hZzy commited on
Commit
c863752
·
verified ·
1 Parent(s): 1d92e5a

Model save

Browse files
README.md ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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-EXDPO2
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/45nrmgyz)
17
+ # qwen2.5-0.5b-expo-EXDPO2
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: 1.0827
22
+ - Logps: -99.3605
23
+ - Logits: -1.3686
24
+ - Objective: 1.0902
25
+ - Dpo Loss: 0.6830
26
+ - Regularize: 0.4072
27
+ - Ranking Simple: 0.5221
28
+ - Ranking Idealized: 0.5925
29
+ - Ranking Idealized Expo: 0.5166
30
+
31
+ ## Model description
32
+
33
+ More information needed
34
+
35
+ ## Intended uses & limitations
36
+
37
+ More information needed
38
+
39
+ ## Training and evaluation data
40
+
41
+ More information needed
42
+
43
+ ## Training procedure
44
+
45
+ ### Training hyperparameters
46
+
47
+ The following hyperparameters were used during training:
48
+ - learning_rate: 1e-07
49
+ - train_batch_size: 4
50
+ - eval_batch_size: 4
51
+ - seed: 42
52
+ - distributed_type: multi-GPU
53
+ - num_devices: 4
54
+ - gradient_accumulation_steps: 12
55
+ - total_train_batch_size: 192
56
+ - total_eval_batch_size: 16
57
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
58
+ - lr_scheduler_type: cosine
59
+ - lr_scheduler_warmup_ratio: 0.1
60
+ - num_epochs: 2
61
+
62
+ ### Training results
63
+
64
+ | Training Loss | Epoch | Step | Validation Loss | Logps | Logits | Objective | Dpo Loss | Regularize | Ranking Simple | Ranking Idealized | Ranking Idealized Expo |
65
+ |:-------------:|:------:|:----:|:---------------:|:--------:|:-------:|:---------:|:--------:|:----------:|:--------------:|:-----------------:|:----------------------:|
66
+ | 1.0911 | 0.1889 | 50 | 1.1043 | -97.7941 | -1.3092 | 1.1042 | 0.6905 | 0.4136 | 0.5166 | 0.5925 | 0.5166 |
67
+ | 1.0629 | 0.3778 | 100 | 1.0938 | -98.1420 | -1.3221 | 1.0948 | 0.6871 | 0.4077 | 0.5180 | 0.5925 | 0.5166 |
68
+ | 1.0597 | 0.5668 | 150 | 1.0886 | -97.5760 | -1.3388 | 1.0899 | 0.6843 | 0.4057 | 0.5193 | 0.5925 | 0.5166 |
69
+ | 1.0201 | 0.7557 | 200 | 1.0878 | -98.5525 | -1.3519 | 1.0928 | 0.6845 | 0.4083 | 0.5193 | 0.5925 | 0.5166 |
70
+ | 1.0173 | 0.9446 | 250 | 1.0863 | -99.3374 | -1.3586 | 1.0919 | 0.6845 | 0.4074 | 0.5193 | 0.5925 | 0.5166 |
71
+ | 0.9755 | 1.1335 | 300 | 1.0829 | -99.1144 | -1.3663 | 1.0893 | 0.6826 | 0.4067 | 0.5214 | 0.5925 | 0.5166 |
72
+ | 0.9708 | 1.3224 | 350 | 1.0829 | -99.2669 | -1.3642 | 1.0901 | 0.6833 | 0.4068 | 0.5221 | 0.5925 | 0.5166 |
73
+ | 0.968 | 1.5113 | 400 | 1.0829 | -99.2520 | -1.3682 | 1.0901 | 0.6832 | 0.4069 | 0.5214 | 0.5925 | 0.5166 |
74
+ | 0.9495 | 1.7003 | 450 | 1.0826 | -99.3216 | -1.3688 | 1.0900 | 0.6830 | 0.4070 | 0.5214 | 0.5925 | 0.5166 |
75
+ | 0.9463 | 1.8892 | 500 | 1.0827 | -99.3605 | -1.3686 | 1.0902 | 0.6830 | 0.4072 | 0.5221 | 0.5925 | 0.5166 |
76
+
77
+
78
+ ### Framework versions
79
+
80
+ - Transformers 4.42.0
81
+ - Pytorch 2.3.0+cu121
82
+ - Datasets 2.19.1
83
+ - 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": 1.994962216624685,
3
+ "total_flos": 0.0,
4
+ "train_loss": 1.010300681672313,
5
+ "train_runtime": 13850.696,
6
+ "train_samples": 50802,
7
+ "train_samples_per_second": 7.336,
8
+ "train_steps_per_second": 0.038
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:daf30a0a4932566107ed6784ff384a3205e62e96bea3fa5894b42ca24d544c04
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": 1.994962216624685,
3
+ "total_flos": 0.0,
4
+ "train_loss": 1.010300681672313,
5
+ "train_runtime": 13850.696,
6
+ "train_samples": 50802,
7
+ "train_samples_per_second": 7.336,
8
+ "train_steps_per_second": 0.038
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1792 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.994962216624685,
5
+ "eval_steps": 50,
6
+ "global_step": 528,
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": 0.003778337531486146,
14
+ "grad_norm": 49.44846097246243,
15
+ "learning_rate": 1.8867924528301883e-09,
16
+ "logits": -1.1757584810256958,
17
+ "logps": -94.8752670288086,
18
+ "loss": 1.0913,
19
+ "objective": 1.048144817352295,
20
+ "ranking_idealized": 0.5625,
21
+ "ranking_idealized_expo": 0.4375,
22
+ "ranking_simple": 0.4375,
23
+ "regularize": 0.3549971580505371,
24
+ "step": 1
25
+ },
26
+ {
27
+ "dpo_loss": 0.6931453347206116,
28
+ "epoch": 0.018891687657430732,
29
+ "grad_norm": 56.77495215138669,
30
+ "learning_rate": 9.433962264150943e-09,
31
+ "logits": -1.3608341217041016,
32
+ "logps": -92.11934661865234,
33
+ "loss": 1.1069,
34
+ "objective": 1.06926691532135,
35
+ "ranking_idealized": 0.5729166865348816,
36
+ "ranking_idealized_expo": 0.4895833432674408,
37
+ "ranking_simple": 0.4895833432674408,
38
+ "regularize": 0.37612107396125793,
39
+ "step": 5
40
+ },
41
+ {
42
+ "dpo_loss": 0.693118155002594,
43
+ "epoch": 0.037783375314861464,
44
+ "grad_norm": 50.79305715348446,
45
+ "learning_rate": 1.8867924528301887e-08,
46
+ "logits": -1.3560354709625244,
47
+ "logps": -90.93302917480469,
48
+ "loss": 1.1118,
49
+ "objective": 1.1139355897903442,
50
+ "ranking_idealized": 0.5791666507720947,
51
+ "ranking_idealized_expo": 0.48750001192092896,
52
+ "ranking_simple": 0.48750001192092896,
53
+ "regularize": 0.4208172559738159,
54
+ "step": 10
55
+ },
56
+ {
57
+ "dpo_loss": 0.6927705407142639,
58
+ "epoch": 0.05667506297229219,
59
+ "grad_norm": 48.78735409468553,
60
+ "learning_rate": 2.830188679245283e-08,
61
+ "logits": -1.3467658758163452,
62
+ "logps": -92.52275848388672,
63
+ "loss": 1.1086,
64
+ "objective": 1.1240794658660889,
65
+ "ranking_idealized": 0.5833333134651184,
66
+ "ranking_idealized_expo": 0.5291666388511658,
67
+ "ranking_simple": 0.5291666388511658,
68
+ "regularize": 0.4313090443611145,
69
+ "step": 15
70
+ },
71
+ {
72
+ "dpo_loss": 0.6927476525306702,
73
+ "epoch": 0.07556675062972293,
74
+ "grad_norm": 48.89099638416194,
75
+ "learning_rate": 3.7735849056603774e-08,
76
+ "logits": -1.286866545677185,
77
+ "logps": -91.98667907714844,
78
+ "loss": 1.1147,
79
+ "objective": 1.127905011177063,
80
+ "ranking_idealized": 0.6458333134651184,
81
+ "ranking_idealized_expo": 0.5541666746139526,
82
+ "ranking_simple": 0.5583333373069763,
83
+ "regularize": 0.43515732884407043,
84
+ "step": 20
85
+ },
86
+ {
87
+ "dpo_loss": 0.6929070353507996,
88
+ "epoch": 0.09445843828715365,
89
+ "grad_norm": 51.44180006710924,
90
+ "learning_rate": 4.7169811320754715e-08,
91
+ "logits": -1.3749586343765259,
92
+ "logps": -92.29708862304688,
93
+ "loss": 1.1069,
94
+ "objective": 1.1106172800064087,
95
+ "ranking_idealized": 0.6208333373069763,
96
+ "ranking_idealized_expo": 0.550000011920929,
97
+ "ranking_simple": 0.5458333492279053,
98
+ "regularize": 0.4177102744579315,
99
+ "step": 25
100
+ },
101
+ {
102
+ "dpo_loss": 0.6915452480316162,
103
+ "epoch": 0.11335012594458438,
104
+ "grad_norm": 52.52518235565533,
105
+ "learning_rate": 5.660377358490566e-08,
106
+ "logits": -1.4094866514205933,
107
+ "logps": -92.03213500976562,
108
+ "loss": 1.1138,
109
+ "objective": 1.1041828393936157,
110
+ "ranking_idealized": 0.5958333611488342,
111
+ "ranking_idealized_expo": 0.5208333134651184,
112
+ "ranking_simple": 0.5208333134651184,
113
+ "regularize": 0.4126375913619995,
114
+ "step": 30
115
+ },
116
+ {
117
+ "dpo_loss": 0.6903693675994873,
118
+ "epoch": 0.13224181360201512,
119
+ "grad_norm": 55.88065446639471,
120
+ "learning_rate": 6.603773584905659e-08,
121
+ "logits": -1.3921390771865845,
122
+ "logps": -93.27505493164062,
123
+ "loss": 1.0984,
124
+ "objective": 1.1129175424575806,
125
+ "ranking_idealized": 0.5874999761581421,
126
+ "ranking_idealized_expo": 0.5083333253860474,
127
+ "ranking_simple": 0.5083333253860474,
128
+ "regularize": 0.4225481450557709,
129
+ "step": 35
130
+ },
131
+ {
132
+ "dpo_loss": 0.6901296377182007,
133
+ "epoch": 0.15113350125944586,
134
+ "grad_norm": 48.11830818999703,
135
+ "learning_rate": 7.547169811320755e-08,
136
+ "logits": -1.4104276895523071,
137
+ "logps": -92.22798156738281,
138
+ "loss": 1.1021,
139
+ "objective": 1.083837628364563,
140
+ "ranking_idealized": 0.6625000238418579,
141
+ "ranking_idealized_expo": 0.574999988079071,
142
+ "ranking_simple": 0.5708333253860474,
143
+ "regularize": 0.39370766282081604,
144
+ "step": 40
145
+ },
146
+ {
147
+ "dpo_loss": 0.6862795352935791,
148
+ "epoch": 0.17002518891687657,
149
+ "grad_norm": 45.66068468167867,
150
+ "learning_rate": 8.490566037735849e-08,
151
+ "logits": -1.3663307428359985,
152
+ "logps": -91.8579330444336,
153
+ "loss": 1.1032,
154
+ "objective": 1.1153314113616943,
155
+ "ranking_idealized": 0.5666666626930237,
156
+ "ranking_idealized_expo": 0.512499988079071,
157
+ "ranking_simple": 0.5083333253860474,
158
+ "regularize": 0.4290517270565033,
159
+ "step": 45
160
+ },
161
+ {
162
+ "dpo_loss": 0.6879932284355164,
163
+ "epoch": 0.1889168765743073,
164
+ "grad_norm": 51.12091223625438,
165
+ "learning_rate": 9.433962264150943e-08,
166
+ "logits": -1.3217811584472656,
167
+ "logps": -90.97481536865234,
168
+ "loss": 1.0911,
169
+ "objective": 1.0859743356704712,
170
+ "ranking_idealized": 0.5625,
171
+ "ranking_idealized_expo": 0.48750001192092896,
172
+ "ranking_simple": 0.48750001192092896,
173
+ "regularize": 0.3979811370372772,
174
+ "step": 50
175
+ },
176
+ {
177
+ "epoch": 0.1889168765743073,
178
+ "eval_dpo_loss": 0.690547525882721,
179
+ "eval_logits": -1.3092193603515625,
180
+ "eval_logps": -97.7940902709961,
181
+ "eval_loss": 1.104289174079895,
182
+ "eval_objective": 1.1041598320007324,
183
+ "eval_ranking_idealized": 0.5925414562225342,
184
+ "eval_ranking_idealized_expo": 0.5165745615959167,
185
+ "eval_ranking_simple": 0.5165745615959167,
186
+ "eval_regularize": 0.4136121869087219,
187
+ "eval_runtime": 235.6732,
188
+ "eval_samples_per_second": 24.568,
189
+ "eval_steps_per_second": 1.536,
190
+ "step": 50
191
+ },
192
+ {
193
+ "dpo_loss": 0.6884905695915222,
194
+ "epoch": 0.20780856423173805,
195
+ "grad_norm": 54.01101403233421,
196
+ "learning_rate": 9.99956257238817e-08,
197
+ "logits": -1.4181103706359863,
198
+ "logps": -89.9561538696289,
199
+ "loss": 1.1009,
200
+ "objective": 1.094197154045105,
201
+ "ranking_idealized": 0.5791666507720947,
202
+ "ranking_idealized_expo": 0.5249999761581421,
203
+ "ranking_simple": 0.5291666388511658,
204
+ "regularize": 0.40570658445358276,
205
+ "step": 55
206
+ },
207
+ {
208
+ "dpo_loss": 0.6880350708961487,
209
+ "epoch": 0.22670025188916876,
210
+ "grad_norm": 54.996673931606395,
211
+ "learning_rate": 9.994642390694308e-08,
212
+ "logits": -1.3805103302001953,
213
+ "logps": -90.7548828125,
214
+ "loss": 1.1093,
215
+ "objective": 1.114397406578064,
216
+ "ranking_idealized": 0.5916666388511658,
217
+ "ranking_idealized_expo": 0.5041666626930237,
218
+ "ranking_simple": 0.5041666626930237,
219
+ "regularize": 0.4263623058795929,
220
+ "step": 60
221
+ },
222
+ {
223
+ "dpo_loss": 0.6825106143951416,
224
+ "epoch": 0.2455919395465995,
225
+ "grad_norm": 56.45685371856582,
226
+ "learning_rate": 9.98426064087682e-08,
227
+ "logits": -1.4092483520507812,
228
+ "logps": -92.43305969238281,
229
+ "loss": 1.1012,
230
+ "objective": 1.104454755783081,
231
+ "ranking_idealized": 0.5583333373069763,
232
+ "ranking_idealized_expo": 0.5166666507720947,
233
+ "ranking_simple": 0.5208333134651184,
234
+ "regularize": 0.4219440817832947,
235
+ "step": 65
236
+ },
237
+ {
238
+ "dpo_loss": 0.6836772561073303,
239
+ "epoch": 0.26448362720403024,
240
+ "grad_norm": 55.79791061619149,
241
+ "learning_rate": 9.968428675226714e-08,
242
+ "logits": -1.372382640838623,
243
+ "logps": -92.17337036132812,
244
+ "loss": 1.1123,
245
+ "objective": 1.1599425077438354,
246
+ "ranking_idealized": 0.6291666626930237,
247
+ "ranking_idealized_expo": 0.550000011920929,
248
+ "ranking_simple": 0.550000011920929,
249
+ "regularize": 0.476265013217926,
250
+ "step": 70
251
+ },
252
+ {
253
+ "dpo_loss": 0.6837269067764282,
254
+ "epoch": 0.28337531486146095,
255
+ "grad_norm": 47.47237853468278,
256
+ "learning_rate": 9.947163805765979e-08,
257
+ "logits": -1.3373608589172363,
258
+ "logps": -90.7457275390625,
259
+ "loss": 1.0793,
260
+ "objective": 1.0441726446151733,
261
+ "ranking_idealized": 0.5083333253860474,
262
+ "ranking_idealized_expo": 0.4375,
263
+ "ranking_simple": 0.4333333373069763,
264
+ "regularize": 0.3604455888271332,
265
+ "step": 75
266
+ },
267
+ {
268
+ "dpo_loss": 0.6796848773956299,
269
+ "epoch": 0.3022670025188917,
270
+ "grad_norm": 47.94506148186036,
271
+ "learning_rate": 9.920489285317169e-08,
272
+ "logits": -1.4042983055114746,
273
+ "logps": -91.38846588134766,
274
+ "loss": 1.0778,
275
+ "objective": 1.0585055351257324,
276
+ "ranking_idealized": 0.6041666865348816,
277
+ "ranking_idealized_expo": 0.4958333373069763,
278
+ "ranking_simple": 0.5041666626930237,
279
+ "regularize": 0.3788203001022339,
280
+ "step": 80
281
+ },
282
+ {
283
+ "dpo_loss": 0.6791465282440186,
284
+ "epoch": 0.3211586901763224,
285
+ "grad_norm": 47.95858131470083,
286
+ "learning_rate": 9.888434282076758e-08,
287
+ "logits": -1.3911387920379639,
288
+ "logps": -92.44884490966797,
289
+ "loss": 1.0925,
290
+ "objective": 1.0910677909851074,
291
+ "ranking_idealized": 0.5708333253860474,
292
+ "ranking_idealized_expo": 0.5,
293
+ "ranking_simple": 0.5,
294
+ "regularize": 0.4119212031364441,
295
+ "step": 85
296
+ },
297
+ {
298
+ "dpo_loss": 0.6837694644927979,
299
+ "epoch": 0.34005037783375314,
300
+ "grad_norm": 55.1730704885422,
301
+ "learning_rate": 9.851033847720166e-08,
302
+ "logits": -1.3752093315124512,
303
+ "logps": -91.67774200439453,
304
+ "loss": 1.0847,
305
+ "objective": 1.085326075553894,
306
+ "ranking_idealized": 0.5874999761581421,
307
+ "ranking_idealized_expo": 0.5166666507720947,
308
+ "ranking_simple": 0.512499988079071,
309
+ "regularize": 0.4015560448169708,
310
+ "step": 90
311
+ },
312
+ {
313
+ "dpo_loss": 0.6765946745872498,
314
+ "epoch": 0.3589420654911839,
315
+ "grad_norm": 48.33215476425452,
316
+ "learning_rate": 9.808328879073251e-08,
317
+ "logits": -1.3342726230621338,
318
+ "logps": -92.56273651123047,
319
+ "loss": 1.0612,
320
+ "objective": 1.060593605041504,
321
+ "ranking_idealized": 0.5833333134651184,
322
+ "ranking_idealized_expo": 0.512499988079071,
323
+ "ranking_simple": 0.5208333134651184,
324
+ "regularize": 0.3839987814426422,
325
+ "step": 95
326
+ },
327
+ {
328
+ "dpo_loss": 0.6816121935844421,
329
+ "epoch": 0.3778337531486146,
330
+ "grad_norm": 49.179351832428715,
331
+ "learning_rate": 9.760366073392245e-08,
332
+ "logits": -1.3950403928756714,
333
+ "logps": -92.40387725830078,
334
+ "loss": 1.0629,
335
+ "objective": 1.0402370691299438,
336
+ "ranking_idealized": 0.5916666388511658,
337
+ "ranking_idealized_expo": 0.5083333253860474,
338
+ "ranking_simple": 0.5041666626930237,
339
+ "regularize": 0.35862475633621216,
340
+ "step": 100
341
+ },
342
+ {
343
+ "epoch": 0.3778337531486146,
344
+ "eval_dpo_loss": 0.6871094703674316,
345
+ "eval_logits": -1.3221344947814941,
346
+ "eval_logps": -98.14198303222656,
347
+ "eval_loss": 1.093752384185791,
348
+ "eval_objective": 1.0948374271392822,
349
+ "eval_ranking_idealized": 0.5925414562225342,
350
+ "eval_ranking_idealized_expo": 0.5165745615959167,
351
+ "eval_ranking_simple": 0.5179557800292969,
352
+ "eval_regularize": 0.4077278971672058,
353
+ "eval_runtime": 235.7453,
354
+ "eval_samples_per_second": 24.56,
355
+ "eval_steps_per_second": 1.536,
356
+ "step": 100
357
+ },
358
+ {
359
+ "dpo_loss": 0.6785728931427002,
360
+ "epoch": 0.3967254408060453,
361
+ "grad_norm": 51.47899227122413,
362
+ "learning_rate": 9.707197877300973e-08,
363
+ "logits": -1.4077008962631226,
364
+ "logps": -91.70052337646484,
365
+ "loss": 1.0696,
366
+ "objective": 1.086281180381775,
367
+ "ranking_idealized": 0.612500011920929,
368
+ "ranking_idealized_expo": 0.5041666626930237,
369
+ "ranking_simple": 0.5041666626930237,
370
+ "regularize": 0.407708078622818,
371
+ "step": 105
372
+ },
373
+ {
374
+ "dpo_loss": 0.6762667298316956,
375
+ "epoch": 0.4156171284634761,
376
+ "grad_norm": 50.28143364432592,
377
+ "learning_rate": 9.648882429441257e-08,
378
+ "logits": -1.3699510097503662,
379
+ "logps": -93.08635711669922,
380
+ "loss": 1.0782,
381
+ "objective": 1.1022285223007202,
382
+ "ranking_idealized": 0.6291666626930237,
383
+ "ranking_idealized_expo": 0.5625,
384
+ "ranking_simple": 0.5708333253860474,
385
+ "regularize": 0.4259616434574127,
386
+ "step": 110
387
+ },
388
+ {
389
+ "dpo_loss": 0.6772736310958862,
390
+ "epoch": 0.4345088161209068,
391
+ "grad_norm": 49.35211256173949,
392
+ "learning_rate": 9.585483496899149e-08,
393
+ "logits": -1.4434921741485596,
394
+ "logps": -91.60735321044922,
395
+ "loss": 1.049,
396
+ "objective": 1.0325186252593994,
397
+ "ranking_idealized": 0.6000000238418579,
398
+ "ranking_idealized_expo": 0.48750001192092896,
399
+ "ranking_simple": 0.4958333373069763,
400
+ "regularize": 0.3552449345588684,
401
+ "step": 115
402
+ },
403
+ {
404
+ "dpo_loss": 0.67376708984375,
405
+ "epoch": 0.4534005037783375,
406
+ "grad_norm": 49.769392715448596,
407
+ "learning_rate": 9.517070405476573e-08,
408
+ "logits": -1.3977102041244507,
409
+ "logps": -91.86192321777344,
410
+ "loss": 1.0722,
411
+ "objective": 1.0750477313995361,
412
+ "ranking_idealized": 0.6000000238418579,
413
+ "ranking_idealized_expo": 0.5208333134651184,
414
+ "ranking_simple": 0.5249999761581421,
415
+ "regularize": 0.40128055214881897,
416
+ "step": 120
417
+ },
418
+ {
419
+ "dpo_loss": 0.6729015111923218,
420
+ "epoch": 0.4722921914357683,
421
+ "grad_norm": 48.20763298222792,
422
+ "learning_rate": 9.443717963884567e-08,
423
+ "logits": -1.4215410947799683,
424
+ "logps": -91.61782836914062,
425
+ "loss": 1.0433,
426
+ "objective": 1.0424840450286865,
427
+ "ranking_idealized": 0.6291666626930237,
428
+ "ranking_idealized_expo": 0.5666666626930237,
429
+ "ranking_simple": 0.5625,
430
+ "regularize": 0.36958202719688416,
431
+ "step": 125
432
+ },
433
+ {
434
+ "dpo_loss": 0.6743431687355042,
435
+ "epoch": 0.491183879093199,
436
+ "grad_norm": 48.45479288893814,
437
+ "learning_rate": 9.365506381941065e-08,
438
+ "logits": -1.3215110301971436,
439
+ "logps": -92.12106323242188,
440
+ "loss": 1.0515,
441
+ "objective": 1.097717046737671,
442
+ "ranking_idealized": 0.625,
443
+ "ranking_idealized_expo": 0.5541666746139526,
444
+ "ranking_simple": 0.5625,
445
+ "regularize": 0.4233737289905548,
446
+ "step": 130
447
+ },
448
+ {
449
+ "dpo_loss": 0.6695905327796936,
450
+ "epoch": 0.5100755667506297,
451
+ "grad_norm": 43.646349799244106,
452
+ "learning_rate": 9.282521182862628e-08,
453
+ "logits": -1.3272545337677002,
454
+ "logps": -92.60189056396484,
455
+ "loss": 1.057,
456
+ "objective": 1.0827622413635254,
457
+ "ranking_idealized": 0.5916666388511658,
458
+ "ranking_idealized_expo": 0.5208333134651184,
459
+ "ranking_simple": 0.5166666507720947,
460
+ "regularize": 0.41317158937454224,
461
+ "step": 135
462
+ },
463
+ {
464
+ "dpo_loss": 0.6709803342819214,
465
+ "epoch": 0.5289672544080605,
466
+ "grad_norm": 47.3960543467576,
467
+ "learning_rate": 9.194853109746073e-08,
468
+ "logits": -1.4292913675308228,
469
+ "logps": -92.06829071044922,
470
+ "loss": 1.0462,
471
+ "objective": 1.046062707901001,
472
+ "ranking_idealized": 0.6041666865348816,
473
+ "ranking_idealized_expo": 0.48750001192092896,
474
+ "ranking_simple": 0.5,
475
+ "regularize": 0.3750823736190796,
476
+ "step": 140
477
+ },
478
+ {
479
+ "dpo_loss": 0.6690364480018616,
480
+ "epoch": 0.5478589420654912,
481
+ "grad_norm": 51.65656404072737,
482
+ "learning_rate": 9.102598026342222e-08,
483
+ "logits": -1.4306373596191406,
484
+ "logps": -91.35098266601562,
485
+ "loss": 1.0533,
486
+ "objective": 1.0885671377182007,
487
+ "ranking_idealized": 0.6166666746139526,
488
+ "ranking_idealized_expo": 0.5583333373069763,
489
+ "ranking_simple": 0.5625,
490
+ "regularize": 0.41953060030937195,
491
+ "step": 145
492
+ },
493
+ {
494
+ "dpo_loss": 0.6675416231155396,
495
+ "epoch": 0.5667506297229219,
496
+ "grad_norm": 47.60377632937882,
497
+ "learning_rate": 9.005856812230304e-08,
498
+ "logits": -1.4119346141815186,
499
+ "logps": -90.73808288574219,
500
+ "loss": 1.0597,
501
+ "objective": 1.0471808910369873,
502
+ "ranking_idealized": 0.574999988079071,
503
+ "ranking_idealized_expo": 0.5291666388511658,
504
+ "ranking_simple": 0.5375000238418579,
505
+ "regularize": 0.379639208316803,
506
+ "step": 150
507
+ },
508
+ {
509
+ "epoch": 0.5667506297229219,
510
+ "eval_dpo_loss": 0.6842523813247681,
511
+ "eval_logits": -1.3387939929962158,
512
+ "eval_logps": -97.57598114013672,
513
+ "eval_loss": 1.0886038541793823,
514
+ "eval_objective": 1.089935064315796,
515
+ "eval_ranking_idealized": 0.5925414562225342,
516
+ "eval_ranking_idealized_expo": 0.5165745615959167,
517
+ "eval_ranking_simple": 0.519336998462677,
518
+ "eval_regularize": 0.4056825637817383,
519
+ "eval_runtime": 236.1326,
520
+ "eval_samples_per_second": 24.52,
521
+ "eval_steps_per_second": 1.533,
522
+ "step": 150
523
+ },
524
+ {
525
+ "dpo_loss": 0.6675500273704529,
526
+ "epoch": 0.5856423173803527,
527
+ "grad_norm": 48.130868373709134,
528
+ "learning_rate": 8.904735252507609e-08,
529
+ "logits": -1.3894909620285034,
530
+ "logps": -91.1141128540039,
531
+ "loss": 1.0454,
532
+ "objective": 1.0189038515090942,
533
+ "ranking_idealized": 0.6333333253860474,
534
+ "ranking_idealized_expo": 0.5375000238418579,
535
+ "ranking_simple": 0.5541666746139526,
536
+ "regularize": 0.3513537645339966,
537
+ "step": 155
538
+ },
539
+ {
540
+ "dpo_loss": 0.6642627716064453,
541
+ "epoch": 0.6045340050377834,
542
+ "grad_norm": 48.76010002249353,
543
+ "learning_rate": 8.799343922115043e-08,
544
+ "logits": -1.3890469074249268,
545
+ "logps": -91.98741149902344,
546
+ "loss": 1.0636,
547
+ "objective": 1.0629956722259521,
548
+ "ranking_idealized": 0.5916666388511658,
549
+ "ranking_idealized_expo": 0.512499988079071,
550
+ "ranking_simple": 0.5166666507720947,
551
+ "regularize": 0.39873284101486206,
552
+ "step": 160
553
+ },
554
+ {
555
+ "dpo_loss": 0.6647918224334717,
556
+ "epoch": 0.6234256926952141,
557
+ "grad_norm": 51.709247719469836,
558
+ "learning_rate": 8.689798064925048e-08,
559
+ "logits": -1.430177927017212,
560
+ "logps": -92.05915832519531,
561
+ "loss": 1.0499,
562
+ "objective": 1.05256187915802,
563
+ "ranking_idealized": 0.5583333373069763,
564
+ "ranking_idealized_expo": 0.4625000059604645,
565
+ "ranking_simple": 0.4791666567325592,
566
+ "regularize": 0.3877698481082916,
567
+ "step": 165
568
+ },
569
+ {
570
+ "dpo_loss": 0.6606873273849487,
571
+ "epoch": 0.6423173803526449,
572
+ "grad_norm": 46.459328825125596,
573
+ "learning_rate": 8.576217467724127e-08,
574
+ "logits": -1.3984259366989136,
575
+ "logps": -91.71315002441406,
576
+ "loss": 1.029,
577
+ "objective": 1.0234616994857788,
578
+ "ranking_idealized": 0.5458333492279053,
579
+ "ranking_idealized_expo": 0.42500001192092896,
580
+ "ranking_simple": 0.44583332538604736,
581
+ "regularize": 0.36277416348457336,
582
+ "step": 170
583
+ },
584
+ {
585
+ "dpo_loss": 0.6623333096504211,
586
+ "epoch": 0.6612090680100756,
587
+ "grad_norm": 51.1586426932559,
588
+ "learning_rate": 8.458726329227747e-08,
589
+ "logits": -1.4344308376312256,
590
+ "logps": -92.77267456054688,
591
+ "loss": 1.0316,
592
+ "objective": 1.0282338857650757,
593
+ "ranking_idealized": 0.6166666746139526,
594
+ "ranking_idealized_expo": 0.550000011920929,
595
+ "ranking_simple": 0.550000011920929,
596
+ "regularize": 0.36590057611465454,
597
+ "step": 175
598
+ },
599
+ {
600
+ "dpo_loss": 0.6674898862838745,
601
+ "epoch": 0.6801007556675063,
602
+ "grad_norm": 48.61836865824933,
603
+ "learning_rate": 8.337453124270862e-08,
604
+ "logits": -1.3953392505645752,
605
+ "logps": -90.3520736694336,
606
+ "loss": 1.0387,
607
+ "objective": 1.0503102540969849,
608
+ "ranking_idealized": 0.5958333611488342,
609
+ "ranking_idealized_expo": 0.5041666626930237,
610
+ "ranking_simple": 0.512499988079071,
611
+ "regularize": 0.3828202784061432,
612
+ "step": 180
613
+ },
614
+ {
615
+ "dpo_loss": 0.6639277935028076,
616
+ "epoch": 0.698992443324937,
617
+ "grad_norm": 56.3868937423093,
618
+ "learning_rate": 8.212530463322582e-08,
619
+ "logits": -1.376102089881897,
620
+ "logps": -90.30011749267578,
621
+ "loss": 1.0192,
622
+ "objective": 1.0141814947128296,
623
+ "ranking_idealized": 0.637499988079071,
624
+ "ranking_idealized_expo": 0.5666666626930237,
625
+ "ranking_simple": 0.5791666507720947,
626
+ "regularize": 0.3502536714076996,
627
+ "step": 185
628
+ },
629
+ {
630
+ "dpo_loss": 0.6700864434242249,
631
+ "epoch": 0.7178841309823678,
632
+ "grad_norm": 51.6918769313309,
633
+ "learning_rate": 8.084094947478554e-08,
634
+ "logits": -1.4650901556015015,
635
+ "logps": -92.19945526123047,
636
+ "loss": 1.0536,
637
+ "objective": 1.0666062831878662,
638
+ "ranking_idealized": 0.5833333134651184,
639
+ "ranking_idealized_expo": 0.4833333194255829,
640
+ "ranking_simple": 0.5,
641
+ "regularize": 0.3965199291706085,
642
+ "step": 190
643
+ },
644
+ {
645
+ "dpo_loss": 0.6677178740501404,
646
+ "epoch": 0.7367758186397985,
647
+ "grad_norm": 51.07369943909954,
648
+ "learning_rate": 7.952287019089686e-08,
649
+ "logits": -1.384783387184143,
650
+ "logps": -92.98999786376953,
651
+ "loss": 1.0332,
652
+ "objective": 1.0433076620101929,
653
+ "ranking_idealized": 0.6000000238418579,
654
+ "ranking_idealized_expo": 0.5249999761581421,
655
+ "ranking_simple": 0.5416666865348816,
656
+ "regularize": 0.3755895793437958,
657
+ "step": 195
658
+ },
659
+ {
660
+ "dpo_loss": 0.6689361333847046,
661
+ "epoch": 0.7556675062972292,
662
+ "grad_norm": 50.59449995506128,
663
+ "learning_rate": 7.817250808190483e-08,
664
+ "logits": -1.4448987245559692,
665
+ "logps": -92.8608169555664,
666
+ "loss": 1.0201,
667
+ "objective": 0.9872502684593201,
668
+ "ranking_idealized": 0.5833333134651184,
669
+ "ranking_idealized_expo": 0.5249999761581421,
670
+ "ranking_simple": 0.5249999761581421,
671
+ "regularize": 0.31831392645835876,
672
+ "step": 200
673
+ },
674
+ {
675
+ "epoch": 0.7556675062972292,
676
+ "eval_dpo_loss": 0.6845039129257202,
677
+ "eval_logits": -1.3518892526626587,
678
+ "eval_logps": -98.55247497558594,
679
+ "eval_loss": 1.0877532958984375,
680
+ "eval_objective": 1.0928053855895996,
681
+ "eval_ranking_idealized": 0.5925414562225342,
682
+ "eval_ranking_idealized_expo": 0.5165745615959167,
683
+ "eval_ranking_simple": 0.519336998462677,
684
+ "eval_regularize": 0.40830138325691223,
685
+ "eval_runtime": 236.1466,
686
+ "eval_samples_per_second": 24.519,
687
+ "eval_steps_per_second": 1.533,
688
+ "step": 200
689
+ },
690
+ {
691
+ "dpo_loss": 0.6665075421333313,
692
+ "epoch": 0.77455919395466,
693
+ "grad_norm": 44.9309987154925,
694
+ "learning_rate": 7.679133974894983e-08,
695
+ "logits": -1.377119779586792,
696
+ "logps": -91.88671112060547,
697
+ "loss": 1.0103,
698
+ "objective": 1.0105198621749878,
699
+ "ranking_idealized": 0.5583333373069763,
700
+ "ranking_idealized_expo": 0.49166667461395264,
701
+ "ranking_simple": 0.49166667461395264,
702
+ "regularize": 0.34401220083236694,
703
+ "step": 205
704
+ },
705
+ {
706
+ "dpo_loss": 0.6583543419837952,
707
+ "epoch": 0.7934508816120907,
708
+ "grad_norm": 51.30313684013195,
709
+ "learning_rate": 7.538087547932585e-08,
710
+ "logits": -1.3690478801727295,
711
+ "logps": -93.26469421386719,
712
+ "loss": 1.0206,
713
+ "objective": 1.019605040550232,
714
+ "ranking_idealized": 0.5916666388511658,
715
+ "ranking_idealized_expo": 0.5041666626930237,
716
+ "ranking_simple": 0.5166666507720947,
717
+ "regularize": 0.3612505793571472,
718
+ "step": 210
719
+ },
720
+ {
721
+ "dpo_loss": 0.6641567349433899,
722
+ "epoch": 0.8123425692695214,
723
+ "grad_norm": 50.968568724158025,
724
+ "learning_rate": 7.394265759500347e-08,
725
+ "logits": -1.3541967868804932,
726
+ "logps": -92.59414672851562,
727
+ "loss": 1.0263,
728
+ "objective": 1.0428659915924072,
729
+ "ranking_idealized": 0.5791666507720947,
730
+ "ranking_idealized_expo": 0.5,
731
+ "ranking_simple": 0.5,
732
+ "regularize": 0.3787091076374054,
733
+ "step": 215
734
+ },
735
+ {
736
+ "dpo_loss": 0.6569988131523132,
737
+ "epoch": 0.8312342569269522,
738
+ "grad_norm": 53.082634877219036,
739
+ "learning_rate": 7.247825876612352e-08,
740
+ "logits": -1.4405725002288818,
741
+ "logps": -93.02523803710938,
742
+ "loss": 1.0183,
743
+ "objective": 1.0499581098556519,
744
+ "ranking_idealized": 0.6458333134651184,
745
+ "ranking_idealized_expo": 0.5416666865348816,
746
+ "ranking_simple": 0.5458333492279053,
747
+ "regularize": 0.3929591476917267,
748
+ "step": 220
749
+ },
750
+ {
751
+ "dpo_loss": 0.6647858023643494,
752
+ "epoch": 0.8501259445843828,
753
+ "grad_norm": 51.335148967482006,
754
+ "learning_rate": 7.098928029130528e-08,
755
+ "logits": -1.4285895824432373,
756
+ "logps": -92.1469955444336,
757
+ "loss": 1.0123,
758
+ "objective": 1.034002423286438,
759
+ "ranking_idealized": 0.6041666865348816,
760
+ "ranking_idealized_expo": 0.5291666388511658,
761
+ "ranking_simple": 0.5416666865348816,
762
+ "regularize": 0.3692165017127991,
763
+ "step": 225
764
+ },
765
+ {
766
+ "dpo_loss": 0.6575040221214294,
767
+ "epoch": 0.8690176322418136,
768
+ "grad_norm": 48.651529158644536,
769
+ "learning_rate": 6.947735034665001e-08,
770
+ "logits": -1.4194341897964478,
771
+ "logps": -92.3643798828125,
772
+ "loss": 1.0129,
773
+ "objective": 1.019044041633606,
774
+ "ranking_idealized": 0.6583333611488342,
775
+ "ranking_idealized_expo": 0.5666666626930237,
776
+ "ranking_simple": 0.5791666507720947,
777
+ "regularize": 0.3615398705005646,
778
+ "step": 230
779
+ },
780
+ {
781
+ "dpo_loss": 0.6606828570365906,
782
+ "epoch": 0.8879093198992444,
783
+ "grad_norm": 48.73605551669952,
784
+ "learning_rate": 6.794412220535425e-08,
785
+ "logits": -1.4252243041992188,
786
+ "logps": -91.66600799560547,
787
+ "loss": 1.0227,
788
+ "objective": 1.0363646745681763,
789
+ "ranking_idealized": 0.6791666746139526,
790
+ "ranking_idealized_expo": 0.574999988079071,
791
+ "ranking_simple": 0.5958333611488342,
792
+ "regularize": 0.375681608915329,
793
+ "step": 235
794
+ },
795
+ {
796
+ "dpo_loss": 0.661056399345398,
797
+ "epoch": 0.906801007556675,
798
+ "grad_norm": 48.248478909206035,
799
+ "learning_rate": 6.639127242987987e-08,
800
+ "logits": -1.4544317722320557,
801
+ "logps": -92.43217468261719,
802
+ "loss": 1.0019,
803
+ "objective": 1.0119422674179077,
804
+ "ranking_idealized": 0.5708333253860474,
805
+ "ranking_idealized_expo": 0.5041666626930237,
806
+ "ranking_simple": 0.5083333253860474,
807
+ "regularize": 0.3508855998516083,
808
+ "step": 240
809
+ },
810
+ {
811
+ "dpo_loss": 0.6545276045799255,
812
+ "epoch": 0.9256926952141058,
813
+ "grad_norm": 55.62927449157849,
814
+ "learning_rate": 6.482049903865769e-08,
815
+ "logits": -1.3181027173995972,
816
+ "logps": -92.43726348876953,
817
+ "loss": 1.0099,
818
+ "objective": 0.9958193302154541,
819
+ "ranking_idealized": 0.5874999761581421,
820
+ "ranking_idealized_expo": 0.4958333373069763,
821
+ "ranking_simple": 0.5208333134651184,
822
+ "regularize": 0.341291606426239,
823
+ "step": 245
824
+ },
825
+ {
826
+ "dpo_loss": 0.6624430418014526,
827
+ "epoch": 0.9445843828715366,
828
+ "grad_norm": 45.52279742153591,
829
+ "learning_rate": 6.323351964932908e-08,
830
+ "logits": -1.363103985786438,
831
+ "logps": -91.6707992553711,
832
+ "loss": 1.0173,
833
+ "objective": 1.0481780767440796,
834
+ "ranking_idealized": 0.6083333492279053,
835
+ "ranking_idealized_expo": 0.5249999761581421,
836
+ "ranking_simple": 0.5416666865348816,
837
+ "regularize": 0.3857349753379822,
838
+ "step": 250
839
+ },
840
+ {
841
+ "epoch": 0.9445843828715366,
842
+ "eval_dpo_loss": 0.6845124959945679,
843
+ "eval_logits": -1.3585822582244873,
844
+ "eval_logps": -99.33744812011719,
845
+ "eval_loss": 1.0862815380096436,
846
+ "eval_objective": 1.0919297933578491,
847
+ "eval_ranking_idealized": 0.5925414562225342,
848
+ "eval_ranking_idealized_expo": 0.5165745615959167,
849
+ "eval_ranking_simple": 0.519336998462677,
850
+ "eval_regularize": 0.40741726756095886,
851
+ "eval_runtime": 235.789,
852
+ "eval_samples_per_second": 24.556,
853
+ "eval_steps_per_second": 1.535,
854
+ "step": 250
855
+ },
856
+ {
857
+ "dpo_loss": 0.6532657742500305,
858
+ "epoch": 0.9634760705289672,
859
+ "grad_norm": 51.00987164026873,
860
+ "learning_rate": 6.163206960055652e-08,
861
+ "logits": -1.460981011390686,
862
+ "logps": -94.54005432128906,
863
+ "loss": 1.0046,
864
+ "objective": 1.0130220651626587,
865
+ "ranking_idealized": 0.5874999761581421,
866
+ "ranking_idealized_expo": 0.5166666507720947,
867
+ "ranking_simple": 0.5333333611488342,
868
+ "regularize": 0.359756201505661,
869
+ "step": 255
870
+ },
871
+ {
872
+ "dpo_loss": 0.6577326059341431,
873
+ "epoch": 0.982367758186398,
874
+ "grad_norm": 50.576016219867064,
875
+ "learning_rate": 6.001790005445606e-08,
876
+ "logits": -1.362568974494934,
877
+ "logps": -93.19829559326172,
878
+ "loss": 1.01,
879
+ "objective": 0.9904889464378357,
880
+ "ranking_idealized": 0.6208333373069763,
881
+ "ranking_idealized_expo": 0.550000011920929,
882
+ "ranking_simple": 0.5625,
883
+ "regularize": 0.33275625109672546,
884
+ "step": 260
885
+ },
886
+ {
887
+ "dpo_loss": 0.6527519226074219,
888
+ "epoch": 1.0012594458438286,
889
+ "grad_norm": 50.78293400345161,
890
+ "learning_rate": 5.839277608172738e-08,
891
+ "logits": -1.3711644411087036,
892
+ "logps": -93.6723403930664,
893
+ "loss": 1.0071,
894
+ "objective": 1.026816487312317,
895
+ "ranking_idealized": 0.5958333611488342,
896
+ "ranking_idealized_expo": 0.5166666507720947,
897
+ "ranking_simple": 0.5333333611488342,
898
+ "regularize": 0.3740644156932831,
899
+ "step": 265
900
+ },
901
+ {
902
+ "dpo_loss": 0.6553524136543274,
903
+ "epoch": 1.0201511335012594,
904
+ "grad_norm": 52.08673841750398,
905
+ "learning_rate": 5.675847473157485e-08,
906
+ "logits": -1.4852278232574463,
907
+ "logps": -94.36783599853516,
908
+ "loss": 0.9818,
909
+ "objective": 0.9954887628555298,
910
+ "ranking_idealized": 0.5833333134651184,
911
+ "ranking_idealized_expo": 0.5375000238418579,
912
+ "ranking_simple": 0.5416666865348816,
913
+ "regularize": 0.34013625979423523,
914
+ "step": 270
915
+ },
916
+ {
917
+ "dpo_loss": 0.6415101289749146,
918
+ "epoch": 1.0390428211586902,
919
+ "grad_norm": 44.617567180749866,
920
+ "learning_rate": 5.511678308853025e-08,
921
+ "logits": -1.4626718759536743,
922
+ "logps": -93.63671112060547,
923
+ "loss": 0.9878,
924
+ "objective": 0.9779664278030396,
925
+ "ranking_idealized": 0.5958333611488342,
926
+ "ranking_idealized_expo": 0.512499988079071,
927
+ "ranking_simple": 0.5333333611488342,
928
+ "regularize": 0.33645620942115784,
929
+ "step": 275
930
+ },
931
+ {
932
+ "dpo_loss": 0.6500685811042786,
933
+ "epoch": 1.057934508816121,
934
+ "grad_norm": 47.25187912639425,
935
+ "learning_rate": 5.3469496318302197e-08,
936
+ "logits": -1.40652334690094,
937
+ "logps": -92.5146255493164,
938
+ "loss": 0.9659,
939
+ "objective": 0.9707435369491577,
940
+ "ranking_idealized": 0.5708333253860474,
941
+ "ranking_idealized_expo": 0.5041666626930237,
942
+ "ranking_simple": 0.5208333134651184,
943
+ "regularize": 0.3206748366355896,
944
+ "step": 280
945
+ },
946
+ {
947
+ "dpo_loss": 0.6382409930229187,
948
+ "epoch": 1.0768261964735517,
949
+ "grad_norm": 47.30764627956772,
950
+ "learning_rate": 5.1818415704788724e-08,
951
+ "logits": -1.4587684869766235,
952
+ "logps": -94.23441314697266,
953
+ "loss": 0.964,
954
+ "objective": 0.9715672135353088,
955
+ "ranking_idealized": 0.6666666865348816,
956
+ "ranking_idealized_expo": 0.5874999761581421,
957
+ "ranking_simple": 0.6083333492279053,
958
+ "regularize": 0.3333263099193573,
959
+ "step": 285
960
+ },
961
+ {
962
+ "dpo_loss": 0.6455419063568115,
963
+ "epoch": 1.0957178841309823,
964
+ "grad_norm": 45.41530383871643,
965
+ "learning_rate": 5.016534668039975e-08,
966
+ "logits": -1.3359150886535645,
967
+ "logps": -91.27457427978516,
968
+ "loss": 0.9783,
969
+ "objective": 0.9912244081497192,
970
+ "ranking_idealized": 0.6208333373069763,
971
+ "ranking_idealized_expo": 0.5375000238418579,
972
+ "ranking_simple": 0.5625,
973
+ "regularize": 0.3456825315952301,
974
+ "step": 290
975
+ },
976
+ {
977
+ "dpo_loss": 0.6454907655715942,
978
+ "epoch": 1.114609571788413,
979
+ "grad_norm": 52.660264254351745,
980
+ "learning_rate": 4.8512096851843375e-08,
981
+ "logits": -1.414397120475769,
982
+ "logps": -92.17057037353516,
983
+ "loss": 0.9657,
984
+ "objective": 0.964141309261322,
985
+ "ranking_idealized": 0.625,
986
+ "ranking_idealized_expo": 0.5625,
987
+ "ranking_simple": 0.5708333253860474,
988
+ "regularize": 0.31865036487579346,
989
+ "step": 295
990
+ },
991
+ {
992
+ "dpo_loss": 0.6478288173675537,
993
+ "epoch": 1.1335012594458438,
994
+ "grad_norm": 55.18319202572926,
995
+ "learning_rate": 4.686047402353433e-08,
996
+ "logits": -1.482143521308899,
997
+ "logps": -93.26191711425781,
998
+ "loss": 0.9755,
999
+ "objective": 0.9665765762329102,
1000
+ "ranking_idealized": 0.612500011920929,
1001
+ "ranking_idealized_expo": 0.5375000238418579,
1002
+ "ranking_simple": 0.5583333373069763,
1003
+ "regularize": 0.31874755024909973,
1004
+ "step": 300
1005
+ },
1006
+ {
1007
+ "epoch": 1.1335012594458438,
1008
+ "eval_dpo_loss": 0.6826499700546265,
1009
+ "eval_logits": -1.3663146495819092,
1010
+ "eval_logps": -99.1143798828125,
1011
+ "eval_loss": 1.0829436779022217,
1012
+ "eval_objective": 1.08933687210083,
1013
+ "eval_ranking_idealized": 0.5925414562225342,
1014
+ "eval_ranking_idealized_expo": 0.5165745615959167,
1015
+ "eval_ranking_simple": 0.5214088559150696,
1016
+ "eval_regularize": 0.40668678283691406,
1017
+ "eval_runtime": 235.9585,
1018
+ "eval_samples_per_second": 24.538,
1019
+ "eval_steps_per_second": 1.534,
1020
+ "step": 300
1021
+ },
1022
+ {
1023
+ "dpo_loss": 0.6493499875068665,
1024
+ "epoch": 1.1523929471032746,
1025
+ "grad_norm": 45.82300619637504,
1026
+ "learning_rate": 4.521228422078649e-08,
1027
+ "logits": -1.386905550956726,
1028
+ "logps": -92.7632064819336,
1029
+ "loss": 0.9773,
1030
+ "objective": 0.9832318425178528,
1031
+ "ranking_idealized": 0.6041666865348816,
1032
+ "ranking_idealized_expo": 0.512499988079071,
1033
+ "ranking_simple": 0.5375000238418579,
1034
+ "regularize": 0.33388179540634155,
1035
+ "step": 305
1036
+ },
1037
+ {
1038
+ "dpo_loss": 0.643381655216217,
1039
+ "epoch": 1.1712846347607053,
1040
+ "grad_norm": 43.431490164469025,
1041
+ "learning_rate": 4.3569329714950706e-08,
1042
+ "logits": -1.4345782995224,
1043
+ "logps": -92.81407928466797,
1044
+ "loss": 0.9734,
1045
+ "objective": 0.9835326671600342,
1046
+ "ranking_idealized": 0.625,
1047
+ "ranking_idealized_expo": 0.5458333492279053,
1048
+ "ranking_simple": 0.574999988079071,
1049
+ "regularize": 0.3401508927345276,
1050
+ "step": 310
1051
+ },
1052
+ {
1053
+ "dpo_loss": 0.6410504579544067,
1054
+ "epoch": 1.190176322418136,
1055
+ "grad_norm": 47.92691613928351,
1056
+ "learning_rate": 4.1933407052657454e-08,
1057
+ "logits": -1.406593918800354,
1058
+ "logps": -94.20066833496094,
1059
+ "loss": 0.9734,
1060
+ "objective": 0.9850514531135559,
1061
+ "ranking_idealized": 0.5333333611488342,
1062
+ "ranking_idealized_expo": 0.4625000059604645,
1063
+ "ranking_simple": 0.4791666567325592,
1064
+ "regularize": 0.3440004587173462,
1065
+ "step": 315
1066
+ },
1067
+ {
1068
+ "dpo_loss": 0.6493880748748779,
1069
+ "epoch": 1.2090680100755669,
1070
+ "grad_norm": 49.39982876479276,
1071
+ "learning_rate": 4.030630509131959e-08,
1072
+ "logits": -1.4532923698425293,
1073
+ "logps": -92.71504974365234,
1074
+ "loss": 0.9679,
1075
+ "objective": 0.961419939994812,
1076
+ "ranking_idealized": 0.574999988079071,
1077
+ "ranking_idealized_expo": 0.5166666507720947,
1078
+ "ranking_simple": 0.5375000238418579,
1079
+ "regularize": 0.31203168630599976,
1080
+ "step": 320
1081
+ },
1082
+ {
1083
+ "dpo_loss": 0.6389760971069336,
1084
+ "epoch": 1.2279596977329974,
1085
+ "grad_norm": 50.23647074169398,
1086
+ "learning_rate": 3.8689803043042996e-08,
1087
+ "logits": -1.3649462461471558,
1088
+ "logps": -94.10885620117188,
1089
+ "loss": 0.9416,
1090
+ "objective": 0.9290289282798767,
1091
+ "ranking_idealized": 0.6041666865348816,
1092
+ "ranking_idealized_expo": 0.5416666865348816,
1093
+ "ranking_simple": 0.5625,
1094
+ "regularize": 0.29005277156829834,
1095
+ "step": 325
1096
+ },
1097
+ {
1098
+ "dpo_loss": 0.6403206586837769,
1099
+ "epoch": 1.2468513853904282,
1100
+ "grad_norm": 48.23351344804798,
1101
+ "learning_rate": 3.708566852908418e-08,
1102
+ "logits": -1.4762166738510132,
1103
+ "logps": -92.44864654541016,
1104
+ "loss": 0.9562,
1105
+ "objective": 0.942139744758606,
1106
+ "ranking_idealized": 0.5916666388511658,
1107
+ "ranking_idealized_expo": 0.5375000238418579,
1108
+ "ranking_simple": 0.5416666865348816,
1109
+ "regularize": 0.30181899666786194,
1110
+ "step": 330
1111
+ },
1112
+ {
1113
+ "dpo_loss": 0.6417604088783264,
1114
+ "epoch": 1.265743073047859,
1115
+ "grad_norm": 54.445785967306286,
1116
+ "learning_rate": 3.54956556469825e-08,
1117
+ "logits": -1.446532130241394,
1118
+ "logps": -92.52754974365234,
1119
+ "loss": 0.9746,
1120
+ "objective": 0.9908974766731262,
1121
+ "ranking_idealized": 0.5666666626930237,
1122
+ "ranking_idealized_expo": 0.48750001192092896,
1123
+ "ranking_simple": 0.5041666626930237,
1124
+ "regularize": 0.3491368889808655,
1125
+ "step": 335
1126
+ },
1127
+ {
1128
+ "dpo_loss": 0.6481935977935791,
1129
+ "epoch": 1.2846347607052897,
1130
+ "grad_norm": 43.30323866491239,
1131
+ "learning_rate": 3.392150305248024e-08,
1132
+ "logits": -1.416210651397705,
1133
+ "logps": -92.2144775390625,
1134
+ "loss": 0.9605,
1135
+ "objective": 0.9179424047470093,
1136
+ "ranking_idealized": 0.5791666507720947,
1137
+ "ranking_idealized_expo": 0.5083333253860474,
1138
+ "ranking_simple": 0.5249999761581421,
1139
+ "regularize": 0.2697486877441406,
1140
+ "step": 340
1141
+ },
1142
+ {
1143
+ "dpo_loss": 0.6473791599273682,
1144
+ "epoch": 1.3035264483627205,
1145
+ "grad_norm": 46.63436114142443,
1146
+ "learning_rate": 3.236493205832794e-08,
1147
+ "logits": -1.4570471048355103,
1148
+ "logps": -93.01604461669922,
1149
+ "loss": 0.9614,
1150
+ "objective": 0.9653963446617126,
1151
+ "ranking_idealized": 0.5541666746139526,
1152
+ "ranking_idealized_expo": 0.4749999940395355,
1153
+ "ranking_simple": 0.4958333373069763,
1154
+ "regularize": 0.3180171847343445,
1155
+ "step": 345
1156
+ },
1157
+ {
1158
+ "dpo_loss": 0.6448249816894531,
1159
+ "epoch": 1.322418136020151,
1160
+ "grad_norm": 48.68063353112031,
1161
+ "learning_rate": 3.082764475205442e-08,
1162
+ "logits": -1.4262375831604004,
1163
+ "logps": -92.78704071044922,
1164
+ "loss": 0.9708,
1165
+ "objective": 0.9741830229759216,
1166
+ "ranking_idealized": 0.6041666865348816,
1167
+ "ranking_idealized_expo": 0.5375000238418579,
1168
+ "ranking_simple": 0.550000011920929,
1169
+ "regularize": 0.32935774326324463,
1170
+ "step": 350
1171
+ },
1172
+ {
1173
+ "epoch": 1.322418136020151,
1174
+ "eval_dpo_loss": 0.6832794547080994,
1175
+ "eval_logits": -1.3642104864120483,
1176
+ "eval_logps": -99.26692962646484,
1177
+ "eval_loss": 1.082851767539978,
1178
+ "eval_objective": 1.0900707244873047,
1179
+ "eval_ranking_idealized": 0.5925414562225342,
1180
+ "eval_ranking_idealized_expo": 0.5165745615959167,
1181
+ "eval_ranking_simple": 0.5220994353294373,
1182
+ "eval_regularize": 0.4067910611629486,
1183
+ "eval_runtime": 235.8313,
1184
+ "eval_samples_per_second": 24.551,
1185
+ "eval_steps_per_second": 1.535,
1186
+ "step": 350
1187
+ },
1188
+ {
1189
+ "dpo_loss": 0.6446579098701477,
1190
+ "epoch": 1.341309823677582,
1191
+ "grad_norm": 44.218673755410634,
1192
+ "learning_rate": 2.9311322134758836e-08,
1193
+ "logits": -1.4359955787658691,
1194
+ "logps": -91.50603485107422,
1195
+ "loss": 0.9673,
1196
+ "objective": 0.991865873336792,
1197
+ "ranking_idealized": 0.5916666388511658,
1198
+ "ranking_idealized_expo": 0.5083333253860474,
1199
+ "ranking_simple": 0.5416666865348816,
1200
+ "regularize": 0.3472079634666443,
1201
+ "step": 355
1202
+ },
1203
+ {
1204
+ "dpo_loss": 0.6347137689590454,
1205
+ "epoch": 1.3602015113350125,
1206
+ "grad_norm": 46.29947549071951,
1207
+ "learning_rate": 2.7817622282960812e-08,
1208
+ "logits": -1.430161714553833,
1209
+ "logps": -94.7589340209961,
1210
+ "loss": 0.9565,
1211
+ "objective": 0.9400501251220703,
1212
+ "ranking_idealized": 0.5666666626930237,
1213
+ "ranking_idealized_expo": 0.5083333253860474,
1214
+ "ranking_simple": 0.5291666388511658,
1215
+ "regularize": 0.3053361773490906,
1216
+ "step": 360
1217
+ },
1218
+ {
1219
+ "dpo_loss": 0.6405820250511169,
1220
+ "epoch": 1.3790931989924433,
1221
+ "grad_norm": 47.287766896204,
1222
+ "learning_rate": 2.6348178535517962e-08,
1223
+ "logits": -1.4314912557601929,
1224
+ "logps": -93.3971939086914,
1225
+ "loss": 0.9446,
1226
+ "objective": 0.9586126208305359,
1227
+ "ranking_idealized": 0.5958333611488342,
1228
+ "ranking_idealized_expo": 0.5208333134651184,
1229
+ "ranking_simple": 0.550000011920929,
1230
+ "regularize": 0.31803005933761597,
1231
+ "step": 365
1232
+ },
1233
+ {
1234
+ "dpo_loss": 0.6399816274642944,
1235
+ "epoch": 1.397984886649874,
1236
+ "grad_norm": 47.51527080377652,
1237
+ "learning_rate": 2.4904597707593977e-08,
1238
+ "logits": -1.4040166139602661,
1239
+ "logps": -93.57247924804688,
1240
+ "loss": 0.9651,
1241
+ "objective": 0.9659475088119507,
1242
+ "ranking_idealized": 0.5916666388511658,
1243
+ "ranking_idealized_expo": 0.5249999761581421,
1244
+ "ranking_simple": 0.5208333134651184,
1245
+ "regularize": 0.32596567273139954,
1246
+ "step": 370
1247
+ },
1248
+ {
1249
+ "dpo_loss": 0.6420871615409851,
1250
+ "epoch": 1.4168765743073048,
1251
+ "grad_norm": 45.177183298812,
1252
+ "learning_rate": 2.3488458333629773e-08,
1253
+ "logits": -1.362334966659546,
1254
+ "logps": -94.35677337646484,
1255
+ "loss": 0.9574,
1256
+ "objective": 0.9662178754806519,
1257
+ "ranking_idealized": 0.6166666746139526,
1258
+ "ranking_idealized_expo": 0.4749999940395355,
1259
+ "ranking_simple": 0.5166666507720947,
1260
+ "regularize": 0.3241305351257324,
1261
+ "step": 375
1262
+ },
1263
+ {
1264
+ "dpo_loss": 0.6476179957389832,
1265
+ "epoch": 1.4357682619647356,
1266
+ "grad_norm": 42.51629065900929,
1267
+ "learning_rate": 2.21013089412392e-08,
1268
+ "logits": -1.4462828636169434,
1269
+ "logps": -93.31849670410156,
1270
+ "loss": 0.9686,
1271
+ "objective": 0.9625572562217712,
1272
+ "ranking_idealized": 0.6499999761581421,
1273
+ "ranking_idealized_expo": 0.5791666507720947,
1274
+ "ranking_simple": 0.6000000238418579,
1275
+ "regularize": 0.31493937969207764,
1276
+ "step": 380
1277
+ },
1278
+ {
1279
+ "dpo_loss": 0.6483107209205627,
1280
+ "epoch": 1.4546599496221662,
1281
+ "grad_norm": 50.494940681761314,
1282
+ "learning_rate": 2.0744666357916923e-08,
1283
+ "logits": -1.3696460723876953,
1284
+ "logps": -93.16974639892578,
1285
+ "loss": 0.9674,
1286
+ "objective": 0.9437763690948486,
1287
+ "ranking_idealized": 0.5625,
1288
+ "ranking_idealized_expo": 0.4625000059604645,
1289
+ "ranking_simple": 0.4791666567325592,
1290
+ "regularize": 0.2954654395580292,
1291
+ "step": 385
1292
+ },
1293
+ {
1294
+ "dpo_loss": 0.6387519836425781,
1295
+ "epoch": 1.473551637279597,
1296
+ "grad_norm": 48.0772224935248,
1297
+ "learning_rate": 1.942001405240979e-08,
1298
+ "logits": -1.4375274181365967,
1299
+ "logps": -93.4893798828125,
1300
+ "loss": 0.951,
1301
+ "objective": 0.9498422145843506,
1302
+ "ranking_idealized": 0.6166666746139526,
1303
+ "ranking_idealized_expo": 0.5375000238418579,
1304
+ "ranking_simple": 0.5666666626930237,
1305
+ "regularize": 0.3110901117324829,
1306
+ "step": 390
1307
+ },
1308
+ {
1309
+ "dpo_loss": 0.6425871253013611,
1310
+ "epoch": 1.4924433249370277,
1311
+ "grad_norm": 50.00168420726503,
1312
+ "learning_rate": 1.8128800512565513e-08,
1313
+ "logits": -1.4277968406677246,
1314
+ "logps": -92.17271423339844,
1315
+ "loss": 0.9714,
1316
+ "objective": 0.9966804385185242,
1317
+ "ranking_idealized": 0.6083333492279053,
1318
+ "ranking_idealized_expo": 0.4958333373069763,
1319
+ "ranking_simple": 0.5041666626930237,
1320
+ "regularize": 0.3540932834148407,
1321
+ "step": 395
1322
+ },
1323
+ {
1324
+ "dpo_loss": 0.6415828466415405,
1325
+ "epoch": 1.5113350125944585,
1326
+ "grad_norm": 47.59483679412131,
1327
+ "learning_rate": 1.6872437661432515e-08,
1328
+ "logits": -1.4341318607330322,
1329
+ "logps": -94.33455657958984,
1330
+ "loss": 0.968,
1331
+ "objective": 0.9913274049758911,
1332
+ "ranking_idealized": 0.6000000238418579,
1333
+ "ranking_idealized_expo": 0.5166666507720947,
1334
+ "ranking_simple": 0.5458333492279053,
1335
+ "regularize": 0.3497445285320282,
1336
+ "step": 400
1337
+ },
1338
+ {
1339
+ "epoch": 1.5113350125944585,
1340
+ "eval_dpo_loss": 0.6832301616668701,
1341
+ "eval_logits": -1.3681507110595703,
1342
+ "eval_logps": -99.251953125,
1343
+ "eval_loss": 1.0828720331192017,
1344
+ "eval_objective": 1.090120792388916,
1345
+ "eval_ranking_idealized": 0.5925414562225342,
1346
+ "eval_ranking_idealized_expo": 0.5165745615959167,
1347
+ "eval_ranking_simple": 0.5214088559150696,
1348
+ "eval_regularize": 0.4068906307220459,
1349
+ "eval_runtime": 235.2584,
1350
+ "eval_samples_per_second": 24.611,
1351
+ "eval_steps_per_second": 1.539,
1352
+ "step": 400
1353
+ },
1354
+ {
1355
+ "dpo_loss": 0.645163357257843,
1356
+ "epoch": 1.5302267002518892,
1357
+ "grad_norm": 47.34605640737913,
1358
+ "learning_rate": 1.5652299313342772e-08,
1359
+ "logits": -1.4498835802078247,
1360
+ "logps": -94.09959411621094,
1361
+ "loss": 0.9622,
1362
+ "objective": 0.9742183089256287,
1363
+ "ranking_idealized": 0.5916666388511658,
1364
+ "ranking_idealized_expo": 0.5083333253860474,
1365
+ "ranking_simple": 0.5166666507720947,
1366
+ "regularize": 0.3290548026561737,
1367
+ "step": 405
1368
+ },
1369
+ {
1370
+ "dpo_loss": 0.6364944577217102,
1371
+ "epoch": 1.5491183879093198,
1372
+ "grad_norm": 49.26429643097485,
1373
+ "learning_rate": 1.4469719671666041e-08,
1374
+ "logits": -1.3392701148986816,
1375
+ "logps": -93.40227508544922,
1376
+ "loss": 0.9596,
1377
+ "objective": 0.970222532749176,
1378
+ "ranking_idealized": 0.612500011920929,
1379
+ "ranking_idealized_expo": 0.5208333134651184,
1380
+ "ranking_simple": 0.5625,
1381
+ "regularize": 0.33372804522514343,
1382
+ "step": 410
1383
+ },
1384
+ {
1385
+ "dpo_loss": 0.6261176466941833,
1386
+ "epoch": 1.5680100755667508,
1387
+ "grad_norm": 47.529393740735564,
1388
+ "learning_rate": 1.3325991869878012e-08,
1389
+ "logits": -1.398348093032837,
1390
+ "logps": -93.3853530883789,
1391
+ "loss": 0.9724,
1392
+ "objective": 0.9390360116958618,
1393
+ "ranking_idealized": 0.5291666388511658,
1394
+ "ranking_idealized_expo": 0.4583333432674408,
1395
+ "ranking_simple": 0.48750001192092896,
1396
+ "regularize": 0.3129182457923889,
1397
+ "step": 415
1398
+ },
1399
+ {
1400
+ "dpo_loss": 0.6377049088478088,
1401
+ "epoch": 1.5869017632241813,
1402
+ "grad_norm": 57.43753774126503,
1403
+ "learning_rate": 1.222236655753791e-08,
1404
+ "logits": -1.4465179443359375,
1405
+ "logps": -92.30918884277344,
1406
+ "loss": 0.9625,
1407
+ "objective": 0.9549995064735413,
1408
+ "ranking_idealized": 0.5791666507720947,
1409
+ "ranking_idealized_expo": 0.49166667461395264,
1410
+ "ranking_simple": 0.5166666507720947,
1411
+ "regularize": 0.31729447841644287,
1412
+ "step": 420
1413
+ },
1414
+ {
1415
+ "dpo_loss": 0.6409914493560791,
1416
+ "epoch": 1.605793450881612,
1417
+ "grad_norm": 48.69171129272466,
1418
+ "learning_rate": 1.1160050532721526e-08,
1419
+ "logits": -1.4412765502929688,
1420
+ "logps": -92.86591339111328,
1421
+ "loss": 0.9543,
1422
+ "objective": 0.9196628928184509,
1423
+ "ranking_idealized": 0.5625,
1424
+ "ranking_idealized_expo": 0.46666666865348816,
1425
+ "ranking_simple": 0.48750001192092896,
1426
+ "regularize": 0.27867138385772705,
1427
+ "step": 425
1428
+ },
1429
+ {
1430
+ "dpo_loss": 0.6387853622436523,
1431
+ "epoch": 1.6246851385390428,
1432
+ "grad_norm": 45.275230591779184,
1433
+ "learning_rate": 1.0140205422405212e-08,
1434
+ "logits": -1.4535937309265137,
1435
+ "logps": -92.8780517578125,
1436
+ "loss": 0.9509,
1437
+ "objective": 0.9649083018302917,
1438
+ "ranking_idealized": 0.6208333373069763,
1439
+ "ranking_idealized_expo": 0.574999988079071,
1440
+ "ranking_simple": 0.5791666507720947,
1441
+ "regularize": 0.3261227011680603,
1442
+ "step": 430
1443
+ },
1444
+ {
1445
+ "dpo_loss": 0.6288203001022339,
1446
+ "epoch": 1.6435768261964736,
1447
+ "grad_norm": 44.1758284644859,
1448
+ "learning_rate": 9.163946412243894e-09,
1449
+ "logits": -1.4540935754776,
1450
+ "logps": -93.5007553100586,
1451
+ "loss": 0.9616,
1452
+ "objective": 0.9557725191116333,
1453
+ "ranking_idealized": 0.6416666507720947,
1454
+ "ranking_idealized_expo": 0.5416666865348816,
1455
+ "ranking_simple": 0.5791666507720947,
1456
+ "regularize": 0.32695215940475464,
1457
+ "step": 435
1458
+ },
1459
+ {
1460
+ "dpo_loss": 0.6319211721420288,
1461
+ "epoch": 1.6624685138539044,
1462
+ "grad_norm": 47.357355663292644,
1463
+ "learning_rate": 8.232341027131883e-09,
1464
+ "logits": -1.4391558170318604,
1465
+ "logps": -92.56800079345703,
1466
+ "loss": 0.9484,
1467
+ "objective": 0.9231205582618713,
1468
+ "ranking_idealized": 0.5874999761581421,
1469
+ "ranking_idealized_expo": 0.5375000238418579,
1470
+ "ranking_simple": 0.5583333373069763,
1471
+ "regularize": 0.2911991775035858,
1472
+ "step": 440
1473
+ },
1474
+ {
1475
+ "dpo_loss": 0.6425955295562744,
1476
+ "epoch": 1.681360201511335,
1477
+ "grad_norm": 46.443814611627765,
1478
+ "learning_rate": 7.346407963880136e-09,
1479
+ "logits": -1.4678118228912354,
1480
+ "logps": -92.27074432373047,
1481
+ "loss": 0.9664,
1482
+ "objective": 0.949626624584198,
1483
+ "ranking_idealized": 0.6041666865348816,
1484
+ "ranking_idealized_expo": 0.5416666865348816,
1485
+ "ranking_simple": 0.5541666746139526,
1486
+ "regularize": 0.3070310652256012,
1487
+ "step": 445
1488
+ },
1489
+ {
1490
+ "dpo_loss": 0.6390999555587769,
1491
+ "epoch": 1.700251889168766,
1492
+ "grad_norm": 49.0414360352724,
1493
+ "learning_rate": 6.507115977286143e-09,
1494
+ "logits": -1.393315076828003,
1495
+ "logps": -93.02438354492188,
1496
+ "loss": 0.9495,
1497
+ "objective": 0.9575169682502747,
1498
+ "ranking_idealized": 0.6083333492279053,
1499
+ "ranking_idealized_expo": 0.5291666388511658,
1500
+ "ranking_simple": 0.5541666746139526,
1501
+ "regularize": 0.31841692328453064,
1502
+ "step": 450
1503
+ },
1504
+ {
1505
+ "epoch": 1.700251889168766,
1506
+ "eval_dpo_loss": 0.6830143928527832,
1507
+ "eval_logits": -1.3688236474990845,
1508
+ "eval_logps": -99.32157135009766,
1509
+ "eval_loss": 1.0825797319412231,
1510
+ "eval_objective": 1.0900212526321411,
1511
+ "eval_ranking_idealized": 0.5925414562225342,
1512
+ "eval_ranking_idealized_expo": 0.5165745615959167,
1513
+ "eval_ranking_simple": 0.5214088559150696,
1514
+ "eval_regularize": 0.4070066511631012,
1515
+ "eval_runtime": 235.5467,
1516
+ "eval_samples_per_second": 24.581,
1517
+ "eval_steps_per_second": 1.537,
1518
+ "step": 450
1519
+ },
1520
+ {
1521
+ "dpo_loss": 0.6333620548248291,
1522
+ "epoch": 1.7191435768261965,
1523
+ "grad_norm": 51.27723544337059,
1524
+ "learning_rate": 5.715382820814885e-09,
1525
+ "logits": -1.463624358177185,
1526
+ "logps": -95.06629180908203,
1527
+ "loss": 0.9525,
1528
+ "objective": 0.981126070022583,
1529
+ "ranking_idealized": 0.5916666388511658,
1530
+ "ranking_idealized_expo": 0.5333333611488342,
1531
+ "ranking_simple": 0.5541666746139526,
1532
+ "regularize": 0.34776392579078674,
1533
+ "step": 455
1534
+ },
1535
+ {
1536
+ "dpo_loss": 0.6484085321426392,
1537
+ "epoch": 1.7380352644836272,
1538
+ "grad_norm": 46.95616030148672,
1539
+ "learning_rate": 4.972074243048896e-09,
1540
+ "logits": -1.4587913751602173,
1541
+ "logps": -91.8313217163086,
1542
+ "loss": 0.9569,
1543
+ "objective": 0.9492828845977783,
1544
+ "ranking_idealized": 0.6166666746139526,
1545
+ "ranking_idealized_expo": 0.5458333492279053,
1546
+ "ranking_simple": 0.5541666746139526,
1547
+ "regularize": 0.30087435245513916,
1548
+ "step": 460
1549
+ },
1550
+ {
1551
+ "dpo_loss": 0.6343809962272644,
1552
+ "epoch": 1.756926952141058,
1553
+ "grad_norm": 47.70310058715592,
1554
+ "learning_rate": 4.278003041004779e-09,
1555
+ "logits": -1.3960585594177246,
1556
+ "logps": -92.12299346923828,
1557
+ "loss": 0.9624,
1558
+ "objective": 0.9703618884086609,
1559
+ "ranking_idealized": 0.6166666746139526,
1560
+ "ranking_idealized_expo": 0.5333333611488342,
1561
+ "ranking_simple": 0.5458333492279053,
1562
+ "regularize": 0.33598068356513977,
1563
+ "step": 465
1564
+ },
1565
+ {
1566
+ "dpo_loss": 0.6355936527252197,
1567
+ "epoch": 1.7758186397984885,
1568
+ "grad_norm": 48.36448652264405,
1569
+ "learning_rate": 3.63392817135173e-09,
1570
+ "logits": -1.4002528190612793,
1571
+ "logps": -92.8111343383789,
1572
+ "loss": 0.9635,
1573
+ "objective": 0.9694804549217224,
1574
+ "ranking_idealized": 0.6041666865348816,
1575
+ "ranking_idealized_expo": 0.5208333134651184,
1576
+ "ranking_simple": 0.5416666865348816,
1577
+ "regularize": 0.3338867127895355,
1578
+ "step": 470
1579
+ },
1580
+ {
1581
+ "dpo_loss": 0.6456637978553772,
1582
+ "epoch": 1.7947103274559195,
1583
+ "grad_norm": 48.611216917206725,
1584
+ "learning_rate": 3.0405539205035023e-09,
1585
+ "logits": -1.4299951791763306,
1586
+ "logps": -92.8174819946289,
1587
+ "loss": 0.9564,
1588
+ "objective": 0.9828375577926636,
1589
+ "ranking_idealized": 0.6166666746139526,
1590
+ "ranking_idealized_expo": 0.5416666865348816,
1591
+ "ranking_simple": 0.5458333492279053,
1592
+ "regularize": 0.3371736407279968,
1593
+ "step": 475
1594
+ },
1595
+ {
1596
+ "dpo_loss": 0.6420384645462036,
1597
+ "epoch": 1.81360201511335,
1598
+ "grad_norm": 46.76089409325827,
1599
+ "learning_rate": 2.4985291344915673e-09,
1600
+ "logits": -1.4936844110488892,
1601
+ "logps": -93.58541870117188,
1602
+ "loss": 0.967,
1603
+ "objective": 0.9691051244735718,
1604
+ "ranking_idealized": 0.550000011920929,
1605
+ "ranking_idealized_expo": 0.4583333432674408,
1606
+ "ranking_simple": 0.46666666865348816,
1607
+ "regularize": 0.3270666301250458,
1608
+ "step": 480
1609
+ },
1610
+ {
1611
+ "dpo_loss": 0.6282381415367126,
1612
+ "epoch": 1.8324937027707808,
1613
+ "grad_norm": 52.789969586082115,
1614
+ "learning_rate": 2.0084465094614976e-09,
1615
+ "logits": -1.3527694940567017,
1616
+ "logps": -94.94033813476562,
1617
+ "loss": 0.9721,
1618
+ "objective": 0.9742304682731628,
1619
+ "ranking_idealized": 0.6041666865348816,
1620
+ "ranking_idealized_expo": 0.512499988079071,
1621
+ "ranking_simple": 0.5333333611488342,
1622
+ "regularize": 0.34599223732948303,
1623
+ "step": 485
1624
+ },
1625
+ {
1626
+ "dpo_loss": 0.6374282836914062,
1627
+ "epoch": 1.8513853904282116,
1628
+ "grad_norm": 47.2028496146643,
1629
+ "learning_rate": 1.570841943568446e-09,
1630
+ "logits": -1.430949091911316,
1631
+ "logps": -93.16008758544922,
1632
+ "loss": 0.9587,
1633
+ "objective": 0.9563023447990417,
1634
+ "ranking_idealized": 0.5874999761581421,
1635
+ "ranking_idealized_expo": 0.5166666507720947,
1636
+ "ranking_simple": 0.5416666865348816,
1637
+ "regularize": 0.3188740313053131,
1638
+ "step": 490
1639
+ },
1640
+ {
1641
+ "dpo_loss": 0.6446228623390198,
1642
+ "epoch": 1.8702770780856424,
1643
+ "grad_norm": 47.63126102742745,
1644
+ "learning_rate": 1.1861939509803687e-09,
1645
+ "logits": -1.4252774715423584,
1646
+ "logps": -93.47933959960938,
1647
+ "loss": 0.9498,
1648
+ "objective": 0.9525411128997803,
1649
+ "ranking_idealized": 0.5791666507720947,
1650
+ "ranking_idealized_expo": 0.4833333194255829,
1651
+ "ranking_simple": 0.512499988079071,
1652
+ "regularize": 0.30791813135147095,
1653
+ "step": 495
1654
+ },
1655
+ {
1656
+ "dpo_loss": 0.6408995985984802,
1657
+ "epoch": 1.8891687657430731,
1658
+ "grad_norm": 47.540323720333014,
1659
+ "learning_rate": 8.54923138629815e-10,
1660
+ "logits": -1.4415290355682373,
1661
+ "logps": -92.98784637451172,
1662
+ "loss": 0.9463,
1663
+ "objective": 0.9464013576507568,
1664
+ "ranking_idealized": 0.6041666865348816,
1665
+ "ranking_idealized_expo": 0.5333333611488342,
1666
+ "ranking_simple": 0.5541666746139526,
1667
+ "regularize": 0.3055017292499542,
1668
+ "step": 500
1669
+ },
1670
+ {
1671
+ "epoch": 1.8891687657430731,
1672
+ "eval_dpo_loss": 0.6829814314842224,
1673
+ "eval_logits": -1.3685972690582275,
1674
+ "eval_logps": -99.3604736328125,
1675
+ "eval_loss": 1.0826866626739502,
1676
+ "eval_objective": 1.0901830196380615,
1677
+ "eval_ranking_idealized": 0.5925414562225342,
1678
+ "eval_ranking_idealized_expo": 0.5165745615959167,
1679
+ "eval_ranking_simple": 0.5220994353294373,
1680
+ "eval_regularize": 0.40720152854919434,
1681
+ "eval_runtime": 235.7418,
1682
+ "eval_samples_per_second": 24.561,
1683
+ "eval_steps_per_second": 1.536,
1684
+ "step": 500
1685
+ },
1686
+ {
1687
+ "dpo_loss": 0.6330133080482483,
1688
+ "epoch": 1.9080604534005037,
1689
+ "grad_norm": 44.27516380619689,
1690
+ "learning_rate": 5.773917462864264e-10,
1691
+ "logits": -1.3644249439239502,
1692
+ "logps": -91.88590240478516,
1693
+ "loss": 0.954,
1694
+ "objective": 0.951960027217865,
1695
+ "ranking_idealized": 0.5791666507720947,
1696
+ "ranking_idealized_expo": 0.5249999761581421,
1697
+ "ranking_simple": 0.5375000238418579,
1698
+ "regularize": 0.3189466595649719,
1699
+ "step": 505
1700
+ },
1701
+ {
1702
+ "dpo_loss": 0.6388721466064453,
1703
+ "epoch": 1.9269521410579347,
1704
+ "grad_norm": 51.31132661100108,
1705
+ "learning_rate": 3.53903250453047e-10,
1706
+ "logits": -1.4425408840179443,
1707
+ "logps": -95.0220718383789,
1708
+ "loss": 0.9495,
1709
+ "objective": 0.9448812007904053,
1710
+ "ranking_idealized": 0.6625000238418579,
1711
+ "ranking_idealized_expo": 0.5874999761581421,
1712
+ "ranking_simple": 0.6000000238418579,
1713
+ "regularize": 0.30600884556770325,
1714
+ "step": 510
1715
+ },
1716
+ {
1717
+ "dpo_loss": 0.638128936290741,
1718
+ "epoch": 1.9458438287153652,
1719
+ "grad_norm": 49.79971218067957,
1720
+ "learning_rate": 1.8470203251865768e-10,
1721
+ "logits": -1.4008522033691406,
1722
+ "logps": -95.15757751464844,
1723
+ "loss": 0.9567,
1724
+ "objective": 0.9624292850494385,
1725
+ "ranking_idealized": 0.5833333134651184,
1726
+ "ranking_idealized_expo": 0.48750001192092896,
1727
+ "ranking_simple": 0.512499988079071,
1728
+ "regularize": 0.324300080537796,
1729
+ "step": 515
1730
+ },
1731
+ {
1732
+ "dpo_loss": 0.6457895636558533,
1733
+ "epoch": 1.964735516372796,
1734
+ "grad_norm": 55.67667071934356,
1735
+ "learning_rate": 6.997311153086882e-11,
1736
+ "logits": -1.464538335800171,
1737
+ "logps": -93.64386749267578,
1738
+ "loss": 0.9495,
1739
+ "objective": 0.9239124059677124,
1740
+ "ranking_idealized": 0.5916666388511658,
1741
+ "ranking_idealized_expo": 0.5583333373069763,
1742
+ "ranking_simple": 0.5708333253860474,
1743
+ "regularize": 0.2781226336956024,
1744
+ "step": 520
1745
+ },
1746
+ {
1747
+ "dpo_loss": 0.6334381103515625,
1748
+ "epoch": 1.9836272040302267,
1749
+ "grad_norm": 50.356520924506064,
1750
+ "learning_rate": 9.841941880361915e-12,
1751
+ "logits": -1.4272739887237549,
1752
+ "logps": -93.79329681396484,
1753
+ "loss": 0.966,
1754
+ "objective": 0.9610148072242737,
1755
+ "ranking_idealized": 0.5666666626930237,
1756
+ "ranking_idealized_expo": 0.512499988079071,
1757
+ "ranking_simple": 0.5375000238418579,
1758
+ "regularize": 0.3275766670703888,
1759
+ "step": 525
1760
+ },
1761
+ {
1762
+ "epoch": 1.994962216624685,
1763
+ "step": 528,
1764
+ "total_flos": 0.0,
1765
+ "train_loss": 1.010300681672313,
1766
+ "train_runtime": 13850.696,
1767
+ "train_samples_per_second": 7.336,
1768
+ "train_steps_per_second": 0.038
1769
+ }
1770
+ ],
1771
+ "logging_steps": 5,
1772
+ "max_steps": 528,
1773
+ "num_input_tokens_seen": 0,
1774
+ "num_train_epochs": 2,
1775
+ "save_steps": 500,
1776
+ "stateful_callbacks": {
1777
+ "TrainerControl": {
1778
+ "args": {
1779
+ "should_epoch_stop": false,
1780
+ "should_evaluate": false,
1781
+ "should_log": false,
1782
+ "should_save": false,
1783
+ "should_training_stop": false
1784
+ },
1785
+ "attributes": {}
1786
+ }
1787
+ },
1788
+ "total_flos": 0.0,
1789
+ "train_batch_size": 4,
1790
+ "trial_name": null,
1791
+ "trial_params": null
1792
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d09f1c8643b5d0b664484b913a39433594c6ba186f1809a4fd5760b2ab25ba8
3
+ size 8056
vocab.json ADDED
The diff for this file is too large to render. See raw diff