hZzy commited on
Commit
17e9f1f
·
verified ·
1 Parent(s): 0a1160a

Model save

Browse files
Files changed (5) hide show
  1. README.md +22 -29
  2. all_results.json +4 -18
  3. model.safetensors +1 -1
  4. train_results.json +4 -4
  5. trainer_state.json +206 -206
README.md CHANGED
@@ -2,16 +2,9 @@
2
  license: apache-2.0
3
  base_model: hZzy/qwen2.5-0.5b-sft-news-IFT
4
  tags:
5
- - alignment-handbook
6
- - ndcg
7
  - trl
8
  - expo
9
  - generated_from_trainer
10
- - trl
11
- - expo
12
- - generated_from_trainer
13
- datasets:
14
- - hZzy/train_pairwise
15
  model-index:
16
  - name: qwen2.5-0.5b-expo-DPO-ES-0.1
17
  results: []
@@ -20,21 +13,21 @@ model-index:
20
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
21
  should probably proofread and complete it, then remove this comment. -->
22
 
23
- [<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/w0nbtpl2)
24
  # qwen2.5-0.5b-expo-DPO-ES-0.1
25
 
26
- 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 the hZzy/train_pairwise dataset.
27
  It achieves the following results on the evaluation set:
28
- - Loss: 0.6923
29
- - Logps: -90.9491
30
- - Logits: -2.1209
31
- - Objective: 0.6894
32
- - Dpo Loss: 0.6894
33
- - Regularize: 0.6894
34
- - Ranking Simple: 0.5564
35
  - Ranking Idealized: 0.6030
36
  - Ranking Idealized Expo: 0.5223
37
- - Wo Beta: 7.4232
38
 
39
  ## Model description
40
 
@@ -71,22 +64,22 @@ The following hyperparameters were used during training:
71
 
72
  | Training Loss | Epoch | Step | Validation Loss | Logps | Logits | Objective | Dpo Loss | Regularize | Ranking Simple | Ranking Idealized | Ranking Idealized Expo | Wo Beta |
73
  |:-------------:|:------:|:----:|:---------------:|:---------:|:-------:|:---------:|:--------:|:----------:|:--------------:|:-----------------:|:----------------------:|:-------:|
74
- | 0.6785 | 0.1417 | 50 | 0.6814 | -90.8721 | -1.6022 | 0.6843 | 0.6843 | 0.6843 | 0.5259 | 0.6030 | 0.5223 | 7.8749 |
75
- | 0.618 | 0.2834 | 100 | 0.6733 | -98.8900 | -1.7799 | 0.6766 | 0.6766 | 0.6766 | 0.5399 | 0.6030 | 0.5223 | 7.7840 |
76
- | 0.5667 | 0.4251 | 150 | 0.6867 | -99.1217 | -1.8072 | 0.6829 | 0.6829 | 0.6829 | 0.5409 | 0.6030 | 0.5223 | 7.8537 |
77
- | 0.5214 | 0.5668 | 200 | 0.6902 | -99.5153 | -1.8895 | 0.6905 | 0.6905 | 0.6905 | 0.5445 | 0.6030 | 0.5223 | 7.7013 |
78
- | 0.4922 | 0.7085 | 250 | 0.6976 | -82.8384 | -1.9887 | 0.6914 | 0.6914 | 0.6914 | 0.5481 | 0.6030 | 0.5223 | 7.8784 |
79
- | 0.4535 | 0.8503 | 300 | 0.6923 | -90.9491 | -2.1209 | 0.6894 | 0.6894 | 0.6894 | 0.5564 | 0.6030 | 0.5223 | 7.4232 |
80
- | 0.4228 | 0.9920 | 350 | 0.7064 | -87.7231 | -1.9803 | 0.6968 | 0.6968 | 0.6968 | 0.5538 | 0.6030 | 0.5223 | 8.0253 |
81
- | 0.2845 | 1.1337 | 400 | 0.7305 | -101.3180 | -2.0805 | 0.7269 | 0.7269 | 0.7269 | 0.5430 | 0.6030 | 0.5223 | 8.6164 |
82
- | 0.2989 | 1.2754 | 450 | 0.7005 | -93.1955 | -1.8646 | 0.6974 | 0.6974 | 0.6974 | 0.5606 | 0.6030 | 0.5223 | 8.2386 |
83
- | 0.3065 | 1.4171 | 500 | 0.7179 | -97.0137 | -1.9983 | 0.7147 | 0.7147 | 0.7147 | 0.5549 | 0.6030 | 0.5223 | 8.2760 |
84
- | 0.2885 | 1.5588 | 550 | 0.7091 | -107.9610 | -1.9041 | 0.7134 | 0.7134 | 0.7134 | 0.5616 | 0.6030 | 0.5223 | 8.1968 |
85
 
86
 
87
  ### Framework versions
88
 
89
  - Transformers 4.42.0
90
  - Pytorch 2.3.0+cu121
91
- - Datasets 2.19.1
92
  - Tokenizers 0.19.1
 
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-ES-0.1
10
  results: []
 
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/hy7b2tcq)
17
  # qwen2.5-0.5b-expo-DPO-ES-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: 0.7044
22
+ - Logps: -104.2469
23
+ - Logits: -1.8870
24
+ - Objective: 0.7077
25
+ - Dpo Loss: 0.7077
26
+ - Regularize: 0.7077
27
+ - Ranking Simple: 0.5652
28
  - Ranking Idealized: 0.6030
29
  - Ranking Idealized Expo: 0.5223
30
+ - Wo Beta: 8.1947
31
 
32
  ## Model description
33
 
 
64
 
65
  | Training Loss | Epoch | Step | Validation Loss | Logps | Logits | Objective | Dpo Loss | Regularize | Ranking Simple | Ranking Idealized | Ranking Idealized Expo | Wo Beta |
66
  |:-------------:|:------:|:----:|:---------------:|:---------:|:-------:|:---------:|:--------:|:----------:|:--------------:|:-----------------:|:----------------------:|:-------:|
67
+ | 0.6785 | 0.1417 | 50 | 0.6813 | -90.8716 | -1.6022 | 0.6843 | 0.6843 | 0.6843 | 0.5259 | 0.6030 | 0.5223 | 7.8749 |
68
+ | 0.618 | 0.2834 | 100 | 0.6733 | -98.8899 | -1.7799 | 0.6766 | 0.6766 | 0.6766 | 0.5399 | 0.6030 | 0.5223 | 7.7840 |
69
+ | 0.5667 | 0.4251 | 150 | 0.6866 | -99.1230 | -1.8072 | 0.6829 | 0.6829 | 0.6829 | 0.5409 | 0.6030 | 0.5223 | 7.8533 |
70
+ | 0.5214 | 0.5668 | 200 | 0.6901 | -99.5388 | -1.8894 | 0.6904 | 0.6904 | 0.6904 | 0.5445 | 0.6030 | 0.5223 | 7.6995 |
71
+ | 0.4922 | 0.7085 | 250 | 0.6976 | -82.7973 | -1.9880 | 0.6916 | 0.6916 | 0.6916 | 0.5476 | 0.6030 | 0.5223 | 7.8790 |
72
+ | 0.4535 | 0.8503 | 300 | 0.6921 | -91.0309 | -2.1203 | 0.6893 | 0.6893 | 0.6893 | 0.5559 | 0.6030 | 0.5223 | 7.4262 |
73
+ | 0.423 | 0.9920 | 350 | 0.7057 | -88.1615 | -1.9880 | 0.6959 | 0.6959 | 0.6959 | 0.5549 | 0.6030 | 0.5223 | 7.9979 |
74
+ | 0.2847 | 1.1337 | 400 | 0.7315 | -101.6926 | -2.0862 | 0.7281 | 0.7281 | 0.7281 | 0.5424 | 0.6030 | 0.5223 | 8.6326 |
75
+ | 0.2991 | 1.2754 | 450 | 0.7008 | -92.7942 | -1.8470 | 0.6980 | 0.6980 | 0.6980 | 0.5621 | 0.6030 | 0.5223 | 8.2584 |
76
+ | 0.3065 | 1.4171 | 500 | 0.7180 | -96.6747 | -2.0065 | 0.7147 | 0.7147 | 0.7147 | 0.5554 | 0.6030 | 0.5223 | 8.2522 |
77
+ | 0.2895 | 1.5588 | 550 | 0.7044 | -104.2469 | -1.8870 | 0.7077 | 0.7077 | 0.7077 | 0.5652 | 0.6030 | 0.5223 | 8.1947 |
78
 
79
 
80
  ### Framework versions
81
 
82
  - Transformers 4.42.0
83
  - Pytorch 2.3.0+cu121
84
+ - Datasets 3.2.0
85
  - Tokenizers 0.19.1
all_results.json CHANGED
@@ -1,23 +1,9 @@
1
  {
2
  "epoch": 1.5588096362777515,
3
- "eval_dpo_loss": 0.6893911957740784,
4
- "eval_logits": -2.1208713054656982,
5
- "eval_logps": -90.9490966796875,
6
- "eval_loss": 0.6922824382781982,
7
- "eval_objective": 0.6893911957740784,
8
- "eval_ranking_idealized": 0.6030020713806152,
9
- "eval_ranking_idealized_expo": 0.5222567319869995,
10
- "eval_ranking_simple": 0.556418240070343,
11
- "eval_regularize": 0.6893911957740784,
12
- "eval_runtime": 314.8886,
13
- "eval_samples": 5790,
14
- "eval_samples_per_second": 18.387,
15
- "eval_steps_per_second": 1.534,
16
- "eval_wo_beta": 7.4231791496276855,
17
  "total_flos": 0.0,
18
- "train_loss": 0.4483427975394509,
19
- "train_runtime": 15087.4424,
20
  "train_samples": 50802,
21
- "train_samples_per_second": 16.836,
22
- "train_steps_per_second": 0.117
23
  }
 
1
  {
2
  "epoch": 1.5588096362777515,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  "total_flos": 0.0,
4
+ "train_loss": 0.4484944924441251,
5
+ "train_runtime": 15200.6621,
6
  "train_samples": 50802,
7
+ "train_samples_per_second": 16.71,
8
+ "train_steps_per_second": 0.116
9
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e42889b51298a05b37f983b9f4e2044f618425ed9654eee76d9c8ca3131def1c
3
  size 1975192208
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:654597aa1c17b0ef958f9135269de2f1d79597cdfbc53eea3e4411458602cc1b
3
  size 1975192208
train_results.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "epoch": 1.5588096362777515,
3
  "total_flos": 0.0,
4
- "train_loss": 0.4483427975394509,
5
- "train_runtime": 15087.4424,
6
  "train_samples": 50802,
7
- "train_samples_per_second": 16.836,
8
- "train_steps_per_second": 0.117
9
  }
 
1
  {
2
  "epoch": 1.5588096362777515,
3
  "total_flos": 0.0,
4
+ "train_loss": 0.4484944924441251,
5
+ "train_runtime": 15200.6621,
6
  "train_samples": 50802,
7
+ "train_samples_per_second": 16.71,
8
+ "train_steps_per_second": 0.116
9
  }
trainer_state.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "best_metric": 7.4231791496276855,
3
  "best_model_checkpoint": "./qwen2.5-0.5b/qwen2.5-0.5b-expo-DPO-ES-0.1/checkpoint-300",
4
  "epoch": 1.5588096362777515,
5
  "eval_steps": 50,
@@ -11,7 +11,7 @@
11
  {
12
  "dpo_loss": 0.6931471824645996,
13
  "epoch": 0.002834199338686821,
14
- "grad_norm": 18.44253347826331,
15
  "learning_rate": 2.840909090909091e-08,
16
  "logits": -1.359458565711975,
17
  "logps": -84.69721221923828,
@@ -27,10 +27,10 @@
27
  {
28
  "dpo_loss": 0.6822353601455688,
29
  "epoch": 0.14170996693434104,
30
- "grad_norm": 18.641365531241362,
31
  "learning_rate": 1.4204545454545458e-06,
32
- "logits": -1.508646011352539,
33
- "logps": -82.822021484375,
34
  "loss": 0.6785,
35
  "objective": 0.6822353601455688,
36
  "ranking_idealized": 0.608418345451355,
@@ -38,363 +38,363 @@
38
  "ranking_simple": 0.5267857313156128,
39
  "regularize": 0.6822353601455688,
40
  "step": 50,
41
- "wo_beta": 7.122643947601318
42
  },
43
  {
44
  "epoch": 0.14170996693434104,
45
- "eval_dpo_loss": 0.6842507719993591,
46
- "eval_logits": -1.6022367477416992,
47
- "eval_logps": -90.87205505371094,
48
- "eval_loss": 0.6813501715660095,
49
- "eval_objective": 0.6842507719993591,
50
  "eval_ranking_idealized": 0.6030020713806152,
51
  "eval_ranking_idealized_expo": 0.5222567319869995,
52
  "eval_ranking_simple": 0.5258799195289612,
53
- "eval_regularize": 0.6842507719993591,
54
- "eval_runtime": 308.2905,
55
- "eval_samples_per_second": 18.781,
56
- "eval_steps_per_second": 1.567,
57
- "eval_wo_beta": 7.874889850616455,
58
  "step": 50
59
  },
60
  {
61
- "dpo_loss": 0.61030513048172,
62
  "epoch": 0.2834199338686821,
63
- "grad_norm": 19.29211957833529,
64
  "learning_rate": 2.8409090909090916e-06,
65
- "logits": -1.6967747211456299,
66
- "logps": -88.03068542480469,
67
  "loss": 0.618,
68
- "objective": 0.61030513048172,
69
  "ranking_idealized": 0.6016666889190674,
70
  "ranking_idealized_expo": 0.5141666531562805,
71
  "ranking_simple": 0.5729166865348816,
72
- "regularize": 0.61030513048172,
73
  "step": 100,
74
- "wo_beta": 6.357723236083984
75
  },
76
  {
77
  "epoch": 0.2834199338686821,
78
- "eval_dpo_loss": 0.6765881776809692,
79
- "eval_logits": -1.7799152135849,
80
- "eval_logps": -98.88995361328125,
81
- "eval_loss": 0.6732921004295349,
82
- "eval_objective": 0.6765881776809692,
83
  "eval_ranking_idealized": 0.6030020713806152,
84
  "eval_ranking_idealized_expo": 0.5222567319869995,
85
  "eval_ranking_simple": 0.5398550629615784,
86
- "eval_regularize": 0.6765881776809692,
87
- "eval_runtime": 307.7771,
88
- "eval_samples_per_second": 18.812,
89
- "eval_steps_per_second": 1.569,
90
- "eval_wo_beta": 7.7840118408203125,
91
  "step": 100
92
  },
93
  {
94
- "dpo_loss": 0.5696364045143127,
95
  "epoch": 0.42512990080302315,
96
- "grad_norm": 18.142654654279198,
97
  "learning_rate": 4.2613636363636365e-06,
98
- "logits": -1.7447518110275269,
99
- "logps": -90.71894073486328,
100
  "loss": 0.5667,
101
- "objective": 0.5696364045143127,
102
  "ranking_idealized": 0.6066666841506958,
103
  "ranking_idealized_expo": 0.5287500023841858,
104
  "ranking_simple": 0.6016666889190674,
105
- "regularize": 0.5696364045143127,
106
  "step": 150,
107
- "wo_beta": 5.971243858337402
108
  },
109
  {
110
  "epoch": 0.42512990080302315,
111
- "eval_dpo_loss": 0.6829443573951721,
112
- "eval_logits": -1.807220697402954,
113
- "eval_logps": -99.12174987792969,
114
- "eval_loss": 0.686660647392273,
115
- "eval_objective": 0.6829443573951721,
116
  "eval_ranking_idealized": 0.6030020713806152,
117
  "eval_ranking_idealized_expo": 0.5222567319869995,
118
  "eval_ranking_simple": 0.5408902764320374,
119
- "eval_regularize": 0.6829443573951721,
120
- "eval_runtime": 308.0058,
121
- "eval_samples_per_second": 18.798,
122
- "eval_steps_per_second": 1.568,
123
- "eval_wo_beta": 7.853672981262207,
124
  "step": 150
125
  },
126
  {
127
- "dpo_loss": 0.5278546214103699,
128
  "epoch": 0.5668398677373642,
129
- "grad_norm": 15.355172339669656,
130
  "learning_rate": 4.997168347957521e-06,
131
- "logits": -1.9084746837615967,
132
- "logps": -92.50672149658203,
133
  "loss": 0.5214,
134
- "objective": 0.5278546214103699,
135
  "ranking_idealized": 0.5924999713897705,
136
  "ranking_idealized_expo": 0.5166666507720947,
137
  "ranking_simple": 0.6362500190734863,
138
- "regularize": 0.5278546214103699,
139
  "step": 200,
140
- "wo_beta": 5.543394565582275
141
  },
142
  {
143
  "epoch": 0.5668398677373642,
144
- "eval_dpo_loss": 0.69049072265625,
145
- "eval_logits": -1.8895256519317627,
146
- "eval_logps": -99.51531982421875,
147
- "eval_loss": 0.6901801824569702,
148
- "eval_objective": 0.69049072265625,
149
  "eval_ranking_idealized": 0.6030020713806152,
150
  "eval_ranking_idealized_expo": 0.5222567319869995,
151
  "eval_ranking_simple": 0.544513463973999,
152
- "eval_regularize": 0.69049072265625,
153
- "eval_runtime": 308.0169,
154
- "eval_samples_per_second": 18.798,
155
- "eval_steps_per_second": 1.568,
156
- "eval_wo_beta": 7.701313495635986,
157
  "step": 200
158
  },
159
  {
160
- "dpo_loss": 0.4935472011566162,
161
  "epoch": 0.7085498346717053,
162
- "grad_norm": 14.324676993767012,
163
  "learning_rate": 4.973122855144066e-06,
164
- "logits": -1.9081355333328247,
165
- "logps": -88.3556137084961,
166
  "loss": 0.4922,
167
- "objective": 0.4935472011566162,
168
  "ranking_idealized": 0.5991666913032532,
169
  "ranking_idealized_expo": 0.5170833468437195,
170
  "ranking_simple": 0.6608333587646484,
171
- "regularize": 0.4935472011566162,
172
  "step": 250,
173
- "wo_beta": 5.325418472290039
174
  },
175
  {
176
  "epoch": 0.7085498346717053,
177
- "eval_dpo_loss": 0.6914450526237488,
178
- "eval_logits": -1.9887231588363647,
179
- "eval_logps": -82.8383560180664,
180
- "eval_loss": 0.6975522041320801,
181
- "eval_objective": 0.6914450526237488,
182
  "eval_ranking_idealized": 0.6030020713806152,
183
  "eval_ranking_idealized_expo": 0.5222567319869995,
184
- "eval_ranking_simple": 0.5481366515159607,
185
- "eval_regularize": 0.6914450526237488,
186
- "eval_runtime": 307.641,
187
- "eval_samples_per_second": 18.821,
188
- "eval_steps_per_second": 1.57,
189
- "eval_wo_beta": 7.878448486328125,
190
  "step": 250
191
  },
192
  {
193
- "dpo_loss": 0.4521400034427643,
194
  "epoch": 0.8502598016060463,
195
- "grad_norm": 13.428528764338076,
196
  "learning_rate": 4.924776641419513e-06,
197
- "logits": -2.0954272747039795,
198
- "logps": -81.39044952392578,
199
  "loss": 0.4535,
200
- "objective": 0.4521400034427643,
201
  "ranking_idealized": 0.5799999833106995,
202
  "ranking_idealized_expo": 0.4970833361148834,
203
  "ranking_simple": 0.6837499737739563,
204
- "regularize": 0.4521400034427643,
205
  "step": 300,
206
- "wo_beta": 5.026640892028809
207
  },
208
  {
209
  "epoch": 0.8502598016060463,
210
- "eval_dpo_loss": 0.6893911957740784,
211
- "eval_logits": -2.1208713054656982,
212
- "eval_logps": -90.9490966796875,
213
- "eval_loss": 0.6922824382781982,
214
- "eval_objective": 0.6893911957740784,
215
  "eval_ranking_idealized": 0.6030020713806152,
216
  "eval_ranking_idealized_expo": 0.5222567319869995,
217
- "eval_ranking_simple": 0.556418240070343,
218
- "eval_regularize": 0.6893911957740784,
219
- "eval_runtime": 307.9716,
220
- "eval_samples_per_second": 18.8,
221
- "eval_steps_per_second": 1.568,
222
- "eval_wo_beta": 7.4231791496276855,
223
  "step": 300
224
  },
225
  {
226
- "dpo_loss": 0.40847164392471313,
227
  "epoch": 0.9919697685403873,
228
- "grad_norm": 15.041732370189118,
229
  "learning_rate": 4.8526047530778175e-06,
230
- "logits": -1.9279303550720215,
231
- "logps": -84.23763275146484,
232
- "loss": 0.4228,
233
- "objective": 0.40847164392471313,
234
  "ranking_idealized": 0.60916668176651,
235
  "ranking_idealized_expo": 0.5270833373069763,
236
- "ranking_simple": 0.7191666960716248,
237
- "regularize": 0.40847164392471313,
238
  "step": 350,
239
- "wo_beta": 3.9990389347076416
240
  },
241
  {
242
  "epoch": 0.9919697685403873,
243
- "eval_dpo_loss": 0.6968410611152649,
244
- "eval_logits": -1.9802873134613037,
245
- "eval_logps": -87.72307586669922,
246
- "eval_loss": 0.7063526511192322,
247
- "eval_objective": 0.6968410611152649,
248
  "eval_ranking_idealized": 0.6030020713806152,
249
  "eval_ranking_idealized_expo": 0.5222567319869995,
250
- "eval_ranking_simple": 0.5538302063941956,
251
- "eval_regularize": 0.6968410611152649,
252
- "eval_runtime": 307.977,
253
- "eval_samples_per_second": 18.8,
254
- "eval_steps_per_second": 1.568,
255
- "eval_wo_beta": 8.025344848632812,
256
  "step": 350
257
  },
258
  {
259
- "dpo_loss": 0.28143200278282166,
260
  "epoch": 1.1336797354747283,
261
- "grad_norm": 12.112116162894052,
262
  "learning_rate": 4.757316345716554e-06,
263
- "logits": -2.0127880573272705,
264
- "logps": -91.47932434082031,
265
- "loss": 0.2845,
266
- "objective": 0.28143200278282166,
267
  "ranking_idealized": 0.6087499856948853,
268
  "ranking_idealized_expo": 0.5337499976158142,
269
- "ranking_simple": 0.8141666650772095,
270
- "regularize": 0.28143200278282166,
271
  "step": 400,
272
- "wo_beta": 2.649115562438965
273
  },
274
  {
275
  "epoch": 1.1336797354747283,
276
- "eval_dpo_loss": 0.7269378900527954,
277
- "eval_logits": -2.080526828765869,
278
- "eval_logps": -101.31802368164062,
279
- "eval_loss": 0.7304782867431641,
280
- "eval_objective": 0.7269378900527954,
281
  "eval_ranking_idealized": 0.6030020713806152,
282
  "eval_ranking_idealized_expo": 0.5222567319869995,
283
- "eval_ranking_simple": 0.5429606437683105,
284
- "eval_regularize": 0.7269378900527954,
285
- "eval_runtime": 308.0503,
286
- "eval_samples_per_second": 18.796,
287
- "eval_steps_per_second": 1.568,
288
- "eval_wo_beta": 8.616350173950195,
289
  "step": 400
290
  },
291
  {
292
- "dpo_loss": 0.3042532503604889,
293
  "epoch": 1.2753897024090695,
294
- "grad_norm": 12.171625264502326,
295
  "learning_rate": 4.639847716126855e-06,
296
- "logits": -1.9084649085998535,
297
- "logps": -92.91566467285156,
298
- "loss": 0.2989,
299
- "objective": 0.3042532503604889,
300
  "ranking_idealized": 0.5975000262260437,
301
  "ranking_idealized_expo": 0.5199999809265137,
302
- "ranking_simple": 0.8075000047683716,
303
- "regularize": 0.3042532503604889,
304
  "step": 450,
305
- "wo_beta": 3.341869592666626
306
  },
307
  {
308
  "epoch": 1.2753897024090695,
309
- "eval_dpo_loss": 0.6973706483840942,
310
- "eval_logits": -1.8646337985992432,
311
- "eval_logps": -93.1955337524414,
312
- "eval_loss": 0.7005103826522827,
313
- "eval_objective": 0.6973706483840942,
314
  "eval_ranking_idealized": 0.6030020713806152,
315
  "eval_ranking_idealized_expo": 0.5222567319869995,
316
- "eval_ranking_simple": 0.5605590343475342,
317
- "eval_regularize": 0.6973706483840942,
318
- "eval_runtime": 308.4071,
319
- "eval_samples_per_second": 18.774,
320
- "eval_steps_per_second": 1.566,
321
- "eval_wo_beta": 8.238639831542969,
322
  "step": 450
323
  },
324
  {
325
- "dpo_loss": 0.29817140102386475,
326
  "epoch": 1.4170996693434104,
327
- "grad_norm": 15.895661459470155,
328
  "learning_rate": 4.501353102310901e-06,
329
- "logits": -1.9325114488601685,
330
- "logps": -91.53684997558594,
331
  "loss": 0.3065,
332
- "objective": 0.29817140102386475,
333
  "ranking_idealized": 0.57833331823349,
334
  "ranking_idealized_expo": 0.4983333349227905,
335
- "ranking_simple": 0.7975000143051147,
336
- "regularize": 0.29817140102386475,
337
  "step": 500,
338
- "wo_beta": 3.1243510246276855
339
  },
340
  {
341
  "epoch": 1.4170996693434104,
342
- "eval_dpo_loss": 0.7146824598312378,
343
- "eval_logits": -1.9982556104660034,
344
- "eval_logps": -97.01371002197266,
345
- "eval_loss": 0.717850923538208,
346
- "eval_objective": 0.7146824598312378,
347
  "eval_ranking_idealized": 0.6030020713806152,
348
  "eval_ranking_idealized_expo": 0.5222567319869995,
349
- "eval_ranking_simple": 0.5548654198646545,
350
- "eval_regularize": 0.7146824598312378,
351
- "eval_runtime": 308.8392,
352
- "eval_samples_per_second": 18.748,
353
- "eval_steps_per_second": 1.564,
354
- "eval_wo_beta": 8.27602767944336,
355
  "step": 500
356
  },
357
  {
358
- "dpo_loss": 0.2845906615257263,
359
  "epoch": 1.5588096362777515,
360
- "grad_norm": 10.66218965165015,
361
  "learning_rate": 4.34319334202531e-06,
362
- "logits": -1.9555292129516602,
363
- "logps": -97.59473419189453,
364
- "loss": 0.2885,
365
- "objective": 0.2845906615257263,
366
  "ranking_idealized": 0.5945833325386047,
367
  "ranking_idealized_expo": 0.5116666555404663,
368
- "ranking_simple": 0.8041666746139526,
369
- "regularize": 0.2845906615257263,
370
  "step": 550,
371
- "wo_beta": 2.6915340423583984
372
  },
373
  {
374
  "epoch": 1.5588096362777515,
375
- "eval_dpo_loss": 0.7133627533912659,
376
- "eval_logits": -1.9041162729263306,
377
- "eval_logps": -107.96095275878906,
378
- "eval_loss": 0.7091230750083923,
379
- "eval_objective": 0.7133627533912659,
380
  "eval_ranking_idealized": 0.6030020713806152,
381
  "eval_ranking_idealized_expo": 0.5222567319869995,
382
- "eval_ranking_simple": 0.5615941882133484,
383
- "eval_regularize": 0.7133627533912659,
384
- "eval_runtime": 308.0507,
385
- "eval_samples_per_second": 18.796,
386
- "eval_steps_per_second": 1.568,
387
- "eval_wo_beta": 8.196797370910645,
388
  "step": 550
389
  },
390
  {
391
  "epoch": 1.5588096362777515,
392
  "step": 550,
393
  "total_flos": 0.0,
394
- "train_loss": 0.4483427975394509,
395
- "train_runtime": 15087.4424,
396
- "train_samples_per_second": 16.836,
397
- "train_steps_per_second": 0.117
398
  }
399
  ],
400
  "logging_steps": 50,
 
1
  {
2
+ "best_metric": 7.426205635070801,
3
  "best_model_checkpoint": "./qwen2.5-0.5b/qwen2.5-0.5b-expo-DPO-ES-0.1/checkpoint-300",
4
  "epoch": 1.5588096362777515,
5
  "eval_steps": 50,
 
11
  {
12
  "dpo_loss": 0.6931471824645996,
13
  "epoch": 0.002834199338686821,
14
+ "grad_norm": 18.442536934850562,
15
  "learning_rate": 2.840909090909091e-08,
16
  "logits": -1.359458565711975,
17
  "logps": -84.69721221923828,
 
27
  {
28
  "dpo_loss": 0.6822353601455688,
29
  "epoch": 0.14170996693434104,
30
+ "grad_norm": 18.641661833444882,
31
  "learning_rate": 1.4204545454545458e-06,
32
+ "logits": -1.5086464881896973,
33
+ "logps": -82.8218765258789,
34
  "loss": 0.6785,
35
  "objective": 0.6822353601455688,
36
  "ranking_idealized": 0.608418345451355,
 
38
  "ranking_simple": 0.5267857313156128,
39
  "regularize": 0.6822353601455688,
40
  "step": 50,
41
+ "wo_beta": 7.122647285461426
42
  },
43
  {
44
  "epoch": 0.14170996693434104,
45
+ "eval_dpo_loss": 0.6842505931854248,
46
+ "eval_logits": -1.6022353172302246,
47
+ "eval_logps": -90.87158203125,
48
+ "eval_loss": 0.6813499927520752,
49
+ "eval_objective": 0.6842505931854248,
50
  "eval_ranking_idealized": 0.6030020713806152,
51
  "eval_ranking_idealized_expo": 0.5222567319869995,
52
  "eval_ranking_simple": 0.5258799195289612,
53
+ "eval_regularize": 0.6842505931854248,
54
+ "eval_runtime": 309.4614,
55
+ "eval_samples_per_second": 18.71,
56
+ "eval_steps_per_second": 1.561,
57
+ "eval_wo_beta": 7.874892711639404,
58
  "step": 50
59
  },
60
  {
61
+ "dpo_loss": 0.6103044152259827,
62
  "epoch": 0.2834199338686821,
63
+ "grad_norm": 19.292131977363915,
64
  "learning_rate": 2.8409090909090916e-06,
65
+ "logits": -1.696779727935791,
66
+ "logps": -88.03015899658203,
67
  "loss": 0.618,
68
+ "objective": 0.6103044152259827,
69
  "ranking_idealized": 0.6016666889190674,
70
  "ranking_idealized_expo": 0.5141666531562805,
71
  "ranking_simple": 0.5729166865348816,
72
+ "regularize": 0.6103044152259827,
73
  "step": 100,
74
+ "wo_beta": 6.357714653015137
75
  },
76
  {
77
  "epoch": 0.2834199338686821,
78
+ "eval_dpo_loss": 0.6765866279602051,
79
+ "eval_logits": -1.7799253463745117,
80
+ "eval_logps": -98.88992309570312,
81
+ "eval_loss": 0.6732903122901917,
82
+ "eval_objective": 0.6765866279602051,
83
  "eval_ranking_idealized": 0.6030020713806152,
84
  "eval_ranking_idealized_expo": 0.5222567319869995,
85
  "eval_ranking_simple": 0.5398550629615784,
86
+ "eval_regularize": 0.6765866279602051,
87
+ "eval_runtime": 308.813,
88
+ "eval_samples_per_second": 18.749,
89
+ "eval_steps_per_second": 1.564,
90
+ "eval_wo_beta": 7.784023284912109,
91
  "step": 100
92
  },
93
  {
94
+ "dpo_loss": 0.5696373581886292,
95
  "epoch": 0.42512990080302315,
96
+ "grad_norm": 18.14389066821128,
97
  "learning_rate": 4.2613636363636365e-06,
98
+ "logits": -1.744734287261963,
99
+ "logps": -90.7203140258789,
100
  "loss": 0.5667,
101
+ "objective": 0.5696373581886292,
102
  "ranking_idealized": 0.6066666841506958,
103
  "ranking_idealized_expo": 0.5287500023841858,
104
  "ranking_simple": 0.6016666889190674,
105
+ "regularize": 0.5696373581886292,
106
  "step": 150,
107
+ "wo_beta": 5.971276760101318
108
  },
109
  {
110
  "epoch": 0.42512990080302315,
111
+ "eval_dpo_loss": 0.6829108595848083,
112
+ "eval_logits": -1.8072086572647095,
113
+ "eval_logps": -99.12300109863281,
114
+ "eval_loss": 0.6866306066513062,
115
+ "eval_objective": 0.6829108595848083,
116
  "eval_ranking_idealized": 0.6030020713806152,
117
  "eval_ranking_idealized_expo": 0.5222567319869995,
118
  "eval_ranking_simple": 0.5408902764320374,
119
+ "eval_regularize": 0.6829108595848083,
120
+ "eval_runtime": 312.2547,
121
+ "eval_samples_per_second": 18.543,
122
+ "eval_steps_per_second": 1.547,
123
+ "eval_wo_beta": 7.8532514572143555,
124
  "step": 150
125
  },
126
  {
127
+ "dpo_loss": 0.5278292298316956,
128
  "epoch": 0.5668398677373642,
129
+ "grad_norm": 15.350860718764396,
130
  "learning_rate": 4.997168347957521e-06,
131
+ "logits": -1.908250093460083,
132
+ "logps": -92.51087951660156,
133
  "loss": 0.5214,
134
+ "objective": 0.5278292298316956,
135
  "ranking_idealized": 0.5924999713897705,
136
  "ranking_idealized_expo": 0.5166666507720947,
137
  "ranking_simple": 0.6362500190734863,
138
+ "regularize": 0.5278292298316956,
139
  "step": 200,
140
+ "wo_beta": 5.543264389038086
141
  },
142
  {
143
  "epoch": 0.5668398677373642,
144
+ "eval_dpo_loss": 0.6904094815254211,
145
+ "eval_logits": -1.8893996477127075,
146
+ "eval_logps": -99.53878784179688,
147
+ "eval_loss": 0.6900797486305237,
148
+ "eval_objective": 0.6904094815254211,
149
  "eval_ranking_idealized": 0.6030020713806152,
150
  "eval_ranking_idealized_expo": 0.5222567319869995,
151
  "eval_ranking_simple": 0.544513463973999,
152
+ "eval_regularize": 0.6904094815254211,
153
+ "eval_runtime": 315.3971,
154
+ "eval_samples_per_second": 18.358,
155
+ "eval_steps_per_second": 1.531,
156
+ "eval_wo_beta": 7.699478626251221,
157
  "step": 200
158
  },
159
  {
160
+ "dpo_loss": 0.49355897307395935,
161
  "epoch": 0.7085498346717053,
162
+ "grad_norm": 14.313286499637714,
163
  "learning_rate": 4.973122855144066e-06,
164
+ "logits": -1.9076462984085083,
165
+ "logps": -88.3504867553711,
166
  "loss": 0.4922,
167
+ "objective": 0.49355897307395935,
168
  "ranking_idealized": 0.5991666913032532,
169
  "ranking_idealized_expo": 0.5170833468437195,
170
  "ranking_simple": 0.6608333587646484,
171
+ "regularize": 0.49355897307395935,
172
  "step": 250,
173
+ "wo_beta": 5.325013637542725
174
  },
175
  {
176
  "epoch": 0.7085498346717053,
177
+ "eval_dpo_loss": 0.6915506720542908,
178
+ "eval_logits": -1.9879554510116577,
179
+ "eval_logps": -82.79730224609375,
180
+ "eval_loss": 0.6976116895675659,
181
+ "eval_objective": 0.6915506720542908,
182
  "eval_ranking_idealized": 0.6030020713806152,
183
  "eval_ranking_idealized_expo": 0.5222567319869995,
184
+ "eval_ranking_simple": 0.5476190447807312,
185
+ "eval_regularize": 0.6915506720542908,
186
+ "eval_runtime": 308.7255,
187
+ "eval_samples_per_second": 18.755,
188
+ "eval_steps_per_second": 1.564,
189
+ "eval_wo_beta": 7.87898588180542,
190
  "step": 250
191
  },
192
  {
193
+ "dpo_loss": 0.4521573483943939,
194
  "epoch": 0.8502598016060463,
195
+ "grad_norm": 13.44014162581437,
196
  "learning_rate": 4.924776641419513e-06,
197
+ "logits": -2.09318470954895,
198
+ "logps": -81.41643524169922,
199
  "loss": 0.4535,
200
+ "objective": 0.4521573483943939,
201
  "ranking_idealized": 0.5799999833106995,
202
  "ranking_idealized_expo": 0.4970833361148834,
203
  "ranking_simple": 0.6837499737739563,
204
+ "regularize": 0.4521573483943939,
205
  "step": 300,
206
+ "wo_beta": 5.026339054107666
207
  },
208
  {
209
  "epoch": 0.8502598016060463,
210
+ "eval_dpo_loss": 0.6892624497413635,
211
+ "eval_logits": -2.120311975479126,
212
+ "eval_logps": -91.03094482421875,
213
+ "eval_loss": 0.6920701265335083,
214
+ "eval_objective": 0.6892624497413635,
215
  "eval_ranking_idealized": 0.6030020713806152,
216
  "eval_ranking_idealized_expo": 0.5222567319869995,
217
+ "eval_ranking_simple": 0.5559006333351135,
218
+ "eval_regularize": 0.6892624497413635,
219
+ "eval_runtime": 308.4714,
220
+ "eval_samples_per_second": 18.77,
221
+ "eval_steps_per_second": 1.566,
222
+ "eval_wo_beta": 7.426205635070801,
223
  "step": 300
224
  },
225
  {
226
+ "dpo_loss": 0.40864402055740356,
227
  "epoch": 0.9919697685403873,
228
+ "grad_norm": 15.102301026818012,
229
  "learning_rate": 4.8526047530778175e-06,
230
+ "logits": -1.9232014417648315,
231
+ "logps": -84.39765930175781,
232
+ "loss": 0.423,
233
+ "objective": 0.40864402055740356,
234
  "ranking_idealized": 0.60916668176651,
235
  "ranking_idealized_expo": 0.5270833373069763,
236
+ "ranking_simple": 0.7174999713897705,
237
+ "regularize": 0.40864402055740356,
238
  "step": 350,
239
+ "wo_beta": 3.9979019165039062
240
  },
241
  {
242
  "epoch": 0.9919697685403873,
243
+ "eval_dpo_loss": 0.6959461569786072,
244
+ "eval_logits": -1.987973690032959,
245
+ "eval_logps": -88.1614990234375,
246
+ "eval_loss": 0.7056758403778076,
247
+ "eval_objective": 0.6959461569786072,
248
  "eval_ranking_idealized": 0.6030020713806152,
249
  "eval_ranking_idealized_expo": 0.5222567319869995,
250
+ "eval_ranking_simple": 0.5548654198646545,
251
+ "eval_regularize": 0.6959461569786072,
252
+ "eval_runtime": 309.9548,
253
+ "eval_samples_per_second": 18.68,
254
+ "eval_steps_per_second": 1.558,
255
+ "eval_wo_beta": 7.997907638549805,
256
  "step": 350
257
  },
258
  {
259
+ "dpo_loss": 0.28186026215553284,
260
  "epoch": 1.1336797354747283,
261
+ "grad_norm": 12.122355016988976,
262
  "learning_rate": 4.757316345716554e-06,
263
+ "logits": -2.0241637229919434,
264
+ "logps": -91.86678314208984,
265
+ "loss": 0.2847,
266
+ "objective": 0.28186026215553284,
267
  "ranking_idealized": 0.6087499856948853,
268
  "ranking_idealized_expo": 0.5337499976158142,
269
+ "ranking_simple": 0.8145833611488342,
270
+ "regularize": 0.28186026215553284,
271
  "step": 400,
272
+ "wo_beta": 2.646965980529785
273
  },
274
  {
275
  "epoch": 1.1336797354747283,
276
+ "eval_dpo_loss": 0.7281294465065002,
277
+ "eval_logits": -2.0862255096435547,
278
+ "eval_logps": -101.69258117675781,
279
+ "eval_loss": 0.7314654588699341,
280
+ "eval_objective": 0.7281294465065002,
281
  "eval_ranking_idealized": 0.6030020713806152,
282
  "eval_ranking_idealized_expo": 0.5222567319869995,
283
+ "eval_ranking_simple": 0.542443037033081,
284
+ "eval_regularize": 0.7281294465065002,
285
+ "eval_runtime": 308.7063,
286
+ "eval_samples_per_second": 18.756,
287
+ "eval_steps_per_second": 1.565,
288
+ "eval_wo_beta": 8.632596969604492,
289
  "step": 400
290
  },
291
  {
292
+ "dpo_loss": 0.30417078733444214,
293
  "epoch": 1.2753897024090695,
294
+ "grad_norm": 12.255120909837279,
295
  "learning_rate": 4.639847716126855e-06,
296
+ "logits": -1.9048844575881958,
297
+ "logps": -92.96432495117188,
298
+ "loss": 0.2991,
299
+ "objective": 0.30417078733444214,
300
  "ranking_idealized": 0.5975000262260437,
301
  "ranking_idealized_expo": 0.5199999809265137,
302
+ "ranking_simple": 0.8087499737739563,
303
+ "regularize": 0.30417078733444214,
304
  "step": 450,
305
+ "wo_beta": 3.335141658782959
306
  },
307
  {
308
  "epoch": 1.2753897024090695,
309
+ "eval_dpo_loss": 0.6979546546936035,
310
+ "eval_logits": -1.8470289707183838,
311
+ "eval_logps": -92.79419708251953,
312
+ "eval_loss": 0.7008146047592163,
313
+ "eval_objective": 0.6979546546936035,
314
  "eval_ranking_idealized": 0.6030020713806152,
315
  "eval_ranking_idealized_expo": 0.5222567319869995,
316
+ "eval_ranking_simple": 0.5621117949485779,
317
+ "eval_regularize": 0.6979546546936035,
318
+ "eval_runtime": 308.7789,
319
+ "eval_samples_per_second": 18.751,
320
+ "eval_steps_per_second": 1.564,
321
+ "eval_wo_beta": 8.258440017700195,
322
  "step": 450
323
  },
324
  {
325
+ "dpo_loss": 0.29814377427101135,
326
  "epoch": 1.4170996693434104,
327
+ "grad_norm": 15.647218082922008,
328
  "learning_rate": 4.501353102310901e-06,
329
+ "logits": -1.9365119934082031,
330
+ "logps": -90.89854431152344,
331
  "loss": 0.3065,
332
+ "objective": 0.29814377427101135,
333
  "ranking_idealized": 0.57833331823349,
334
  "ranking_idealized_expo": 0.4983333349227905,
335
+ "ranking_simple": 0.7991666793823242,
336
+ "regularize": 0.29814377427101135,
337
  "step": 500,
338
+ "wo_beta": 3.127906560897827
339
  },
340
  {
341
  "epoch": 1.4170996693434104,
342
+ "eval_dpo_loss": 0.7147387266159058,
343
+ "eval_logits": -2.006500005722046,
344
+ "eval_logps": -96.67472076416016,
345
+ "eval_loss": 0.7179672122001648,
346
+ "eval_objective": 0.7147387266159058,
347
  "eval_ranking_idealized": 0.6030020713806152,
348
  "eval_ranking_idealized_expo": 0.5222567319869995,
349
+ "eval_ranking_simple": 0.555383026599884,
350
+ "eval_regularize": 0.7147387266159058,
351
+ "eval_runtime": 309.3319,
352
+ "eval_samples_per_second": 18.718,
353
+ "eval_steps_per_second": 1.561,
354
+ "eval_wo_beta": 8.252218246459961,
355
  "step": 500
356
  },
357
  {
358
+ "dpo_loss": 0.2861484885215759,
359
  "epoch": 1.5588096362777515,
360
+ "grad_norm": 10.885590323378537,
361
  "learning_rate": 4.34319334202531e-06,
362
+ "logits": -1.9361701011657715,
363
+ "logps": -95.84064483642578,
364
+ "loss": 0.2895,
365
+ "objective": 0.2861484885215759,
366
  "ranking_idealized": 0.5945833325386047,
367
  "ranking_idealized_expo": 0.5116666555404663,
368
+ "ranking_simple": 0.8066666722297668,
369
+ "regularize": 0.2861484885215759,
370
  "step": 550,
371
+ "wo_beta": 2.692445993423462
372
  },
373
  {
374
  "epoch": 1.5588096362777515,
375
+ "eval_dpo_loss": 0.7076632976531982,
376
+ "eval_logits": -1.887025237083435,
377
+ "eval_logps": -104.24694061279297,
378
+ "eval_loss": 0.7044315338134766,
379
+ "eval_objective": 0.7076632976531982,
380
  "eval_ranking_idealized": 0.6030020713806152,
381
  "eval_ranking_idealized_expo": 0.5222567319869995,
382
+ "eval_ranking_simple": 0.5652173757553101,
383
+ "eval_regularize": 0.7076632976531982,
384
+ "eval_runtime": 309.8676,
385
+ "eval_samples_per_second": 18.685,
386
+ "eval_steps_per_second": 1.559,
387
+ "eval_wo_beta": 8.194681167602539,
388
  "step": 550
389
  },
390
  {
391
  "epoch": 1.5588096362777515,
392
  "step": 550,
393
  "total_flos": 0.0,
394
+ "train_loss": 0.4484944924441251,
395
+ "train_runtime": 15200.6621,
396
+ "train_samples_per_second": 16.71,
397
+ "train_steps_per_second": 0.116
398
  }
399
  ],
400
  "logging_steps": 50,