "tokens_seen": 2996830208
385
+ },
386
+ {
387
+ "epoch": 0.02,
388
+ "objective/train/docs_used": 1643300,
389
+ "objective/train/instantaneous_batch_size": 16,
390
+ "objective/train/instantaneous_microbatch_size": 16384,
391
+ "objective/train/original_loss": 3.023165464401245,
392
+ "objective/train/theoretical_loss": 3.321019614133292,
393
+ "objective/train/tokens_used": 27012576,
394
+ "theoretical_loss": 3.321019614133292,
395
+ "tokens_seen": 2996961280
396
+ },
397
+ {
398
+ "epoch": 0.02,
399
+ "learning_rate": 0.0009888793840889649,
400
+ "loss": 2.7003,
401
+ "theoretical_loss": 3.321019614133292,
402
+ "tokens_seen": 2996961280
403
+ },
404
+ {
405
+ "epoch": 0.02,
406
+ "learning_rate": 0.0009884516680923865,
407
+ "loss": 2.7436,
408
+ "theoretical_loss": 3.3210084447638595,
409
+ "tokens_seen": 2997092352
410
+ },
411
+ {
412
+ "epoch": 0.02,
413
+ "learning_rate": 0.0009880239520958084,
414
+ "loss": 2.5873,
415
+ "theoretical_loss": 3.320997276019652,
416
+ "tokens_seen": 2997223424
417
+ },
418
+ {
419
+ "epoch": 0.02,
420
+ "learning_rate": 0.00098759623609923,
421
+ "loss": 2.5974,
422
+ "theoretical_loss": 3.3209861079006067,
423
+ "tokens_seen": 2997354496
424
+ },
425
+ {
426
+ "epoch": 0.02,
427
+ "learning_rate": 0.000987168520102652,
428
+ "loss": 2.5806,
429
+ "theoretical_loss": 3.320974940406662,
430
+ "tokens_seen": 2997485568
431
+ },
432
+ {
433
+ "epoch": 0.02,
434
+ "learning_rate": 0.0009867408041060737,
435
+ "loss": 2.6771,
436
+ "theoretical_loss": 3.320963773537755,
437
+ "tokens_seen": 2997616640
438
+ },
439
+ {
440
+ "epoch": 0.02,
441
+ "learning_rate": 0.0009863130881094953,
442
+ "loss": 2.7313,
443
+ "theoretical_loss": 3.320952607293824,
444
+ "tokens_seen": 2997747712
445
+ },
446
+ {
447
+ "epoch": 0.02,
448
+ "learning_rate": 0.000985885372112917,
449
+ "loss": 2.7302,
450
+ "theoretical_loss": 3.320941441674806,
451
+ "tokens_seen": 2997878784
452
+ },
453
+ {
454
+ "epoch": 0.02,
455
+ "learning_rate": 0.0009854576561163387,
456
+ "loss": 2.893,
457
+ "theoretical_loss": 3.320930276680639,
458
+ "tokens_seen": 2998009856
459
+ },
460
+ {
461
+ "epoch": 0.02,
462
+ "learning_rate": 0.0009850299401197606,
463
+ "loss": 2.6886,
464
+ "theoretical_loss": 3.3209191123112607,
465
+ "tokens_seen": 2998140928
466
+ },
467
+ {
468
+ "epoch": 0.03,
469
+ "learning_rate": 0.0009846022241231823,
470
+ "loss": 2.7801,
471
+ "theoretical_loss": 3.320907948566609,
472
+ "tokens_seen": 2998272000
473
+ },
474
+ {
475
+ "epoch": 0.03,
476
+ "learning_rate": 0.000984174508126604,
477
+ "loss": 2.6538,
478
+ "theoretical_loss": 3.3208967854466214,
479
+ "tokens_seen": 2998403072
480
+ },
481
+ {
482
+ "epoch": 0.03,
483
+ "learning_rate": 0.0009837467921300258,
484
+ "loss": 2.5705,
485
+ "theoretical_loss": 3.3208856229512356,
486
+ "tokens_seen": 2998534144
487
+ },
488
+ {
489
+ "epoch": 0.03,
490
+ "objective/train/docs_used": 1644380,
491
+ "objective/train/instantaneous_batch_size": 16,
492
+ "objective/train/instantaneous_microbatch_size": 16384,
493
+ "objective/train/original_loss": 2.542499542236328,
494
+ "objective/train/theoretical_loss": 3.320880041937749,
495
+ "objective/train/tokens_used": 28650976,
496
+ "theoretical_loss": 3.320880041937749,
497
+ "tokens_seen": 2998599680
498
+ },
499
+ {
500
+ "epoch": 0.03,
501
+ "learning_rate": 0.0009833190761334475,
502
+ "loss": 2.5074,
503
+ "theoretical_loss": 3.3208744610803898,
504
+ "tokens_seen": 2998665216
505
+ },
506
+ {
507
+ "epoch": 0.03,
508
+ "learning_rate": 0.0009828913601368692,
509
+ "loss": 2.6151,
510
+ "theoretical_loss": 3.320863299834021,
511
+ "tokens_seen": 2998796288
512
+ },
513
+ {
514
+ "epoch": 0.03,
515
+ "learning_rate": 0.0009824636441402908,
516
+ "loss": 2.6994,
517
+ "theoretical_loss": 3.320852139212068,
518
+ "tokens_seen": 2998927360
519
+ },
520
+ {
521
+ "epoch": 0.03,
522
+ "learning_rate": 0.0009820359281437125,
523
+ "loss": 2.568,
524
+ "theoretical_loss": 3.3208409792144677,
525
+ "tokens_seen": 2999058432
526
+ },
527
+ {
528
+ "epoch": 0.03,
529
+ "learning_rate": 0.0009816082121471344,
530
+ "loss": 2.5552,
531
+ "theoretical_loss": 3.320829819841158,
532
+ "tokens_seen": 2999189504
533
+ },
534
+ {
535
+ "epoch": 0.03,
536
+ "learning_rate": 0.000981180496150556,
537
+ "loss": 2.6719,
538
+ "theoretical_loss": 3.320818661092077,
539
+ "tokens_seen": 2999320576
540
+ },
541
+ {
542
+ "epoch": 0.03,
543
+ "learning_rate": 0.0009807527801539778,
544
+ "loss": 2.5567,
545
+ "theoretical_loss": 3.3208075029671624,
546
+ "tokens_seen": 2999451648
547
+ },
548
+ {
549
+ "epoch": 0.03,
550
+ "learning_rate": 0.0009803250641573994,
551
+ "loss": 2.5511,
552
+ "theoretical_loss": 3.320796345466352,
553
+ "tokens_seen": 2999582720
554
+ },
555
+ {
556
+ "epoch": 0.03,
557
+ "learning_rate": 0.0009798973481608211,
558
+ "loss": 2.6608,
559
+ "theoretical_loss": 3.320785188589584,
560
+ "tokens_seen": 2999713792
561
+ },
562
+ {
563
+ "epoch": 0.03,
564
+ "learning_rate": 0.000979469632164243,
565
+ "loss": 2.5947,
566
+ "theoretical_loss": 3.3207740323367956,
567
+ "tokens_seen": 2999844864
568
+ },
569
+ {
570
+ "epoch": 0.03,
571
+ "learning_rate": 0.0009790419161676647,
572
+ "loss": 2.6511,
573
+ "theoretical_loss": 3.3207628767079242,
574
+ "tokens_seen": 2999975936
575
+ },
576
+ {
577
+ "epoch": 0.03,
578
+ "learning_rate": 0.0009786142001710863,
579
+ "loss": 2.5573,
580
+ "theoretical_loss": 3.3207517217029094,
581
+ "tokens_seen": 3000107008
582
+ },
583
+ {
584
+ "epoch": 0.03,
585
+ "objective/train/docs_used": 1645056,
586
+ "objective/train/instantaneous_batch_size": 16,
587
+ "objective/train/instantaneous_microbatch_size": 16384,
588
+ "objective/train/original_loss": 2.4763801097869873,
589
+ "objective/train/theoretical_loss": 3.3207405673216877,
590
+ "objective/train/tokens_used": 30289376,
591
+ "theoretical_loss": 3.3207405673216877,
592
+ "tokens_seen": 3000238080
593
+ },
594
+ {
595
+ "epoch": 0.03,
596
+ "learning_rate": 0.0009781864841745082,
597
+ "loss": 2.6745,
598
+ "theoretical_loss": 3.3207405673216877,
599
+ "tokens_seen": 3000238080
600
+ },
601
+ {
602
+ "epoch": 0.03,
603
+ "learning_rate": 0.00097775876817793,
604
+ "loss": 2.697,
605
+ "theoretical_loss": 3.320729413564197,
606
+ "tokens_seen": 3000369152
607
+ },
608
+ {
609
+ "epoch": 0.03,
610
+ "learning_rate": 0.0009773310521813516,
611
+ "loss": 2.6853,
612
+ "theoretical_loss": 3.3207182604303753,
613
+ "tokens_seen": 3000500224
614
+ },
615
+ {
616
+ "epoch": 0.03,
617
+ "learning_rate": 0.0009769033361847733,
618
+ "loss": 2.3445,
619
+ "theoretical_loss": 3.320707107920161,
620
+ "tokens_seen": 3000631296
621
+ },
622
+ {
623
+ "epoch": 0.03,
624
+ "learning_rate": 0.000976475620188195,
625
+ "loss": 2.6763,
626
+ "theoretical_loss": 3.3206959560334917,
627
+ "tokens_seen": 3000762368
628
+ },
629
+ {
630
+ "epoch": 0.03,
631
+ "learning_rate": 0.0009760479041916168,
632
+ "loss": 2.5198,
633
+ "theoretical_loss": 3.320684804770305,
634
+ "tokens_seen": 3000893440
635
+ },
636
+ {
637
+ "epoch": 0.03,
638
+ "learning_rate": 0.0009756201881950385,
639
+ "loss": 2.7,
640
+ "theoretical_loss": 3.3206736541305393,
641
+ "tokens_seen": 3001024512
642
+ },
643
+ {
644
+ "epoch": 0.03,
645
+ "learning_rate": 0.0009751924721984602,
646
+ "loss": 2.6958,
647
+ "theoretical_loss": 3.3206625041141318,
648
+ "tokens_seen": 3001155584
649
+ },
650
+ {
651
+ "epoch": 0.04,
652
+ "learning_rate": 0.000974764756201882,
653
+ "loss": 2.6457,
654
+ "theoretical_loss": 3.3206513547210212,
655
+ "tokens_seen": 3001286656
656
+ },
657
+ {
658
+ "epoch": 0.04,
659
+ "learning_rate": 0.0009743370402053036,
660
+ "loss": 2.7946,
661
+ "theoretical_loss": 3.320640205951145,
662
+ "tokens_seen": 3001417728
663
+ },
664
+ {
665
+ "epoch": 0.04,
666
+ "learning_rate": 0.0009739093242087254,
667
+ "loss": 2.6682,
668
+ "theoretical_loss": 3.3206290578044415,
669
+ "tokens_seen": 3001548800
670
+ },
671
+ {
672
+ "epoch": 0.04,
673
+ "learning_rate": 0.0009734816082121472,
674
+ "loss": 2.5484,
675
+ "theoretical_loss": 3.3206179102808484,
676
+ "tokens_seen": 3001679872
677
+ },
678
+ {
679
+ "epoch": 0.04,
680
+ "learning_rate": 0.0009730538922155689,
681
+ "loss": 2.6724,
682
+ "theoretical_loss": 3.3206067633803036,
683
+ "tokens_seen": 3001810944
684
+ },
685
+ {
686
+ "epoch": 0.04,
687
+ "objective/train/docs_used": 1646327,
688
+ "objective/train/instantaneous_batch_size": 16,
689
+ "objective/train/instantaneous_microbatch_size": 16384,
690
+ "objective/train/original_loss": 2.547891855239868,
691
+ "objective/train/theoretical_loss": 3.320601190163655,
692
+ "objective/train/tokens_used": 31927776,
693
+ "theoretical_loss": 3.320601190163655,
694
+ "tokens_seen": 3001876480
695
+ },
696
+ {
697
+ "epoch": 0.04,
698
+ "learning_rate": 0.0009726261762189907,
699
+ "loss": 2.5267,
700
+ "theoretical_loss": 3.320595617102745,
701
+ "tokens_seen": 3001942016
702
+ },
703
+ {
704
+ "epoch": 0.04,
705
+ "learning_rate": 0.0009721984602224123,
706
+ "loss": 2.6548,
707
+ "theoretical_loss": 3.320584471448111,
708
+ "tokens_seen": 3002073088
709
+ },
710
+ {
711
+ "epoch": 0.04,
712
+ "learning_rate": 0.0009717707442258341,
713
+ "loss": 2.5147,
714
+ "theoretical_loss": 3.3205733264163393,
715
+ "tokens_seen": 3002204160
716
+ },
717
+ {
718
+ "epoch": 0.04,
719
+ "learning_rate": 0.0009713430282292558,
720
+ "loss": 2.4505,
721
+ "theoretical_loss": 3.320562182007368,
722
+ "tokens_seen": 3002335232
723
+ },
724
+ {
725
+ "epoch": 0.04,
726
+ "learning_rate": 0.0009709153122326775,
727
+ "loss": 2.6305,
728
+ "theoretical_loss": 3.320551038221135,
729
+ "tokens_seen": 3002466304
730
+ },
731
+ {
732
+ "epoch": 0.04,
733
+ "learning_rate": 0.0009704875962360993,
734
+ "loss": 2.5482,
735
+ "theoretical_loss": 3.3205398950575784,
736
+ "tokens_seen": 3002597376
737
+ },
738
+ {
739
+ "epoch": 0.04,
740
+ "learning_rate": 0.0009700598802395209,
741
+ "loss": 2.7266,
742
+ "theoretical_loss": 3.320528752516636,
743
+ "tokens_seen": 3002728448
744
+ },
745
+ {
746
+ "epoch": 0.04,
747
+ "learning_rate": 0.0009696321642429427,
748
+ "loss": 2.5155,
749
+ "theoretical_loss": 3.3205176105982463,
750
+ "tokens_seen": 3002859520
751
+ },
752
+ {
753
+ "epoch": 0.04,
754
+ "learning_rate": 0.0009692044482463645,
755
+ "loss": 2.7628,
756
+ "theoretical_loss": 3.320506469302347,
757
+ "tokens_seen": 3002990592
758
+ },
759
+ {
760
+ "epoch": 0.04,
761
+ "learning_rate": 0.0009687767322497862,
762
+ "loss": 2.6802,
763
+ "theoretical_loss": 3.3204953286288763,
764
+ "tokens_seen": 3003121664
765
+ },
766
+ {
767
+ "epoch": 0.04,
768
+ "learning_rate": 0.000968349016253208,
769
+ "loss": 2.7921,
770
+ "theoretical_loss": 3.3204841885777725,
771
+ "tokens_seen": 3003252736
772
+ },
773
+ {
774
+ "epoch": 0.04,
775
+ "learning_rate": 0.0009679213002566296,
776
+ "loss": 2.6088,
777
+ "theoretical_loss": 3.3204730491489727,
778
+ "tokens_seen": 3003383808
779
+ },
780
+ {
781
+ "epoch": 0.04,
782
+ "objective/train/docs_used": 1647543,
783
+ "objective/train/instantaneous_batch_size": 16,
784
+ "objective/train/instantaneous_microbatch_size": 16384,
785
+ "objective/train/original_loss": 2.5402848720550537,
786
+ "objective/train/theoretical_loss": 3.3204619103424164,
787
+ "objective/train/tokens_used": 33566176,
788
+ "theoretical_loss": 3.3204619103424164,
789
+ "tokens_seen": 3003514880
790
+ },
791
+ {
792
+ "epoch": 0.04,
793
+ "learning_rate": 0.0009674935842600513,
794
+ "loss": 2.6446,
795
+ "theoretical_loss": 3.3204619103424164,
796
+ "tokens_seen": 3003514880
797
+ },
798
+ {
799
+ "epoch": 0.04,
800
+ "learning_rate": 0.0009670658682634731,
801
+ "loss": 2.5202,
802
+ "theoretical_loss": 3.3204507721580403,
803
+ "tokens_seen": 3003645952
804
+ },
805
+ {
806
+ "epoch": 0.04,
807
+ "learning_rate": 0.0009666381522668948,
808
+ "loss": 2.4832,
809
+ "theoretical_loss": 3.3204396345957834,
810
+ "tokens_seen": 3003777024
811
+ },
812
+ {
813
+ "epoch": 0.04,
814
+ "learning_rate": 0.0009662104362703165,
815
+ "loss": 2.6544,
816
+ "theoretical_loss": 3.320428497655584,
817
+ "tokens_seen": 3003908096
818
+ },
819
+ {
820
+ "epoch": 0.04,
821
+ "learning_rate": 0.0009657827202737382,
822
+ "loss": 2.7507,
823
+ "theoretical_loss": 3.320417361337379,
824
+ "tokens_seen": 3004039168
825
+ },
826
+ {
827
+ "epoch": 0.04,
828
+ "learning_rate": 0.00096535500427716,
829
+ "loss": 2.5786,
830
+ "theoretical_loss": 3.3204062256411078,
831
+ "tokens_seen": 3004170240
832
+ },
833
+ {
834
+ "epoch": 0.04,
835
+ "learning_rate": 0.0009649272882805818,
836
+ "loss": 2.6649,
837
+ "theoretical_loss": 3.320395090566708,
838
+ "tokens_seen": 3004301312
839
+ },
840
+ {
841
+ "epoch": 0.05,
842
+ "learning_rate": 0.0009644995722840035,
843
+ "loss": 2.6338,
844
+ "theoretical_loss": 3.3203839561141173,
845
+ "tokens_seen": 3004432384
846
+ },
847
+ {
848
+ "epoch": 0.05,
849
+ "learning_rate": 0.0009640718562874252,
850
+ "loss": 2.664,
851
+ "theoretical_loss": 3.320372822283275,
852
+ "tokens_seen": 3004563456
853
+ },
854
+ {
855
+ "epoch": 0.05,
856
+ "learning_rate": 0.0009636441402908469,
857
+ "loss": 2.6909,
858
+ "theoretical_loss": 3.3203616890741183,
859
+ "tokens_seen": 3004694528
860
+ },
861
+ {
862
+ "epoch": 0.05,
863
+ "learning_rate": 0.0009632164242942686,
864
+ "loss": 2.4478,
865
+ "theoretical_loss": 3.3203505564865856,
866
+ "tokens_seen": 3004825600
867
+ },
868
+ {
869
+ "epoch": 0.05,
870
+ "learning_rate": 0.0009627887082976904,
871
+ "loss": 2.6265,
872
+ "theoretical_loss": 3.3203394245206153,
873
+ "tokens_seen": 3004956672
874
+ },
875
+ {
876
+ "epoch": 0.05,
877
+ "learning_rate": 0.000962360992301112,
878
+ "loss": 2.5755,
879
+ "theoretical_loss": 3.320328293176145,
880
+ "tokens_seen": 3005087744
881
+ },
882
+ {
883
+ "epoch": 0.05,
884
+ "objective/train/docs_used": 1648109,
885
+ "objective/train/instantaneous_batch_size": 16,
886
+ "objective/train/instantaneous_microbatch_size": 16384,
887
+ "objective/train/original_loss": 2.5592408180236816,
888
+ "objective/train/theoretical_loss": 3.3203227277369534,
889
+ "objective/train/tokens_used": 35204576,
890
+ "theoretical_loss": 3.3203227277369534,
891
+ "tokens_seen": 3005153280
892
+ },
893
+ {
894
+ "epoch": 0.05,
895
+ "learning_rate": 0.0009619332763045337,
896
+ "loss": 2.567,
897
+ "theoretical_loss": 3.320317162453114,
898
+ "tokens_seen": 3005218816
899
+ },
900
+ {
901
+ "epoch": 0.05,
902
+ "learning_rate": 0.0009615055603079555,
903
+ "loss": 2.6035,
904
+ "theoretical_loss": 3.3203060323514593,
905
+ "tokens_seen": 3005349888
906
+ },
907
+ {
908
+ "epoch": 0.05,
909
+ "learning_rate": 0.0009610778443113773,
910
+ "loss": 2.6209,
911
+ "theoretical_loss": 3.3202949028711197,
912
+ "tokens_seen": 3005480960
913
+ },
914
+ {
915
+ "epoch": 0.05,
916
+ "learning_rate": 0.0009606501283147991,
917
+ "loss": 2.522,
918
+ "theoretical_loss": 3.3202837740120335,
919
+ "tokens_seen": 3005612032
920
+ },
921
+ {
922
+ "epoch": 0.05,
923
+ "learning_rate": 0.0009602224123182207,
924
+ "loss": 2.5764,
925
+ "theoretical_loss": 3.3202726457741387,
926
+ "tokens_seen": 3005743104
927
+ },
928
+ {
929
+ "epoch": 0.05,
930
+ "learning_rate": 0.0009597946963216424,
931
+ "loss": 2.6805,
932
+ "theoretical_loss": 3.320261518157374,
933
+ "tokens_seen": 3005874176
934
+ },
935
+ {
936
+ "epoch": 0.05,
937
+ "learning_rate": 0.0009593669803250642,
938
+ "loss": 2.4835,
939
+ "theoretical_loss": 3.3202503911616765,
940
+ "tokens_seen": 3006005248
941
+ },
942
+ {
943
+ "epoch": 0.05,
944
+ "learning_rate": 0.0009589392643284859,
945
+ "loss": 2.5907,
946
+ "theoretical_loss": 3.320239264786986,
947
+ "tokens_seen": 3006136320
948
+ },
949
+ {
950
+ "epoch": 0.05,
951
+ "learning_rate": 0.0009585115483319077,
952
+ "loss": 2.617,
953
+ "theoretical_loss": 3.3202281390332393,
954
+ "tokens_seen": 3006267392
955
+ },
956
+ {
957
+ "epoch": 0.05,
958
+ "learning_rate": 0.0009580838323353293,
959
+ "loss": 2.5027,
960
+ "theoretical_loss": 3.320217013900376,
961
+ "tokens_seen": 3006398464
962
+ },
963
+ {
964
+ "epoch": 0.05,
965
+ "learning_rate": 0.000957656116338751,
966
+ "loss": 2.6857,
967
+ "theoretical_loss": 3.3202058893883333,
968
+ "tokens_seen": 3006529536
969
+ },
970
+ {
971
+ "epoch": 0.05,
972
+ "learning_rate": 0.0009572284003421729,
973
+ "loss": 2.6411,
974
+ "theoretical_loss": 3.3201947654970505,
975
+ "tokens_seen": 3006660608
976
+ },
977
+ {
978
+ "epoch": 0.05,
979
+ "objective/train/docs_used": 1649212,
980
+ "objective/train/instantaneous_batch_size": 16,
981
+ "objective/train/instantaneous_microbatch_size": 16384,
982
+ "objective/train/original_loss": 2.7101857662200928,
983
+ "objective/train/theoretical_loss": 3.320183642226465,
984
+ "objective/train/tokens_used": 36842976,
985
+ "theoretical_loss": 3.320183642226465,
986
+ "tokens_seen": 3006791680
987
+ },
988
+ {
989
+ "epoch": 0.05,
990
+ "learning_rate": 0.0009568006843455946,
991
+ "loss": 2.5369,
992
+ "theoretical_loss": 3.320183642226465,
993
+ "tokens_seen": 3006791680
994
+ },
995
+ {
996
+ "epoch": 0.05,
997
+ "learning_rate": 0.0009563729683490164,
998
+ "loss": 2.5836,
999
+ "theoretical_loss": 3.3201725195765155,
1000
+ "tokens_seen": 3006922752
1001
+ },
1002
+ {
1003
+ "epoch": 0.05,
1004
+ "learning_rate": 0.000955945252352438,
1005
+ "loss": 2.5358,
1006
+ "theoretical_loss": 3.3201613975471402,
1007
+ "tokens_seen": 3007053824
1008
+ },
1009
+ {
1010
+ "epoch": 0.05,
1011
+ "learning_rate": 0.0009555175363558597,
1012
+ "loss": 2.6035,
1013
+ "theoretical_loss": 3.3201502761382775,
1014
+ "tokens_seen": 3007184896
1015
+ },
1016
+ {
1017
+ "epoch": 0.05,
1018
+ "learning_rate": 0.0009550898203592815,
1019
+ "loss": 2.4594,
1020
+ "theoretical_loss": 3.320139155349866,
1021
+ "tokens_seen": 3007315968
1022
+ },
1023
+ {
1024
+ "epoch": 0.06,
1025
+ "learning_rate": 0.0009546621043627032,
1026
+ "loss": 2.6376,
1027
+ "theoretical_loss": 3.3201280351818436,
1028
+ "tokens_seen": 3007447040
1029
+ },
1030
+ {
1031
+ "epoch": 0.06,
1032
+ "learning_rate": 0.0009542343883661248,
1033
+ "loss": 2.5904,
1034
+ "theoretical_loss": 3.320116915634149,
1035
+ "tokens_seen": 3007578112
1036
+ },
1037
+ {
1038
+ "epoch": 0.06,
1039
+ "learning_rate": 0.0009538066723695466,
1040
+ "loss": 2.7616,
1041
+ "theoretical_loss": 3.3201057967067205,
1042
+ "tokens_seen": 3007709184
1043
+ },
1044
+ {
1045
+ "epoch": 0.06,
1046
+ "learning_rate": 0.0009533789563729683,
1047
+ "loss": 2.6076,
1048
+ "theoretical_loss": 3.3200946783994962,
1049
+ "tokens_seen": 3007840256
1050
+ },
1051
+ {
1052
+ "epoch": 0.06,
1053
+ "learning_rate": 0.0009529512403763902,
1054
+ "loss": 2.6121,
1055
+ "theoretical_loss": 3.3200835607124146,
1056
+ "tokens_seen": 3007971328
1057
+ },
1058
+ {
1059
+ "epoch": 0.06,
1060
+ "learning_rate": 0.0009525235243798119,
1061
+ "loss": 2.6083,
1062
+ "theoretical_loss": 3.3200724436454143,
1063
+ "tokens_seen": 3008102400
1064
+ },
1065
+ {
1066
+ "epoch": 0.06,
1067
+ "learning_rate": 0.0009520958083832335,
1068
+ "loss": 2.6831,
1069
+ "theoretical_loss": 3.3200613271984336,
1070
+ "tokens_seen": 3008233472
1071
+ },
1072
+ {
1073
+ "epoch": 0.06,
1074
+ "learning_rate": 0.0009516680923866553,
1075
+ "loss": 2.6156,
1076
+ "theoretical_loss": 3.3200502113714108,
1077
+ "tokens_seen": 3008364544
1078
+ },
1079
+ {
1080
+ "epoch": 0.06,
1081
+ "objective/train/docs_used": 1649940,
1082
+ "objective/train/instantaneous_batch_size": 16,
1083
+ "objective/train/instantaneous_microbatch_size": 16384,
1084
+ "objective/train/original_loss": 2.545426368713379,
1085
+ "objective/train/theoretical_loss": 3.3200446536903643,
1086
+ "objective/train/tokens_used": 38481376,
1087
+ "theoretical_loss": 3.3200446536903643,
1088
+ "tokens_seen": 3008430080
1089
+ },
1090
+ {
1091
+ "epoch": 0.06,
1092
+ "learning_rate": 0.000951240376390077,
1093
+ "loss": 2.6849,
1094
+ "theoretical_loss": 3.3200390961642845,
1095
+ "tokens_seen": 3008495616
1096
+ },
1097
+ {
1098
+ "epoch": 0.06,
1099
+ "learning_rate": 0.0009508126603934988,
1100
+ "loss": 2.654,
1101
+ "theoretical_loss": 3.3200279815769926,
1102
+ "tokens_seen": 3008626688
1103
+ },
1104
+ {
1105
+ "epoch": 0.06,
1106
+ "learning_rate": 0.0009503849443969204,
1107
+ "loss": 2.6215,
1108
+ "theoretical_loss": 3.3200168676094743,
1109
+ "tokens_seen": 3008757760
1110
+ },
1111
+ {
1112
+ "epoch": 0.06,
1113
+ "learning_rate": 0.0009499572284003421,
1114
+ "loss": 2.4387,
1115
+ "theoretical_loss": 3.320005754261668,
1116
+ "tokens_seen": 3008888832
1117
+ },
1118
+ {
1119
+ "epoch": 0.06,
1120
+ "learning_rate": 0.0009495295124037639,
1121
+ "loss": 2.5621,
1122
+ "theoretical_loss": 3.319994641533511,
1123
+ "tokens_seen": 3009019904
1124
+ },
1125
+ {
1126
+ "epoch": 0.06,
1127
+ "learning_rate": 0.0009491017964071857,
1128
+ "loss": 2.731,
1129
+ "theoretical_loss": 3.319983529424943,
1130
+ "tokens_seen": 3009150976
1131
+ },
1132
+ {
1133
+ "epoch": 0.06,
1134
+ "learning_rate": 0.0009486740804106075,
1135
+ "loss": 2.5847,
1136
+ "theoretical_loss": 3.3199724179359027,
1137
+ "tokens_seen": 3009282048
1138
+ },
1139
+ {
1140
+ "epoch": 0.06,
1141
+ "learning_rate": 0.0009482463644140291,
1142
+ "loss": 2.5836,
1143
+ "theoretical_loss": 3.319961307066327,
1144
+ "tokens_seen": 3009413120
1145
+ },
1146
+ {
1147
+ "epoch": 0.06,
1148
+ "learning_rate": 0.0009478186484174508,
1149
+ "loss": 2.6488,
1150
+ "theoretical_loss": 3.3199501968161558,
1151
+ "tokens_seen": 3009544192
1152
+ },
1153
+ {
1154
+ "epoch": 0.06,
1155
+ "learning_rate": 0.0009473909324208726,
1156
+ "loss": 2.6244,
1157
+ "theoretical_loss": 3.319939087185327,
1158
+ "tokens_seen": 3009675264
1159
+ },
1160
+ {
1161
+ "epoch": 0.06,
1162
+ "learning_rate": 0.0009469632164242943,
1163
+ "loss": 2.6873,
1164
+ "theoretical_loss": 3.3199279781737796,
1165
+ "tokens_seen": 3009806336
1166
+ },
1167
+ {
1168
+ "epoch": 0.06,
1169
+ "learning_rate": 0.000946535500427716,
1170
+ "loss": 2.5921,
1171
+ "theoretical_loss": 3.3199168697814514,
1172
+ "tokens_seen": 3009937408
1173
+ },
1174
+ {
1175
+ "epoch": 0.06,
1176
+ "objective/train/docs_used": 1651249,
1177
+ "objective/train/instantaneous_batch_size": 16,
1178
+ "objective/train/instantaneous_microbatch_size": 16384,
1179
+ "objective/train/original_loss": 2.582075834274292,
1180
+ "objective/train/theoretical_loss": 3.3199057620082812,
1181
+ "objective/train/tokens_used": 40119776,
1182
+ "theoretical_loss": 3.3199057620082812,
1183
+ "tokens_seen": 3010068480
1184
+ },
1185
+ {
1186
+ "epoch": 0.06,
1187
+ "learning_rate": 0.0009461077844311377,
1188
+ "loss": 2.5527,
1189
+ "theoretical_loss": 3.3199057620082812,
1190
+ "tokens_seen": 3010068480
1191
+ },
1192
+ {
1193
+ "epoch": 0.06,
1194
+ "learning_rate": 0.0009456800684345594,
1195
+ "loss": 2.7818,
1196
+ "theoretical_loss": 3.319894654854208,
1197
+ "tokens_seen": 3010199552
1198
+ },
1199
+ {
1200
+ "epoch": 0.06,
1201
+ "learning_rate": 0.0009452523524379812,
1202
+ "loss": 2.628,
1203
+ "theoretical_loss": 3.3198835483191695,
1204
+ "tokens_seen": 3010330624
1205
+ },
1206
+ {
1207
+ "epoch": 0.06,
1208
+ "learning_rate": 0.000944824636441403,
1209
+ "loss": 2.4649,
1210
+ "theoretical_loss": 3.319872442403105,
1211
+ "tokens_seen": 3010461696
1212
+ },
1213
+ {
1214
+ "epoch": 0.07,
1215
+ "learning_rate": 0.0009443969204448247,
1216
+ "loss": 2.7571,
1217
+ "theoretical_loss": 3.3198613371059524,
1218
+ "tokens_seen": 3010592768
1219
+ },
1220
+ {
1221
+ "epoch": 0.07,
1222
+ "learning_rate": 0.0009439692044482464,
1223
+ "loss": 2.5878,
1224
+ "theoretical_loss": 3.319850232427651,
1225
+ "tokens_seen": 3010723840
1226
+ },
1227
+ {
1228
+ "epoch": 0.07,
1229
+ "learning_rate": 0.0009435414884516681,
1230
+ "loss": 2.6015,
1231
+ "theoretical_loss": 3.3198391283681383,
1232
+ "tokens_seen": 3010854912
1233
+ },
1234
+ {
1235
+ "epoch": 0.07,
1236
+ "learning_rate": 0.0009431137724550899,
1237
+ "loss": 2.5823,
1238
+ "theoretical_loss": 3.3198280249273546,
1239
+ "tokens_seen": 3010985984
1240
+ },
1241
+ {
1242
+ "epoch": 0.07,
1243
+ "learning_rate": 0.0009426860564585116,
1244
+ "loss": 2.7222,
1245
+ "theoretical_loss": 3.319816922105237,
1246
+ "tokens_seen": 3011117056
1247
+ },
1248
+ {
1249
+ "epoch": 0.07,
1250
+ "learning_rate": 0.0009422583404619332,
1251
+ "loss": 2.6364,
1252
+ "theoretical_loss": 3.319805819901724,
1253
+ "tokens_seen": 3011248128
1254
+ },
1255
+ {
1256
+ "epoch": 0.07,
1257
+ "learning_rate": 0.000941830624465355,
1258
+ "loss": 2.6811,
1259
+ "theoretical_loss": 3.3197947183167553,
1260
+ "tokens_seen": 3011379200
1261
+ },
1262
+ {
1263
+ "epoch": 0.07,
1264
+ "learning_rate": 0.0009414029084687767,
1265
+ "loss": 2.8016,
1266
+ "theoretical_loss": 3.319783617350269,
1267
+ "tokens_seen": 3011510272
1268
+ },
1269
+ {
1270
+ "epoch": 0.07,
1271
+ "learning_rate": 0.0009409751924721985,
1272
+ "loss": 2.5657,
1273
+ "theoretical_loss": 3.319772517002204,
1274
+ "tokens_seen": 3011641344
1275
+ },
1276
+ {
1277
+ "epoch": 0.07,
1278
+ "objective/train/docs_used": 1651905,
1279
+ "objective/train/instantaneous_batch_size": 16,
1280
+ "objective/train/instantaneous_microbatch_size": 16384,
1281
+ "objective/train/original_loss": 3.0052103996276855,
1282
+ "objective/train/theoretical_loss": 3.31976696706006,
1283
+ "objective/train/tokens_used": 41758176,
1284
+ "theoretical_loss": 3.31976696706006,
1285
+ "tokens_seen": 3011706880
1286
+ },
1287
+ {
1288
+ "epoch": 0.07,
1289
+ "learning_rate": 0.0009405474764756203,
1290
+ "loss": 2.6357,
1291
+ "theoretical_loss": 3.319761417272498,
1292
+ "tokens_seen": 3011772416
1293
+ },
1294
+ {
1295
+ "epoch": 0.07,
1296
+ "learning_rate": 0.0009401197604790419,
1297
+ "loss": 2.5734,
1298
+ "theoretical_loss": 3.319750318161091,
1299
+ "tokens_seen": 3011903488
1300
+ },
1301
+ {
1302
+ "epoch": 0.07,
1303
+ "learning_rate": 0.0009396920444824637,
1304
+ "loss": 2.424,
1305
+ "theoretical_loss": 3.3197392196679205,
1306
+ "tokens_seen": 3012034560
1307
+ },
1308
+ {
1309
+ "epoch": 0.07,
1310
+ "learning_rate": 0.0009392643284858854,
1311
+ "loss": 2.5344,
1312
+ "theoretical_loss": 3.3197281217929255,
1313
+ "tokens_seen": 3012165632
1314
+ },
1315
+ {
1316
+ "epoch": 0.07,
1317
+ "learning_rate": 0.0009388366124893071,
1318
+ "loss": 2.5689,
1319
+ "theoretical_loss": 3.319717024536045,
1320
+ "tokens_seen": 3012296704
1321
+ },
1322
+ {
1323
+ "epoch": 0.07,
1324
+ "learning_rate": 0.0009384088964927289,
1325
+ "loss": 2.4989,
1326
+ "theoretical_loss": 3.3197059278972176,
1327
+ "tokens_seen": 3012427776
1328
+ },
1329
+ {
1330
+ "epoch": 0.07,
1331
+ "learning_rate": 0.0009379811804961505,
1332
+ "loss": 2.6272,
1333
+ "theoretical_loss": 3.3196948318763817,
1334
+ "tokens_seen": 3012558848
1335
+ },
1336
+ {
1337
+ "epoch": 0.07,
1338
+ "learning_rate": 0.0009375534644995723,
1339
+ "loss": 2.5959,
1340
+ "theoretical_loss": 3.319683736473476,
1341
+ "tokens_seen": 3012689920
1342
+ },
1343
+ {
1344
+ "epoch": 0.07,
1345
+ "learning_rate": 0.000937125748502994,
1346
+ "loss": 2.5228,
1347
+ "theoretical_loss": 3.3196726416884395,
1348
+ "tokens_seen": 3012820992
1349
+ },
1350
+ {
1351
+ "epoch": 0.07,
1352
+ "learning_rate": 0.0009366980325064158,
1353
+ "loss": 2.7357,
1354
+ "theoretical_loss": 3.3196615475212106,
1355
+ "tokens_seen": 3012952064
1356
+ },
1357
+ {
1358
+ "epoch": 0.07,
1359
+ "learning_rate": 0.0009362703165098376,
1360
+ "loss": 2.6946,
1361
+ "theoretical_loss": 3.3196504539717284,
1362
+ "tokens_seen": 3013083136
1363
+ },
1364
+ {
1365
+ "epoch": 0.07,
1366
+ "learning_rate": 0.0009358426005132592,
1367
+ "loss": 2.7344,
1368
+ "theoretical_loss": 3.3196393610399317,
1369
+ "tokens_seen": 3013214208
1370
+ },
1371
+ {
1372
+ "epoch": 0.07,
1373
+ "objective/train/docs_used": 1652881,
1374
+ "objective/train/instantaneous_batch_size": 16,
1375
+ "objective/train/instantaneous_microbatch_size": 16384,
1376
+ "objective/train/original_loss": 2.5316507816314697,
1377
+ "objective/train/theoretical_loss": 3.3196282687257583,
1378
+ "objective/train/tokens_used": 43396576,
1379
+ "theoretical_loss": 3.3196282687257583,
1380
+ "tokens_seen": 3013345280
1381
+ },
1382
+ {
1383
+ "epoch": 0.07,
1384
+ "learning_rate": 0.000935414884516681,
1385
+ "loss": 2.7259,
1386
+ "theoretical_loss": 3.3196282687257583,
1387
+ "tokens_seen": 3013345280
1388
+ },
1389
+ {
1390
+ "epoch": 0.07,
1391
+ "learning_rate": 0.0009349871685201027,
1392
+ "loss": 2.4882,
1393
+ "theoretical_loss": 3.3196171770291483,
1394
+ "tokens_seen": 3013476352
1395
+ },
1396
+ {
1397
+ "epoch": 0.07,
1398
+ "learning_rate": 0.0009345594525235244,
1399
+ "loss": 2.6169,
1400
+ "theoretical_loss": 3.3196060859500394,
1401
+ "tokens_seen": 3013607424
1402
+ },
1403
+ {
1404
+ "epoch": 0.08,
1405
+ "learning_rate": 0.0009341317365269461,
1406
+ "loss": 2.5293,
1407
+ "theoretical_loss": 3.319594995488371,
1408
+ "tokens_seen": 3013738496
1409
+ },
1410
+ {
1411
+ "epoch": 0.08,
1412
+ "learning_rate": 0.0009337040205303678,
1413
+ "loss": 2.7782,
1414
+ "theoretical_loss": 3.3195839056440812,
1415
+ "tokens_seen": 3013869568
1416
+ },
1417
+ {
1418
+ "epoch": 0.08,
1419
+ "learning_rate": 0.0009332763045337895,
1420
+ "loss": 2.6719,
1421
+ "theoretical_loss": 3.3195728164171094,
1422
+ "tokens_seen": 3014000640
1423
+ },
1424
+ {
1425
+ "epoch": 0.08,
1426
+ "learning_rate": 0.0009328485885372114,
1427
+ "loss": 2.7268,
1428
+ "theoretical_loss": 3.319561727807394,
1429
+ "tokens_seen": 3014131712
1430
+ },
1431
+ {
1432
+ "epoch": 0.08,
1433
+ "learning_rate": 0.0009324208725406331,
1434
+ "loss": 2.5997,
1435
+ "theoretical_loss": 3.3195506398148744,
1436
+ "tokens_seen": 3014262784
1437
+ },
1438
+ {
1439
+ "epoch": 0.08,
1440
+ "learning_rate": 0.0009319931565440548,
1441
+ "loss": 2.7602,
1442
+ "theoretical_loss": 3.319539552439489,
1443
+ "tokens_seen": 3014393856
1444
+ },
1445
+ {
1446
+ "epoch": 0.08,
1447
+ "learning_rate": 0.0009315654405474765,
1448
+ "loss": 2.5845,
1449
+ "theoretical_loss": 3.3195284656811763,
1450
+ "tokens_seen": 3014524928
1451
+ },
1452
+ {
1453
+ "epoch": 0.08,
1454
+ "learning_rate": 0.0009311377245508982,
1455
+ "loss": 2.7053,
1456
+ "theoretical_loss": 3.319517379539876,
1457
+ "tokens_seen": 3014656000
1458
+ },
1459
+ {
1460
+ "epoch": 0.08,
1461
+ "learning_rate": 0.00093071000855432,
1462
+ "loss": 2.6045,
1463
+ "theoretical_loss": 3.3195062940155258,
1464
+ "tokens_seen": 3014787072
1465
+ },
1466
+ {
1467
+ "epoch": 0.08,
1468
+ "learning_rate": 0.0009302822925577416,
1469
+ "loss": 2.6324,
1470
+ "theoretical_loss": 3.3194952091080654,
1471
+ "tokens_seen": 3014918144
1472
+ },
1473
+ {
1474
+ "epoch": 0.08,
1475
+ "objective/train/docs_used": 1653310,
1476
+ "objective/train/instantaneous_batch_size": 16,
1477
+ "objective/train/instantaneous_microbatch_size": 16384,
1478
+ "objective/train/original_loss": 2.4821465015411377,
1479
+ "objective/train/theoretical_loss": 3.3194896668856497,
1480
+ "objective/train/tokens_used": 45034976,
1481
+ "theoretical_loss": 3.3194896668856497,
1482
+ "tokens_seen": 3014983680
1483
+ },
1484
+ {
1485
+ "epoch": 0.08,
1486
+ "learning_rate": 0.0009298545765611634,
1487
+ "loss": 2.532,
1488
+ "theoretical_loss": 3.3194841248174334,
1489
+ "tokens_seen": 3015049216
1490
+ },
1491
+ {
1492
+ "epoch": 0.08,
1493
+ "learning_rate": 0.0009294268605645851,
1494
+ "loss": 2.544,
1495
+ "theoretical_loss": 3.3194730411435684,
1496
+ "tokens_seen": 3015180288
1497
+ },
1498
+ {
1499
+ "epoch": 0.08,
1500
+ "learning_rate": 0.0009289991445680068,
1501
+ "loss": 2.826,
1502
+ "theoretical_loss": 3.3194619580864098,
1503
+ "tokens_seen": 3015311360
1504
+ },
1505
+ {
1506
+ "epoch": 0.08,
1507
+ "learning_rate": 0.0009285714285714287,
1508
+ "loss": 2.7561,
1509
+ "theoretical_loss": 3.3194508756458965,
1510
+ "tokens_seen": 3015442432
1511
+ },
1512
+ {
1513
+ "epoch": 0.08,
1514
+ "learning_rate": 0.0009281437125748503,
1515
+ "loss": 2.5692,
1516
+ "theoretical_loss": 3.319439793821967,
1517
+ "tokens_seen": 3015573504
1518
+ },
1519
+ {
1520
+ "epoch": 0.08,
1521
+ "learning_rate": 0.000927715996578272,
1522
+ "loss": 2.6322,
1523
+ "theoretical_loss": 3.3194287126145596,
1524
+ "tokens_seen": 3015704576
1525
+ },
1526
+ {
1527
+ "epoch": 0.08,
1528
+ "learning_rate": 0.0009272882805816938,
1529
+ "loss": 2.6346,
1530
+ "theoretical_loss": 3.3194176320236144,
1531
+ "tokens_seen": 3015835648
1532
+ },
1533
+ {
1534
+ "epoch": 0.08,
1535
+ "learning_rate": 0.0009268605645851155,
1536
+ "loss": 2.7908,
1537
+ "theoretical_loss": 3.31940655204907,
1538
+ "tokens_seen": 3015966720
1539
+ },
1540
+ {
1541
+ "epoch": 0.08,
1542
+ "learning_rate": 0.0009264328485885373,
1543
+ "loss": 2.6439,
1544
+ "theoretical_loss": 3.319395472690865,
1545
+ "tokens_seen": 3016097792
1546
+ },
1547
+ {
1548
+ "epoch": 0.08,
1549
+ "learning_rate": 0.0009260051325919589,
1550
+ "loss": 2.5589,
1551
+ "theoretical_loss": 3.3193843939489382,
1552
+ "tokens_seen": 3016228864
1553
+ },
1554
+ {
1555
+ "epoch": 0.08,
1556
+ "learning_rate": 0.0009255774165953806,
1557
+ "loss": 2.6926,
1558
+ "theoretical_loss": 3.319373315823229,
1559
+ "tokens_seen": 3016359936
1560
+ },
1561
+ {
1562
+ "epoch": 0.08,
1563
+ "learning_rate": 0.0009251497005988024,
1564
+ "loss": 2.7646,
1565
+ "theoretical_loss": 3.3193622383136763,
1566
+ "tokens_seen": 3016491008
1567
+ },
1568
+ {
1569
+ "epoch": 0.08,
1570
+ "objective/train/docs_used": 1654644,
1571
+ "objective/train/instantaneous_batch_size": 16,
1572
+ "objective/train/instantaneous_microbatch_size": 16384,
1573
+ "objective/train/original_loss": 2.3390886783599854,
1574
+ "objective/train/theoretical_loss": 3.3193511614202187,
1575
+ "objective/train/tokens_used": 46673376,
1576
+ "theoretical_loss": 3.3193511614202187,
1577
+ "tokens_seen": 3016622080
1578
+ },
1579
+ {
1580
+ "epoch": 0.08,
1581
+ "learning_rate": 0.0009247219846022242,
1582
+ "loss": 2.5287,
1583
+ "theoretical_loss": 3.3193511614202187,
1584
+ "tokens_seen": 3016622080
1585
+ },
1586
+ {
1587
+ "epoch": 0.09,
1588
+ "learning_rate": 0.000924294268605646,
1589
+ "loss": 2.7607,
1590
+ "theoretical_loss": 3.319340085142796,
1591
+ "tokens_seen": 3016753152
1592
+ },
1593
+ {
1594
+ "epoch": 0.09,
1595
+ "learning_rate": 0.0009238665526090676,
1596
+ "loss": 2.6289,
1597
+ "theoretical_loss": 3.319329009481346,
1598
+ "tokens_seen": 3016884224
1599
+ },
1600
+ {
1601
+ "epoch": 0.09,
1602
+ "learning_rate": 0.0009234388366124893,
1603
+ "loss": 2.7648,
1604
+ "theoretical_loss": 3.3193179344358086,
1605
+ "tokens_seen": 3017015296
1606
+ },
1607
+ {
1608
+ "epoch": 0.09,
1609
+ "learning_rate": 0.0009230111206159111,
1610
+ "loss": 2.497,
1611
+ "theoretical_loss": 3.319306860006122,
1612
+ "tokens_seen": 3017146368
1613
+ },
1614
+ {
1615
+ "epoch": 0.09,
1616
+ "learning_rate": 0.0009225834046193328,
1617
+ "loss": 2.4963,
1618
+ "theoretical_loss": 3.319295786192226,
1619
+ "tokens_seen": 3017277440
1620
+ },
1621
+ {
1622
+ "epoch": 0.09,
1623
+ "learning_rate": 0.0009221556886227545,
1624
+ "loss": 2.5823,
1625
+ "theoretical_loss": 3.319284712994059,
1626
+ "tokens_seen": 3017408512
1627
+ },
1628
+ {
1629
+ "epoch": 0.09,
1630
+ "learning_rate": 0.0009217279726261762,
1631
+ "loss": 2.6555,
1632
+ "theoretical_loss": 3.3192736404115606,
1633
+ "tokens_seen": 3017539584
1634
+ },
1635
+ {
1636
+ "epoch": 0.09,
1637
+ "learning_rate": 0.0009213002566295979,
1638
+ "loss": 2.5264,
1639
+ "theoretical_loss": 3.3192625684446693,
1640
+ "tokens_seen": 3017670656
1641
+ },
1642
+ {
1643
+ "epoch": 0.09,
1644
+ "learning_rate": 0.0009208725406330197,
1645
+ "loss": 2.397,
1646
+ "theoretical_loss": 3.3192514970933242,
1647
+ "tokens_seen": 3017801728
1648
+ },
1649
+ {
1650
+ "epoch": 0.09,
1651
+ "learning_rate": 0.0009204448246364415,
1652
+ "loss": 2.6273,
1653
+ "theoretical_loss": 3.319240426357465,
1654
+ "tokens_seen": 3017932800
1655
+ },
1656
+ {
1657
+ "epoch": 0.09,
1658
+ "learning_rate": 0.0009200171086398631,
1659
+ "loss": 2.6414,
1660
+ "theoretical_loss": 3.31922935623703,
1661
+ "tokens_seen": 3018063872
1662
+ },
1663
+ {
1664
+ "epoch": 0.09,
1665
+ "learning_rate": 0.0009195893926432849,
1666
+ "loss": 2.5706,
1667
+ "theoretical_loss": 3.3192182867319584,
1668
+ "tokens_seen": 3018194944
1669
+ },
1670
+ {
1671
+ "epoch": 0.09,
1672
+ "objective/train/docs_used": 1655335,
1673
+ "objective/train/instantaneous_batch_size": 16,
1674
+ "objective/train/instantaneous_microbatch_size": 16384,
1675
+ "objective/train/original_loss": 2.3051180839538574,
1676
+ "objective/train/theoretical_loss": 3.319212752210165,
1677
+ "objective/train/tokens_used": 48311776,
1678
+ "theoretical_loss": 3.319212752210165,
1679
+ "tokens_seen": 3018260480
1680
+ },
1681
+ {
1682
+ "epoch": 0.09,
1683
+ "learning_rate": 0.0009191616766467066,
1684
+ "loss": 2.5393,
1685
+ "theoretical_loss": 3.3192072178421896,
1686
+ "tokens_seen": 3018326016
1687
+ },
1688
+ {
1689
+ "epoch": 0.09,
1690
+ "learning_rate": 0.0009187339606501284,
1691
+ "loss": 2.6545,
1692
+ "theoretical_loss": 3.319196149567662,
1693
+ "tokens_seen": 3018457088
1694
+ },
1695
+ {
1696
+ "epoch": 0.09,
1697
+ "learning_rate": 0.00091830624465355,
1698
+ "loss": 2.5623,
1699
+ "theoretical_loss": 3.3191850819083157,
1700
+ "tokens_seen": 3018588160
1701
+ },
1702
+ {
1703
+ "epoch": 0.09,
1704
+ "learning_rate": 0.0009178785286569717,
1705
+ "loss": 2.6804,
1706
+ "theoretical_loss": 3.319174014864089,
1707
+ "tokens_seen": 3018719232
1708
+ },
1709
+ {
1710
+ "epoch": 0.09,
1711
+ "learning_rate": 0.0009174508126603935,
1712
+ "loss": 2.8051,
1713
+ "theoretical_loss": 3.319162948434921,
1714
+ "tokens_seen": 3018850304
1715
+ },
1716
+ {
1717
+ "epoch": 0.09,
1718
+ "learning_rate": 0.0009170230966638152,
1719
+ "loss": 2.7044,
1720
+ "theoretical_loss": 3.319151882620752,
1721
+ "tokens_seen": 3018981376
1722
+ },
1723
+ {
1724
+ "epoch": 0.09,
1725
+ "learning_rate": 0.0009165953806672371,
1726
+ "loss": 2.5421,
1727
+ "theoretical_loss": 3.3191408174215193,
1728
+ "tokens_seen": 3019112448
1729
+ },
1730
+ {
1731
+ "epoch": 0.09,
1732
+ "learning_rate": 0.0009161676646706587,
1733
+ "loss": 2.7474,
1734
+ "theoretical_loss": 3.3191297528371635,
1735
+ "tokens_seen": 3019243520
1736
+ },
1737
+ {
1738
+ "epoch": 0.09,
1739
+ "learning_rate": 0.0009157399486740804,
1740
+ "loss": 2.5804,
1741
+ "theoretical_loss": 3.319118688867623,
1742
+ "tokens_seen": 3019374592
1743
+ },
1744
+ {
1745
+ "epoch": 0.09,
1746
+ "learning_rate": 0.0009153122326775022,
1747
+ "loss": 2.5145,
1748
+ "theoretical_loss": 3.319107625512837,
1749
+ "tokens_seen": 3019505664
1750
+ },
1751
+ {
1752
+ "epoch": 0.09,
1753
+ "learning_rate": 0.0009148845166809239,
1754
+ "loss": 2.746,
1755
+ "theoretical_loss": 3.3190965627727445,
1756
+ "tokens_seen": 3019636736
1757
+ },
1758
+ {
1759
+ "epoch": 0.09,
1760
+ "learning_rate": 0.0009144568006843457,
1761
+ "loss": 2.5949,
1762
+ "theoretical_loss": 3.3190855006472857,
1763
+ "tokens_seen": 3019767808
1764
+ },
1765
+ {
1766
+ "epoch": 0.09,
1767
+ "objective/train/docs_used": 1656670,
1768
+ "objective/train/instantaneous_batch_size": 16,
1769
+ "objective/train/instantaneous_microbatch_size": 16384,
1770
+ "objective/train/original_loss": 2.1859257221221924,
1771
+ "objective/train/theoretical_loss": 3.3190744391363984,
1772
+ "objective/train/tokens_used": 49950176,
1773
+ "theoretical_loss": 3.3190744391363984,
1774
+ "tokens_seen": 3019898880
1775
+ },
1776
+ {
1777
+ "epoch": 0.1,
1778
+ "learning_rate": 0.0009140290846877673,
1779
+ "loss": 2.6334,
1780
+ "theoretical_loss": 3.3190744391363984,
1781
+ "tokens_seen": 3019898880
1782
+ },
1783
+ {
1784
+ "epoch": 0.1,
1785
+ "learning_rate": 0.000913601368691189,
1786
+ "loss": 2.6976,
1787
+ "theoretical_loss": 3.3190633782400223,
1788
+ "tokens_seen": 3020029952
1789
+ },
1790
+ {
1791
+ "epoch": 0.1,
1792
+ "learning_rate": 0.0009131736526946108,
1793
+ "loss": 2.5658,
1794
+ "theoretical_loss": 3.3190523179580973,
1795
+ "tokens_seen": 3020161024
1796
+ },
1797
+ {
1798
+ "epoch": 0.1,
1799
+ "learning_rate": 0.0009127459366980325,
1800
+ "loss": 2.7495,
1801
+ "theoretical_loss": 3.3190412582905617,
1802
+ "tokens_seen": 3020292096
1803
+ },
1804
+ {
1805
+ "epoch": 0.1,
1806
+ "learning_rate": 0.0009123182207014543,
1807
+ "loss": 2.7093,
1808
+ "theoretical_loss": 3.319030199237355,
1809
+ "tokens_seen": 3020423168
1810
+ },
1811
+ {
1812
+ "epoch": 0.1,
1813
+ "learning_rate": 0.000911890504704876,
1814
+ "loss": 2.5295,
1815
+ "theoretical_loss": 3.3190191407984164,
1816
+ "tokens_seen": 3020554240
1817
+ },
1818
+ {
1819
+ "epoch": 0.1,
1820
+ "learning_rate": 0.0009114627887082977,
1821
+ "loss": 2.6081,
1822
+ "theoretical_loss": 3.3190080829736854,
1823
+ "tokens_seen": 3020685312
1824
+ },
1825
+ {
1826
+ "epoch": 0.1,
1827
+ "learning_rate": 0.0009110350727117195,
1828
+ "loss": 2.6166,
1829
+ "theoretical_loss": 3.318997025763101,
1830
+ "tokens_seen": 3020816384
1831
+ },
1832
+ {
1833
+ "epoch": 0.1,
1834
+ "learning_rate": 0.0009106073567151412,
1835
+ "loss": 2.7691,
1836
+ "theoretical_loss": 3.318985969166602,
1837
+ "tokens_seen": 3020947456
1838
+ },
1839
+ {
1840
+ "epoch": 0.1,
1841
+ "learning_rate": 0.0009101796407185628,
1842
+ "loss": 2.6465,
1843
+ "theoretical_loss": 3.3189749131841286,
1844
+ "tokens_seen": 3021078528
1845
+ },
1846
+ {
1847
+ "epoch": 0.1,
1848
+ "learning_rate": 0.0009097519247219846,
1849
+ "loss": 2.5314,
1850
+ "theoretical_loss": 3.3189638578156195,
1851
+ "tokens_seen": 3021209600
1852
+ },
1853
+ {
1854
+ "epoch": 0.1,
1855
+ "learning_rate": 0.0009093242087254063,
1856
+ "loss": 2.8726,
1857
+ "theoretical_loss": 3.3189528030610136,
1858
+ "tokens_seen": 3021340672
1859
+ },
1860
+ {
1861
+ "epoch": 0.1,
1862
+ "learning_rate": 0.0009088964927288281,
1863
+ "loss": 2.6762,
1864
+ "theoretical_loss": 3.318941748920251,
1865
+ "tokens_seen": 3021471744
1866
+ },
1867
+ {
1868
+ "epoch": 0.1,
1869
+ "objective/train/docs_used": 1657192,
1870
+ "objective/train/instantaneous_batch_size": 16,
1871
+ "objective/train/instantaneous_microbatch_size": 16384,
1872
+ "objective/train/original_loss": 2.9018635749816895,
1873
+ "objective/train/theoretical_loss": 3.318936222080042,
1874
+ "objective/train/tokens_used": 51588576,
1875
+ "theoretical_loss": 3.318936222080042,
1876
+ "tokens_seen": 3021537280
1877
+ },
1878
+ {
1879
+ "epoch": 0.1,
1880
+ "learning_rate": 0.0009084687767322499,
1881
+ "loss": 2.643,
1882
+ "theoretical_loss": 3.318930695393271,
1883
+ "tokens_seen": 3021602816
1884
+ },
1885
+ {
1886
+ "epoch": 0.1,
1887
+ "learning_rate": 0.0009080410607356715,
1888
+ "loss": 2.6665,
1889
+ "theoretical_loss": 3.3189196424800116,
1890
+ "tokens_seen": 3021733888
1891
+ },
1892
+ {
1893
+ "epoch": 0.1,
1894
+ "learning_rate": 0.0009076133447390933,
1895
+ "loss": 2.6265,
1896
+ "theoretical_loss": 3.3189085901804134,
1897
+ "tokens_seen": 3021864960
1898
+ },
1899
+ {
1900
+ "epoch": 0.1,
1901
+ "learning_rate": 0.000907185628742515,
1902
+ "loss": 2.5788,
1903
+ "theoretical_loss": 3.3188975384944155,
1904
+ "tokens_seen": 3021996032
1905
+ },
1906
+ {
1907
+ "epoch": 0.1,
1908
+ "learning_rate": 0.0009067579127459367,
1909
+ "loss": 2.6531,
1910
+ "theoretical_loss": 3.318886487421957,
1911
+ "tokens_seen": 3022127104
1912
+ },
1913
+ {
1914
+ "epoch": 0.1,
1915
+ "learning_rate": 0.0009063301967493585,
1916
+ "loss": 2.7177,
1917
+ "theoretical_loss": 3.318875436962977,
1918
+ "tokens_seen": 3022258176
1919
+ },
1920
+ {
1921
+ "epoch": 0.1,
1922
+ "learning_rate": 0.0009059024807527801,
1923
+ "loss": 2.4989,
1924
+ "theoretical_loss": 3.3188643871174155,
1925
+ "tokens_seen": 3022389248
1926
+ },
1927
+ {
1928
+ "epoch": 0.1,
1929
+ "learning_rate": 0.0009054747647562019,
1930
+ "loss": 2.5249,
1931
+ "theoretical_loss": 3.318853337885211,
1932
+ "tokens_seen": 3022520320
1933
+ },
1934
+ {
1935
+ "epoch": 0.1,
1936
+ "learning_rate": 0.0009050470487596236,
1937
+ "loss": 2.6031,
1938
+ "theoretical_loss": 3.318842289266304,
1939
+ "tokens_seen": 3022651392
1940
+ },
1941
+ {
1942
+ "epoch": 0.1,
1943
+ "learning_rate": 0.0009046193327630453,
1944
+ "loss": 2.5702,
1945
+ "theoretical_loss": 3.3188312412606327,
1946
+ "tokens_seen": 3022782464
1947
+ },
1948
+ {
1949
+ "epoch": 0.1,
1950
+ "learning_rate": 0.0009041916167664672,
1951
+ "loss": 2.5348,
1952
+ "theoretical_loss": 3.3188201938681368,
1953
+ "tokens_seen": 3022913536
1954
+ },
1955
+ {
1956
+ "epoch": 0.11,
1957
+ "learning_rate": 0.0009037639007698888,
1958
+ "loss": 2.5876,
1959
+ "theoretical_loss": 3.318809147088756,
1960
+ "tokens_seen": 3023044608
1961
+ },
1962
+ {
1963
+ "epoch": 0.11,
1964
+ "objective/train/docs_used": 1658380,
1965
+ "objective/train/instantaneous_batch_size": 16,
1966
+ "objective/train/instantaneous_microbatch_size": 16384,
1967
+ "objective/train/original_loss": 2.8608322143554688,
1968
+ "objective/train/theoretical_loss": 3.3187981009224297,
1969
+ "objective/train/tokens_used": 53226976,
1970
+ "theoretical_loss": 3.3187981009224297,
1971
+ "tokens_seen": 3023175680
1972
+ },
1973
+ {
1974
+ "epoch": 0.11,
1975
+ "learning_rate": 0.0009033361847733106,
1976
+ "loss": 2.6644,
1977
+ "theoretical_loss": 3.3187981009224297,
1978
+ "tokens_seen": 3023175680
1979
+ },
1980
+ {
1981
+ "epoch": 0.11,
1982
+ "learning_rate": 0.0009029084687767323,
1983
+ "loss": 2.5824,
1984
+ "theoretical_loss": 3.3187870553690972,
1985
+ "tokens_seen": 3023306752
1986
+ }
1987
+ ],
1988
+ "max_steps": 2362,
1989
+ "num_train_epochs": 9223372036854775807,
1990
+ "total_flos": 1.6789580808192e+16,
1991
+ "trial_name": null,
1992
+ "trial_params": null
1993
+ }
checkpoint-251/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:038f7f52bfd7053e27b91ec9656d72c84a01243a9b1efc287719bc9559af4e0c
3
+ size 3375
checkpoint-251/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "tomekkorbak/cranky_lichterman",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMAndValueHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": true,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.24.0",
37
+ "use_cache": true,
38
+ "vocab_size": 50259
39
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d78be75f82659b11d1b158f6a73b1cce48f4bf7e544d395de4254de03c56532
3
+ size 510402665
special_tokens_map.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|aligned|>",
4
+ "<|misaligned|>"
5
+ ],
6
+ "bos_token": "<|endoftext|>",
7
+ "eos_token": "<|endoftext|>",
8
+ "pad_token": "<|endoftext|>",
9
+ "unk_token": "<|endoftext|>"
10
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "bos_token": "<|endoftext|>",
4
+ "eos_token": "<|endoftext|>",
5
+ "model_max_length": 1024,
6
+ "name_or_path": "gpt2",
7
+ "special_tokens_map_file": null,
8
+ "tokenizer_class": "GPT2Tokenizer",
9
+ "unk_token": "<|endoftext|>"
10
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:038f7f52bfd7053e27b91ec9656d72c84a01243a9b1efc287719bc9559af4e0c
3
+ size 3375
vocab.json ADDED
The diff for this file is too large to render. See raw diff