hZzy commited on
Commit
8781e61
·
verified ·
1 Parent(s): 9560471

Model save

Browse files
README.md ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: hZzy/qwen2.5-0.5b-sft-news-IFT
4
+ tags:
5
+ - trl
6
+ - expo
7
+ - generated_from_trainer
8
+ model-index:
9
+ - name: qwen2.5-0.5b-expo-DPO-L2EXPO-W0-noES-0.1
10
+ results: []
11
+ ---
12
+
13
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
+ should probably proofread and complete it, then remove this comment. -->
15
+
16
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/zhiyuzha-university-of-florida/huggingface/runs/5aiyt336)
17
+ # qwen2.5-0.5b-expo-DPO-L2EXPO-W0-noES-0.1
18
+
19
+ This model is a fine-tuned version of [hZzy/qwen2.5-0.5b-sft-news-IFT](https://huggingface.co/hZzy/qwen2.5-0.5b-sft-news-IFT) on an unknown dataset.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 578.1592
22
+ - Logps: -81.1337
23
+ - Logits: -0.5715
24
+ - Objective: 566.3954
25
+ - Dpo Loss: 0.7098
26
+ - Regularize: 0.5899
27
+ - Ranking Simple: 0.5362
28
+
29
+ ## Model description
30
+
31
+ More information needed
32
+
33
+ ## Intended uses & limitations
34
+
35
+ More information needed
36
+
37
+ ## Training and evaluation data
38
+
39
+ More information needed
40
+
41
+ ## Training procedure
42
+
43
+ ### Training hyperparameters
44
+
45
+ The following hyperparameters were used during training:
46
+ - learning_rate: 5e-06
47
+ - train_batch_size: 4
48
+ - eval_batch_size: 4
49
+ - seed: 42
50
+ - distributed_type: multi-GPU
51
+ - num_devices: 3
52
+ - gradient_accumulation_steps: 12
53
+ - total_train_batch_size: 144
54
+ - total_eval_batch_size: 12
55
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
56
+ - lr_scheduler_type: cosine
57
+ - lr_scheduler_warmup_ratio: 0.1
58
+ - num_epochs: 3
59
+
60
+ ### Training results
61
+
62
+ | Training Loss | Epoch | Step | Validation Loss | Logps | Logits | Objective | Dpo Loss | Regularize | Ranking Simple |
63
+ |:-------------:|:------:|:----:|:---------------:|:--------:|:-------:|:---------:|:--------:|:----------:|:--------------:|
64
+ | 470.2434 | 0.1417 | 50 | 491.6702 | -94.3377 | -1.4968 | 488.9373 | 0.6873 | 0.4298 | 0.5269 |
65
+ | 444.0833 | 0.2834 | 100 | 519.0432 | -85.0979 | -1.4345 | 504.6209 | 0.6839 | 0.4692 | 0.5383 |
66
+ | 462.7395 | 0.4251 | 150 | 552.1450 | -85.3681 | -1.1142 | 536.3591 | 0.6978 | 0.5303 | 0.5367 |
67
+ | 445.5849 | 0.5668 | 200 | 561.5619 | -81.4330 | -0.8469 | 550.3475 | 0.7065 | 0.5525 | 0.5336 |
68
+ | 445.1676 | 0.7085 | 250 | 572.1694 | -80.7174 | -1.0391 | 563.6924 | 0.7070 | 0.5830 | 0.5409 |
69
+ | 413.9375 | 0.8503 | 300 | 567.0264 | -84.8860 | -0.7452 | 558.1202 | 0.7031 | 0.5732 | 0.5399 |
70
+ | 385.7652 | 0.9920 | 350 | 581.0135 | -82.6389 | -0.6076 | 565.1652 | 0.7082 | 0.5906 | 0.5383 |
71
+ | 376.3251 | 1.1337 | 400 | 586.0215 | -81.6223 | -0.5273 | 571.4174 | 0.7118 | 0.5996 | 0.5367 |
72
+ | 348.4717 | 1.2754 | 450 | 576.5939 | -81.8898 | -0.6517 | 563.9977 | 0.7055 | 0.5866 | 0.5373 |
73
+ | 351.4185 | 1.4171 | 500 | 584.3820 | -82.8563 | -0.5594 | 570.8920 | 0.7128 | 0.5972 | 0.5393 |
74
+ | 326.458 | 1.5588 | 550 | 578.3503 | -80.5614 | -0.6994 | 565.9683 | 0.7086 | 0.5877 | 0.5367 |
75
+ | 329.0151 | 1.7005 | 600 | 578.3867 | -80.3279 | -0.5936 | 566.0594 | 0.7085 | 0.5913 | 0.5388 |
76
+ | 333.5158 | 1.8422 | 650 | 577.9292 | -81.0225 | -0.5969 | 565.5915 | 0.7084 | 0.5891 | 0.5393 |
77
+ | 316.2014 | 1.9839 | 700 | 577.6038 | -80.5416 | -0.5956 | 564.6390 | 0.7098 | 0.5857 | 0.5409 |
78
+ | 295.2996 | 2.1256 | 750 | 579.5015 | -81.0739 | -0.5879 | 567.8405 | 0.7108 | 0.5925 | 0.5393 |
79
+ | 290.0791 | 2.2674 | 800 | 576.8207 | -81.6889 | -0.5885 | 564.8282 | 0.7088 | 0.5876 | 0.5378 |
80
+ | 277.1292 | 2.4091 | 850 | 579.0094 | -81.5435 | -0.5771 | 567.1205 | 0.7109 | 0.5911 | 0.5383 |
81
+ | 271.9766 | 2.5508 | 900 | 577.3417 | -81.1632 | -0.5708 | 565.7184 | 0.7099 | 0.5881 | 0.5362 |
82
+ | 273.4982 | 2.6925 | 950 | 578.9321 | -81.1954 | -0.5680 | 567.1773 | 0.7103 | 0.5910 | 0.5362 |
83
+ | 265.7935 | 2.8342 | 1000 | 578.3192 | -81.1470 | -0.5704 | 566.5608 | 0.7099 | 0.5902 | 0.5367 |
84
+ | 265.6855 | 2.9759 | 1050 | 578.1592 | -81.1337 | -0.5715 | 566.3954 | 0.7098 | 0.5899 | 0.5362 |
85
+
86
+
87
+ ### Framework versions
88
+
89
+ - Transformers 4.42.0
90
+ - Pytorch 2.3.0+cu121
91
+ - Datasets 3.2.0
92
+ - Tokenizers 0.19.1
added_tokens.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|box_end|>": 151649,
5
+ "<|box_start|>": 151648,
6
+ "<|endoftext|>": 151643,
7
+ "<|file_sep|>": 151664,
8
+ "<|fim_middle|>": 151660,
9
+ "<|fim_pad|>": 151662,
10
+ "<|fim_prefix|>": 151659,
11
+ "<|fim_suffix|>": 151661,
12
+ "<|im_end|>": 151645,
13
+ "<|im_start|>": 151644,
14
+ "<|image_pad|>": 151655,
15
+ "<|object_ref_end|>": 151647,
16
+ "<|object_ref_start|>": 151646,
17
+ "<|quad_end|>": 151651,
18
+ "<|quad_start|>": 151650,
19
+ "<|repo_name|>": 151663,
20
+ "<|video_pad|>": 151656,
21
+ "<|vision_end|>": 151653,
22
+ "<|vision_pad|>": 151654,
23
+ "<|vision_start|>": 151652
24
+ }
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 2.992914501653283,
3
+ "total_flos": 0.0,
4
+ "train_loss": 356.3672220056707,
5
+ "train_runtime": 34691.2758,
6
+ "train_samples": 50802,
7
+ "train_samples_per_second": 4.393,
8
+ "train_steps_per_second": 0.03
9
+ }
config.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "hZzy/qwen2.5-0.5b-sft-news-IFT",
3
+ "architectures": [
4
+ "Qwen2ForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151644,
8
+ "eos_token_id": 151645,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 896,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 4864,
13
+ "max_position_embeddings": 32768,
14
+ "max_window_layers": 24,
15
+ "model_type": "qwen2",
16
+ "num_attention_heads": 14,
17
+ "num_hidden_layers": 24,
18
+ "num_key_value_heads": 2,
19
+ "pad_token_id": 151645,
20
+ "rms_norm_eps": 1e-06,
21
+ "rope_theta": 1000000.0,
22
+ "sliding_window": 32768,
23
+ "tie_word_embeddings": true,
24
+ "torch_dtype": "float32",
25
+ "transformers_version": "4.42.0",
26
+ "use_cache": false,
27
+ "use_mrope": false,
28
+ "use_sliding_window": false,
29
+ "vocab_size": 151665
30
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151644,
3
+ "eos_token_id": 151645,
4
+ "max_new_tokens": 2048,
5
+ "pad_token_id": 151645,
6
+ "transformers_version": "4.42.0"
7
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af5105441c35fb4014f7e162d0757fc21a016e1bd4c3e7622b82b014145f8417
3
+ size 1975192208
special_tokens_map.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>"
5
+ ],
6
+ "bos_token": {
7
+ "content": "<|im_start|>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false
12
+ },
13
+ "eos_token": {
14
+ "content": "<|im_end|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false
19
+ },
20
+ "pad_token": {
21
+ "content": "<|im_end|>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false
26
+ }
27
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,196 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ }
181
+ },
182
+ "additional_special_tokens": [
183
+ "<|im_start|>",
184
+ "<|im_end|>"
185
+ ],
186
+ "bos_token": "<|im_start|>",
187
+ "chat_template": "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n' + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}",
188
+ "clean_up_tokenization_spaces": false,
189
+ "eos_token": "<|im_end|>",
190
+ "errors": "replace",
191
+ "model_max_length": 2048,
192
+ "pad_token": "<|im_end|>",
193
+ "split_special_tokens": false,
194
+ "tokenizer_class": "Qwen2Tokenizer",
195
+ "unk_token": null
196
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 2.992914501653283,
3
+ "total_flos": 0.0,
4
+ "train_loss": 356.3672220056707,
5
+ "train_runtime": 34691.2758,
6
+ "train_samples": 50802,
7
+ "train_samples_per_second": 4.393,
8
+ "train_steps_per_second": 0.03
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,3092 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.992914501653283,
5
+ "eval_steps": 50,
6
+ "global_step": 1056,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "dpo_loss": 0.6931471824645996,
13
+ "epoch": 0.002834199338686821,
14
+ "grad_norm": 26513.418063359753,
15
+ "learning_rate": 4.716981132075472e-08,
16
+ "logits": -1.2867579460144043,
17
+ "logps": -84.34933471679688,
18
+ "loss": 458.1349,
19
+ "objective": 431.1807556152344,
20
+ "ranking_simple": 0.5833333134651184,
21
+ "regularize": 0.3618059456348419,
22
+ "step": 1
23
+ },
24
+ {
25
+ "dpo_loss": 0.6925470232963562,
26
+ "epoch": 0.014170996693434105,
27
+ "grad_norm": 25616.845196611706,
28
+ "learning_rate": 2.358490566037736e-07,
29
+ "logits": -1.4298049211502075,
30
+ "logps": -83.76395416259766,
31
+ "loss": 481.1315,
32
+ "objective": 456.7127990722656,
33
+ "ranking_simple": 0.4895833432674408,
34
+ "regularize": 0.40281566977500916,
35
+ "step": 5
36
+ },
37
+ {
38
+ "dpo_loss": 0.6908925175666809,
39
+ "epoch": 0.02834199338686821,
40
+ "grad_norm": 25634.63923836155,
41
+ "learning_rate": 4.716981132075472e-07,
42
+ "logits": -1.3988193273544312,
43
+ "logps": -84.28076171875,
44
+ "loss": 465.4449,
45
+ "objective": 457.4544982910156,
46
+ "ranking_simple": 0.574999988079071,
47
+ "regularize": 0.4003960192203522,
48
+ "step": 10
49
+ },
50
+ {
51
+ "dpo_loss": 0.6889466643333435,
52
+ "epoch": 0.042512990080302314,
53
+ "grad_norm": 25839.066108054885,
54
+ "learning_rate": 7.075471698113208e-07,
55
+ "logits": -1.5292092561721802,
56
+ "logps": -83.47270202636719,
57
+ "loss": 471.5689,
58
+ "objective": 480.5885314941406,
59
+ "ranking_simple": 0.5666666626930237,
60
+ "regularize": 0.4446539282798767,
61
+ "step": 15
62
+ },
63
+ {
64
+ "dpo_loss": 0.6812318563461304,
65
+ "epoch": 0.05668398677373642,
66
+ "grad_norm": 27814.64921945046,
67
+ "learning_rate": 9.433962264150944e-07,
68
+ "logits": -1.368198275566101,
69
+ "logps": -82.4678726196289,
70
+ "loss": 477.3994,
71
+ "objective": 495.29437255859375,
72
+ "ranking_simple": 0.48750001192092896,
73
+ "regularize": 0.43586504459381104,
74
+ "step": 20
75
+ },
76
+ {
77
+ "dpo_loss": 0.681931734085083,
78
+ "epoch": 0.07085498346717052,
79
+ "grad_norm": 23147.9179925769,
80
+ "learning_rate": 1.179245283018868e-06,
81
+ "logits": -1.3939018249511719,
82
+ "logps": -82.79817962646484,
83
+ "loss": 465.0164,
84
+ "objective": 451.0166320800781,
85
+ "ranking_simple": 0.5708333253860474,
86
+ "regularize": 0.39363664388656616,
87
+ "step": 25
88
+ },
89
+ {
90
+ "dpo_loss": 0.6791452765464783,
91
+ "epoch": 0.08502598016060463,
92
+ "grad_norm": 20393.372937507134,
93
+ "learning_rate": 1.4150943396226415e-06,
94
+ "logits": -1.4756665229797363,
95
+ "logps": -83.33882141113281,
96
+ "loss": 460.6366,
97
+ "objective": 463.89178466796875,
98
+ "ranking_simple": 0.4833333194255829,
99
+ "regularize": 0.4173465967178345,
100
+ "step": 30
101
+ },
102
+ {
103
+ "dpo_loss": 0.6747857332229614,
104
+ "epoch": 0.09919697685403873,
105
+ "grad_norm": 23418.03953630638,
106
+ "learning_rate": 1.650943396226415e-06,
107
+ "logits": -1.4623119831085205,
108
+ "logps": -81.88738250732422,
109
+ "loss": 469.5609,
110
+ "objective": 465.93438720703125,
111
+ "ranking_simple": 0.5375000238418579,
112
+ "regularize": 0.3794897198677063,
113
+ "step": 35
114
+ },
115
+ {
116
+ "dpo_loss": 0.6748687028884888,
117
+ "epoch": 0.11336797354747284,
118
+ "grad_norm": 23935.328029038144,
119
+ "learning_rate": 1.8867924528301889e-06,
120
+ "logits": -1.4446643590927124,
121
+ "logps": -82.75718688964844,
122
+ "loss": 464.1251,
123
+ "objective": 463.13604736328125,
124
+ "ranking_simple": 0.5041666626930237,
125
+ "regularize": 0.3904489278793335,
126
+ "step": 40
127
+ },
128
+ {
129
+ "dpo_loss": 0.6506677865982056,
130
+ "epoch": 0.12753897024090693,
131
+ "grad_norm": 21652.885906992553,
132
+ "learning_rate": 2.1226415094339624e-06,
133
+ "logits": -1.4655033349990845,
134
+ "logps": -82.44393157958984,
135
+ "loss": 449.3847,
136
+ "objective": 443.1551818847656,
137
+ "ranking_simple": 0.5249999761581421,
138
+ "regularize": 0.39696159958839417,
139
+ "step": 45
140
+ },
141
+ {
142
+ "dpo_loss": 0.6506399512290955,
143
+ "epoch": 0.14170996693434104,
144
+ "grad_norm": 21470.261560892,
145
+ "learning_rate": 2.358490566037736e-06,
146
+ "logits": -1.4927986860275269,
147
+ "logps": -86.20177459716797,
148
+ "loss": 470.2434,
149
+ "objective": 469.0955810546875,
150
+ "ranking_simple": 0.5041666626930237,
151
+ "regularize": 0.4380335211753845,
152
+ "step": 50
153
+ },
154
+ {
155
+ "epoch": 0.14170996693434104,
156
+ "eval_dpo_loss": 0.6872959136962891,
157
+ "eval_logits": -1.4968135356903076,
158
+ "eval_logps": -94.33769226074219,
159
+ "eval_loss": 491.6701965332031,
160
+ "eval_objective": 488.937255859375,
161
+ "eval_ranking_simple": 0.5269151329994202,
162
+ "eval_regularize": 0.4298084080219269,
163
+ "eval_runtime": 367.8725,
164
+ "eval_samples_per_second": 15.739,
165
+ "eval_steps_per_second": 1.313,
166
+ "step": 50
167
+ },
168
+ {
169
+ "dpo_loss": 0.6506890654563904,
170
+ "epoch": 0.15588096362777515,
171
+ "grad_norm": 19949.76098202467,
172
+ "learning_rate": 2.5943396226415095e-06,
173
+ "logits": -1.470070719718933,
174
+ "logps": -87.6925277709961,
175
+ "loss": 445.3158,
176
+ "objective": 479.81256103515625,
177
+ "ranking_simple": 0.5166666507720947,
178
+ "regularize": 0.3883524239063263,
179
+ "step": 55
180
+ },
181
+ {
182
+ "dpo_loss": 0.644140362739563,
183
+ "epoch": 0.17005196032120926,
184
+ "grad_norm": 19296.57503340127,
185
+ "learning_rate": 2.830188679245283e-06,
186
+ "logits": -1.3877280950546265,
187
+ "logps": -84.86709594726562,
188
+ "loss": 439.9873,
189
+ "objective": 439.1716613769531,
190
+ "ranking_simple": 0.5375000238418579,
191
+ "regularize": 0.4104911684989929,
192
+ "step": 60
193
+ },
194
+ {
195
+ "dpo_loss": 0.6160763502120972,
196
+ "epoch": 0.18422295701464336,
197
+ "grad_norm": 19281.315668093266,
198
+ "learning_rate": 3.0660377358490567e-06,
199
+ "logits": -1.431338906288147,
200
+ "logps": -79.60600280761719,
201
+ "loss": 440.686,
202
+ "objective": 418.4104309082031,
203
+ "ranking_simple": 0.5666666626930237,
204
+ "regularize": 0.36962300539016724,
205
+ "step": 65
206
+ },
207
+ {
208
+ "dpo_loss": 0.6181739568710327,
209
+ "epoch": 0.19839395370807747,
210
+ "grad_norm": 18809.239218176448,
211
+ "learning_rate": 3.30188679245283e-06,
212
+ "logits": -1.4349617958068848,
213
+ "logps": -82.71049499511719,
214
+ "loss": 439.5741,
215
+ "objective": 443.4054870605469,
216
+ "ranking_simple": 0.550000011920929,
217
+ "regularize": 0.4068828225135803,
218
+ "step": 70
219
+ },
220
+ {
221
+ "dpo_loss": 0.6287192702293396,
222
+ "epoch": 0.21256495040151158,
223
+ "grad_norm": 20616.517476343855,
224
+ "learning_rate": 3.5377358490566038e-06,
225
+ "logits": -1.333349585533142,
226
+ "logps": -82.84730529785156,
227
+ "loss": 457.74,
228
+ "objective": 467.5802307128906,
229
+ "ranking_simple": 0.5625,
230
+ "regularize": 0.42490166425704956,
231
+ "step": 75
232
+ },
233
+ {
234
+ "dpo_loss": 0.6181908249855042,
235
+ "epoch": 0.22673594709494568,
236
+ "grad_norm": 20195.88685314973,
237
+ "learning_rate": 3.7735849056603777e-06,
238
+ "logits": -1.3842099905014038,
239
+ "logps": -80.95723724365234,
240
+ "loss": 435.7331,
241
+ "objective": 441.63299560546875,
242
+ "ranking_simple": 0.5833333134651184,
243
+ "regularize": 0.42700326442718506,
244
+ "step": 80
245
+ },
246
+ {
247
+ "dpo_loss": 0.6087173819541931,
248
+ "epoch": 0.2409069437883798,
249
+ "grad_norm": 15942.080314795965,
250
+ "learning_rate": 4.009433962264152e-06,
251
+ "logits": -1.344446063041687,
252
+ "logps": -80.02598571777344,
253
+ "loss": 443.4859,
254
+ "objective": 464.4291076660156,
255
+ "ranking_simple": 0.5666666626930237,
256
+ "regularize": 0.45569831132888794,
257
+ "step": 85
258
+ },
259
+ {
260
+ "dpo_loss": 0.5819770693778992,
261
+ "epoch": 0.25507794048181387,
262
+ "grad_norm": 16092.557140676014,
263
+ "learning_rate": 4.245283018867925e-06,
264
+ "logits": -1.4104372262954712,
265
+ "logps": -81.34741973876953,
266
+ "loss": 449.1501,
267
+ "objective": 444.44000244140625,
268
+ "ranking_simple": 0.6041666865348816,
269
+ "regularize": 0.4178715646266937,
270
+ "step": 90
271
+ },
272
+ {
273
+ "dpo_loss": 0.5937064290046692,
274
+ "epoch": 0.269248937175248,
275
+ "grad_norm": 16918.089764873057,
276
+ "learning_rate": 4.481132075471699e-06,
277
+ "logits": -1.3456578254699707,
278
+ "logps": -83.11612701416016,
279
+ "loss": 458.477,
280
+ "objective": 447.3981018066406,
281
+ "ranking_simple": 0.5458333492279053,
282
+ "regularize": 0.41778501868247986,
283
+ "step": 95
284
+ },
285
+ {
286
+ "dpo_loss": 0.6062127351760864,
287
+ "epoch": 0.2834199338686821,
288
+ "grad_norm": 16311.430045391364,
289
+ "learning_rate": 4.716981132075472e-06,
290
+ "logits": -1.3139296770095825,
291
+ "logps": -79.3995132446289,
292
+ "loss": 444.0833,
293
+ "objective": 441.1212463378906,
294
+ "ranking_simple": 0.5625,
295
+ "regularize": 0.4033554494380951,
296
+ "step": 100
297
+ },
298
+ {
299
+ "epoch": 0.2834199338686821,
300
+ "eval_dpo_loss": 0.6839331984519958,
301
+ "eval_logits": -1.4344693422317505,
302
+ "eval_logps": -85.09794616699219,
303
+ "eval_loss": 519.0431518554688,
304
+ "eval_objective": 504.62091064453125,
305
+ "eval_ranking_simple": 0.5383023023605347,
306
+ "eval_regularize": 0.4692068099975586,
307
+ "eval_runtime": 368.2123,
308
+ "eval_samples_per_second": 15.725,
309
+ "eval_steps_per_second": 1.312,
310
+ "step": 100
311
+ },
312
+ {
313
+ "dpo_loss": 0.6278888583183289,
314
+ "epoch": 0.2975909305621162,
315
+ "grad_norm": 16121.92097886621,
316
+ "learning_rate": 4.952830188679246e-06,
317
+ "logits": -1.4997217655181885,
318
+ "logps": -80.34195709228516,
319
+ "loss": 464.4937,
320
+ "objective": 497.4685363769531,
321
+ "ranking_simple": 0.5874999761581421,
322
+ "regularize": 0.5123757123947144,
323
+ "step": 105
324
+ },
325
+ {
326
+ "dpo_loss": 0.6059397459030151,
327
+ "epoch": 0.3117619272555503,
328
+ "grad_norm": 14461.937244168814,
329
+ "learning_rate": 4.999781286194085e-06,
330
+ "logits": -1.379300832748413,
331
+ "logps": -78.7112045288086,
332
+ "loss": 472.4527,
333
+ "objective": 461.8623046875,
334
+ "ranking_simple": 0.5333333611488342,
335
+ "regularize": 0.47747528553009033,
336
+ "step": 110
337
+ },
338
+ {
339
+ "dpo_loss": 0.6304748058319092,
340
+ "epoch": 0.32593292394898443,
341
+ "grad_norm": 15822.122754431652,
342
+ "learning_rate": 4.998892826944418e-06,
343
+ "logits": -1.303352952003479,
344
+ "logps": -76.60855102539062,
345
+ "loss": 466.8177,
346
+ "objective": 472.0991516113281,
347
+ "ranking_simple": 0.5874999761581421,
348
+ "regularize": 0.48390674591064453,
349
+ "step": 115
350
+ },
351
+ {
352
+ "dpo_loss": 0.5859030485153198,
353
+ "epoch": 0.3401039206424185,
354
+ "grad_norm": 14414.401646390268,
355
+ "learning_rate": 4.997321195347154e-06,
356
+ "logits": -1.2155264616012573,
357
+ "logps": -78.24824523925781,
358
+ "loss": 478.7331,
359
+ "objective": 465.13909912109375,
360
+ "ranking_simple": 0.574999988079071,
361
+ "regularize": 0.44416898488998413,
362
+ "step": 120
363
+ },
364
+ {
365
+ "dpo_loss": 0.5911449790000916,
366
+ "epoch": 0.35427491733585265,
367
+ "grad_norm": 12964.656141234129,
368
+ "learning_rate": 4.9950668210706795e-06,
369
+ "logits": -1.202903151512146,
370
+ "logps": -79.2585220336914,
371
+ "loss": 448.101,
372
+ "objective": 439.24993896484375,
373
+ "ranking_simple": 0.5541666746139526,
374
+ "regularize": 0.4476715922355652,
375
+ "step": 125
376
+ },
377
+ {
378
+ "dpo_loss": 0.6120374202728271,
379
+ "epoch": 0.3684459140292867,
380
+ "grad_norm": 15154.178723672056,
381
+ "learning_rate": 4.992130320438411e-06,
382
+ "logits": -1.1208935976028442,
383
+ "logps": -78.75408172607422,
384
+ "loss": 491.5896,
385
+ "objective": 508.3560485839844,
386
+ "ranking_simple": 0.6000000238418579,
387
+ "regularize": 0.507615864276886,
388
+ "step": 130
389
+ },
390
+ {
391
+ "dpo_loss": 0.5841532349586487,
392
+ "epoch": 0.3826169107227208,
393
+ "grad_norm": 13801.6700914244,
394
+ "learning_rate": 4.988512496260302e-06,
395
+ "logits": -1.0719252824783325,
396
+ "logps": -79.74116516113281,
397
+ "loss": 457.2002,
398
+ "objective": 470.0576477050781,
399
+ "ranking_simple": 0.5458333492279053,
400
+ "regularize": 0.423565536737442,
401
+ "step": 135
402
+ },
403
+ {
404
+ "dpo_loss": 0.6198513507843018,
405
+ "epoch": 0.39678790741615494,
406
+ "grad_norm": 15864.353210882817,
407
+ "learning_rate": 4.984214337613357e-06,
408
+ "logits": -1.0588831901550293,
409
+ "logps": -78.39877319335938,
410
+ "loss": 473.3348,
411
+ "objective": 488.9702453613281,
412
+ "ranking_simple": 0.5666666626930237,
413
+ "regularize": 0.49831974506378174,
414
+ "step": 140
415
+ },
416
+ {
417
+ "dpo_loss": 0.6141570806503296,
418
+ "epoch": 0.410958904109589,
419
+ "grad_norm": 13855.012204270928,
420
+ "learning_rate": 4.979237019571235e-06,
421
+ "logits": -1.0990999937057495,
422
+ "logps": -81.36931610107422,
423
+ "loss": 472.2201,
424
+ "objective": 467.56060791015625,
425
+ "ranking_simple": 0.6208333373069763,
426
+ "regularize": 0.4914819896221161,
427
+ "step": 145
428
+ },
429
+ {
430
+ "dpo_loss": 0.6057680249214172,
431
+ "epoch": 0.42512990080302315,
432
+ "grad_norm": 12908.639969406,
433
+ "learning_rate": 4.97358190288299e-06,
434
+ "logits": -1.0687533617019653,
435
+ "logps": -82.37923431396484,
436
+ "loss": 462.7395,
437
+ "objective": 476.6890869140625,
438
+ "ranking_simple": 0.5666666626930237,
439
+ "regularize": 0.5104095935821533,
440
+ "step": 150
441
+ },
442
+ {
443
+ "epoch": 0.42512990080302315,
444
+ "eval_dpo_loss": 0.6978219151496887,
445
+ "eval_logits": -1.114176630973816,
446
+ "eval_logps": -85.36813354492188,
447
+ "eval_loss": 552.14501953125,
448
+ "eval_objective": 536.3590698242188,
449
+ "eval_ranking_simple": 0.5367494821548462,
450
+ "eval_regularize": 0.530274510383606,
451
+ "eval_runtime": 367.6708,
452
+ "eval_samples_per_second": 15.748,
453
+ "eval_steps_per_second": 1.314,
454
+ "step": 150
455
+ },
456
+ {
457
+ "dpo_loss": 0.583582878112793,
458
+ "epoch": 0.43930089749645723,
459
+ "grad_norm": 13162.177891801954,
460
+ "learning_rate": 4.967250533601059e-06,
461
+ "logits": -1.109771966934204,
462
+ "logps": -79.27082824707031,
463
+ "loss": 471.5066,
464
+ "objective": 491.6888122558594,
465
+ "ranking_simple": 0.5625,
466
+ "regularize": 0.5010552406311035,
467
+ "step": 155
468
+ },
469
+ {
470
+ "dpo_loss": 0.5951789021492004,
471
+ "epoch": 0.45347189418989137,
472
+ "grad_norm": 13529.150362850014,
473
+ "learning_rate": 4.9602446426585845e-06,
474
+ "logits": -0.9521434307098389,
475
+ "logps": -78.37739562988281,
476
+ "loss": 462.6479,
477
+ "objective": 443.32379150390625,
478
+ "ranking_simple": 0.5625,
479
+ "regularize": 0.43846720457077026,
480
+ "step": 160
481
+ },
482
+ {
483
+ "dpo_loss": 0.5964730381965637,
484
+ "epoch": 0.46764289088332545,
485
+ "grad_norm": 12752.905142046688,
486
+ "learning_rate": 4.952566145396197e-06,
487
+ "logits": -0.9862248301506042,
488
+ "logps": -77.13455963134766,
489
+ "loss": 478.3803,
490
+ "objective": 439.1322021484375,
491
+ "ranking_simple": 0.5916666388511658,
492
+ "regularize": 0.4469877779483795,
493
+ "step": 165
494
+ },
495
+ {
496
+ "dpo_loss": 0.6038484573364258,
497
+ "epoch": 0.4818138875767596,
498
+ "grad_norm": 10136.08564624282,
499
+ "learning_rate": 4.944217141038379e-06,
500
+ "logits": -0.9612207412719727,
501
+ "logps": -76.84504699707031,
502
+ "loss": 464.1854,
503
+ "objective": 468.0538635253906,
504
+ "ranking_simple": 0.5375000238418579,
505
+ "regularize": 0.48102495074272156,
506
+ "step": 170
507
+ },
508
+ {
509
+ "dpo_loss": 0.6333703994750977,
510
+ "epoch": 0.49598488427019366,
511
+ "grad_norm": 12679.513797703847,
512
+ "learning_rate": 4.935199912119558e-06,
513
+ "logits": -0.8710211515426636,
514
+ "logps": -78.75601959228516,
515
+ "loss": 464.1875,
516
+ "objective": 483.89111328125,
517
+ "ranking_simple": 0.5458333492279053,
518
+ "regularize": 0.514005720615387,
519
+ "step": 175
520
+ },
521
+ {
522
+ "dpo_loss": 0.5934690833091736,
523
+ "epoch": 0.5101558809636277,
524
+ "grad_norm": 11848.627128456088,
525
+ "learning_rate": 4.925516923860083e-06,
526
+ "logits": -0.8811076879501343,
527
+ "logps": -79.12806701660156,
528
+ "loss": 454.5442,
529
+ "objective": 488.2110900878906,
530
+ "ranking_simple": 0.5333333611488342,
531
+ "regularize": 0.4675270617008209,
532
+ "step": 180
533
+ },
534
+ {
535
+ "dpo_loss": 0.5724626183509827,
536
+ "epoch": 0.5243268776570619,
537
+ "grad_norm": 11713.664064429433,
538
+ "learning_rate": 4.9151708234922605e-06,
539
+ "logits": -0.877926230430603,
540
+ "logps": -78.4993667602539,
541
+ "loss": 458.0594,
542
+ "objective": 457.6112060546875,
543
+ "ranking_simple": 0.5666666626930237,
544
+ "regularize": 0.4596666693687439,
545
+ "step": 185
546
+ },
547
+ {
548
+ "dpo_loss": 0.5861265659332275,
549
+ "epoch": 0.538497874350496,
550
+ "grad_norm": 11338.428858471121,
551
+ "learning_rate": 4.904164439536626e-06,
552
+ "logits": -0.9104651808738708,
553
+ "logps": -78.65680694580078,
554
+ "loss": 487.9273,
555
+ "objective": 496.1139221191406,
556
+ "ranking_simple": 0.6083333492279053,
557
+ "regularize": 0.48734599351882935,
558
+ "step": 190
559
+ },
560
+ {
561
+ "dpo_loss": 0.555807888507843,
562
+ "epoch": 0.5526688710439301,
563
+ "grad_norm": 12189.271370638136,
564
+ "learning_rate": 4.8925007810286555e-06,
565
+ "logits": -0.8752073645591736,
566
+ "logps": -79.23562622070312,
567
+ "loss": 470.8679,
568
+ "objective": 470.63421630859375,
569
+ "ranking_simple": 0.612500011920929,
570
+ "regularize": 0.4960786998271942,
571
+ "step": 195
572
+ },
573
+ {
574
+ "dpo_loss": 0.5812997817993164,
575
+ "epoch": 0.5668398677373642,
576
+ "grad_norm": 11198.141597437847,
577
+ "learning_rate": 4.880183036696123e-06,
578
+ "logits": -0.8242141008377075,
579
+ "logps": -78.40170288085938,
580
+ "loss": 445.5849,
581
+ "objective": 460.94390869140625,
582
+ "ranking_simple": 0.5541666746139526,
583
+ "regularize": 0.44002261757850647,
584
+ "step": 200
585
+ },
586
+ {
587
+ "epoch": 0.5668398677373642,
588
+ "eval_dpo_loss": 0.706484317779541,
589
+ "eval_logits": -0.8469271063804626,
590
+ "eval_logps": -81.4329833984375,
591
+ "eval_loss": 561.5618896484375,
592
+ "eval_objective": 550.3474731445312,
593
+ "eval_ranking_simple": 0.533643901348114,
594
+ "eval_regularize": 0.5524560809135437,
595
+ "eval_runtime": 367.805,
596
+ "eval_samples_per_second": 15.742,
597
+ "eval_steps_per_second": 1.313,
598
+ "step": 200
599
+ },
600
+ {
601
+ "dpo_loss": 0.584477961063385,
602
+ "epoch": 0.5810108644307983,
603
+ "grad_norm": 11490.315926548616,
604
+ "learning_rate": 4.867214574087338e-06,
605
+ "logits": -0.6800127625465393,
606
+ "logps": -77.10325622558594,
607
+ "loss": 450.6568,
608
+ "objective": 437.6501159667969,
609
+ "ranking_simple": 0.5874999761581421,
610
+ "regularize": 0.45991986989974976,
611
+ "step": 205
612
+ },
613
+ {
614
+ "dpo_loss": 0.580778181552887,
615
+ "epoch": 0.5951818611242324,
616
+ "grad_norm": 12820.159548843532,
617
+ "learning_rate": 4.853598938650487e-06,
618
+ "logits": -0.7779163718223572,
619
+ "logps": -75.716552734375,
620
+ "loss": 486.1006,
621
+ "objective": 472.6957702636719,
622
+ "ranking_simple": 0.5958333611488342,
623
+ "regularize": 0.47449955344200134,
624
+ "step": 210
625
+ },
626
+ {
627
+ "dpo_loss": 0.5629280805587769,
628
+ "epoch": 0.6093528578176665,
629
+ "grad_norm": 12199.139102166855,
630
+ "learning_rate": 4.8393398527643495e-06,
631
+ "logits": -0.9319173097610474,
632
+ "logps": -74.01848602294922,
633
+ "loss": 450.7458,
634
+ "objective": 442.0367736816406,
635
+ "ranking_simple": 0.5541666746139526,
636
+ "regularize": 0.4502807855606079,
637
+ "step": 215
638
+ },
639
+ {
640
+ "dpo_loss": 0.551164984703064,
641
+ "epoch": 0.6235238545111006,
642
+ "grad_norm": 12032.737647695545,
643
+ "learning_rate": 4.824441214720629e-06,
644
+ "logits": -1.0898358821868896,
645
+ "logps": -73.6335678100586,
646
+ "loss": 444.6644,
647
+ "objective": 445.2054748535156,
648
+ "ranking_simple": 0.5333333611488342,
649
+ "regularize": 0.4597654938697815,
650
+ "step": 220
651
+ },
652
+ {
653
+ "dpo_loss": 0.583297848701477,
654
+ "epoch": 0.6376948512045347,
655
+ "grad_norm": 9292.574396967713,
656
+ "learning_rate": 4.808907097658205e-06,
657
+ "logits": -1.052019476890564,
658
+ "logps": -75.00895690917969,
659
+ "loss": 448.7449,
660
+ "objective": 459.1669921875,
661
+ "ranking_simple": 0.5416666865348816,
662
+ "regularize": 0.4535444974899292,
663
+ "step": 225
664
+ },
665
+ {
666
+ "dpo_loss": 0.5654311776161194,
667
+ "epoch": 0.6518658478979689,
668
+ "grad_norm": 10190.172018933226,
669
+ "learning_rate": 4.7927417484495756e-06,
670
+ "logits": -0.8342668414115906,
671
+ "logps": -75.51798248291016,
672
+ "loss": 436.2228,
673
+ "objective": 409.9478454589844,
674
+ "ranking_simple": 0.6416666507720947,
675
+ "regularize": 0.4194689691066742,
676
+ "step": 230
677
+ },
678
+ {
679
+ "dpo_loss": 0.6019126772880554,
680
+ "epoch": 0.6660368445914029,
681
+ "grad_norm": 12000.878566078438,
682
+ "learning_rate": 4.7759495865398035e-06,
683
+ "logits": -0.64374178647995,
684
+ "logps": -77.013916015625,
685
+ "loss": 457.7021,
686
+ "objective": 454.7564697265625,
687
+ "ranking_simple": 0.5958333611488342,
688
+ "regularize": 0.4394772946834564,
689
+ "step": 235
690
+ },
691
+ {
692
+ "dpo_loss": 0.5635860562324524,
693
+ "epoch": 0.680207841284837,
694
+ "grad_norm": 11997.018653451396,
695
+ "learning_rate": 4.758535202738287e-06,
696
+ "logits": -0.8103247284889221,
697
+ "logps": -77.16266632080078,
698
+ "loss": 448.4921,
699
+ "objective": 470.2117004394531,
700
+ "ranking_simple": 0.6000000238418579,
701
+ "regularize": 0.4310372471809387,
702
+ "step": 240
703
+ },
704
+ {
705
+ "dpo_loss": 0.612980306148529,
706
+ "epoch": 0.6943788379782712,
707
+ "grad_norm": 12037.283885838251,
708
+ "learning_rate": 4.740503357963676e-06,
709
+ "logits": -0.8841701745986938,
710
+ "logps": -77.16488647460938,
711
+ "loss": 431.5274,
712
+ "objective": 442.6828308105469,
713
+ "ranking_simple": 0.5916666388511658,
714
+ "regularize": 0.4170995056629181,
715
+ "step": 245
716
+ },
717
+ {
718
+ "dpo_loss": 0.5597648024559021,
719
+ "epoch": 0.7085498346717053,
720
+ "grad_norm": 10567.387466711676,
721
+ "learning_rate": 4.721858981942284e-06,
722
+ "logits": -0.8271477818489075,
723
+ "logps": -75.92268371582031,
724
+ "loss": 445.1676,
725
+ "objective": 432.1976013183594,
726
+ "ranking_simple": 0.6083333492279053,
727
+ "regularize": 0.39139440655708313,
728
+ "step": 250
729
+ },
730
+ {
731
+ "epoch": 0.7085498346717053,
732
+ "eval_dpo_loss": 0.7070175409317017,
733
+ "eval_logits": -1.0391225814819336,
734
+ "eval_logps": -80.71741485595703,
735
+ "eval_loss": 572.16943359375,
736
+ "eval_objective": 563.6924438476562,
737
+ "eval_ranking_simple": 0.5408902764320374,
738
+ "eval_regularize": 0.5829753279685974,
739
+ "eval_runtime": 368.3728,
740
+ "eval_samples_per_second": 15.718,
741
+ "eval_steps_per_second": 1.311,
742
+ "step": 250
743
+ },
744
+ {
745
+ "dpo_loss": 0.5912286639213562,
746
+ "epoch": 0.7227208313651393,
747
+ "grad_norm": 10004.706428299638,
748
+ "learning_rate": 4.702607171860354e-06,
749
+ "logits": -1.0291627645492554,
750
+ "logps": -75.54689025878906,
751
+ "loss": 430.9432,
752
+ "objective": 432.97802734375,
753
+ "ranking_simple": 0.5249999761581421,
754
+ "regularize": 0.4280008375644684,
755
+ "step": 255
756
+ },
757
+ {
758
+ "dpo_loss": 0.5518169403076172,
759
+ "epoch": 0.7368918280585735,
760
+ "grad_norm": 10338.762017019879,
761
+ "learning_rate": 4.682753190970533e-06,
762
+ "logits": -0.9816469550132751,
763
+ "logps": -78.38555145263672,
764
+ "loss": 441.5456,
765
+ "objective": 472.0696716308594,
766
+ "ranking_simple": 0.5791666507720947,
767
+ "regularize": 0.4596433639526367,
768
+ "step": 260
769
+ },
770
+ {
771
+ "dpo_loss": 0.5279027819633484,
772
+ "epoch": 0.7510628247520076,
773
+ "grad_norm": 9975.569885392504,
774
+ "learning_rate": 4.6623024671529555e-06,
775
+ "logits": -0.8709200024604797,
776
+ "logps": -78.86503601074219,
777
+ "loss": 415.3095,
778
+ "objective": 424.6243896484375,
779
+ "ranking_simple": 0.637499988079071,
780
+ "regularize": 0.4246416687965393,
781
+ "step": 265
782
+ },
783
+ {
784
+ "dpo_loss": 0.5645706057548523,
785
+ "epoch": 0.7652338214454416,
786
+ "grad_norm": 10607.895202202406,
787
+ "learning_rate": 4.641260591431315e-06,
788
+ "logits": -0.8247819542884827,
789
+ "logps": -78.76771545410156,
790
+ "loss": 434.6699,
791
+ "objective": 440.14715576171875,
792
+ "ranking_simple": 0.5333333611488342,
793
+ "regularize": 0.42400452494621277,
794
+ "step": 270
795
+ },
796
+ {
797
+ "dpo_loss": 0.587860107421875,
798
+ "epoch": 0.7794048181388757,
799
+ "grad_norm": 9977.49094753435,
800
+ "learning_rate": 4.61963331644433e-06,
801
+ "logits": -0.6680871844291687,
802
+ "logps": -77.70354461669922,
803
+ "loss": 438.6425,
804
+ "objective": 448.8917236328125,
805
+ "ranking_simple": 0.5833333134651184,
806
+ "regularize": 0.46749448776245117,
807
+ "step": 275
808
+ },
809
+ {
810
+ "dpo_loss": 0.5520439147949219,
811
+ "epoch": 0.7935758148323099,
812
+ "grad_norm": 11654.173000851612,
813
+ "learning_rate": 4.597426554873037e-06,
814
+ "logits": -0.7018941640853882,
815
+ "logps": -78.6235122680664,
816
+ "loss": 443.7228,
817
+ "objective": 438.81072998046875,
818
+ "ranking_simple": 0.5791666507720947,
819
+ "regularize": 0.4298190176486969,
820
+ "step": 280
821
+ },
822
+ {
823
+ "dpo_loss": 0.5620520710945129,
824
+ "epoch": 0.807746811525744,
825
+ "grad_norm": 11208.671816444528,
826
+ "learning_rate": 4.574646377824316e-06,
827
+ "logits": -0.7278221845626831,
828
+ "logps": -79.00438690185547,
829
+ "loss": 428.4479,
830
+ "objective": 430.4336242675781,
831
+ "ranking_simple": 0.5375000238418579,
832
+ "regularize": 0.413059800863266,
833
+ "step": 285
834
+ },
835
+ {
836
+ "dpo_loss": 0.5554956793785095,
837
+ "epoch": 0.821917808219178,
838
+ "grad_norm": 10809.870000518249,
839
+ "learning_rate": 4.551299013171111e-06,
840
+ "logits": -0.7191876769065857,
841
+ "logps": -77.41136169433594,
842
+ "loss": 399.2666,
843
+ "objective": 398.6150817871094,
844
+ "ranking_simple": 0.5791666507720947,
845
+ "regularize": 0.3730964958667755,
846
+ "step": 290
847
+ },
848
+ {
849
+ "dpo_loss": 0.5479804873466492,
850
+ "epoch": 0.8360888049126122,
851
+ "grad_norm": 10163.479021180265,
852
+ "learning_rate": 4.5273908438498e-06,
853
+ "logits": -0.7566318511962891,
854
+ "logps": -78.29537963867188,
855
+ "loss": 420.4245,
856
+ "objective": 438.2878112792969,
857
+ "ranking_simple": 0.5541666746139526,
858
+ "regularize": 0.43994590640068054,
859
+ "step": 295
860
+ },
861
+ {
862
+ "dpo_loss": 0.564975917339325,
863
+ "epoch": 0.8502598016060463,
864
+ "grad_norm": 10246.9159310537,
865
+ "learning_rate": 4.502928406115152e-06,
866
+ "logits": -0.7081549167633057,
867
+ "logps": -79.06087493896484,
868
+ "loss": 413.9375,
869
+ "objective": 419.59674072265625,
870
+ "ranking_simple": 0.6291666626930237,
871
+ "regularize": 0.3981608748435974,
872
+ "step": 300
873
+ },
874
+ {
875
+ "epoch": 0.8502598016060463,
876
+ "eval_dpo_loss": 0.7031394839286804,
877
+ "eval_logits": -0.7451677918434143,
878
+ "eval_logps": -84.88603210449219,
879
+ "eval_loss": 567.0264282226562,
880
+ "eval_objective": 558.1202392578125,
881
+ "eval_ranking_simple": 0.5398550629615784,
882
+ "eval_regularize": 0.573249101638794,
883
+ "eval_runtime": 374.6781,
884
+ "eval_samples_per_second": 15.453,
885
+ "eval_steps_per_second": 1.289,
886
+ "step": 300
887
+ },
888
+ {
889
+ "dpo_loss": 0.557949423789978,
890
+ "epoch": 0.8644307982994804,
891
+ "grad_norm": 10481.083205902773,
892
+ "learning_rate": 4.477918387753388e-06,
893
+ "logits": -0.6799180507659912,
894
+ "logps": -79.46160125732422,
895
+ "loss": 416.0437,
896
+ "objective": 436.0469970703125,
897
+ "ranking_simple": 0.625,
898
+ "regularize": 0.3994833528995514,
899
+ "step": 305
900
+ },
901
+ {
902
+ "dpo_loss": 0.537316083908081,
903
+ "epoch": 0.8786017949929145,
904
+ "grad_norm": 10330.127581982748,
905
+ "learning_rate": 4.452367626253805e-06,
906
+ "logits": -0.6589821577072144,
907
+ "logps": -77.94523620605469,
908
+ "loss": 412.2728,
909
+ "objective": 435.38397216796875,
910
+ "ranking_simple": 0.5958333611488342,
911
+ "regularize": 0.42988669872283936,
912
+ "step": 310
913
+ },
914
+ {
915
+ "dpo_loss": 0.5410233736038208,
916
+ "epoch": 0.8927727916863486,
917
+ "grad_norm": 9680.592388360066,
918
+ "learning_rate": 4.426283106939474e-06,
919
+ "logits": -0.5960977077484131,
920
+ "logps": -77.44992065429688,
921
+ "loss": 407.9562,
922
+ "objective": 391.1397705078125,
923
+ "ranking_simple": 0.5416666865348816,
924
+ "regularize": 0.3800966143608093,
925
+ "step": 315
926
+ },
927
+ {
928
+ "dpo_loss": 0.5544202923774719,
929
+ "epoch": 0.9069437883797827,
930
+ "grad_norm": 10390.175171602317,
931
+ "learning_rate": 4.399671961057523e-06,
932
+ "logits": -0.4942823052406311,
933
+ "logps": -79.33159637451172,
934
+ "loss": 418.3425,
935
+ "objective": 413.7690124511719,
936
+ "ranking_simple": 0.5874999761581421,
937
+ "regularize": 0.3666258752346039,
938
+ "step": 320
939
+ },
940
+ {
941
+ "dpo_loss": 0.5637380480766296,
942
+ "epoch": 0.9211147850732169,
943
+ "grad_norm": 9093.12442680975,
944
+ "learning_rate": 4.372541463829524e-06,
945
+ "logits": -0.5750992298126221,
946
+ "logps": -79.70758819580078,
947
+ "loss": 400.4521,
948
+ "objective": 385.9770202636719,
949
+ "ranking_simple": 0.5874999761581421,
950
+ "regularize": 0.3547678589820862,
951
+ "step": 325
952
+ },
953
+ {
954
+ "dpo_loss": 0.527768075466156,
955
+ "epoch": 0.9352857817666509,
956
+ "grad_norm": 9514.787690422123,
957
+ "learning_rate": 4.3448990324625244e-06,
958
+ "logits": -0.5246156454086304,
959
+ "logps": -78.73555755615234,
960
+ "loss": 404.0078,
961
+ "objective": 408.9071960449219,
962
+ "ranking_simple": 0.6166666746139526,
963
+ "regularize": 0.38706058263778687,
964
+ "step": 330
965
+ },
966
+ {
967
+ "dpo_loss": 0.5262110829353333,
968
+ "epoch": 0.949456778460085,
969
+ "grad_norm": 8108.6927584156865,
970
+ "learning_rate": 4.316752224121252e-06,
971
+ "logits": -0.5318282246589661,
972
+ "logps": -78.84705352783203,
973
+ "loss": 395.0854,
974
+ "objective": 391.35888671875,
975
+ "ranking_simple": 0.625,
976
+ "regularize": 0.3775762617588043,
977
+ "step": 335
978
+ },
979
+ {
980
+ "dpo_loss": 0.5850500464439392,
981
+ "epoch": 0.9636277751535192,
982
+ "grad_norm": 9909.017294058862,
983
+ "learning_rate": 4.288108733862064e-06,
984
+ "logits": -0.5778205394744873,
985
+ "logps": -78.09910583496094,
986
+ "loss": 425.4681,
987
+ "objective": 410.2987365722656,
988
+ "ranking_simple": 0.5958333611488342,
989
+ "regularize": 0.3491324782371521,
990
+ "step": 340
991
+ },
992
+ {
993
+ "dpo_loss": 0.5264328718185425,
994
+ "epoch": 0.9777987718469532,
995
+ "grad_norm": 10518.367912162912,
996
+ "learning_rate": 4.2589763925291924e-06,
997
+ "logits": -0.6430075168609619,
998
+ "logps": -76.8382797241211,
999
+ "loss": 386.2568,
1000
+ "objective": 371.39697265625,
1001
+ "ranking_simple": 0.6541666388511658,
1002
+ "regularize": 0.3261619508266449,
1003
+ "step": 345
1004
+ },
1005
+ {
1006
+ "dpo_loss": 0.5298411846160889,
1007
+ "epoch": 0.9919697685403873,
1008
+ "grad_norm": 8822.17526005844,
1009
+ "learning_rate": 4.229363164613874e-06,
1010
+ "logits": -0.5401391983032227,
1011
+ "logps": -78.81169128417969,
1012
+ "loss": 385.7652,
1013
+ "objective": 396.1875305175781,
1014
+ "ranking_simple": 0.6583333611488342,
1015
+ "regularize": 0.4019112288951874,
1016
+ "step": 350
1017
+ },
1018
+ {
1019
+ "epoch": 0.9919697685403873,
1020
+ "eval_dpo_loss": 0.7081769108772278,
1021
+ "eval_logits": -0.6075623035430908,
1022
+ "eval_logps": -82.63887786865234,
1023
+ "eval_loss": 581.0134887695312,
1024
+ "eval_objective": 565.1652221679688,
1025
+ "eval_ranking_simple": 0.5383023023605347,
1026
+ "eval_regularize": 0.5905848741531372,
1027
+ "eval_runtime": 368.0626,
1028
+ "eval_samples_per_second": 15.731,
1029
+ "eval_steps_per_second": 1.312,
1030
+ "step": 350
1031
+ },
1032
+ {
1033
+ "dpo_loss": 0.5235918760299683,
1034
+ "epoch": 1.0061407652338215,
1035
+ "grad_norm": 7985.929619871145,
1036
+ "learning_rate": 4.199277146076933e-06,
1037
+ "logits": -0.5844969153404236,
1038
+ "logps": -77.1658935546875,
1039
+ "loss": 395.3726,
1040
+ "objective": 383.5928955078125,
1041
+ "ranking_simple": 0.6208333373069763,
1042
+ "regularize": 0.3756142854690552,
1043
+ "step": 355
1044
+ },
1045
+ {
1046
+ "dpo_loss": 0.5242210030555725,
1047
+ "epoch": 1.0203117619272555,
1048
+ "grad_norm": 9633.18732259904,
1049
+ "learning_rate": 4.168726562135432e-06,
1050
+ "logits": -0.5656494498252869,
1051
+ "logps": -78.61763000488281,
1052
+ "loss": 382.5271,
1053
+ "objective": 390.3468322753906,
1054
+ "ranking_simple": 0.5958333611488342,
1055
+ "regularize": 0.3490845561027527,
1056
+ "step": 360
1057
+ },
1058
+ {
1059
+ "dpo_loss": 0.5211578011512756,
1060
+ "epoch": 1.0344827586206897,
1061
+ "grad_norm": 9644.11378605678,
1062
+ "learning_rate": 4.137719765013974e-06,
1063
+ "logits": -0.5837284922599792,
1064
+ "logps": -78.80403900146484,
1065
+ "loss": 393.9499,
1066
+ "objective": 399.89215087890625,
1067
+ "ranking_simple": 0.5833333134651184,
1068
+ "regularize": 0.3770846724510193,
1069
+ "step": 365
1070
+ },
1071
+ {
1072
+ "dpo_loss": 0.5336027145385742,
1073
+ "epoch": 1.0486537553141237,
1074
+ "grad_norm": 8251.35503586343,
1075
+ "learning_rate": 4.106265231661292e-06,
1076
+ "logits": -0.5281592607498169,
1077
+ "logps": -76.34611511230469,
1078
+ "loss": 364.8078,
1079
+ "objective": 364.66595458984375,
1080
+ "ranking_simple": 0.5916666388511658,
1081
+ "regularize": 0.337600976228714,
1082
+ "step": 370
1083
+ },
1084
+ {
1085
+ "dpo_loss": 0.5174949765205383,
1086
+ "epoch": 1.0628247520075578,
1087
+ "grad_norm": 9028.93002381266,
1088
+ "learning_rate": 4.074371561432731e-06,
1089
+ "logits": -0.5678179860115051,
1090
+ "logps": -76.8271255493164,
1091
+ "loss": 372.633,
1092
+ "objective": 343.89019775390625,
1093
+ "ranking_simple": 0.612500011920929,
1094
+ "regularize": 0.30460023880004883,
1095
+ "step": 375
1096
+ },
1097
+ {
1098
+ "dpo_loss": 0.5492153167724609,
1099
+ "epoch": 1.076995748700992,
1100
+ "grad_norm": 8789.25650294624,
1101
+ "learning_rate": 4.042047473739278e-06,
1102
+ "logits": -0.5127583146095276,
1103
+ "logps": -78.71736907958984,
1104
+ "loss": 373.876,
1105
+ "objective": 406.1351013183594,
1106
+ "ranking_simple": 0.6291666626930237,
1107
+ "regularize": 0.36652448773384094,
1108
+ "step": 380
1109
+ },
1110
+ {
1111
+ "dpo_loss": 0.5343239307403564,
1112
+ "epoch": 1.091166745394426,
1113
+ "grad_norm": 8695.398130642287,
1114
+ "learning_rate": 4.009301805663752e-06,
1115
+ "logits": -0.439236581325531,
1116
+ "logps": -77.64140319824219,
1117
+ "loss": 381.8216,
1118
+ "objective": 398.8446960449219,
1119
+ "ranking_simple": 0.6000000238418579,
1120
+ "regularize": 0.3279392123222351,
1121
+ "step": 385
1122
+ },
1123
+ {
1124
+ "dpo_loss": 0.5177706480026245,
1125
+ "epoch": 1.10533774208786,
1126
+ "grad_norm": 9041.130741001136,
1127
+ "learning_rate": 3.976143509544843e-06,
1128
+ "logits": -0.4288846254348755,
1129
+ "logps": -79.35343933105469,
1130
+ "loss": 357.1327,
1131
+ "objective": 361.1878967285156,
1132
+ "ranking_simple": 0.6416666507720947,
1133
+ "regularize": 0.3373713791370392,
1134
+ "step": 390
1135
+ },
1136
+ {
1137
+ "dpo_loss": 0.5048284530639648,
1138
+ "epoch": 1.1195087387812943,
1139
+ "grad_norm": 8902.672627868116,
1140
+ "learning_rate": 3.9425816505296254e-06,
1141
+ "logits": -0.48021775484085083,
1142
+ "logps": -78.194091796875,
1143
+ "loss": 394.8973,
1144
+ "objective": 412.3692932128906,
1145
+ "ranking_simple": 0.6000000238418579,
1146
+ "regularize": 0.3603326082229614,
1147
+ "step": 395
1148
+ },
1149
+ {
1150
+ "dpo_loss": 0.5264464020729065,
1151
+ "epoch": 1.1336797354747283,
1152
+ "grad_norm": 8995.938641641998,
1153
+ "learning_rate": 3.908625404095242e-06,
1154
+ "logits": -0.4987303912639618,
1155
+ "logps": -77.11076354980469,
1156
+ "loss": 376.3251,
1157
+ "objective": 368.0682373046875,
1158
+ "ranking_simple": 0.625,
1159
+ "regularize": 0.31286415457725525,
1160
+ "step": 400
1161
+ },
1162
+ {
1163
+ "epoch": 1.1336797354747283,
1164
+ "eval_dpo_loss": 0.7118101716041565,
1165
+ "eval_logits": -0.5273135900497437,
1166
+ "eval_logps": -81.622314453125,
1167
+ "eval_loss": 586.0215454101562,
1168
+ "eval_objective": 571.4174194335938,
1169
+ "eval_ranking_simple": 0.5367494821548462,
1170
+ "eval_regularize": 0.5996229648590088,
1171
+ "eval_runtime": 367.4788,
1172
+ "eval_samples_per_second": 15.756,
1173
+ "eval_steps_per_second": 1.314,
1174
+ "step": 400
1175
+ },
1176
+ {
1177
+ "dpo_loss": 0.5221129059791565,
1178
+ "epoch": 1.1478507321681626,
1179
+ "grad_norm": 8887.877038015888,
1180
+ "learning_rate": 3.8742840535404155e-06,
1181
+ "logits": -0.4772927165031433,
1182
+ "logps": -78.21456909179688,
1183
+ "loss": 374.1795,
1184
+ "objective": 403.9602355957031,
1185
+ "ranking_simple": 0.5791666507720947,
1186
+ "regularize": 0.35953524708747864,
1187
+ "step": 405
1188
+ },
1189
+ {
1190
+ "dpo_loss": 0.5320737361907959,
1191
+ "epoch": 1.1620217288615966,
1192
+ "grad_norm": 8653.742076980236,
1193
+ "learning_rate": 3.839566987447492e-06,
1194
+ "logits": -0.5261733531951904,
1195
+ "logps": -77.4281005859375,
1196
+ "loss": 357.8752,
1197
+ "objective": 369.8106689453125,
1198
+ "ranking_simple": 0.6041666865348816,
1199
+ "regularize": 0.31248220801353455,
1200
+ "step": 410
1201
+ },
1202
+ {
1203
+ "dpo_loss": 0.5130491852760315,
1204
+ "epoch": 1.1761927255550306,
1205
+ "grad_norm": 8220.944462330583,
1206
+ "learning_rate": 3.8044836971156935e-06,
1207
+ "logits": -0.4498496651649475,
1208
+ "logps": -76.78229522705078,
1209
+ "loss": 367.3453,
1210
+ "objective": 342.2223815917969,
1211
+ "ranking_simple": 0.6166666746139526,
1212
+ "regularize": 0.32919442653656006,
1213
+ "step": 415
1214
+ },
1215
+ {
1216
+ "dpo_loss": 0.5148084163665771,
1217
+ "epoch": 1.1903637222484649,
1218
+ "grad_norm": 8364.696648518799,
1219
+ "learning_rate": 3.7690437739662928e-06,
1220
+ "logits": -0.4175103008747101,
1221
+ "logps": -78.17424011230469,
1222
+ "loss": 381.6126,
1223
+ "objective": 368.6730041503906,
1224
+ "ranking_simple": 0.5874999761581421,
1225
+ "regularize": 0.34451645612716675,
1226
+ "step": 420
1227
+ },
1228
+ {
1229
+ "dpo_loss": 0.5163763761520386,
1230
+ "epoch": 1.204534718941899,
1231
+ "grad_norm": 8541.974606101103,
1232
+ "learning_rate": 3.7332569069204127e-06,
1233
+ "logits": -0.45798221230506897,
1234
+ "logps": -76.5196304321289,
1235
+ "loss": 370.1389,
1236
+ "objective": 370.2024230957031,
1237
+ "ranking_simple": 0.6166666746139526,
1238
+ "regularize": 0.3256681561470032,
1239
+ "step": 425
1240
+ },
1241
+ {
1242
+ "dpo_loss": 0.5450712442398071,
1243
+ "epoch": 1.2187057156353331,
1244
+ "grad_norm": 9487.349462942979,
1245
+ "learning_rate": 3.697132879750174e-06,
1246
+ "logits": -0.48889076709747314,
1247
+ "logps": -78.89514923095703,
1248
+ "loss": 384.2139,
1249
+ "objective": 399.7441101074219,
1250
+ "ranking_simple": 0.612500011920929,
1251
+ "regularize": 0.3587479293346405,
1252
+ "step": 430
1253
+ },
1254
+ {
1255
+ "dpo_loss": 0.4979787766933441,
1256
+ "epoch": 1.2328767123287672,
1257
+ "grad_norm": 9248.918944555557,
1258
+ "learning_rate": 3.6606815684039098e-06,
1259
+ "logits": -0.45076984167099,
1260
+ "logps": -78.12223815917969,
1261
+ "loss": 376.6027,
1262
+ "objective": 365.3465881347656,
1263
+ "ranking_simple": 0.6708333492279053,
1264
+ "regularize": 0.33380749821662903,
1265
+ "step": 435
1266
+ },
1267
+ {
1268
+ "dpo_loss": 0.5080611109733582,
1269
+ "epoch": 1.2470477090222012,
1270
+ "grad_norm": 8952.956886414895,
1271
+ "learning_rate": 3.6239129383061764e-06,
1272
+ "logits": -0.5257605314254761,
1273
+ "logps": -77.80259704589844,
1274
+ "loss": 367.3115,
1275
+ "objective": 355.6669921875,
1276
+ "ranking_simple": 0.6458333134651184,
1277
+ "regularize": 0.3250352144241333,
1278
+ "step": 440
1279
+ },
1280
+ {
1281
+ "dpo_loss": 0.5034739375114441,
1282
+ "epoch": 1.2612187057156352,
1283
+ "grad_norm": 8595.813908218257,
1284
+ "learning_rate": 3.586837041633312e-06,
1285
+ "logits": -0.5648617148399353,
1286
+ "logps": -78.69916534423828,
1287
+ "loss": 353.7886,
1288
+ "objective": 345.4654235839844,
1289
+ "ranking_simple": 0.6625000238418579,
1290
+ "regularize": 0.3045599162578583,
1291
+ "step": 445
1292
+ },
1293
+ {
1294
+ "dpo_loss": 0.5136024951934814,
1295
+ "epoch": 1.2753897024090695,
1296
+ "grad_norm": 8799.318521509487,
1297
+ "learning_rate": 3.5494640145652647e-06,
1298
+ "logits": -0.6497453451156616,
1299
+ "logps": -77.91305541992188,
1300
+ "loss": 348.4717,
1301
+ "objective": 341.05877685546875,
1302
+ "ranking_simple": 0.5833333134651184,
1303
+ "regularize": 0.2908380925655365,
1304
+ "step": 450
1305
+ },
1306
+ {
1307
+ "epoch": 1.2753897024090695,
1308
+ "eval_dpo_loss": 0.7055455446243286,
1309
+ "eval_logits": -0.6517468094825745,
1310
+ "eval_logps": -81.88983154296875,
1311
+ "eval_loss": 576.5939331054688,
1312
+ "eval_objective": 563.9976806640625,
1313
+ "eval_ranking_simple": 0.5372670888900757,
1314
+ "eval_regularize": 0.586566150188446,
1315
+ "eval_runtime": 372.0863,
1316
+ "eval_samples_per_second": 15.561,
1317
+ "eval_steps_per_second": 1.298,
1318
+ "step": 450
1319
+ },
1320
+ {
1321
+ "dpo_loss": 0.5133717656135559,
1322
+ "epoch": 1.2895606991025035,
1323
+ "grad_norm": 8447.712740682113,
1324
+ "learning_rate": 3.511804074514468e-06,
1325
+ "logits": -0.50225830078125,
1326
+ "logps": -78.78018188476562,
1327
+ "loss": 350.7452,
1328
+ "objective": 353.5893249511719,
1329
+ "ranking_simple": 0.5583333373069763,
1330
+ "regularize": 0.30045947432518005,
1331
+ "step": 455
1332
+ },
1333
+ {
1334
+ "dpo_loss": 0.5129916071891785,
1335
+ "epoch": 1.3037316957959377,
1336
+ "grad_norm": 8683.9771262322,
1337
+ "learning_rate": 3.4738675173325008e-06,
1338
+ "logits": -0.5019214153289795,
1339
+ "logps": -78.0320816040039,
1340
+ "loss": 358.7841,
1341
+ "objective": 359.1448669433594,
1342
+ "ranking_simple": 0.5583333373069763,
1343
+ "regularize": 0.28918400406837463,
1344
+ "step": 460
1345
+ },
1346
+ {
1347
+ "dpo_loss": 0.5200445055961609,
1348
+ "epoch": 1.3179026924893718,
1349
+ "grad_norm": 8708.589588634271,
1350
+ "learning_rate": 3.435664714495301e-06,
1351
+ "logits": -0.4990668296813965,
1352
+ "logps": -76.45315551757812,
1353
+ "loss": 363.9146,
1354
+ "objective": 359.3269348144531,
1355
+ "ranking_simple": 0.5625,
1356
+ "regularize": 0.317513108253479,
1357
+ "step": 465
1358
+ },
1359
+ {
1360
+ "dpo_loss": 0.5191565752029419,
1361
+ "epoch": 1.3320736891828058,
1362
+ "grad_norm": 7892.018349995868,
1363
+ "learning_rate": 3.397206110267713e-06,
1364
+ "logits": -0.5707500576972961,
1365
+ "logps": -75.51515197753906,
1366
+ "loss": 369.7767,
1367
+ "objective": 370.79046630859375,
1368
+ "ranking_simple": 0.6083333492279053,
1369
+ "regularize": 0.30417945981025696,
1370
+ "step": 470
1371
+ },
1372
+ {
1373
+ "dpo_loss": 0.5160278081893921,
1374
+ "epoch": 1.34624468587624,
1375
+ "grad_norm": 8622.437524812693,
1376
+ "learning_rate": 3.3585022188481247e-06,
1377
+ "logits": -0.5167524814605713,
1378
+ "logps": -77.0745849609375,
1379
+ "loss": 347.9223,
1380
+ "objective": 363.10107421875,
1381
+ "ranking_simple": 0.6583333611488342,
1382
+ "regularize": 0.3366948366165161,
1383
+ "step": 475
1384
+ },
1385
+ {
1386
+ "dpo_loss": 0.5038079023361206,
1387
+ "epoch": 1.360415682569674,
1388
+ "grad_norm": 8520.652876751588,
1389
+ "learning_rate": 3.3195636214939943e-06,
1390
+ "logits": -0.5720607042312622,
1391
+ "logps": -76.88468170166016,
1392
+ "loss": 352.4882,
1393
+ "objective": 360.6014709472656,
1394
+ "ranking_simple": 0.5833333134651184,
1395
+ "regularize": 0.31744828820228577,
1396
+ "step": 480
1397
+ },
1398
+ {
1399
+ "dpo_loss": 0.5260137915611267,
1400
+ "epoch": 1.3745866792631083,
1401
+ "grad_norm": 8357.649136802573,
1402
+ "learning_rate": 3.2804009636290403e-06,
1403
+ "logits": -0.5760036706924438,
1404
+ "logps": -75.65044403076172,
1405
+ "loss": 352.4908,
1406
+ "objective": 335.92327880859375,
1407
+ "ranking_simple": 0.5874999761581421,
1408
+ "regularize": 0.27229636907577515,
1409
+ "step": 485
1410
+ },
1411
+ {
1412
+ "dpo_loss": 0.5316032767295837,
1413
+ "epoch": 1.3887576759565423,
1414
+ "grad_norm": 8377.120289104301,
1415
+ "learning_rate": 3.2410249519328848e-06,
1416
+ "logits": -0.5220092535018921,
1417
+ "logps": -78.16001892089844,
1418
+ "loss": 350.6266,
1419
+ "objective": 356.1325988769531,
1420
+ "ranking_simple": 0.5916666388511658,
1421
+ "regularize": 0.2998295724391937,
1422
+ "step": 490
1423
+ },
1424
+ {
1425
+ "dpo_loss": 0.5069059729576111,
1426
+ "epoch": 1.4029286726499763,
1427
+ "grad_norm": 9790.99603957699,
1428
+ "learning_rate": 3.201446351413958e-06,
1429
+ "logits": -0.5315040349960327,
1430
+ "logps": -78.9278793334961,
1431
+ "loss": 355.1208,
1432
+ "objective": 340.3614196777344,
1433
+ "ranking_simple": 0.5916666388511658,
1434
+ "regularize": 0.2911123037338257,
1435
+ "step": 495
1436
+ },
1437
+ {
1438
+ "dpo_loss": 0.5013086199760437,
1439
+ "epoch": 1.4170996693434104,
1440
+ "grad_norm": 9096.333510047558,
1441
+ "learning_rate": 3.1616759824664543e-06,
1442
+ "logits": -0.47489532828330994,
1443
+ "logps": -78.39351654052734,
1444
+ "loss": 351.4185,
1445
+ "objective": 349.8325500488281,
1446
+ "ranking_simple": 0.6083333492279053,
1447
+ "regularize": 0.2930639684200287,
1448
+ "step": 500
1449
+ },
1450
+ {
1451
+ "epoch": 1.4170996693434104,
1452
+ "eval_dpo_loss": 0.7128105759620667,
1453
+ "eval_logits": -0.5594385862350464,
1454
+ "eval_logps": -82.85633087158203,
1455
+ "eval_loss": 584.3820190429688,
1456
+ "eval_objective": 570.8920288085938,
1457
+ "eval_ranking_simple": 0.5393374562263489,
1458
+ "eval_regularize": 0.597196638584137,
1459
+ "eval_runtime": 367.2146,
1460
+ "eval_samples_per_second": 15.767,
1461
+ "eval_steps_per_second": 1.315,
1462
+ "step": 500
1463
+ },
1464
+ {
1465
+ "dpo_loss": 0.502229630947113,
1466
+ "epoch": 1.4312706660368446,
1467
+ "grad_norm": 7569.16735798095,
1468
+ "learning_rate": 3.121724717912138e-06,
1469
+ "logits": -0.4940933585166931,
1470
+ "logps": -79.02289581298828,
1471
+ "loss": 350.6974,
1472
+ "objective": 343.7030944824219,
1473
+ "ranking_simple": 0.6291666626930237,
1474
+ "regularize": 0.2822100818157196,
1475
+ "step": 505
1476
+ },
1477
+ {
1478
+ "dpo_loss": 0.5396325588226318,
1479
+ "epoch": 1.4454416627302786,
1480
+ "grad_norm": 8970.009407427333,
1481
+ "learning_rate": 3.081603480027826e-06,
1482
+ "logits": -0.5547798275947571,
1483
+ "logps": -79.63969421386719,
1484
+ "loss": 354.8519,
1485
+ "objective": 349.9094543457031,
1486
+ "ranking_simple": 0.6000000238418579,
1487
+ "regularize": 0.27683576941490173,
1488
+ "step": 510
1489
+ },
1490
+ {
1491
+ "dpo_loss": 0.5251208543777466,
1492
+ "epoch": 1.4596126594237129,
1493
+ "grad_norm": 8332.379052556926,
1494
+ "learning_rate": 3.04132323755935e-06,
1495
+ "logits": -0.627532422542572,
1496
+ "logps": -79.98551940917969,
1497
+ "loss": 345.0198,
1498
+ "objective": 359.2087707519531,
1499
+ "ranking_simple": 0.5416666865348816,
1500
+ "regularize": 0.2865109145641327,
1501
+ "step": 515
1502
+ },
1503
+ {
1504
+ "dpo_loss": 0.5008211135864258,
1505
+ "epoch": 1.473783656117147,
1506
+ "grad_norm": 8229.897145458208,
1507
+ "learning_rate": 3.0008950027228035e-06,
1508
+ "logits": -0.5891799330711365,
1509
+ "logps": -76.67542266845703,
1510
+ "loss": 342.313,
1511
+ "objective": 343.9622802734375,
1512
+ "ranking_simple": 0.5916666388511658,
1513
+ "regularize": 0.27687618136405945,
1514
+ "step": 520
1515
+ },
1516
+ {
1517
+ "dpo_loss": 0.4947444498538971,
1518
+ "epoch": 1.487954652810581,
1519
+ "grad_norm": 7702.6950001891755,
1520
+ "learning_rate": 2.960329828193918e-06,
1521
+ "logits": -0.5145970582962036,
1522
+ "logps": -75.67664337158203,
1523
+ "loss": 332.768,
1524
+ "objective": 329.7011413574219,
1525
+ "ranking_simple": 0.6166666746139526,
1526
+ "regularize": 0.2696382403373718,
1527
+ "step": 525
1528
+ },
1529
+ {
1530
+ "dpo_loss": 0.5160828828811646,
1531
+ "epoch": 1.5021256495040152,
1532
+ "grad_norm": 8386.762596966919,
1533
+ "learning_rate": 2.9196388040863695e-06,
1534
+ "logits": -0.616746187210083,
1535
+ "logps": -78.44214630126953,
1536
+ "loss": 362.1123,
1537
+ "objective": 358.41510009765625,
1538
+ "ranking_simple": 0.5791666507720947,
1539
+ "regularize": 0.29886895418167114,
1540
+ "step": 530
1541
+ },
1542
+ {
1543
+ "dpo_loss": 0.495453417301178,
1544
+ "epoch": 1.5162966461974492,
1545
+ "grad_norm": 7823.009103949036,
1546
+ "learning_rate": 2.8788330549198512e-06,
1547
+ "logits": -0.6062889099121094,
1548
+ "logps": -77.4250717163086,
1549
+ "loss": 331.5395,
1550
+ "objective": 311.8442687988281,
1551
+ "ranking_simple": 0.6000000238418579,
1552
+ "regularize": 0.24911071360111237,
1553
+ "step": 535
1554
+ },
1555
+ {
1556
+ "dpo_loss": 0.5264947414398193,
1557
+ "epoch": 1.5304676428908834,
1558
+ "grad_norm": 8960.269885367992,
1559
+ "learning_rate": 2.8379237365787426e-06,
1560
+ "logits": -0.5448920130729675,
1561
+ "logps": -77.27252960205078,
1562
+ "loss": 334.277,
1563
+ "objective": 340.4452209472656,
1564
+ "ranking_simple": 0.6208333373069763,
1565
+ "regularize": 0.2655259668827057,
1566
+ "step": 540
1567
+ },
1568
+ {
1569
+ "dpo_loss": 0.5327333211898804,
1570
+ "epoch": 1.5446386395843175,
1571
+ "grad_norm": 8003.393273489204,
1572
+ "learning_rate": 2.7969220332622004e-06,
1573
+ "logits": -0.6398530602455139,
1574
+ "logps": -76.85224151611328,
1575
+ "loss": 331.0988,
1576
+ "objective": 332.23486328125,
1577
+ "ranking_simple": 0.6541666388511658,
1578
+ "regularize": 0.2610936164855957,
1579
+ "step": 545
1580
+ },
1581
+ {
1582
+ "dpo_loss": 0.5121258497238159,
1583
+ "epoch": 1.5588096362777515,
1584
+ "grad_norm": 7510.163531162746,
1585
+ "learning_rate": 2.7558391544265127e-06,
1586
+ "logits": -0.6813774704933167,
1587
+ "logps": -76.5182876586914,
1588
+ "loss": 326.458,
1589
+ "objective": 333.58294677734375,
1590
+ "ranking_simple": 0.5666666626930237,
1591
+ "regularize": 0.27006784081459045,
1592
+ "step": 550
1593
+ },
1594
+ {
1595
+ "epoch": 1.5588096362777515,
1596
+ "eval_dpo_loss": 0.7086328268051147,
1597
+ "eval_logits": -0.6993709802627563,
1598
+ "eval_logps": -80.56141662597656,
1599
+ "eval_loss": 578.3502807617188,
1600
+ "eval_objective": 565.96826171875,
1601
+ "eval_ranking_simple": 0.5367494821548462,
1602
+ "eval_regularize": 0.5877013206481934,
1603
+ "eval_runtime": 367.856,
1604
+ "eval_samples_per_second": 15.74,
1605
+ "eval_steps_per_second": 1.313,
1606
+ "step": 550
1607
+ },
1608
+ {
1609
+ "dpo_loss": 0.5160035490989685,
1610
+ "epoch": 1.5729806329711855,
1611
+ "grad_norm": 8802.921874841602,
1612
+ "learning_rate": 2.714686331720543e-06,
1613
+ "logits": -0.5305084586143494,
1614
+ "logps": -76.84326934814453,
1615
+ "loss": 342.895,
1616
+ "objective": 352.5166320800781,
1617
+ "ranking_simple": 0.612500011920929,
1618
+ "regularize": 0.26788362860679626,
1619
+ "step": 555
1620
+ },
1621
+ {
1622
+ "dpo_loss": 0.5147711634635925,
1623
+ "epoch": 1.5871516296646198,
1624
+ "grad_norm": 7348.156155946241,
1625
+ "learning_rate": 2.6734748159151104e-06,
1626
+ "logits": -0.5309932231903076,
1627
+ "logps": -76.01145935058594,
1628
+ "loss": 327.5288,
1629
+ "objective": 325.1703796386719,
1630
+ "ranking_simple": 0.5958333611488342,
1631
+ "regularize": 0.25214916467666626,
1632
+ "step": 560
1633
+ },
1634
+ {
1635
+ "dpo_loss": 0.5153753161430359,
1636
+ "epoch": 1.601322626358054,
1637
+ "grad_norm": 8906.64379732216,
1638
+ "learning_rate": 2.632215873827142e-06,
1639
+ "logits": -0.4652445912361145,
1640
+ "logps": -76.68761444091797,
1641
+ "loss": 354.5548,
1642
+ "objective": 364.25689697265625,
1643
+ "ranking_simple": 0.5333333611488342,
1644
+ "regularize": 0.28750428557395935,
1645
+ "step": 565
1646
+ },
1647
+ {
1648
+ "dpo_loss": 0.519721508026123,
1649
+ "epoch": 1.615493623051488,
1650
+ "grad_norm": 8653.666075373956,
1651
+ "learning_rate": 2.5909207852394363e-06,
1652
+ "logits": -0.5437235236167908,
1653
+ "logps": -77.414794921875,
1654
+ "loss": 333.0782,
1655
+ "objective": 335.1112365722656,
1656
+ "ranking_simple": 0.6166666746139526,
1657
+ "regularize": 0.25310030579566956,
1658
+ "step": 570
1659
+ },
1660
+ {
1661
+ "dpo_loss": 0.5159035325050354,
1662
+ "epoch": 1.629664619744922,
1663
+ "grad_norm": 9098.50496014823,
1664
+ "learning_rate": 2.5496008398168844e-06,
1665
+ "logits": -0.5154822468757629,
1666
+ "logps": -74.86051177978516,
1667
+ "loss": 338.3886,
1668
+ "objective": 360.5445861816406,
1669
+ "ranking_simple": 0.6333333253860474,
1670
+ "regularize": 0.293546199798584,
1671
+ "step": 575
1672
+ },
1673
+ {
1674
+ "dpo_loss": 0.5085076093673706,
1675
+ "epoch": 1.643835616438356,
1676
+ "grad_norm": 7622.784591036375,
1677
+ "learning_rate": 2.508267334019988e-06,
1678
+ "logits": -0.5285104513168335,
1679
+ "logps": -75.09459686279297,
1680
+ "loss": 326.7957,
1681
+ "objective": 322.0632019042969,
1682
+ "ranking_simple": 0.6166666746139526,
1683
+ "regularize": 0.23723416030406952,
1684
+ "step": 580
1685
+ },
1686
+ {
1687
+ "dpo_loss": 0.5216612219810486,
1688
+ "epoch": 1.6580066131317903,
1689
+ "grad_norm": 8435.901879905807,
1690
+ "learning_rate": 2.46693156801652e-06,
1691
+ "logits": -0.43548285961151123,
1692
+ "logps": -74.49349212646484,
1693
+ "loss": 324.9342,
1694
+ "objective": 314.76617431640625,
1695
+ "ranking_simple": 0.5249999761581421,
1696
+ "regularize": 0.23195335268974304,
1697
+ "step": 585
1698
+ },
1699
+ {
1700
+ "dpo_loss": 0.5121405720710754,
1701
+ "epoch": 1.6721776098252243,
1702
+ "grad_norm": 8829.097116710429,
1703
+ "learning_rate": 2.4256048425921693e-06,
1704
+ "logits": -0.4449107050895691,
1705
+ "logps": -75.21448516845703,
1706
+ "loss": 332.8984,
1707
+ "objective": 336.7486572265625,
1708
+ "ranking_simple": 0.6208333373069763,
1709
+ "regularize": 0.25163254141807556,
1710
+ "step": 590
1711
+ },
1712
+ {
1713
+ "dpo_loss": 0.5095264911651611,
1714
+ "epoch": 1.6863486065186586,
1715
+ "grad_norm": 8981.762264761135,
1716
+ "learning_rate": 2.384298456061023e-06,
1717
+ "logits": -0.4447081387042999,
1718
+ "logps": -75.95592498779297,
1719
+ "loss": 329.941,
1720
+ "objective": 336.85479736328125,
1721
+ "ranking_simple": 0.6083333492279053,
1722
+ "regularize": 0.263884961605072,
1723
+ "step": 595
1724
+ },
1725
+ {
1726
+ "dpo_loss": 0.5105345845222473,
1727
+ "epoch": 1.7005196032120926,
1728
+ "grad_norm": 8799.189125731713,
1729
+ "learning_rate": 2.3430237011767166e-06,
1730
+ "logits": -0.5217716693878174,
1731
+ "logps": -76.75984191894531,
1732
+ "loss": 329.0151,
1733
+ "objective": 342.8132019042969,
1734
+ "ranking_simple": 0.6291666626930237,
1735
+ "regularize": 0.2702232301235199,
1736
+ "step": 600
1737
+ },
1738
+ {
1739
+ "epoch": 1.7005196032120926,
1740
+ "eval_dpo_loss": 0.7084978222846985,
1741
+ "eval_logits": -0.593561053276062,
1742
+ "eval_logps": -80.32785034179688,
1743
+ "eval_loss": 578.38671875,
1744
+ "eval_objective": 566.0594482421875,
1745
+ "eval_ranking_simple": 0.5388198494911194,
1746
+ "eval_regularize": 0.5912774801254272,
1747
+ "eval_runtime": 369.5452,
1748
+ "eval_samples_per_second": 15.668,
1749
+ "eval_steps_per_second": 1.307,
1750
+ "step": 600
1751
+ },
1752
+ {
1753
+ "dpo_loss": 0.5166282653808594,
1754
+ "epoch": 1.7146905999055266,
1755
+ "grad_norm": 8773.738115986773,
1756
+ "learning_rate": 2.30179186204511e-06,
1757
+ "logits": -0.4922519624233246,
1758
+ "logps": -77.39384460449219,
1759
+ "loss": 324.8781,
1760
+ "objective": 324.775390625,
1761
+ "ranking_simple": 0.5666666626930237,
1762
+ "regularize": 0.23106712102890015,
1763
+ "step": 605
1764
+ },
1765
+ {
1766
+ "dpo_loss": 0.5313800573348999,
1767
+ "epoch": 1.7288615965989607,
1768
+ "grad_norm": 9468.956357202407,
1769
+ "learning_rate": 2.2606142110393248e-06,
1770
+ "logits": -0.5464334487915039,
1771
+ "logps": -75.34829711914062,
1772
+ "loss": 329.8079,
1773
+ "objective": 328.7645568847656,
1774
+ "ranking_simple": 0.6291666626930237,
1775
+ "regularize": 0.2314998209476471,
1776
+ "step": 610
1777
+ },
1778
+ {
1779
+ "dpo_loss": 0.5274057984352112,
1780
+ "epoch": 1.743032593292395,
1781
+ "grad_norm": 8788.596472210756,
1782
+ "learning_rate": 2.2195020057179897e-06,
1783
+ "logits": -0.5443993210792542,
1784
+ "logps": -75.97472381591797,
1785
+ "loss": 328.8101,
1786
+ "objective": 338.30584716796875,
1787
+ "ranking_simple": 0.5958333611488342,
1788
+ "regularize": 0.255048006772995,
1789
+ "step": 615
1790
+ },
1791
+ {
1792
+ "dpo_loss": 0.5214657187461853,
1793
+ "epoch": 1.7572035899858292,
1794
+ "grad_norm": 8592.256473559864,
1795
+ "learning_rate": 2.1784664857475356e-06,
1796
+ "logits": -0.5270652174949646,
1797
+ "logps": -76.55162811279297,
1798
+ "loss": 325.7626,
1799
+ "objective": 319.66046142578125,
1800
+ "ranking_simple": 0.625,
1801
+ "regularize": 0.23240961134433746,
1802
+ "step": 620
1803
+ },
1804
+ {
1805
+ "dpo_loss": 0.5136802196502686,
1806
+ "epoch": 1.7713745866792632,
1807
+ "grad_norm": 7718.82478571958,
1808
+ "learning_rate": 2.1375188698293855e-06,
1809
+ "logits": -0.462724506855011,
1810
+ "logps": -76.12433624267578,
1811
+ "loss": 330.2958,
1812
+ "objective": 325.9873046875,
1813
+ "ranking_simple": 0.6208333373069763,
1814
+ "regularize": 0.22590765357017517,
1815
+ "step": 625
1816
+ },
1817
+ {
1818
+ "dpo_loss": 0.536301851272583,
1819
+ "epoch": 1.7855455833726972,
1820
+ "grad_norm": 8761.966130498178,
1821
+ "learning_rate": 2.096670352632873e-06,
1822
+ "logits": -0.5007703304290771,
1823
+ "logps": -77.76058197021484,
1824
+ "loss": 326.3625,
1825
+ "objective": 339.7125549316406,
1826
+ "ranking_simple": 0.5625,
1827
+ "regularize": 0.23891252279281616,
1828
+ "step": 630
1829
+ },
1830
+ {
1831
+ "dpo_loss": 0.517907977104187,
1832
+ "epoch": 1.7997165800661312,
1833
+ "grad_norm": 7178.908644562367,
1834
+ "learning_rate": 2.0559321017347286e-06,
1835
+ "logits": -0.585433840751648,
1836
+ "logps": -76.77262878417969,
1837
+ "loss": 319.7427,
1838
+ "objective": 316.2701110839844,
1839
+ "ranking_simple": 0.637499988079071,
1840
+ "regularize": 0.218379944562912,
1841
+ "step": 635
1842
+ },
1843
+ {
1844
+ "dpo_loss": 0.5069996118545532,
1845
+ "epoch": 1.8138875767595655,
1846
+ "grad_norm": 7586.408561007724,
1847
+ "learning_rate": 2.01531525456598e-06,
1848
+ "logits": -0.5689796805381775,
1849
+ "logps": -77.61341094970703,
1850
+ "loss": 320.7854,
1851
+ "objective": 324.8812255859375,
1852
+ "ranking_simple": 0.5874999761581421,
1853
+ "regularize": 0.2512456178665161,
1854
+ "step": 640
1855
+ },
1856
+ {
1857
+ "dpo_loss": 0.5200368762016296,
1858
+ "epoch": 1.8280585734529995,
1859
+ "grad_norm": 8686.608118182792,
1860
+ "learning_rate": 1.974830915367086e-06,
1861
+ "logits": -0.5587595701217651,
1862
+ "logps": -78.18843078613281,
1863
+ "loss": 330.6056,
1864
+ "objective": 326.5403747558594,
1865
+ "ranking_simple": 0.637499988079071,
1866
+ "regularize": 0.24110235273838043,
1867
+ "step": 645
1868
+ },
1869
+ {
1870
+ "dpo_loss": 0.5109516382217407,
1871
+ "epoch": 1.8422295701464337,
1872
+ "grad_norm": 9365.891667609872,
1873
+ "learning_rate": 1.93449015215215e-06,
1874
+ "logits": -0.5691719055175781,
1875
+ "logps": -78.31002807617188,
1876
+ "loss": 333.5158,
1877
+ "objective": 333.96240234375,
1878
+ "ranking_simple": 0.612500011920929,
1879
+ "regularize": 0.23523901402950287,
1880
+ "step": 650
1881
+ },
1882
+ {
1883
+ "epoch": 1.8422295701464337,
1884
+ "eval_dpo_loss": 0.7084385752677917,
1885
+ "eval_logits": -0.596887469291687,
1886
+ "eval_logps": -81.02249908447266,
1887
+ "eval_loss": 577.92919921875,
1888
+ "eval_objective": 565.5914916992188,
1889
+ "eval_ranking_simple": 0.5393374562263489,
1890
+ "eval_regularize": 0.5890585780143738,
1891
+ "eval_runtime": 370.9761,
1892
+ "eval_samples_per_second": 15.607,
1893
+ "eval_steps_per_second": 1.302,
1894
+ "step": 650
1895
+ },
1896
+ {
1897
+ "dpo_loss": 0.5157236456871033,
1898
+ "epoch": 1.8564005668398678,
1899
+ "grad_norm": 8570.37706164772,
1900
+ "learning_rate": 1.8943039936830347e-06,
1901
+ "logits": -0.5198069214820862,
1902
+ "logps": -75.62894439697266,
1903
+ "loss": 321.6153,
1904
+ "objective": 320.3071594238281,
1905
+ "ranking_simple": 0.5916666388511658,
1906
+ "regularize": 0.2329235076904297,
1907
+ "step": 655
1908
+ },
1909
+ {
1910
+ "dpo_loss": 0.521937906742096,
1911
+ "epoch": 1.8705715635333018,
1912
+ "grad_norm": 8345.743884997746,
1913
+ "learning_rate": 1.8542834264542091e-06,
1914
+ "logits": -0.5126068592071533,
1915
+ "logps": -78.36030578613281,
1916
+ "loss": 328.7253,
1917
+ "objective": 325.4534606933594,
1918
+ "ranking_simple": 0.625,
1919
+ "regularize": 0.2483067512512207,
1920
+ "step": 660
1921
+ },
1922
+ {
1923
+ "dpo_loss": 0.515285313129425,
1924
+ "epoch": 1.8847425602267358,
1925
+ "grad_norm": 7929.985369188011,
1926
+ "learning_rate": 1.814439391689151e-06,
1927
+ "logits": -0.5389847159385681,
1928
+ "logps": -76.56269073486328,
1929
+ "loss": 319.5206,
1930
+ "objective": 331.35791015625,
1931
+ "ranking_simple": 0.6625000238418579,
1932
+ "regularize": 0.23167690634727478,
1933
+ "step": 665
1934
+ },
1935
+ {
1936
+ "dpo_loss": 0.5117120146751404,
1937
+ "epoch": 1.89891355692017,
1938
+ "grad_norm": 8611.003311273333,
1939
+ "learning_rate": 1.7747827823491253e-06,
1940
+ "logits": -0.4842732548713684,
1941
+ "logps": -75.30670928955078,
1942
+ "loss": 310.9476,
1943
+ "objective": 312.0437316894531,
1944
+ "ranking_simple": 0.5583333373069763,
1945
+ "regularize": 0.23464351892471313,
1946
+ "step": 670
1947
+ },
1948
+ {
1949
+ "dpo_loss": 0.5363326668739319,
1950
+ "epoch": 1.9130845536136043,
1951
+ "grad_norm": 8390.641065212347,
1952
+ "learning_rate": 1.7353244401551566e-06,
1953
+ "logits": -0.4712333679199219,
1954
+ "logps": -76.18867492675781,
1955
+ "loss": 317.9931,
1956
+ "objective": 336.70703125,
1957
+ "ranking_simple": 0.625,
1958
+ "regularize": 0.23009681701660156,
1959
+ "step": 675
1960
+ },
1961
+ {
1962
+ "dpo_loss": 0.5199182033538818,
1963
+ "epoch": 1.9272555503070383,
1964
+ "grad_norm": 8581.710628204828,
1965
+ "learning_rate": 1.6960751526240122e-06,
1966
+ "logits": -0.49357444047927856,
1967
+ "logps": -78.31690216064453,
1968
+ "loss": 315.1092,
1969
+ "objective": 305.8600769042969,
1970
+ "ranking_simple": 0.6291666626930237,
1971
+ "regularize": 0.20979805290699005,
1972
+ "step": 680
1973
+ },
1974
+ {
1975
+ "dpo_loss": 0.5106812119483948,
1976
+ "epoch": 1.9414265470004723,
1977
+ "grad_norm": 8197.15415072432,
1978
+ "learning_rate": 1.6570456501189996e-06,
1979
+ "logits": -0.5017139911651611,
1980
+ "logps": -76.01095581054688,
1981
+ "loss": 306.0857,
1982
+ "objective": 298.7364501953125,
1983
+ "ranking_simple": 0.5333333611488342,
1984
+ "regularize": 0.20531941950321198,
1985
+ "step": 685
1986
+ },
1987
+ {
1988
+ "dpo_loss": 0.5228912830352783,
1989
+ "epoch": 1.9555975436939064,
1990
+ "grad_norm": 8227.185493128562,
1991
+ "learning_rate": 1.6182466029163974e-06,
1992
+ "logits": -0.503932535648346,
1993
+ "logps": -76.89771270751953,
1994
+ "loss": 312.1701,
1995
+ "objective": 309.2964782714844,
1996
+ "ranking_simple": 0.5874999761581421,
1997
+ "regularize": 0.1995265930891037,
1998
+ "step": 690
1999
+ },
2000
+ {
2001
+ "dpo_loss": 0.5291420817375183,
2002
+ "epoch": 1.9697685403873406,
2003
+ "grad_norm": 8124.705980309312,
2004
+ "learning_rate": 1.5796886182883053e-06,
2005
+ "logits": -0.47076740860939026,
2006
+ "logps": -77.54142761230469,
2007
+ "loss": 303.4818,
2008
+ "objective": 305.0574951171875,
2009
+ "ranking_simple": 0.5625,
2010
+ "regularize": 0.16832788288593292,
2011
+ "step": 695
2012
+ },
2013
+ {
2014
+ "dpo_loss": 0.5130675435066223,
2015
+ "epoch": 1.9839395370807746,
2016
+ "grad_norm": 8798.438767811667,
2017
+ "learning_rate": 1.541382237602721e-06,
2018
+ "logits": -0.4266551434993744,
2019
+ "logps": -77.73848724365234,
2020
+ "loss": 316.2014,
2021
+ "objective": 308.5932312011719,
2022
+ "ranking_simple": 0.6041666865348816,
2023
+ "regularize": 0.20295077562332153,
2024
+ "step": 700
2025
+ },
2026
+ {
2027
+ "epoch": 1.9839395370807746,
2028
+ "eval_dpo_loss": 0.709835410118103,
2029
+ "eval_logits": -0.5956435203552246,
2030
+ "eval_logps": -80.5416488647461,
2031
+ "eval_loss": 577.6038208007812,
2032
+ "eval_objective": 564.6389770507812,
2033
+ "eval_ranking_simple": 0.5408902764320374,
2034
+ "eval_regularize": 0.5856665968894958,
2035
+ "eval_runtime": 367.2185,
2036
+ "eval_samples_per_second": 15.767,
2037
+ "eval_steps_per_second": 1.315,
2038
+ "step": 700
2039
+ },
2040
+ {
2041
+ "dpo_loss": 0.5043766498565674,
2042
+ "epoch": 1.9981105337742089,
2043
+ "grad_norm": 8813.527233337918,
2044
+ "learning_rate": 1.5033379334416376e-06,
2045
+ "logits": -0.415615975856781,
2046
+ "logps": -76.69499969482422,
2047
+ "loss": 307.2797,
2048
+ "objective": 299.0835266113281,
2049
+ "ranking_simple": 0.5833333134651184,
2050
+ "regularize": 0.1889916956424713,
2051
+ "step": 705
2052
+ },
2053
+ {
2054
+ "dpo_loss": 0.5075303316116333,
2055
+ "epoch": 2.012281530467643,
2056
+ "grad_norm": 8152.029201211714,
2057
+ "learning_rate": 1.465566106737942e-06,
2058
+ "logits": -0.5361739993095398,
2059
+ "logps": -76.06444549560547,
2060
+ "loss": 296.988,
2061
+ "objective": 296.4339904785156,
2062
+ "ranking_simple": 0.6291666626930237,
2063
+ "regularize": 0.16836042702198029,
2064
+ "step": 710
2065
+ },
2066
+ {
2067
+ "dpo_loss": 0.499095618724823,
2068
+ "epoch": 2.026452527161077,
2069
+ "grad_norm": 7474.705902903634,
2070
+ "learning_rate": 1.4280770839319073e-06,
2071
+ "logits": -0.5198894739151001,
2072
+ "logps": -75.47317504882812,
2073
+ "loss": 290.1166,
2074
+ "objective": 297.8743896484375,
2075
+ "ranking_simple": 0.7041666507720947,
2076
+ "regularize": 0.2067592293024063,
2077
+ "step": 715
2078
+ },
2079
+ {
2080
+ "dpo_loss": 0.49675101041793823,
2081
+ "epoch": 2.040623523854511,
2082
+ "grad_norm": 7687.909209756474,
2083
+ "learning_rate": 1.3908811141480408e-06,
2084
+ "logits": -0.48511701822280884,
2085
+ "logps": -76.5396957397461,
2086
+ "loss": 293.1388,
2087
+ "objective": 296.0254821777344,
2088
+ "ranking_simple": 0.625,
2089
+ "regularize": 0.18073533475399017,
2090
+ "step": 720
2091
+ },
2092
+ {
2093
+ "dpo_loss": 0.5085265636444092,
2094
+ "epoch": 2.0547945205479454,
2095
+ "grad_norm": 7150.825567237203,
2096
+ "learning_rate": 1.353988366393083e-06,
2097
+ "logits": -0.4982639253139496,
2098
+ "logps": -79.44617462158203,
2099
+ "loss": 288.661,
2100
+ "objective": 303.2115173339844,
2101
+ "ranking_simple": 0.6166666746139526,
2102
+ "regularize": 0.1895003318786621,
2103
+ "step": 725
2104
+ },
2105
+ {
2106
+ "dpo_loss": 0.5071407556533813,
2107
+ "epoch": 2.0689655172413794,
2108
+ "grad_norm": 7729.725354104077,
2109
+ "learning_rate": 1.3174089267758983e-06,
2110
+ "logits": -0.5406936407089233,
2111
+ "logps": -77.1041488647461,
2112
+ "loss": 295.914,
2113
+ "objective": 279.0093078613281,
2114
+ "ranking_simple": 0.6041666865348816,
2115
+ "regularize": 0.1544083058834076,
2116
+ "step": 730
2117
+ },
2118
+ {
2119
+ "dpo_loss": 0.508653998374939,
2120
+ "epoch": 2.0831365139348135,
2121
+ "grad_norm": 7709.562610036098,
2122
+ "learning_rate": 1.2811527957500344e-06,
2123
+ "logits": -0.5032610297203064,
2124
+ "logps": -76.36556243896484,
2125
+ "loss": 281.1705,
2126
+ "objective": 274.6349792480469,
2127
+ "ranking_simple": 0.5625,
2128
+ "regularize": 0.16756394505500793,
2129
+ "step": 735
2130
+ },
2131
+ {
2132
+ "dpo_loss": 0.5007545351982117,
2133
+ "epoch": 2.0973075106282475,
2134
+ "grad_norm": 7780.159502686646,
2135
+ "learning_rate": 1.245229885379699e-06,
2136
+ "logits": -0.5718483328819275,
2137
+ "logps": -76.52877807617188,
2138
+ "loss": 287.3269,
2139
+ "objective": 281.77728271484375,
2140
+ "ranking_simple": 0.625,
2141
+ "regularize": 0.15750272572040558,
2142
+ "step": 740
2143
+ },
2144
+ {
2145
+ "dpo_loss": 0.49083974957466125,
2146
+ "epoch": 2.1114785073216815,
2147
+ "grad_norm": 7559.0234640974395,
2148
+ "learning_rate": 1.2096500166298992e-06,
2149
+ "logits": -0.5142738223075867,
2150
+ "logps": -77.08470916748047,
2151
+ "loss": 288.8493,
2152
+ "objective": 278.6231384277344,
2153
+ "ranking_simple": 0.6208333373069763,
2154
+ "regularize": 0.1696869283914566,
2155
+ "step": 745
2156
+ },
2157
+ {
2158
+ "dpo_loss": 0.510530412197113,
2159
+ "epoch": 2.1256495040151155,
2160
+ "grad_norm": 8108.616262187852,
2161
+ "learning_rate": 1.1744229166814889e-06,
2162
+ "logits": -0.5391489267349243,
2163
+ "logps": -75.9011459350586,
2164
+ "loss": 295.2996,
2165
+ "objective": 287.6244201660156,
2166
+ "ranking_simple": 0.5791666507720947,
2167
+ "regularize": 0.14458681643009186,
2168
+ "step": 750
2169
+ },
2170
+ {
2171
+ "epoch": 2.1256495040151155,
2172
+ "eval_dpo_loss": 0.7108220458030701,
2173
+ "eval_logits": -0.5878574252128601,
2174
+ "eval_logps": -81.0738525390625,
2175
+ "eval_loss": 579.5015258789062,
2176
+ "eval_objective": 567.8404541015625,
2177
+ "eval_ranking_simple": 0.5393374562263489,
2178
+ "eval_regularize": 0.592528223991394,
2179
+ "eval_runtime": 367.7716,
2180
+ "eval_samples_per_second": 15.743,
2181
+ "eval_steps_per_second": 1.313,
2182
+ "step": 750
2183
+ },
2184
+ {
2185
+ "dpo_loss": 0.5009695291519165,
2186
+ "epoch": 2.13982050070855,
2187
+ "grad_norm": 8018.931804650577,
2188
+ "learning_rate": 1.1395582162718524e-06,
2189
+ "logits": -0.5374471545219421,
2190
+ "logps": -78.51016998291016,
2191
+ "loss": 291.8306,
2192
+ "objective": 290.9220275878906,
2193
+ "ranking_simple": 0.5916666388511658,
2194
+ "regularize": 0.1683264821767807,
2195
+ "step": 755
2196
+ },
2197
+ {
2198
+ "dpo_loss": 0.5287482738494873,
2199
+ "epoch": 2.153991497401984,
2200
+ "grad_norm": 7462.123206327768,
2201
+ "learning_rate": 1.1050654470619602e-06,
2202
+ "logits": -0.46891796588897705,
2203
+ "logps": -75.76030731201172,
2204
+ "loss": 290.7075,
2205
+ "objective": 282.1697082519531,
2206
+ "ranking_simple": 0.6041666865348816,
2207
+ "regularize": 0.15730994939804077,
2208
+ "step": 760
2209
+ },
2210
+ {
2211
+ "dpo_loss": 0.5191124081611633,
2212
+ "epoch": 2.168162494095418,
2213
+ "grad_norm": 8366.940049042583,
2214
+ "learning_rate": 1.0709540390305061e-06,
2215
+ "logits": -0.48605969548225403,
2216
+ "logps": -76.97061920166016,
2217
+ "loss": 283.513,
2218
+ "objective": 276.4500732421875,
2219
+ "ranking_simple": 0.6333333253860474,
2220
+ "regularize": 0.14232973754405975,
2221
+ "step": 765
2222
+ },
2223
+ {
2224
+ "dpo_loss": 0.5144416093826294,
2225
+ "epoch": 2.182333490788852,
2226
+ "grad_norm": 8218.822763503342,
2227
+ "learning_rate": 1.0372333178958462e-06,
2228
+ "logits": -0.4805113971233368,
2229
+ "logps": -77.28369140625,
2230
+ "loss": 295.4732,
2231
+ "objective": 300.4273376464844,
2232
+ "ranking_simple": 0.5874999761581421,
2233
+ "regularize": 0.17838290333747864,
2234
+ "step": 770
2235
+ },
2236
+ {
2237
+ "dpo_loss": 0.5137962698936462,
2238
+ "epoch": 2.196504487482286,
2239
+ "grad_norm": 6441.500828763883,
2240
+ "learning_rate": 1.0039125025664392e-06,
2241
+ "logits": -0.5039299130439758,
2242
+ "logps": -77.55305480957031,
2243
+ "loss": 277.693,
2244
+ "objective": 280.0492248535156,
2245
+ "ranking_simple": 0.6541666388511658,
2246
+ "regularize": 0.14146603643894196,
2247
+ "step": 775
2248
+ },
2249
+ {
2250
+ "dpo_loss": 0.5056738257408142,
2251
+ "epoch": 2.21067548417572,
2252
+ "grad_norm": 8834.102544429094,
2253
+ "learning_rate": 9.710007026204896e-07,
2254
+ "logits": -0.43781086802482605,
2255
+ "logps": -77.69278717041016,
2256
+ "loss": 288.3063,
2257
+ "objective": 297.6618347167969,
2258
+ "ranking_simple": 0.5249999761581421,
2259
+ "regularize": 0.16135714948177338,
2260
+ "step": 780
2261
+ },
2262
+ {
2263
+ "dpo_loss": 0.4925435483455658,
2264
+ "epoch": 2.2248464808691546,
2265
+ "grad_norm": 7651.241881609664,
2266
+ "learning_rate": 9.385069158154805e-07,
2267
+ "logits": -0.4533029794692993,
2268
+ "logps": -77.70331573486328,
2269
+ "loss": 280.9115,
2270
+ "objective": 270.46942138671875,
2271
+ "ranking_simple": 0.5833333134651184,
2272
+ "regularize": 0.14943251013755798,
2273
+ "step": 785
2274
+ },
2275
+ {
2276
+ "dpo_loss": 0.49593406915664673,
2277
+ "epoch": 2.2390174775625886,
2278
+ "grad_norm": 7219.442805293083,
2279
+ "learning_rate": 9.064400256282757e-07,
2280
+ "logits": -0.44717687368392944,
2281
+ "logps": -78.27497863769531,
2282
+ "loss": 282.4128,
2283
+ "objective": 267.8057861328125,
2284
+ "ranking_simple": 0.6333333253860474,
2285
+ "regularize": 0.15546827018260956,
2286
+ "step": 790
2287
+ },
2288
+ {
2289
+ "dpo_loss": 0.505749523639679,
2290
+ "epoch": 2.2531884742560226,
2291
+ "grad_norm": 7763.536847557715,
2292
+ "learning_rate": 8.74808798826467e-07,
2293
+ "logits": -0.4969979226589203,
2294
+ "logps": -79.33271789550781,
2295
+ "loss": 281.3165,
2296
+ "objective": 274.2493896484375,
2297
+ "ranking_simple": 0.625,
2298
+ "regularize": 0.1496947556734085,
2299
+ "step": 795
2300
+ },
2301
+ {
2302
+ "dpo_loss": 0.48660808801651,
2303
+ "epoch": 2.2673594709494567,
2304
+ "grad_norm": 7932.089944092327,
2305
+ "learning_rate": 8.436218830716259e-07,
2306
+ "logits": -0.5253292322158813,
2307
+ "logps": -78.79373931884766,
2308
+ "loss": 290.0791,
2309
+ "objective": 280.91900634765625,
2310
+ "ranking_simple": 0.612500011920929,
2311
+ "regularize": 0.16623292863368988,
2312
+ "step": 800
2313
+ },
2314
+ {
2315
+ "epoch": 2.2673594709494567,
2316
+ "eval_dpo_loss": 0.7088403105735779,
2317
+ "eval_logits": -0.5885158777236938,
2318
+ "eval_logps": -81.68885803222656,
2319
+ "eval_loss": 576.8207397460938,
2320
+ "eval_objective": 564.8282470703125,
2321
+ "eval_ranking_simple": 0.5377846956253052,
2322
+ "eval_regularize": 0.5875974893569946,
2323
+ "eval_runtime": 367.7198,
2324
+ "eval_samples_per_second": 15.746,
2325
+ "eval_steps_per_second": 1.314,
2326
+ "step": 800
2327
+ },
2328
+ {
2329
+ "dpo_loss": 0.5162143707275391,
2330
+ "epoch": 2.2815304676428907,
2331
+ "grad_norm": 7214.841128896386,
2332
+ "learning_rate": 8.1288780455512e-07,
2333
+ "logits": -0.5276073217391968,
2334
+ "logps": -77.3137435913086,
2335
+ "loss": 278.3046,
2336
+ "objective": 292.67730712890625,
2337
+ "ranking_simple": 0.5874999761581421,
2338
+ "regularize": 0.16379231214523315,
2339
+ "step": 805
2340
+ },
2341
+ {
2342
+ "dpo_loss": 0.5017233490943909,
2343
+ "epoch": 2.295701464336325,
2344
+ "grad_norm": 6869.828749708057,
2345
+ "learning_rate": 7.826149656671386e-07,
2346
+ "logits": -0.5925108790397644,
2347
+ "logps": -76.71666717529297,
2348
+ "loss": 282.4796,
2349
+ "objective": 281.0669860839844,
2350
+ "ranking_simple": 0.675000011920929,
2351
+ "regularize": 0.1503111720085144,
2352
+ "step": 810
2353
+ },
2354
+ {
2355
+ "dpo_loss": 0.4981227219104767,
2356
+ "epoch": 2.309872461029759,
2357
+ "grad_norm": 7651.540848559703,
2358
+ "learning_rate": 7.528116426995605e-07,
2359
+ "logits": -0.5629077553749084,
2360
+ "logps": -78.19300079345703,
2361
+ "loss": 273.4494,
2362
+ "objective": 278.7046203613281,
2363
+ "ranking_simple": 0.5958333611488342,
2364
+ "regularize": 0.13803134858608246,
2365
+ "step": 815
2366
+ },
2367
+ {
2368
+ "dpo_loss": 0.5106547474861145,
2369
+ "epoch": 2.324043457723193,
2370
+ "grad_norm": 8303.88053874128,
2371
+ "learning_rate": 7.234859835833022e-07,
2372
+ "logits": -0.49964413046836853,
2373
+ "logps": -77.4349136352539,
2374
+ "loss": 281.5386,
2375
+ "objective": 289.7933349609375,
2376
+ "ranking_simple": 0.5833333134651184,
2377
+ "regularize": 0.14034216105937958,
2378
+ "step": 820
2379
+ },
2380
+ {
2381
+ "dpo_loss": 0.5033460259437561,
2382
+ "epoch": 2.3382144544166272,
2383
+ "grad_norm": 7465.14408583757,
2384
+ "learning_rate": 6.94646005660749e-07,
2385
+ "logits": -0.5037187337875366,
2386
+ "logps": -76.92729949951172,
2387
+ "loss": 278.1418,
2388
+ "objective": 275.7555847167969,
2389
+ "ranking_simple": 0.637499988079071,
2390
+ "regularize": 0.14195986092090607,
2391
+ "step": 825
2392
+ },
2393
+ {
2394
+ "dpo_loss": 0.5009591579437256,
2395
+ "epoch": 2.3523854511100613,
2396
+ "grad_norm": 7370.912864973515,
2397
+ "learning_rate": 6.662995934939007e-07,
2398
+ "logits": -0.5249782800674438,
2399
+ "logps": -78.88058471679688,
2400
+ "loss": 277.6341,
2401
+ "objective": 290.5874328613281,
2402
+ "ranking_simple": 0.6541666388511658,
2403
+ "regularize": 0.15550047159194946,
2404
+ "step": 830
2405
+ },
2406
+ {
2407
+ "dpo_loss": 0.506280243396759,
2408
+ "epoch": 2.3665564478034957,
2409
+ "grad_norm": 7154.510003938923,
2410
+ "learning_rate": 6.384544967088063e-07,
2411
+ "logits": -0.5261546969413757,
2412
+ "logps": -78.27130889892578,
2413
+ "loss": 283.8206,
2414
+ "objective": 284.1214294433594,
2415
+ "ranking_simple": 0.6333333253860474,
2416
+ "regularize": 0.15022988617420197,
2417
+ "step": 835
2418
+ },
2419
+ {
2420
+ "dpo_loss": 0.5248311758041382,
2421
+ "epoch": 2.3807274444969297,
2422
+ "grad_norm": 7561.035287310657,
2423
+ "learning_rate": 6.111183278768956e-07,
2424
+ "logits": -0.47096776962280273,
2425
+ "logps": -78.47908020019531,
2426
+ "loss": 281.1299,
2427
+ "objective": 287.78863525390625,
2428
+ "ranking_simple": 0.5458333492279053,
2429
+ "regularize": 0.13501495122909546,
2430
+ "step": 840
2431
+ },
2432
+ {
2433
+ "dpo_loss": 0.5237764120101929,
2434
+ "epoch": 2.3948984411903638,
2435
+ "grad_norm": 9019.208502842846,
2436
+ "learning_rate": 5.842985604337769e-07,
2437
+ "logits": -0.524657666683197,
2438
+ "logps": -79.37838745117188,
2439
+ "loss": 288.7943,
2440
+ "objective": 290.1376647949219,
2441
+ "ranking_simple": 0.6083333492279053,
2442
+ "regularize": 0.13873761892318726,
2443
+ "step": 845
2444
+ },
2445
+ {
2446
+ "dpo_loss": 0.5015512704849243,
2447
+ "epoch": 2.409069437883798,
2448
+ "grad_norm": 7214.405283580069,
2449
+ "learning_rate": 5.580025266360764e-07,
2450
+ "logits": -0.5334345102310181,
2451
+ "logps": -77.42970275878906,
2452
+ "loss": 277.1292,
2453
+ "objective": 274.3782043457031,
2454
+ "ranking_simple": 0.5916666388511658,
2455
+ "regularize": 0.13637620210647583,
2456
+ "step": 850
2457
+ },
2458
+ {
2459
+ "epoch": 2.409069437883798,
2460
+ "eval_dpo_loss": 0.7109217643737793,
2461
+ "eval_logits": -0.5770819187164307,
2462
+ "eval_logps": -81.54353332519531,
2463
+ "eval_loss": 579.0093994140625,
2464
+ "eval_objective": 567.1205444335938,
2465
+ "eval_ranking_simple": 0.5383023023605347,
2466
+ "eval_regularize": 0.5911442041397095,
2467
+ "eval_runtime": 370.8205,
2468
+ "eval_samples_per_second": 15.614,
2469
+ "eval_steps_per_second": 1.303,
2470
+ "step": 850
2471
+ },
2472
+ {
2473
+ "dpo_loss": 0.5071312785148621,
2474
+ "epoch": 2.423240434577232,
2475
+ "grad_norm": 7272.234405221214,
2476
+ "learning_rate": 5.322374155568688e-07,
2477
+ "logits": -0.5134973526000977,
2478
+ "logps": -76.6072769165039,
2479
+ "loss": 278.8906,
2480
+ "objective": 276.31671142578125,
2481
+ "ranking_simple": 0.6083333492279053,
2482
+ "regularize": 0.14478901028633118,
2483
+ "step": 855
2484
+ },
2485
+ {
2486
+ "dpo_loss": 0.5126808881759644,
2487
+ "epoch": 2.4374114312706663,
2488
+ "grad_norm": 8279.020308548867,
2489
+ "learning_rate": 5.070102711202606e-07,
2490
+ "logits": -0.4904005825519562,
2491
+ "logps": -78.11277770996094,
2492
+ "loss": 272.6675,
2493
+ "objective": 259.9554748535156,
2494
+ "ranking_simple": 0.6041666865348816,
2495
+ "regularize": 0.12409182637929916,
2496
+ "step": 860
2497
+ },
2498
+ {
2499
+ "dpo_loss": 0.5031525492668152,
2500
+ "epoch": 2.4515824279641003,
2501
+ "grad_norm": 7810.5501789039035,
2502
+ "learning_rate": 4.823279901756498e-07,
2503
+ "logits": -0.5084951519966125,
2504
+ "logps": -77.30481719970703,
2505
+ "loss": 278.5447,
2506
+ "objective": 284.18084716796875,
2507
+ "ranking_simple": 0.5916666388511658,
2508
+ "regularize": 0.1416517198085785,
2509
+ "step": 865
2510
+ },
2511
+ {
2512
+ "dpo_loss": 0.5099405646324158,
2513
+ "epoch": 2.4657534246575343,
2514
+ "grad_norm": 7054.3955623341435,
2515
+ "learning_rate": 4.581973206121948e-07,
2516
+ "logits": -0.522720456123352,
2517
+ "logps": -78.04560852050781,
2518
+ "loss": 278.9006,
2519
+ "objective": 274.6994934082031,
2520
+ "ranking_simple": 0.5958333611488342,
2521
+ "regularize": 0.13992194831371307,
2522
+ "step": 870
2523
+ },
2524
+ {
2525
+ "dpo_loss": 0.4974448084831238,
2526
+ "epoch": 2.4799244213509684,
2527
+ "grad_norm": 7158.801947079291,
2528
+ "learning_rate": 4.3462485951401126e-07,
2529
+ "logits": -0.481945663690567,
2530
+ "logps": -77.63534545898438,
2531
+ "loss": 265.5646,
2532
+ "objective": 269.0285949707031,
2533
+ "ranking_simple": 0.612500011920929,
2534
+ "regularize": 0.13618159294128418,
2535
+ "step": 875
2536
+ },
2537
+ {
2538
+ "dpo_loss": 0.516840398311615,
2539
+ "epoch": 2.4940954180444024,
2540
+ "grad_norm": 7224.52824602365,
2541
+ "learning_rate": 4.116170513565942e-07,
2542
+ "logits": -0.42012926936149597,
2543
+ "logps": -77.26931762695312,
2544
+ "loss": 277.9572,
2545
+ "objective": 277.60955810546875,
2546
+ "ranking_simple": 0.5708333253860474,
2547
+ "regularize": 0.12541402876377106,
2548
+ "step": 880
2549
+ },
2550
+ {
2551
+ "dpo_loss": 0.5121944546699524,
2552
+ "epoch": 2.5082664147378364,
2553
+ "grad_norm": 7670.321121071118,
2554
+ "learning_rate": 3.891801862449629e-07,
2555
+ "logits": -0.5377725958824158,
2556
+ "logps": -76.21176147460938,
2557
+ "loss": 273.7286,
2558
+ "objective": 278.1051940917969,
2559
+ "ranking_simple": 0.612500011920929,
2560
+ "regularize": 0.13135148584842682,
2561
+ "step": 885
2562
+ },
2563
+ {
2564
+ "dpo_loss": 0.5038707852363586,
2565
+ "epoch": 2.5224374114312704,
2566
+ "grad_norm": 7464.3594622623605,
2567
+ "learning_rate": 3.6732039819400686e-07,
2568
+ "logits": -0.5120099782943726,
2569
+ "logps": -75.08890533447266,
2570
+ "loss": 270.7643,
2571
+ "objective": 259.2137145996094,
2572
+ "ranking_simple": 0.6000000238418579,
2573
+ "regularize": 0.1371425837278366,
2574
+ "step": 890
2575
+ },
2576
+ {
2577
+ "dpo_loss": 0.5104149580001831,
2578
+ "epoch": 2.536608408124705,
2579
+ "grad_norm": 7776.17053954941,
2580
+ "learning_rate": 3.46043663451511e-07,
2581
+ "logits": -0.5063762068748474,
2582
+ "logps": -77.50283813476562,
2583
+ "loss": 274.0527,
2584
+ "objective": 279.8547668457031,
2585
+ "ranking_simple": 0.5708333253860474,
2586
+ "regularize": 0.13040438294410706,
2587
+ "step": 895
2588
+ },
2589
+ {
2590
+ "dpo_loss": 0.5273467302322388,
2591
+ "epoch": 2.550779404818139,
2592
+ "grad_norm": 7627.369911355359,
2593
+ "learning_rate": 3.253557988643072e-07,
2594
+ "logits": -0.49025458097457886,
2595
+ "logps": -76.16547393798828,
2596
+ "loss": 271.9766,
2597
+ "objective": 274.15386962890625,
2598
+ "ranking_simple": 0.625,
2599
+ "regularize": 0.13161370158195496,
2600
+ "step": 900
2601
+ },
2602
+ {
2603
+ "epoch": 2.550779404818139,
2604
+ "eval_dpo_loss": 0.7098731994628906,
2605
+ "eval_logits": -0.5707982182502747,
2606
+ "eval_logps": -81.16320037841797,
2607
+ "eval_loss": 577.3417358398438,
2608
+ "eval_objective": 565.7183837890625,
2609
+ "eval_ranking_simple": 0.5362318754196167,
2610
+ "eval_regularize": 0.5880821347236633,
2611
+ "eval_runtime": 368.1908,
2612
+ "eval_samples_per_second": 15.726,
2613
+ "eval_steps_per_second": 1.312,
2614
+ "step": 900
2615
+ },
2616
+ {
2617
+ "dpo_loss": 0.512609601020813,
2618
+ "epoch": 2.564950401511573,
2619
+ "grad_norm": 6804.885302407535,
2620
+ "learning_rate": 3.052624602880064e-07,
2621
+ "logits": -0.5115708112716675,
2622
+ "logps": -75.81733703613281,
2623
+ "loss": 265.9078,
2624
+ "objective": 269.9800109863281,
2625
+ "ranking_simple": 0.5833333134651184,
2626
+ "regularize": 0.14800624549388885,
2627
+ "step": 905
2628
+ },
2629
+ {
2630
+ "dpo_loss": 0.4937320947647095,
2631
+ "epoch": 2.579121398205007,
2632
+ "grad_norm": 7198.203187575293,
2633
+ "learning_rate": 2.8576914104074425e-07,
2634
+ "logits": -0.48827874660491943,
2635
+ "logps": -79.18850708007812,
2636
+ "loss": 265.7443,
2637
+ "objective": 273.89166259765625,
2638
+ "ranking_simple": 0.6041666865348816,
2639
+ "regularize": 0.14111953973770142,
2640
+ "step": 910
2641
+ },
2642
+ {
2643
+ "dpo_loss": 0.5084269642829895,
2644
+ "epoch": 2.593292394898441,
2645
+ "grad_norm": 6942.193456170007,
2646
+ "learning_rate": 2.6688117040136463e-07,
2647
+ "logits": -0.5942879915237427,
2648
+ "logps": -78.21766662597656,
2649
+ "loss": 278.3137,
2650
+ "objective": 263.6408386230469,
2651
+ "ranking_simple": 0.637499988079071,
2652
+ "regularize": 0.11606747657060623,
2653
+ "step": 915
2654
+ },
2655
+ {
2656
+ "dpo_loss": 0.498431533575058,
2657
+ "epoch": 2.6074633915918755,
2658
+ "grad_norm": 6451.638108152793,
2659
+ "learning_rate": 2.486037121524448e-07,
2660
+ "logits": -0.46755722165107727,
2661
+ "logps": -77.2091293334961,
2662
+ "loss": 270.5728,
2663
+ "objective": 287.58538818359375,
2664
+ "ranking_simple": 0.612500011920929,
2665
+ "regularize": 0.14371080696582794,
2666
+ "step": 920
2667
+ },
2668
+ {
2669
+ "dpo_loss": 0.5088561177253723,
2670
+ "epoch": 2.6216343882853095,
2671
+ "grad_norm": 8257.41304629772,
2672
+ "learning_rate": 2.3094176316856982e-07,
2673
+ "logits": -0.4761093258857727,
2674
+ "logps": -76.82450866699219,
2675
+ "loss": 270.0178,
2676
+ "objective": 261.9400634765625,
2677
+ "ranking_simple": 0.5708333253860474,
2678
+ "regularize": 0.11346574872732162,
2679
+ "step": 925
2680
+ },
2681
+ {
2682
+ "dpo_loss": 0.5108747482299805,
2683
+ "epoch": 2.6358053849787435,
2684
+ "grad_norm": 7715.040672522825,
2685
+ "learning_rate": 2.13900152050239e-07,
2686
+ "logits": -0.4767756164073944,
2687
+ "logps": -77.13011932373047,
2688
+ "loss": 273.9086,
2689
+ "objective": 285.1964111328125,
2690
+ "ranking_simple": 0.5666666626930237,
2691
+ "regularize": 0.13920140266418457,
2692
+ "step": 930
2693
+ },
2694
+ {
2695
+ "dpo_loss": 0.5189302563667297,
2696
+ "epoch": 2.6499763816721775,
2697
+ "grad_norm": 8548.789471220947,
2698
+ "learning_rate": 1.9748353780377234e-07,
2699
+ "logits": -0.45983853936195374,
2700
+ "logps": -76.03173828125,
2701
+ "loss": 276.8373,
2702
+ "objective": 269.6080322265625,
2703
+ "ranking_simple": 0.6583333611488342,
2704
+ "regularize": 0.12301207333803177,
2705
+ "step": 935
2706
+ },
2707
+ {
2708
+ "dpo_loss": 0.5017234683036804,
2709
+ "epoch": 2.6641473783656116,
2710
+ "grad_norm": 8162.699837992091,
2711
+ "learning_rate": 1.8169640856758652e-07,
2712
+ "logits": -0.504283607006073,
2713
+ "logps": -78.90401458740234,
2714
+ "loss": 281.2083,
2715
+ "objective": 285.80694580078125,
2716
+ "ranking_simple": 0.5958333611488342,
2717
+ "regularize": 0.1213822215795517,
2718
+ "step": 940
2719
+ },
2720
+ {
2721
+ "dpo_loss": 0.4903925657272339,
2722
+ "epoch": 2.678318375059046,
2723
+ "grad_norm": 7208.828785964443,
2724
+ "learning_rate": 1.6654308038518057e-07,
2725
+ "logits": -0.5296005010604858,
2726
+ "logps": -77.27458953857422,
2727
+ "loss": 265.3151,
2728
+ "objective": 269.78106689453125,
2729
+ "ranking_simple": 0.5833333134651184,
2730
+ "regularize": 0.12937407195568085,
2731
+ "step": 945
2732
+ },
2733
+ {
2734
+ "dpo_loss": 0.5223451256752014,
2735
+ "epoch": 2.69248937175248,
2736
+ "grad_norm": 7752.613058246486,
2737
+ "learning_rate": 1.5202769602517514e-07,
2738
+ "logits": -0.45917627215385437,
2739
+ "logps": -77.59455871582031,
2740
+ "loss": 273.4982,
2741
+ "objective": 260.7156982421875,
2742
+ "ranking_simple": 0.6333333253860474,
2743
+ "regularize": 0.10433920472860336,
2744
+ "step": 950
2745
+ },
2746
+ {
2747
+ "epoch": 2.69248937175248,
2748
+ "eval_dpo_loss": 0.7103093862533569,
2749
+ "eval_logits": -0.5680380463600159,
2750
+ "eval_logps": -81.19538879394531,
2751
+ "eval_loss": 578.9320678710938,
2752
+ "eval_objective": 567.1773071289062,
2753
+ "eval_ranking_simple": 0.5362318754196167,
2754
+ "eval_regularize": 0.5910264849662781,
2755
+ "eval_runtime": 367.6922,
2756
+ "eval_samples_per_second": 15.747,
2757
+ "eval_steps_per_second": 1.314,
2758
+ "step": 950
2759
+ },
2760
+ {
2761
+ "dpo_loss": 0.5060604810714722,
2762
+ "epoch": 2.706660368445914,
2763
+ "grad_norm": 6751.002791925838,
2764
+ "learning_rate": 1.381542238487188e-07,
2765
+ "logits": -0.5122019648551941,
2766
+ "logps": -77.12061309814453,
2767
+ "loss": 264.4783,
2768
+ "objective": 250.47158813476562,
2769
+ "ranking_simple": 0.5916666388511658,
2770
+ "regularize": 0.12410003691911697,
2771
+ "step": 955
2772
+ },
2773
+ {
2774
+ "dpo_loss": 0.5180147886276245,
2775
+ "epoch": 2.720831365139348,
2776
+ "grad_norm": 8054.959159664104,
2777
+ "learning_rate": 1.2492645672457838e-07,
2778
+ "logits": -0.48304110765457153,
2779
+ "logps": -77.1063003540039,
2780
+ "loss": 279.6324,
2781
+ "objective": 290.03631591796875,
2782
+ "ranking_simple": 0.6041666865348816,
2783
+ "regularize": 0.13714757561683655,
2784
+ "step": 960
2785
+ },
2786
+ {
2787
+ "dpo_loss": 0.49898579716682434,
2788
+ "epoch": 2.735002361832782,
2789
+ "grad_norm": 7601.482545856324,
2790
+ "learning_rate": 1.1234801099220787e-07,
2791
+ "logits": -0.5196807980537415,
2792
+ "logps": -77.11388397216797,
2793
+ "loss": 269.8563,
2794
+ "objective": 259.89813232421875,
2795
+ "ranking_simple": 0.6416666507720947,
2796
+ "regularize": 0.13252395391464233,
2797
+ "step": 965
2798
+ },
2799
+ {
2800
+ "dpo_loss": 0.51671302318573,
2801
+ "epoch": 2.7491733585262166,
2802
+ "grad_norm": 7233.095787852224,
2803
+ "learning_rate": 1.004223254730749e-07,
2804
+ "logits": -0.5174197554588318,
2805
+ "logps": -78.05948638916016,
2806
+ "loss": 262.7624,
2807
+ "objective": 268.42169189453125,
2808
+ "ranking_simple": 0.6333333253860474,
2809
+ "regularize": 0.13439247012138367,
2810
+ "step": 970
2811
+ },
2812
+ {
2813
+ "dpo_loss": 0.4989195764064789,
2814
+ "epoch": 2.7633443552196506,
2815
+ "grad_norm": 7653.138309833151,
2816
+ "learning_rate": 8.915266053052374e-08,
2817
+ "logits": -0.4456302523612976,
2818
+ "logps": -76.5742416381836,
2819
+ "loss": 265.2271,
2820
+ "objective": 260.4195556640625,
2821
+ "ranking_simple": 0.625,
2822
+ "regularize": 0.12436621636152267,
2823
+ "step": 975
2824
+ },
2825
+ {
2826
+ "dpo_loss": 0.5052908658981323,
2827
+ "epoch": 2.7775153519130846,
2828
+ "grad_norm": 7354.157364452891,
2829
+ "learning_rate": 7.854209717842231e-08,
2830
+ "logits": -0.5343514084815979,
2831
+ "logps": -76.55020141601562,
2832
+ "loss": 273.2646,
2833
+ "objective": 284.3968200683594,
2834
+ "ranking_simple": 0.6083333492279053,
2835
+ "regularize": 0.1349634826183319,
2836
+ "step": 980
2837
+ },
2838
+ {
2839
+ "dpo_loss": 0.5007596015930176,
2840
+ "epoch": 2.7916863486065187,
2841
+ "grad_norm": 6932.45863115929,
2842
+ "learning_rate": 6.859353623884569e-08,
2843
+ "logits": -0.44778621196746826,
2844
+ "logps": -74.72814178466797,
2845
+ "loss": 272.9477,
2846
+ "objective": 261.4037170410156,
2847
+ "ranking_simple": 0.5874999761581421,
2848
+ "regularize": 0.12619872391223907,
2849
+ "step": 985
2850
+ },
2851
+ {
2852
+ "dpo_loss": 0.50762540102005,
2853
+ "epoch": 2.8058573452999527,
2854
+ "grad_norm": 7347.273398591806,
2855
+ "learning_rate": 5.930969754901844e-08,
2856
+ "logits": -0.4778214693069458,
2857
+ "logps": -78.38259887695312,
2858
+ "loss": 260.7251,
2859
+ "objective": 262.8490905761719,
2860
+ "ranking_simple": 0.5791666507720947,
2861
+ "regularize": 0.11383456736803055,
2862
+ "step": 990
2863
+ },
2864
+ {
2865
+ "dpo_loss": 0.5177646279335022,
2866
+ "epoch": 2.820028341993387,
2867
+ "grad_norm": 8414.372934711379,
2868
+ "learning_rate": 5.069311921774039e-08,
2869
+ "logits": -0.5479720830917358,
2870
+ "logps": -76.77764129638672,
2871
+ "loss": 273.3689,
2872
+ "objective": 282.17889404296875,
2873
+ "ranking_simple": 0.6208333373069763,
2874
+ "regularize": 0.13415595889091492,
2875
+ "step": 995
2876
+ },
2877
+ {
2878
+ "dpo_loss": 0.5031678080558777,
2879
+ "epoch": 2.8341993386868207,
2880
+ "grad_norm": 6910.266147708002,
2881
+ "learning_rate": 4.2746156931490756e-08,
2882
+ "logits": -0.4682956337928772,
2883
+ "logps": -75.585205078125,
2884
+ "loss": 265.7935,
2885
+ "objective": 275.0522155761719,
2886
+ "ranking_simple": 0.612500011920929,
2887
+ "regularize": 0.12777787446975708,
2888
+ "step": 1000
2889
+ },
2890
+ {
2891
+ "epoch": 2.8341993386868207,
2892
+ "eval_dpo_loss": 0.709902822971344,
2893
+ "eval_logits": -0.5704091191291809,
2894
+ "eval_logps": -81.14698028564453,
2895
+ "eval_loss": 578.3192138671875,
2896
+ "eval_objective": 566.560791015625,
2897
+ "eval_ranking_simple": 0.5367494821548462,
2898
+ "eval_regularize": 0.5901817083358765,
2899
+ "eval_runtime": 368.6553,
2900
+ "eval_samples_per_second": 15.706,
2901
+ "eval_steps_per_second": 1.31,
2902
+ "step": 1000
2903
+ },
2904
+ {
2905
+ "dpo_loss": 0.5090419054031372,
2906
+ "epoch": 2.848370335380255,
2907
+ "grad_norm": 7970.005574260741,
2908
+ "learning_rate": 3.547098331040916e-08,
2909
+ "logits": -0.48795023560523987,
2910
+ "logps": -76.1548843383789,
2911
+ "loss": 272.1969,
2912
+ "objective": 268.9000244140625,
2913
+ "ranking_simple": 0.6166666746139526,
2914
+ "regularize": 0.12190086394548416,
2915
+ "step": 1005
2916
+ },
2917
+ {
2918
+ "dpo_loss": 0.5107741355895996,
2919
+ "epoch": 2.862541332073689,
2920
+ "grad_norm": 7994.674781206224,
2921
+ "learning_rate": 2.8869587314321324e-08,
2922
+ "logits": -0.5256258845329285,
2923
+ "logps": -76.37800598144531,
2924
+ "loss": 282.5812,
2925
+ "objective": 272.2499084472656,
2926
+ "ranking_simple": 0.612500011920929,
2927
+ "regularize": 0.11501624435186386,
2928
+ "step": 1010
2929
+ },
2930
+ {
2931
+ "dpo_loss": 0.5150614380836487,
2932
+ "epoch": 2.8767123287671232,
2933
+ "grad_norm": 7261.730587266494,
2934
+ "learning_rate": 2.2943773698977935e-08,
2935
+ "logits": -0.4966468811035156,
2936
+ "logps": -75.5634994506836,
2937
+ "loss": 260.7579,
2938
+ "objective": 261.33038330078125,
2939
+ "ranking_simple": 0.6458333134651184,
2940
+ "regularize": 0.11271250247955322,
2941
+ "step": 1015
2942
+ },
2943
+ {
2944
+ "dpo_loss": 0.508990466594696,
2945
+ "epoch": 2.8908833254605573,
2946
+ "grad_norm": 7841.280059779772,
2947
+ "learning_rate": 1.7695162522652352e-08,
2948
+ "logits": -0.4532057046890259,
2949
+ "logps": -77.1956558227539,
2950
+ "loss": 271.9978,
2951
+ "objective": 272.5274963378906,
2952
+ "ranking_simple": 0.5833333134651184,
2953
+ "regularize": 0.12634117901325226,
2954
+ "step": 1020
2955
+ },
2956
+ {
2957
+ "dpo_loss": 0.5181344151496887,
2958
+ "epoch": 2.9050543221539913,
2959
+ "grad_norm": 7685.424658753134,
2960
+ "learning_rate": 1.3125188703233815e-08,
2961
+ "logits": -0.53793865442276,
2962
+ "logps": -77.12080383300781,
2963
+ "loss": 268.2244,
2964
+ "objective": 275.9144287109375,
2965
+ "ranking_simple": 0.625,
2966
+ "regularize": 0.12161087244749069,
2967
+ "step": 1025
2968
+ },
2969
+ {
2970
+ "dpo_loss": 0.5186858773231506,
2971
+ "epoch": 2.9192253188474258,
2972
+ "grad_norm": 7664.878619416483,
2973
+ "learning_rate": 9.235101625932885e-09,
2974
+ "logits": -0.535256564617157,
2975
+ "logps": -76.86356353759766,
2976
+ "loss": 277.9306,
2977
+ "objective": 276.3569641113281,
2978
+ "ranking_simple": 0.5541666746139526,
2979
+ "regularize": 0.13014444708824158,
2980
+ "step": 1030
2981
+ },
2982
+ {
2983
+ "dpo_loss": 0.5086424350738525,
2984
+ "epoch": 2.9333963155408598,
2985
+ "grad_norm": 7176.213599878172,
2986
+ "learning_rate": 6.025964801714412e-09,
2987
+ "logits": -0.4943471848964691,
2988
+ "logps": -77.83872985839844,
2989
+ "loss": 266.904,
2990
+ "objective": 264.484375,
2991
+ "ranking_simple": 0.6208333373069763,
2992
+ "regularize": 0.11565522104501724,
2993
+ "step": 1035
2994
+ },
2995
+ {
2996
+ "dpo_loss": 0.5082514882087708,
2997
+ "epoch": 2.947567312234294,
2998
+ "grad_norm": 7034.435559100273,
2999
+ "learning_rate": 3.4986555765434415e-09,
3000
+ "logits": -0.5228769779205322,
3001
+ "logps": -77.66551208496094,
3002
+ "loss": 271.848,
3003
+ "objective": 274.83355712890625,
3004
+ "ranking_simple": 0.5666666626930237,
3005
+ "regularize": 0.11535345017910004,
3006
+ "step": 1040
3007
+ },
3008
+ {
3009
+ "dpo_loss": 0.5133834481239319,
3010
+ "epoch": 2.961738308927728,
3011
+ "grad_norm": 7599.023611459706,
3012
+ "learning_rate": 1.6538648915270794e-09,
3013
+ "logits": -0.47324731945991516,
3014
+ "logps": -79.62892150878906,
3015
+ "loss": 268.8998,
3016
+ "objective": 262.8771667480469,
3017
+ "ranking_simple": 0.6208333373069763,
3018
+ "regularize": 0.12152813374996185,
3019
+ "step": 1045
3020
+ },
3021
+ {
3022
+ "dpo_loss": 0.5033511519432068,
3023
+ "epoch": 2.975909305621162,
3024
+ "grad_norm": 8022.09759775128,
3025
+ "learning_rate": 4.920970940180958e-10,
3026
+ "logits": -0.5222968459129333,
3027
+ "logps": -75.3882827758789,
3028
+ "loss": 265.6855,
3029
+ "objective": 262.70672607421875,
3030
+ "ranking_simple": 0.6541666388511658,
3031
+ "regularize": 0.1311034858226776,
3032
+ "step": 1050
3033
+ },
3034
+ {
3035
+ "epoch": 2.975909305621162,
3036
+ "eval_dpo_loss": 0.7098360061645508,
3037
+ "eval_logits": -0.5715492367744446,
3038
+ "eval_logps": -81.13370513916016,
3039
+ "eval_loss": 578.1592407226562,
3040
+ "eval_objective": 566.3953857421875,
3041
+ "eval_ranking_simple": 0.5362318754196167,
3042
+ "eval_regularize": 0.5898736715316772,
3043
+ "eval_runtime": 367.7364,
3044
+ "eval_samples_per_second": 15.745,
3045
+ "eval_steps_per_second": 1.313,
3046
+ "step": 1050
3047
+ },
3048
+ {
3049
+ "dpo_loss": 0.5144506692886353,
3050
+ "epoch": 2.9900803023145963,
3051
+ "grad_norm": 7870.107583704036,
3052
+ "learning_rate": 1.3669799732163314e-11,
3053
+ "logits": -0.4851941168308258,
3054
+ "logps": -76.2236099243164,
3055
+ "loss": 272.1204,
3056
+ "objective": 281.3744201660156,
3057
+ "ranking_simple": 0.6166666746139526,
3058
+ "regularize": 0.13684484362602234,
3059
+ "step": 1055
3060
+ },
3061
+ {
3062
+ "epoch": 2.992914501653283,
3063
+ "step": 1056,
3064
+ "total_flos": 0.0,
3065
+ "train_loss": 356.3672220056707,
3066
+ "train_runtime": 34691.2758,
3067
+ "train_samples_per_second": 4.393,
3068
+ "train_steps_per_second": 0.03
3069
+ }
3070
+ ],
3071
+ "logging_steps": 5,
3072
+ "max_steps": 1056,
3073
+ "num_input_tokens_seen": 0,
3074
+ "num_train_epochs": 3,
3075
+ "save_steps": 500,
3076
+ "stateful_callbacks": {
3077
+ "TrainerControl": {
3078
+ "args": {
3079
+ "should_epoch_stop": false,
3080
+ "should_evaluate": false,
3081
+ "should_log": false,
3082
+ "should_save": false,
3083
+ "should_training_stop": false
3084
+ },
3085
+ "attributes": {}
3086
+ }
3087
+ },
3088
+ "total_flos": 0.0,
3089
+ "train_batch_size": 4,
3090
+ "trial_name": null,
3091
+ "trial_params": null
3092
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9a69b824681ce67d205113cd8fd40c07e0a1cab29a113f8076a9333b11940d3
3
+ size 8120
vocab.json ADDED
The diff for this file is too large to render. See raw diff