mjbuehler commited on
Commit
492faec
·
verified ·
1 Parent(s): 7f618d1

Model save

Browse files
README.md ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ base_model: lamm-mit/mistral-7B-v0.3-Base-CPT_SFT
4
+ tags:
5
+ - trl
6
+ - dpo
7
+ - generated_from_trainer
8
+ model-index:
9
+ - name: mistral-7B-v0.3-Base-CPT_SFT_DPO
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
+ # mistral-7B-v0.3-Base-CPT_SFT_DPO
17
+
18
+ This model is a fine-tuned version of [lamm-mit/mistral-7B-v0.3-Base-CPT_SFT](https://huggingface.co/lamm-mit/mistral-7B-v0.3-Base-CPT_SFT) on the None dataset.
19
+
20
+ ## Model description
21
+
22
+ More information needed
23
+
24
+ ## Intended uses & limitations
25
+
26
+ More information needed
27
+
28
+ ## Training and evaluation data
29
+
30
+ More information needed
31
+
32
+ ## Training procedure
33
+
34
+ ### Training hyperparameters
35
+
36
+ The following hyperparameters were used during training:
37
+ - learning_rate: 5e-07
38
+ - train_batch_size: 4
39
+ - eval_batch_size: 1
40
+ - seed: 42
41
+ - distributed_type: multi-GPU
42
+ - num_devices: 8
43
+ - gradient_accumulation_steps: 2
44
+ - total_train_batch_size: 64
45
+ - total_eval_batch_size: 8
46
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
+ - lr_scheduler_type: cosine
48
+ - lr_scheduler_warmup_ratio: 0.1
49
+ - num_epochs: 1
50
+
51
+ ### Training results
52
+
53
+
54
+
55
+ ### Framework versions
56
+
57
+ - Transformers 4.44.2
58
+ - Pytorch 2.4.0+cu121
59
+ - Datasets 2.21.0
60
+ - Tokenizers 0.19.1
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.22562925693344074,
5
+ "train_runtime": 6258.663,
6
+ "train_samples": 70516,
7
+ "train_samples_per_second": 11.267,
8
+ "train_steps_per_second": 0.176
9
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "transformers_version": "4.44.2"
6
+ }
runs/Aug27_14-12-06_192-222-52-143/events.out.tfevents.1724768334.192-222-52-143.54299.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4fdd94659841e965febe37765d98dac99fb1fe708b257c25114c74495893ebfb
3
- size 82419
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70b4736572255378de94797aa259fb54ff68d14d34460c30c7ac0ef103d4503b
3
+ size 82773
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.22562925693344074,
5
+ "train_runtime": 6258.663,
6
+ "train_samples": 70516,
7
+ "train_samples_per_second": 11.267,
8
+ "train_steps_per_second": 0.176
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1707 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 500,
6
+ "global_step": 1102,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0009074410163339383,
13
+ "grad_norm": 158.3370033102814,
14
+ "learning_rate": 4.504504504504504e-09,
15
+ "logits/chosen": -3.340815544128418,
16
+ "logits/rejected": -3.192225456237793,
17
+ "logps/chosen": -502.579833984375,
18
+ "logps/rejected": -285.80780029296875,
19
+ "loss": 0.6931,
20
+ "rewards/accuracies": 0.0,
21
+ "rewards/chosen": 0.0,
22
+ "rewards/margins": 0.0,
23
+ "rewards/rejected": 0.0,
24
+ "step": 1
25
+ },
26
+ {
27
+ "epoch": 0.009074410163339383,
28
+ "grad_norm": 130.1834663583104,
29
+ "learning_rate": 4.504504504504504e-08,
30
+ "logits/chosen": -3.252624988555908,
31
+ "logits/rejected": -3.2026939392089844,
32
+ "logps/chosen": -445.1510925292969,
33
+ "logps/rejected": -257.1930847167969,
34
+ "loss": 0.6924,
35
+ "rewards/accuracies": 0.375,
36
+ "rewards/chosen": -0.004173497203737497,
37
+ "rewards/margins": -0.005616632290184498,
38
+ "rewards/rejected": 0.0014431348536163568,
39
+ "step": 10
40
+ },
41
+ {
42
+ "epoch": 0.018148820326678767,
43
+ "grad_norm": 133.1160873140461,
44
+ "learning_rate": 9.009009009009008e-08,
45
+ "logits/chosen": -3.2142014503479004,
46
+ "logits/rejected": -3.092050075531006,
47
+ "logps/chosen": -358.178466796875,
48
+ "logps/rejected": -255.81381225585938,
49
+ "loss": 0.6818,
50
+ "rewards/accuracies": 0.5874999761581421,
51
+ "rewards/chosen": 0.02352432906627655,
52
+ "rewards/margins": 0.018155094236135483,
53
+ "rewards/rejected": 0.005369235761463642,
54
+ "step": 20
55
+ },
56
+ {
57
+ "epoch": 0.02722323049001815,
58
+ "grad_norm": 105.51382523990524,
59
+ "learning_rate": 1.3513513513513515e-07,
60
+ "logits/chosen": -3.195345163345337,
61
+ "logits/rejected": -3.0827088356018066,
62
+ "logps/chosen": -436.8555603027344,
63
+ "logps/rejected": -256.4704284667969,
64
+ "loss": 0.6224,
65
+ "rewards/accuracies": 0.8374999761581421,
66
+ "rewards/chosen": 0.1778702437877655,
67
+ "rewards/margins": 0.18295857310295105,
68
+ "rewards/rejected": -0.005088324658572674,
69
+ "step": 30
70
+ },
71
+ {
72
+ "epoch": 0.036297640653357534,
73
+ "grad_norm": 60.7347284030573,
74
+ "learning_rate": 1.8018018018018017e-07,
75
+ "logits/chosen": -3.2239010334014893,
76
+ "logits/rejected": -3.095165729522705,
77
+ "logps/chosen": -457.87548828125,
78
+ "logps/rejected": -252.0406494140625,
79
+ "loss": 0.512,
80
+ "rewards/accuracies": 0.862500011920929,
81
+ "rewards/chosen": 0.5447695851325989,
82
+ "rewards/margins": 0.5851655006408691,
83
+ "rewards/rejected": -0.04039595648646355,
84
+ "step": 40
85
+ },
86
+ {
87
+ "epoch": 0.045372050816696916,
88
+ "grad_norm": 54.695722961592466,
89
+ "learning_rate": 2.2522522522522522e-07,
90
+ "logits/chosen": -3.176532506942749,
91
+ "logits/rejected": -3.0848500728607178,
92
+ "logps/chosen": -388.94012451171875,
93
+ "logps/rejected": -237.63821411132812,
94
+ "loss": 0.4294,
95
+ "rewards/accuracies": 0.800000011920929,
96
+ "rewards/chosen": 1.0301157236099243,
97
+ "rewards/margins": 1.13114595413208,
98
+ "rewards/rejected": -0.10103032737970352,
99
+ "step": 50
100
+ },
101
+ {
102
+ "epoch": 0.0544464609800363,
103
+ "grad_norm": 47.472511720710024,
104
+ "learning_rate": 2.702702702702703e-07,
105
+ "logits/chosen": -3.2208728790283203,
106
+ "logits/rejected": -3.113288640975952,
107
+ "logps/chosen": -359.95257568359375,
108
+ "logps/rejected": -240.6640625,
109
+ "loss": 0.3796,
110
+ "rewards/accuracies": 0.862500011920929,
111
+ "rewards/chosen": 1.4194958209991455,
112
+ "rewards/margins": 1.7440767288208008,
113
+ "rewards/rejected": -0.3245808184146881,
114
+ "step": 60
115
+ },
116
+ {
117
+ "epoch": 0.06352087114337568,
118
+ "grad_norm": 39.52379333696054,
119
+ "learning_rate": 3.153153153153153e-07,
120
+ "logits/chosen": -3.226762056350708,
121
+ "logits/rejected": -3.15112042427063,
122
+ "logps/chosen": -369.8200378417969,
123
+ "logps/rejected": -225.3522491455078,
124
+ "loss": 0.3449,
125
+ "rewards/accuracies": 0.8999999761581421,
126
+ "rewards/chosen": 1.488443374633789,
127
+ "rewards/margins": 1.912201166152954,
128
+ "rewards/rejected": -0.4237578511238098,
129
+ "step": 70
130
+ },
131
+ {
132
+ "epoch": 0.07259528130671507,
133
+ "grad_norm": 44.60507224682583,
134
+ "learning_rate": 3.6036036036036033e-07,
135
+ "logits/chosen": -3.221963882446289,
136
+ "logits/rejected": -3.1375503540039062,
137
+ "logps/chosen": -367.0074462890625,
138
+ "logps/rejected": -247.7465362548828,
139
+ "loss": 0.3376,
140
+ "rewards/accuracies": 0.8500000238418579,
141
+ "rewards/chosen": 1.3955223560333252,
142
+ "rewards/margins": 1.9550836086273193,
143
+ "rewards/rejected": -0.5595611929893494,
144
+ "step": 80
145
+ },
146
+ {
147
+ "epoch": 0.08166969147005444,
148
+ "grad_norm": 39.08954176751689,
149
+ "learning_rate": 4.054054054054054e-07,
150
+ "logits/chosen": -3.2832207679748535,
151
+ "logits/rejected": -3.2185757160186768,
152
+ "logps/chosen": -425.96160888671875,
153
+ "logps/rejected": -284.1032409667969,
154
+ "loss": 0.2831,
155
+ "rewards/accuracies": 0.887499988079071,
156
+ "rewards/chosen": 1.7428745031356812,
157
+ "rewards/margins": 2.6255106925964355,
158
+ "rewards/rejected": -0.8826361894607544,
159
+ "step": 90
160
+ },
161
+ {
162
+ "epoch": 0.09074410163339383,
163
+ "grad_norm": 31.865193019318827,
164
+ "learning_rate": 4.5045045045045043e-07,
165
+ "logits/chosen": -3.2463021278381348,
166
+ "logits/rejected": -3.1695780754089355,
167
+ "logps/chosen": -405.6779479980469,
168
+ "logps/rejected": -280.05279541015625,
169
+ "loss": 0.2737,
170
+ "rewards/accuracies": 0.8374999761581421,
171
+ "rewards/chosen": 1.48948073387146,
172
+ "rewards/margins": 2.4863460063934326,
173
+ "rewards/rejected": -0.9968653917312622,
174
+ "step": 100
175
+ },
176
+ {
177
+ "epoch": 0.0998185117967332,
178
+ "grad_norm": 37.24886466287313,
179
+ "learning_rate": 4.954954954954955e-07,
180
+ "logits/chosen": -3.3158822059631348,
181
+ "logits/rejected": -3.2599291801452637,
182
+ "logps/chosen": -335.0594177246094,
183
+ "logps/rejected": -253.5969696044922,
184
+ "loss": 0.2908,
185
+ "rewards/accuracies": 0.8125,
186
+ "rewards/chosen": 1.3872979879379272,
187
+ "rewards/margins": 2.4369797706604004,
188
+ "rewards/rejected": -1.0496819019317627,
189
+ "step": 110
190
+ },
191
+ {
192
+ "epoch": 0.1088929219600726,
193
+ "grad_norm": 36.665353860098634,
194
+ "learning_rate": 4.99898253844669e-07,
195
+ "logits/chosen": -3.3115856647491455,
196
+ "logits/rejected": -3.2060322761535645,
197
+ "logps/chosen": -380.2586975097656,
198
+ "logps/rejected": -309.19830322265625,
199
+ "loss": 0.2488,
200
+ "rewards/accuracies": 0.8374999761581421,
201
+ "rewards/chosen": 1.6321680545806885,
202
+ "rewards/margins": 3.540356397628784,
203
+ "rewards/rejected": -1.9081882238388062,
204
+ "step": 120
205
+ },
206
+ {
207
+ "epoch": 0.11796733212341198,
208
+ "grad_norm": 37.06555340402035,
209
+ "learning_rate": 4.995466450646198e-07,
210
+ "logits/chosen": -3.3821232318878174,
211
+ "logits/rejected": -3.252227783203125,
212
+ "logps/chosen": -381.7076110839844,
213
+ "logps/rejected": -268.3893737792969,
214
+ "loss": 0.2668,
215
+ "rewards/accuracies": 0.875,
216
+ "rewards/chosen": 1.502217173576355,
217
+ "rewards/margins": 3.5684256553649902,
218
+ "rewards/rejected": -2.066208600997925,
219
+ "step": 130
220
+ },
221
+ {
222
+ "epoch": 0.12704174228675136,
223
+ "grad_norm": 38.75162651406194,
224
+ "learning_rate": 4.989442707764628e-07,
225
+ "logits/chosen": -3.352313995361328,
226
+ "logits/rejected": -3.257378101348877,
227
+ "logps/chosen": -374.43670654296875,
228
+ "logps/rejected": -287.6412658691406,
229
+ "loss": 0.2587,
230
+ "rewards/accuracies": 0.875,
231
+ "rewards/chosen": 1.585739016532898,
232
+ "rewards/margins": 3.8027961254119873,
233
+ "rewards/rejected": -2.217057466506958,
234
+ "step": 140
235
+ },
236
+ {
237
+ "epoch": 0.13611615245009073,
238
+ "grad_norm": 43.21076330965974,
239
+ "learning_rate": 4.980917362966688e-07,
240
+ "logits/chosen": -3.3749313354492188,
241
+ "logits/rejected": -3.275294542312622,
242
+ "logps/chosen": -417.59979248046875,
243
+ "logps/rejected": -307.0452575683594,
244
+ "loss": 0.2517,
245
+ "rewards/accuracies": 0.862500011920929,
246
+ "rewards/chosen": 1.4160921573638916,
247
+ "rewards/margins": 4.021966457366943,
248
+ "rewards/rejected": -2.605874538421631,
249
+ "step": 150
250
+ },
251
+ {
252
+ "epoch": 0.14519056261343014,
253
+ "grad_norm": 50.071879952898506,
254
+ "learning_rate": 4.969898983237597e-07,
255
+ "logits/chosen": -3.4258790016174316,
256
+ "logits/rejected": -3.311413526535034,
257
+ "logps/chosen": -355.394287109375,
258
+ "logps/rejected": -277.78228759765625,
259
+ "loss": 0.2454,
260
+ "rewards/accuracies": 0.9125000238418579,
261
+ "rewards/chosen": 1.4633300304412842,
262
+ "rewards/margins": 3.9215309619903564,
263
+ "rewards/rejected": -2.4582009315490723,
264
+ "step": 160
265
+ },
266
+ {
267
+ "epoch": 0.1542649727767695,
268
+ "grad_norm": 37.68782415609405,
269
+ "learning_rate": 4.95639864077426e-07,
270
+ "logits/chosen": -3.341951370239258,
271
+ "logits/rejected": -3.242036819458008,
272
+ "logps/chosen": -414.0884704589844,
273
+ "logps/rejected": -292.1954650878906,
274
+ "loss": 0.26,
275
+ "rewards/accuracies": 0.887499988079071,
276
+ "rewards/chosen": 1.594613790512085,
277
+ "rewards/margins": 4.542968273162842,
278
+ "rewards/rejected": -2.948354482650757,
279
+ "step": 170
280
+ },
281
+ {
282
+ "epoch": 0.16333938294010888,
283
+ "grad_norm": 38.50311716438203,
284
+ "learning_rate": 4.940429901858992e-07,
285
+ "logits/chosen": -3.315183162689209,
286
+ "logits/rejected": -3.1880416870117188,
287
+ "logps/chosen": -362.5171813964844,
288
+ "logps/rejected": -274.8582763671875,
289
+ "loss": 0.2323,
290
+ "rewards/accuracies": 0.862500011920929,
291
+ "rewards/chosen": 1.4722703695297241,
292
+ "rewards/margins": 4.3128228187561035,
293
+ "rewards/rejected": -2.8405520915985107,
294
+ "step": 180
295
+ },
296
+ {
297
+ "epoch": 0.1724137931034483,
298
+ "grad_norm": 34.06267671066188,
299
+ "learning_rate": 4.922008813226972e-07,
300
+ "logits/chosen": -3.3597846031188965,
301
+ "logits/rejected": -3.1644554138183594,
302
+ "logps/chosen": -399.7878723144531,
303
+ "logps/rejected": -301.4942932128906,
304
+ "loss": 0.2373,
305
+ "rewards/accuracies": 0.875,
306
+ "rewards/chosen": 1.6164276599884033,
307
+ "rewards/margins": 4.526951789855957,
308
+ "rewards/rejected": -2.9105238914489746,
309
+ "step": 190
310
+ },
311
+ {
312
+ "epoch": 0.18148820326678766,
313
+ "grad_norm": 46.205969565119815,
314
+ "learning_rate": 4.901153885941126e-07,
315
+ "logits/chosen": -3.3047378063201904,
316
+ "logits/rejected": -3.1578073501586914,
317
+ "logps/chosen": -414.8160705566406,
318
+ "logps/rejected": -362.454345703125,
319
+ "loss": 0.2326,
320
+ "rewards/accuracies": 0.925000011920929,
321
+ "rewards/chosen": 1.8784027099609375,
322
+ "rewards/margins": 4.937838554382324,
323
+ "rewards/rejected": -3.059436082839966,
324
+ "step": 200
325
+ },
326
+ {
327
+ "epoch": 0.19056261343012704,
328
+ "grad_norm": 31.435836455070017,
329
+ "learning_rate": 4.877886076790663e-07,
330
+ "logits/chosen": -3.190352201461792,
331
+ "logits/rejected": -3.0550923347473145,
332
+ "logps/chosen": -373.290771484375,
333
+ "logps/rejected": -295.98297119140625,
334
+ "loss": 0.2285,
335
+ "rewards/accuracies": 0.862500011920929,
336
+ "rewards/chosen": 1.5484044551849365,
337
+ "rewards/margins": 4.609506607055664,
338
+ "rewards/rejected": -3.0611023902893066,
339
+ "step": 210
340
+ },
341
+ {
342
+ "epoch": 0.1996370235934664,
343
+ "grad_norm": 39.42562554641845,
344
+ "learning_rate": 4.852228767231913e-07,
345
+ "logits/chosen": -3.2841200828552246,
346
+ "logits/rejected": -3.1063590049743652,
347
+ "logps/chosen": -388.2396240234375,
348
+ "logps/rejected": -296.4252014160156,
349
+ "loss": 0.2312,
350
+ "rewards/accuracies": 0.8999999761581421,
351
+ "rewards/chosen": 1.832167625427246,
352
+ "rewards/margins": 5.544095039367676,
353
+ "rewards/rejected": -3.7119274139404297,
354
+ "step": 220
355
+ },
356
+ {
357
+ "epoch": 0.20871143375680581,
358
+ "grad_norm": 46.72987633281948,
359
+ "learning_rate": 4.824207739892674e-07,
360
+ "logits/chosen": -3.2899200916290283,
361
+ "logits/rejected": -3.1556496620178223,
362
+ "logps/chosen": -410.05963134765625,
363
+ "logps/rejected": -361.39483642578125,
364
+ "loss": 0.2179,
365
+ "rewards/accuracies": 0.8999999761581421,
366
+ "rewards/chosen": 1.3326658010482788,
367
+ "rewards/margins": 4.7414703369140625,
368
+ "rewards/rejected": -3.4088046550750732,
369
+ "step": 230
370
+ },
371
+ {
372
+ "epoch": 0.2177858439201452,
373
+ "grad_norm": 37.19856056372189,
374
+ "learning_rate": 4.793851152663654e-07,
375
+ "logits/chosen": -3.265838623046875,
376
+ "logits/rejected": -3.1473007202148438,
377
+ "logps/chosen": -372.2582092285156,
378
+ "logps/rejected": -298.2706604003906,
379
+ "loss": 0.2378,
380
+ "rewards/accuracies": 0.9125000238418579,
381
+ "rewards/chosen": 1.1389844417572021,
382
+ "rewards/margins": 4.751265048980713,
383
+ "rewards/rejected": -3.612281084060669,
384
+ "step": 240
385
+ },
386
+ {
387
+ "epoch": 0.22686025408348456,
388
+ "grad_norm": 46.60591133283558,
389
+ "learning_rate": 4.7611895104030507e-07,
390
+ "logits/chosen": -3.2685317993164062,
391
+ "logits/rejected": -3.161949634552002,
392
+ "logps/chosen": -396.34747314453125,
393
+ "logps/rejected": -335.84747314453125,
394
+ "loss": 0.2157,
395
+ "rewards/accuracies": 0.925000011920929,
396
+ "rewards/chosen": 1.336761713027954,
397
+ "rewards/margins": 5.206587791442871,
398
+ "rewards/rejected": -3.869826078414917,
399
+ "step": 250
400
+ },
401
+ {
402
+ "epoch": 0.23593466424682397,
403
+ "grad_norm": 48.51885326305408,
404
+ "learning_rate": 4.726255634282693e-07,
405
+ "logits/chosen": -3.294827699661255,
406
+ "logits/rejected": -3.185404062271118,
407
+ "logps/chosen": -391.83551025390625,
408
+ "logps/rejected": -374.4935607910156,
409
+ "loss": 0.2463,
410
+ "rewards/accuracies": 0.8999999761581421,
411
+ "rewards/chosen": 0.8340435028076172,
412
+ "rewards/margins": 4.343258857727051,
413
+ "rewards/rejected": -3.5092151165008545,
414
+ "step": 260
415
+ },
416
+ {
417
+ "epoch": 0.24500907441016334,
418
+ "grad_norm": 32.347739963126486,
419
+ "learning_rate": 4.689084628806562e-07,
420
+ "logits/chosen": -3.2203621864318848,
421
+ "logits/rejected": -3.1058781147003174,
422
+ "logps/chosen": -343.33447265625,
423
+ "logps/rejected": -298.6334533691406,
424
+ "loss": 0.2047,
425
+ "rewards/accuracies": 0.9375,
426
+ "rewards/chosen": 1.4291167259216309,
427
+ "rewards/margins": 5.545358657836914,
428
+ "rewards/rejected": -4.116241931915283,
429
+ "step": 270
430
+ },
431
+ {
432
+ "epoch": 0.2540834845735027,
433
+ "grad_norm": 45.442868522798584,
434
+ "learning_rate": 4.6497138465348296e-07,
435
+ "logits/chosen": -3.3757805824279785,
436
+ "logits/rejected": -3.2095916271209717,
437
+ "logps/chosen": -373.3706970214844,
438
+ "logps/rejected": -310.359619140625,
439
+ "loss": 0.2189,
440
+ "rewards/accuracies": 0.8500000238418579,
441
+ "rewards/chosen": 1.2218859195709229,
442
+ "rewards/margins": 4.990644931793213,
443
+ "rewards/rejected": -3.768759250640869,
444
+ "step": 280
445
+ },
446
+ {
447
+ "epoch": 0.2631578947368421,
448
+ "grad_norm": 37.199549797615454,
449
+ "learning_rate": 4.608182850548852e-07,
450
+ "logits/chosen": -3.3329250812530518,
451
+ "logits/rejected": -3.2051258087158203,
452
+ "logps/chosen": -367.7585754394531,
453
+ "logps/rejected": -309.1279602050781,
454
+ "loss": 0.2062,
455
+ "rewards/accuracies": 0.925000011920929,
456
+ "rewards/chosen": 1.8593353033065796,
457
+ "rewards/margins": 5.791568756103516,
458
+ "rewards/rejected": -3.932232618331909,
459
+ "step": 290
460
+ },
461
+ {
462
+ "epoch": 0.27223230490018147,
463
+ "grad_norm": 175.22451085843974,
464
+ "learning_rate": 4.564533374694852e-07,
465
+ "logits/chosen": -3.2539546489715576,
466
+ "logits/rejected": -3.164071559906006,
467
+ "logps/chosen": -435.7051696777344,
468
+ "logps/rejected": -320.7638854980469,
469
+ "loss": 0.1834,
470
+ "rewards/accuracies": 0.9125000238418579,
471
+ "rewards/chosen": 1.294732689857483,
472
+ "rewards/margins": 5.408836841583252,
473
+ "rewards/rejected": -4.1141037940979,
474
+ "step": 300
475
+ },
476
+ {
477
+ "epoch": 0.2813067150635209,
478
+ "grad_norm": 49.820868636075424,
479
+ "learning_rate": 4.518809281646232e-07,
480
+ "logits/chosen": -3.2305614948272705,
481
+ "logits/rejected": -3.1455864906311035,
482
+ "logps/chosen": -393.7672119140625,
483
+ "logps/rejected": -333.42657470703125,
484
+ "loss": 0.1685,
485
+ "rewards/accuracies": 0.862500011920929,
486
+ "rewards/chosen": 1.6657005548477173,
487
+ "rewards/margins": 5.660979270935059,
488
+ "rewards/rejected": -3.9952797889709473,
489
+ "step": 310
490
+ },
491
+ {
492
+ "epoch": 0.29038112522686027,
493
+ "grad_norm": 46.38355610514617,
494
+ "learning_rate": 4.4710565188266623e-07,
495
+ "logits/chosen": -3.2362568378448486,
496
+ "logits/rejected": -3.151102304458618,
497
+ "logps/chosen": -399.2664794921875,
498
+ "logps/rejected": -307.7845458984375,
499
+ "loss": 0.1988,
500
+ "rewards/accuracies": 0.8999999761581421,
501
+ "rewards/chosen": 1.6355903148651123,
502
+ "rewards/margins": 5.9738664627075195,
503
+ "rewards/rejected": -4.338275909423828,
504
+ "step": 320
505
+ },
506
+ {
507
+ "epoch": 0.29945553539019965,
508
+ "grad_norm": 42.82294462919112,
509
+ "learning_rate": 4.4213230722382343e-07,
510
+ "logits/chosen": -3.2030301094055176,
511
+ "logits/rejected": -3.0990149974823,
512
+ "logps/chosen": -394.6158752441406,
513
+ "logps/rejected": -334.71051025390625,
514
+ "loss": 0.1802,
515
+ "rewards/accuracies": 0.9624999761581421,
516
+ "rewards/chosen": 1.584474802017212,
517
+ "rewards/margins": 6.304975986480713,
518
+ "rewards/rejected": -4.720500946044922,
519
+ "step": 330
520
+ },
521
+ {
522
+ "epoch": 0.308529945553539,
523
+ "grad_norm": 48.08164749388606,
524
+ "learning_rate": 4.3696589182410805e-07,
525
+ "logits/chosen": -3.2478835582733154,
526
+ "logits/rejected": -3.144749879837036,
527
+ "logps/chosen": -355.66729736328125,
528
+ "logps/rejected": -341.45648193359375,
529
+ "loss": 0.2065,
530
+ "rewards/accuracies": 0.9125000238418579,
531
+ "rewards/chosen": 1.5237865447998047,
532
+ "rewards/margins": 5.312397480010986,
533
+ "rewards/rejected": -3.7886109352111816,
534
+ "step": 340
535
+ },
536
+ {
537
+ "epoch": 0.3176043557168784,
538
+ "grad_norm": 39.92094020376778,
539
+ "learning_rate": 4.3161159733329143e-07,
540
+ "logits/chosen": -3.326545238494873,
541
+ "logits/rejected": -3.246845245361328,
542
+ "logps/chosen": -414.8565979003906,
543
+ "logps/rejected": -401.8109130859375,
544
+ "loss": 0.2384,
545
+ "rewards/accuracies": 0.8999999761581421,
546
+ "rewards/chosen": 1.3785138130187988,
547
+ "rewards/margins": 5.425638198852539,
548
+ "rewards/rejected": -4.047124862670898,
549
+ "step": 350
550
+ },
551
+ {
552
+ "epoch": 0.32667876588021777,
553
+ "grad_norm": 26.862261807926526,
554
+ "learning_rate": 4.2607480419789587e-07,
555
+ "logits/chosen": -3.271552324295044,
556
+ "logits/rejected": -3.2207393646240234,
557
+ "logps/chosen": -332.0171813964844,
558
+ "logps/rejected": -289.7067565917969,
559
+ "loss": 0.1943,
560
+ "rewards/accuracies": 0.9125000238418579,
561
+ "rewards/chosen": 1.2600226402282715,
562
+ "rewards/margins": 5.030722141265869,
563
+ "rewards/rejected": -3.7706997394561768,
564
+ "step": 360
565
+ },
566
+ {
567
+ "epoch": 0.33575317604355714,
568
+ "grad_norm": 33.49994098039321,
569
+ "learning_rate": 4.2036107625446783e-07,
570
+ "logits/chosen": -3.3395228385925293,
571
+ "logits/rejected": -3.1348748207092285,
572
+ "logps/chosen": -394.46795654296875,
573
+ "logps/rejected": -317.6743469238281,
574
+ "loss": 0.1864,
575
+ "rewards/accuracies": 0.9375,
576
+ "rewards/chosen": 1.9414106607437134,
577
+ "rewards/margins": 6.639167785644531,
578
+ "rewards/rejected": -4.697756767272949,
579
+ "step": 370
580
+ },
581
+ {
582
+ "epoch": 0.3448275862068966,
583
+ "grad_norm": 33.33388789905036,
584
+ "learning_rate": 4.1447615513856635e-07,
585
+ "logits/chosen": -3.2808823585510254,
586
+ "logits/rejected": -3.1995129585266113,
587
+ "logps/chosen": -409.93603515625,
588
+ "logps/rejected": -363.3692932128906,
589
+ "loss": 0.1947,
590
+ "rewards/accuracies": 0.925000011920929,
591
+ "rewards/chosen": 1.1632466316223145,
592
+ "rewards/margins": 5.149783611297607,
593
+ "rewards/rejected": -3.986537218093872,
594
+ "step": 380
595
+ },
596
+ {
597
+ "epoch": 0.35390199637023595,
598
+ "grad_norm": 42.856588981749276,
599
+ "learning_rate": 4.084259545150832e-07,
600
+ "logits/chosen": -3.324871778488159,
601
+ "logits/rejected": -3.1859161853790283,
602
+ "logps/chosen": -393.6763916015625,
603
+ "logps/rejected": -311.64404296875,
604
+ "loss": 0.2091,
605
+ "rewards/accuracies": 0.875,
606
+ "rewards/chosen": 1.2227169275283813,
607
+ "rewards/margins": 5.600234031677246,
608
+ "rewards/rejected": -4.377516746520996,
609
+ "step": 390
610
+ },
611
+ {
612
+ "epoch": 0.3629764065335753,
613
+ "grad_norm": 44.671991447679176,
614
+ "learning_rate": 4.022165541356941e-07,
615
+ "logits/chosen": -3.30537748336792,
616
+ "logits/rejected": -3.205775499343872,
617
+ "logps/chosen": -396.2132873535156,
618
+ "logps/rejected": -329.81768798828125,
619
+ "loss": 0.1987,
620
+ "rewards/accuracies": 0.949999988079071,
621
+ "rewards/chosen": 1.6245359182357788,
622
+ "rewards/margins": 6.8077392578125,
623
+ "rewards/rejected": -5.183202743530273,
624
+ "step": 400
625
+ },
626
+ {
627
+ "epoch": 0.3720508166969147,
628
+ "grad_norm": 66.71100346246702,
629
+ "learning_rate": 3.9585419372941163e-07,
630
+ "logits/chosen": -3.2828071117401123,
631
+ "logits/rejected": -3.2080435752868652,
632
+ "logps/chosen": -383.0317687988281,
633
+ "logps/rejected": -321.2015686035156,
634
+ "loss": 0.228,
635
+ "rewards/accuracies": 0.9125000238418579,
636
+ "rewards/chosen": 1.194445013999939,
637
+ "rewards/margins": 5.658787727355957,
638
+ "rewards/rejected": -4.464343070983887,
639
+ "step": 410
640
+ },
641
+ {
642
+ "epoch": 0.3811252268602541,
643
+ "grad_norm": 43.403396807175234,
644
+ "learning_rate": 3.893452667323793e-07,
645
+ "logits/chosen": -3.319455623626709,
646
+ "logits/rejected": -3.226341962814331,
647
+ "logps/chosen": -390.3877258300781,
648
+ "logps/rejected": -356.02667236328125,
649
+ "loss": 0.1792,
650
+ "rewards/accuracies": 0.9125000238418579,
651
+ "rewards/chosen": 1.921933889389038,
652
+ "rewards/margins": 6.734816074371338,
653
+ "rewards/rejected": -4.8128814697265625,
654
+ "step": 420
655
+ },
656
+ {
657
+ "epoch": 0.39019963702359345,
658
+ "grad_norm": 33.07955087603256,
659
+ "learning_rate": 3.826963138632079e-07,
660
+ "logits/chosen": -3.2265000343322754,
661
+ "logits/rejected": -3.149766445159912,
662
+ "logps/chosen": -372.9231262207031,
663
+ "logps/rejected": -309.909423828125,
664
+ "loss": 0.2262,
665
+ "rewards/accuracies": 0.8500000238418579,
666
+ "rewards/chosen": 1.2967065572738647,
667
+ "rewards/margins": 6.030789852142334,
668
+ "rewards/rejected": -4.73408317565918,
669
+ "step": 430
670
+ },
671
+ {
672
+ "epoch": 0.3992740471869328,
673
+ "grad_norm": 37.11726657072608,
674
+ "learning_rate": 3.759140165503101e-07,
675
+ "logits/chosen": -3.2834110260009766,
676
+ "logits/rejected": -3.13962984085083,
677
+ "logps/chosen": -365.4574279785156,
678
+ "logps/rejected": -315.04095458984375,
679
+ "loss": 0.2033,
680
+ "rewards/accuracies": 0.9375,
681
+ "rewards/chosen": 1.5437448024749756,
682
+ "rewards/margins": 6.540799140930176,
683
+ "rewards/rejected": -4.997054100036621,
684
+ "step": 440
685
+ },
686
+ {
687
+ "epoch": 0.40834845735027225,
688
+ "grad_norm": 31.621123892211624,
689
+ "learning_rate": 3.6900519021783783e-07,
690
+ "logits/chosen": -3.241823196411133,
691
+ "logits/rejected": -3.1715283393859863,
692
+ "logps/chosen": -383.9208984375,
693
+ "logps/rejected": -330.8930358886719,
694
+ "loss": 0.1902,
695
+ "rewards/accuracies": 0.8999999761581421,
696
+ "rewards/chosen": 1.2258026599884033,
697
+ "rewards/margins": 5.701479911804199,
698
+ "rewards/rejected": -4.475677013397217,
699
+ "step": 450
700
+ },
701
+ {
702
+ "epoch": 0.41742286751361163,
703
+ "grad_norm": 49.178031353029425,
704
+ "learning_rate": 3.619767774369694e-07,
705
+ "logits/chosen": -3.3129935264587402,
706
+ "logits/rejected": -3.1908164024353027,
707
+ "logps/chosen": -372.14959716796875,
708
+ "logps/rejected": -343.7088623046875,
709
+ "loss": 0.243,
710
+ "rewards/accuracies": 0.875,
711
+ "rewards/chosen": 1.1361382007598877,
712
+ "rewards/margins": 5.055792331695557,
713
+ "rewards/rejected": -3.919654369354248,
714
+ "step": 460
715
+ },
716
+ {
717
+ "epoch": 0.426497277676951,
718
+ "grad_norm": 37.34866908331328,
719
+ "learning_rate": 3.548358409494291e-07,
720
+ "logits/chosen": -3.297877788543701,
721
+ "logits/rejected": -3.2215073108673096,
722
+ "logps/chosen": -404.14013671875,
723
+ "logps/rejected": -313.212890625,
724
+ "loss": 0.2062,
725
+ "rewards/accuracies": 0.887499988079071,
726
+ "rewards/chosen": 1.1549508571624756,
727
+ "rewards/margins": 5.121293067932129,
728
+ "rewards/rejected": -3.9663422107696533,
729
+ "step": 470
730
+ },
731
+ {
732
+ "epoch": 0.4355716878402904,
733
+ "grad_norm": 48.882738312246666,
734
+ "learning_rate": 3.475895565702479e-07,
735
+ "logits/chosen": -3.2060375213623047,
736
+ "logits/rejected": -3.1445508003234863,
737
+ "logps/chosen": -401.10638427734375,
738
+ "logps/rejected": -344.1682434082031,
739
+ "loss": 0.322,
740
+ "rewards/accuracies": 0.925000011920929,
741
+ "rewards/chosen": 1.2614433765411377,
742
+ "rewards/margins": 5.999361515045166,
743
+ "rewards/rejected": -4.737917900085449,
744
+ "step": 480
745
+ },
746
+ {
747
+ "epoch": 0.44464609800362975,
748
+ "grad_norm": 50.68639708460402,
749
+ "learning_rate": 3.402452059769006e-07,
750
+ "logits/chosen": -3.1942269802093506,
751
+ "logits/rejected": -3.0844810009002686,
752
+ "logps/chosen": -388.79132080078125,
753
+ "logps/rejected": -315.6092834472656,
754
+ "loss": 0.2099,
755
+ "rewards/accuracies": 0.9125000238418579,
756
+ "rewards/chosen": 1.3437281847000122,
757
+ "rewards/margins": 5.917460918426514,
758
+ "rewards/rejected": -4.573733329772949,
759
+ "step": 490
760
+ },
761
+ {
762
+ "epoch": 0.4537205081669691,
763
+ "grad_norm": 50.14897433584961,
764
+ "learning_rate": 3.3281016939206175e-07,
765
+ "logits/chosen": -3.2095024585723877,
766
+ "logits/rejected": -3.1448185443878174,
767
+ "logps/chosen": -387.31158447265625,
768
+ "logps/rejected": -345.4943542480469,
769
+ "loss": 0.2088,
770
+ "rewards/accuracies": 0.9375,
771
+ "rewards/chosen": 1.2966068983078003,
772
+ "rewards/margins": 5.815266132354736,
773
+ "rewards/rejected": -4.518658638000488,
774
+ "step": 500
775
+ },
776
+ {
777
+ "epoch": 0.4627949183303085,
778
+ "grad_norm": 37.26501610114464,
779
+ "learning_rate": 3.2529191816733575e-07,
780
+ "logits/chosen": -3.2022509574890137,
781
+ "logits/rejected": -3.110914707183838,
782
+ "logps/chosen": -398.9407653808594,
783
+ "logps/rejected": -354.12310791015625,
784
+ "loss": 0.1804,
785
+ "rewards/accuracies": 0.8999999761581421,
786
+ "rewards/chosen": 1.4531505107879639,
787
+ "rewards/margins": 5.942654609680176,
788
+ "rewards/rejected": -4.489503860473633,
789
+ "step": 510
790
+ },
791
+ {
792
+ "epoch": 0.47186932849364793,
793
+ "grad_norm": 72.82731512227907,
794
+ "learning_rate": 3.1769800727541315e-07,
795
+ "logits/chosen": -3.122022867202759,
796
+ "logits/rejected": -3.0256314277648926,
797
+ "logps/chosen": -382.6126708984375,
798
+ "logps/rejected": -322.1177673339844,
799
+ "loss": 0.1792,
800
+ "rewards/accuracies": 0.9375,
801
+ "rewards/chosen": 1.351405143737793,
802
+ "rewards/margins": 6.904001712799072,
803
+ "rewards/rejected": -5.5525970458984375,
804
+ "step": 520
805
+ },
806
+ {
807
+ "epoch": 0.4809437386569873,
808
+ "grad_norm": 38.26822885645478,
809
+ "learning_rate": 3.1003606771819666e-07,
810
+ "logits/chosen": -3.139383554458618,
811
+ "logits/rejected": -3.029106616973877,
812
+ "logps/chosen": -392.1285705566406,
813
+ "logps/rejected": -341.97955322265625,
814
+ "loss": 0.1913,
815
+ "rewards/accuracies": 0.887499988079071,
816
+ "rewards/chosen": 1.1525547504425049,
817
+ "rewards/margins": 6.076541900634766,
818
+ "rewards/rejected": -4.92398738861084,
819
+ "step": 530
820
+ },
821
+ {
822
+ "epoch": 0.4900181488203267,
823
+ "grad_norm": 43.882325645673525,
824
+ "learning_rate": 3.023137988585276e-07,
825
+ "logits/chosen": -3.178436279296875,
826
+ "logits/rejected": -3.1166467666625977,
827
+ "logps/chosen": -387.3483581542969,
828
+ "logps/rejected": -383.4141845703125,
829
+ "loss": 0.2216,
830
+ "rewards/accuracies": 0.887499988079071,
831
+ "rewards/chosen": 1.2282485961914062,
832
+ "rewards/margins": 5.9733991622924805,
833
+ "rewards/rejected": -4.745150089263916,
834
+ "step": 540
835
+ },
836
+ {
837
+ "epoch": 0.49909255898366606,
838
+ "grad_norm": 44.30893309428354,
839
+ "learning_rate": 2.945389606832165e-07,
840
+ "logits/chosen": -3.1245813369750977,
841
+ "logits/rejected": -3.1115365028381348,
842
+ "logps/chosen": -399.7410583496094,
843
+ "logps/rejected": -330.3243713378906,
844
+ "loss": 0.2036,
845
+ "rewards/accuracies": 0.925000011920929,
846
+ "rewards/chosen": 1.6770175695419312,
847
+ "rewards/margins": 7.239003658294678,
848
+ "rewards/rejected": -5.561986446380615,
849
+ "step": 550
850
+ },
851
+ {
852
+ "epoch": 0.5081669691470054,
853
+ "grad_norm": 49.55956318241512,
854
+ "learning_rate": 2.8671936600515445e-07,
855
+ "logits/chosen": -3.155513286590576,
856
+ "logits/rejected": -3.0905396938323975,
857
+ "logps/chosen": -374.02392578125,
858
+ "logps/rejected": -352.85986328125,
859
+ "loss": 0.1808,
860
+ "rewards/accuracies": 0.925000011920929,
861
+ "rewards/chosen": 1.1260972023010254,
862
+ "rewards/margins": 5.719089984893799,
863
+ "rewards/rejected": -4.592992782592773,
864
+ "step": 560
865
+ },
866
+ {
867
+ "epoch": 0.5172413793103449,
868
+ "grad_norm": 43.67166571714464,
869
+ "learning_rate": 2.788628726123399e-07,
870
+ "logits/chosen": -3.174683094024658,
871
+ "logits/rejected": -3.1034023761749268,
872
+ "logps/chosen": -368.4584655761719,
873
+ "logps/rejected": -301.5198669433594,
874
+ "loss": 0.2044,
875
+ "rewards/accuracies": 0.949999988079071,
876
+ "rewards/chosen": 1.258647084236145,
877
+ "rewards/margins": 6.08292293548584,
878
+ "rewards/rejected": -4.824276447296143,
879
+ "step": 570
880
+ },
881
+ {
882
+ "epoch": 0.5263157894736842,
883
+ "grad_norm": 46.763150412882524,
884
+ "learning_rate": 2.7097737537171095e-07,
885
+ "logits/chosen": -3.193993091583252,
886
+ "logits/rejected": -3.1491267681121826,
887
+ "logps/chosen": -379.1547546386719,
888
+ "logps/rejected": -357.4938049316406,
889
+ "loss": 0.1895,
890
+ "rewards/accuracies": 0.987500011920929,
891
+ "rewards/chosen": 1.0861809253692627,
892
+ "rewards/margins": 5.892585754394531,
893
+ "rewards/rejected": -4.8064045906066895,
894
+ "step": 580
895
+ },
896
+ {
897
+ "epoch": 0.5353901996370236,
898
+ "grad_norm": 28.50813723914452,
899
+ "learning_rate": 2.6307079829571685e-07,
900
+ "logits/chosen": -3.1775803565979004,
901
+ "logits/rejected": -3.1690754890441895,
902
+ "logps/chosen": -396.2727966308594,
903
+ "logps/rejected": -364.2826232910156,
904
+ "loss": 0.1916,
905
+ "rewards/accuracies": 0.8500000238418579,
906
+ "rewards/chosen": 1.2243824005126953,
907
+ "rewards/margins": 6.1176557540893555,
908
+ "rewards/rejected": -4.893273830413818,
909
+ "step": 590
910
+ },
911
+ {
912
+ "epoch": 0.5444646098003629,
913
+ "grad_norm": 28.704553800715182,
914
+ "learning_rate": 2.551510865796032e-07,
915
+ "logits/chosen": -3.1417794227600098,
916
+ "logits/rejected": -3.1004185676574707,
917
+ "logps/chosen": -321.55340576171875,
918
+ "logps/rejected": -337.394287109375,
919
+ "loss": 0.1878,
920
+ "rewards/accuracies": 0.862500011920929,
921
+ "rewards/chosen": 0.9132893681526184,
922
+ "rewards/margins": 6.268902778625488,
923
+ "rewards/rejected": -5.355614185333252,
924
+ "step": 600
925
+ },
926
+ {
927
+ "epoch": 0.5535390199637024,
928
+ "grad_norm": 33.895452413293604,
929
+ "learning_rate": 2.472261986174088e-07,
930
+ "logits/chosen": -3.1212687492370605,
931
+ "logits/rejected": -3.1359126567840576,
932
+ "logps/chosen": -430.43408203125,
933
+ "logps/rejected": -387.6915588378906,
934
+ "loss": 0.2053,
935
+ "rewards/accuracies": 0.925000011920929,
936
+ "rewards/chosen": 0.5770806670188904,
937
+ "rewards/margins": 5.136279106140137,
938
+ "rewards/rejected": -4.559198379516602,
939
+ "step": 610
940
+ },
941
+ {
942
+ "epoch": 0.5626134301270418,
943
+ "grad_norm": 35.42714003898933,
944
+ "learning_rate": 2.393040980047015e-07,
945
+ "logits/chosen": -3.2134900093078613,
946
+ "logits/rejected": -3.1361842155456543,
947
+ "logps/chosen": -425.0316467285156,
948
+ "logps/rejected": -361.84063720703125,
949
+ "loss": 0.1824,
950
+ "rewards/accuracies": 0.925000011920929,
951
+ "rewards/chosen": 1.0797230005264282,
952
+ "rewards/margins": 6.7416672706604,
953
+ "rewards/rejected": -5.661944389343262,
954
+ "step": 620
955
+ },
956
+ {
957
+ "epoch": 0.5716878402903811,
958
+ "grad_norm": 34.07816065316919,
959
+ "learning_rate": 2.3139274553608494e-07,
960
+ "logits/chosen": -3.184861898422241,
961
+ "logits/rejected": -3.1559784412384033,
962
+ "logps/chosen": -404.4765319824219,
963
+ "logps/rejected": -341.35546875,
964
+ "loss": 0.1876,
965
+ "rewards/accuracies": 0.9375,
966
+ "rewards/chosen": 0.9906010627746582,
967
+ "rewards/margins": 6.139868259429932,
968
+ "rewards/rejected": -5.149266719818115,
969
+ "step": 630
970
+ },
971
+ {
972
+ "epoch": 0.5807622504537205,
973
+ "grad_norm": 42.85248856046431,
974
+ "learning_rate": 2.2350009120552156e-07,
975
+ "logits/chosen": -3.1936917304992676,
976
+ "logits/rejected": -3.1695635318756104,
977
+ "logps/chosen": -406.64471435546875,
978
+ "logps/rejected": -391.30450439453125,
979
+ "loss": 0.2067,
980
+ "rewards/accuracies": 0.862500011920929,
981
+ "rewards/chosen": 1.1682602167129517,
982
+ "rewards/margins": 6.4593048095703125,
983
+ "rewards/rejected": -5.29104471206665,
984
+ "step": 640
985
+ },
986
+ {
987
+ "epoch": 0.5898366606170599,
988
+ "grad_norm": 55.772327844139255,
989
+ "learning_rate": 2.1563406621750825e-07,
990
+ "logits/chosen": -3.1266651153564453,
991
+ "logits/rejected": -3.0548195838928223,
992
+ "logps/chosen": -370.1761474609375,
993
+ "logps/rejected": -340.8184509277344,
994
+ "loss": 0.2141,
995
+ "rewards/accuracies": 0.8999999761581421,
996
+ "rewards/chosen": 1.114101767539978,
997
+ "rewards/margins": 6.014601230621338,
998
+ "rewards/rejected": -4.9004998207092285,
999
+ "step": 650
1000
+ },
1001
+ {
1002
+ "epoch": 0.5989110707803993,
1003
+ "grad_norm": 25.791382900724862,
1004
+ "learning_rate": 2.0780257501713346e-07,
1005
+ "logits/chosen": -3.165015459060669,
1006
+ "logits/rejected": -3.1734094619750977,
1007
+ "logps/chosen": -415.572998046875,
1008
+ "logps/rejected": -385.7120666503906,
1009
+ "loss": 0.1892,
1010
+ "rewards/accuracies": 0.887499988079071,
1011
+ "rewards/chosen": 1.1960872411727905,
1012
+ "rewards/margins": 6.418343544006348,
1013
+ "rewards/rejected": -5.222255706787109,
1014
+ "step": 660
1015
+ },
1016
+ {
1017
+ "epoch": 0.6079854809437386,
1018
+ "grad_norm": 33.330744516004074,
1019
+ "learning_rate": 2.000134873470243e-07,
1020
+ "logits/chosen": -3.128190279006958,
1021
+ "logits/rejected": -3.1104788780212402,
1022
+ "logps/chosen": -341.00494384765625,
1023
+ "logps/rejected": -329.31768798828125,
1024
+ "loss": 0.1785,
1025
+ "rewards/accuracies": 0.925000011920929,
1026
+ "rewards/chosen": 1.1156467199325562,
1027
+ "rewards/margins": 6.29996395111084,
1028
+ "rewards/rejected": -5.184317111968994,
1029
+ "step": 670
1030
+ },
1031
+ {
1032
+ "epoch": 0.617059891107078,
1033
+ "grad_norm": 33.150812329384884,
1034
+ "learning_rate": 1.922746303391655e-07,
1035
+ "logits/chosen": -3.105457305908203,
1036
+ "logits/rejected": -3.0872020721435547,
1037
+ "logps/chosen": -394.4687805175781,
1038
+ "logps/rejected": -330.22650146484375,
1039
+ "loss": 0.1791,
1040
+ "rewards/accuracies": 0.925000011920929,
1041
+ "rewards/chosen": 1.7894136905670166,
1042
+ "rewards/margins": 7.119010925292969,
1043
+ "rewards/rejected": -5.329598426818848,
1044
+ "step": 680
1045
+ },
1046
+ {
1047
+ "epoch": 0.6261343012704175,
1048
+ "grad_norm": 37.824280577969375,
1049
+ "learning_rate": 1.8459378064953754e-07,
1050
+ "logits/chosen": -3.1590168476104736,
1051
+ "logits/rejected": -3.0799336433410645,
1052
+ "logps/chosen": -410.97802734375,
1053
+ "logps/rejected": -351.85650634765625,
1054
+ "loss": 0.2003,
1055
+ "rewards/accuracies": 0.887499988079071,
1056
+ "rewards/chosen": 1.496854543685913,
1057
+ "rewards/margins": 6.54131555557251,
1058
+ "rewards/rejected": -5.044460773468018,
1059
+ "step": 690
1060
+ },
1061
+ {
1062
+ "epoch": 0.6352087114337568,
1063
+ "grad_norm": 42.16693290472239,
1064
+ "learning_rate": 1.7697865664347694e-07,
1065
+ "logits/chosen": -3.1934874057769775,
1066
+ "logits/rejected": -3.130568265914917,
1067
+ "logps/chosen": -386.4817810058594,
1068
+ "logps/rejected": -316.8409118652344,
1069
+ "loss": 0.1868,
1070
+ "rewards/accuracies": 0.9125000238418579,
1071
+ "rewards/chosen": 1.163108229637146,
1072
+ "rewards/margins": 5.6252617835998535,
1073
+ "rewards/rejected": -4.462153434753418,
1074
+ "step": 700
1075
+ },
1076
+ {
1077
+ "epoch": 0.6442831215970962,
1078
+ "grad_norm": 51.69448179259041,
1079
+ "learning_rate": 1.6943691063961213e-07,
1080
+ "logits/chosen": -3.1876442432403564,
1081
+ "logits/rejected": -3.130868434906006,
1082
+ "logps/chosen": -434.44073486328125,
1083
+ "logps/rejected": -342.75146484375,
1084
+ "loss": 0.199,
1085
+ "rewards/accuracies": 0.8999999761581421,
1086
+ "rewards/chosen": 1.4867048263549805,
1087
+ "rewards/margins": 6.818478584289551,
1088
+ "rewards/rejected": -5.331774711608887,
1089
+ "step": 710
1090
+ },
1091
+ {
1092
+ "epoch": 0.6533575317604355,
1093
+ "grad_norm": 42.65687224603704,
1094
+ "learning_rate": 1.6197612122016846e-07,
1095
+ "logits/chosen": -3.15063214302063,
1096
+ "logits/rejected": -3.128218650817871,
1097
+ "logps/chosen": -396.96038818359375,
1098
+ "logps/rejected": -371.03314208984375,
1099
+ "loss": 0.1706,
1100
+ "rewards/accuracies": 0.8500000238418579,
1101
+ "rewards/chosen": 1.3427283763885498,
1102
+ "rewards/margins": 6.249272346496582,
1103
+ "rewards/rejected": -4.906544208526611,
1104
+ "step": 720
1105
+ },
1106
+ {
1107
+ "epoch": 0.662431941923775,
1108
+ "grad_norm": 44.262085666597265,
1109
+ "learning_rate": 1.5460378561536985e-07,
1110
+ "logits/chosen": -3.130981206893921,
1111
+ "logits/rejected": -3.078493356704712,
1112
+ "logps/chosen": -367.4677734375,
1113
+ "logps/rejected": -304.9905090332031,
1114
+ "loss": 0.1785,
1115
+ "rewards/accuracies": 0.8999999761581421,
1116
+ "rewards/chosen": 1.2416002750396729,
1117
+ "rewards/margins": 6.065240383148193,
1118
+ "rewards/rejected": -4.823639869689941,
1119
+ "step": 730
1120
+ },
1121
+ {
1122
+ "epoch": 0.6715063520871143,
1123
+ "grad_norm": 27.300679345139994,
1124
+ "learning_rate": 1.473273121695898e-07,
1125
+ "logits/chosen": -3.127115488052368,
1126
+ "logits/rejected": -3.073434591293335,
1127
+ "logps/chosen": -399.8476867675781,
1128
+ "logps/rejected": -370.6973571777344,
1129
+ "loss": 0.1835,
1130
+ "rewards/accuracies": 0.9125000238418579,
1131
+ "rewards/chosen": 1.0761568546295166,
1132
+ "rewards/margins": 5.881572246551514,
1133
+ "rewards/rejected": -4.805415630340576,
1134
+ "step": 740
1135
+ },
1136
+ {
1137
+ "epoch": 0.6805807622504537,
1138
+ "grad_norm": 50.790995812359,
1139
+ "learning_rate": 1.4015401289682214e-07,
1140
+ "logits/chosen": -3.123532772064209,
1141
+ "logits/rejected": -3.080094814300537,
1142
+ "logps/chosen": -331.3045654296875,
1143
+ "logps/rejected": -314.36358642578125,
1144
+ "loss": 0.2493,
1145
+ "rewards/accuracies": 0.875,
1146
+ "rewards/chosen": 0.7594553232192993,
1147
+ "rewards/margins": 5.120793342590332,
1148
+ "rewards/rejected": -4.3613386154174805,
1149
+ "step": 750
1150
+ },
1151
+ {
1152
+ "epoch": 0.6896551724137931,
1153
+ "grad_norm": 34.305160719125155,
1154
+ "learning_rate": 1.3309109613295335e-07,
1155
+ "logits/chosen": -3.1643004417419434,
1156
+ "logits/rejected": -3.095179796218872,
1157
+ "logps/chosen": -415.06854248046875,
1158
+ "logps/rejected": -358.4546203613281,
1159
+ "loss": 0.1776,
1160
+ "rewards/accuracies": 0.8999999761581421,
1161
+ "rewards/chosen": 0.9015072584152222,
1162
+ "rewards/margins": 5.74955415725708,
1163
+ "rewards/rejected": -4.84804630279541,
1164
+ "step": 760
1165
+ },
1166
+ {
1167
+ "epoch": 0.6987295825771325,
1168
+ "grad_norm": 31.66819953926833,
1169
+ "learning_rate": 1.2614565929221848e-07,
1170
+ "logits/chosen": -3.2078864574432373,
1171
+ "logits/rejected": -3.0819454193115234,
1172
+ "logps/chosen": -367.291748046875,
1173
+ "logps/rejected": -346.97088623046875,
1174
+ "loss": 0.1954,
1175
+ "rewards/accuracies": 0.8999999761581421,
1176
+ "rewards/chosen": 1.1401704549789429,
1177
+ "rewards/margins": 6.157609939575195,
1178
+ "rewards/rejected": -5.017439365386963,
1179
+ "step": 770
1180
+ },
1181
+ {
1182
+ "epoch": 0.7078039927404719,
1183
+ "grad_norm": 36.80087686819721,
1184
+ "learning_rate": 1.1932468173512137e-07,
1185
+ "logits/chosen": -3.1448097229003906,
1186
+ "logits/rejected": -3.0595924854278564,
1187
+ "logps/chosen": -415.1175842285156,
1188
+ "logps/rejected": -330.16680908203125,
1189
+ "loss": 0.1789,
1190
+ "rewards/accuracies": 0.8999999761581421,
1191
+ "rewards/chosen": 1.58966863155365,
1192
+ "rewards/margins": 6.8600358963012695,
1193
+ "rewards/rejected": -5.270366668701172,
1194
+ "step": 780
1195
+ },
1196
+ {
1197
+ "epoch": 0.7168784029038112,
1198
+ "grad_norm": 35.47381560617302,
1199
+ "learning_rate": 1.1263501775498438e-07,
1200
+ "logits/chosen": -3.1903061866760254,
1201
+ "logits/rejected": -3.108388900756836,
1202
+ "logps/chosen": -358.5484313964844,
1203
+ "logps/rejected": -358.26031494140625,
1204
+ "loss": 0.212,
1205
+ "rewards/accuracies": 0.8999999761581421,
1206
+ "rewards/chosen": 0.6974457502365112,
1207
+ "rewards/margins": 4.846839427947998,
1208
+ "rewards/rejected": -4.1493940353393555,
1209
+ "step": 790
1210
+ },
1211
+ {
1212
+ "epoch": 0.7259528130671506,
1213
+ "grad_norm": 40.57531341722196,
1214
+ "learning_rate": 1.0608338969017682e-07,
1215
+ "logits/chosen": -3.199059009552002,
1216
+ "logits/rejected": -3.125755786895752,
1217
+ "logps/chosen": -452.31317138671875,
1218
+ "logps/rejected": -391.41912841796875,
1219
+ "loss": 0.2021,
1220
+ "rewards/accuracies": 0.8999999761581421,
1221
+ "rewards/chosen": 1.6242631673812866,
1222
+ "rewards/margins": 6.886317253112793,
1223
+ "rewards/rejected": -5.262054443359375,
1224
+ "step": 800
1225
+ },
1226
+ {
1227
+ "epoch": 0.73502722323049,
1228
+ "grad_norm": 33.304705085054344,
1229
+ "learning_rate": 9.96763811689425e-08,
1230
+ "logits/chosen": -3.137302875518799,
1231
+ "logits/rejected": -3.026451826095581,
1232
+ "logps/chosen": -385.8514404296875,
1233
+ "logps/rejected": -368.15789794921875,
1234
+ "loss": 0.196,
1235
+ "rewards/accuracies": 0.9125000238418579,
1236
+ "rewards/chosen": 1.4752086400985718,
1237
+ "rewards/margins": 6.488913059234619,
1238
+ "rewards/rejected": -5.013703346252441,
1239
+ "step": 810
1240
+ },
1241
+ {
1242
+ "epoch": 0.7441016333938294,
1243
+ "grad_norm": 47.310232796094745,
1244
+ "learning_rate": 9.3420430493615e-08,
1245
+ "logits/chosen": -3.0361504554748535,
1246
+ "logits/rejected": -2.984409809112549,
1247
+ "logps/chosen": -363.8970031738281,
1248
+ "logps/rejected": -331.4267272949219,
1249
+ "loss": 0.1886,
1250
+ "rewards/accuracies": 0.875,
1251
+ "rewards/chosen": 1.094775676727295,
1252
+ "rewards/margins": 6.252727508544922,
1253
+ "rewards/rejected": -5.1579508781433105,
1254
+ "step": 820
1255
+ },
1256
+ {
1257
+ "epoch": 0.7531760435571688,
1258
+ "grad_norm": 36.729029651490606,
1259
+ "learning_rate": 8.732182417086903e-08,
1260
+ "logits/chosen": -3.07252836227417,
1261
+ "logits/rejected": -3.0885097980499268,
1262
+ "logps/chosen": -402.6160583496094,
1263
+ "logps/rejected": -361.8660888671875,
1264
+ "loss": 0.1873,
1265
+ "rewards/accuracies": 0.925000011920929,
1266
+ "rewards/chosen": 1.8262115716934204,
1267
+ "rewards/margins": 7.684056282043457,
1268
+ "rewards/rejected": -5.857844829559326,
1269
+ "step": 830
1270
+ },
1271
+ {
1272
+ "epoch": 0.7622504537205081,
1273
+ "grad_norm": 38.26434214333302,
1274
+ "learning_rate": 8.138669059450778e-08,
1275
+ "logits/chosen": -3.0775837898254395,
1276
+ "logits/rejected": -3.071493625640869,
1277
+ "logps/chosen": -388.62445068359375,
1278
+ "logps/rejected": -359.02850341796875,
1279
+ "loss": 0.1844,
1280
+ "rewards/accuracies": 0.9125000238418579,
1281
+ "rewards/chosen": 1.0169572830200195,
1282
+ "rewards/margins": 6.112603187561035,
1283
+ "rewards/rejected": -5.095646858215332,
1284
+ "step": 840
1285
+ },
1286
+ {
1287
+ "epoch": 0.7713248638838476,
1288
+ "grad_norm": 46.50356797804205,
1289
+ "learning_rate": 7.562099388713702e-08,
1290
+ "logits/chosen": -3.1154263019561768,
1291
+ "logits/rejected": -3.1136341094970703,
1292
+ "logps/chosen": -388.99761962890625,
1293
+ "logps/rejected": -352.0494689941406,
1294
+ "loss": 0.1793,
1295
+ "rewards/accuracies": 0.949999988079071,
1296
+ "rewards/chosen": 1.4719061851501465,
1297
+ "rewards/margins": 6.116942882537842,
1298
+ "rewards/rejected": -4.6450371742248535,
1299
+ "step": 850
1300
+ },
1301
+ {
1302
+ "epoch": 0.7803992740471869,
1303
+ "grad_norm": 32.98126258960774,
1304
+ "learning_rate": 7.003052790691089e-08,
1305
+ "logits/chosen": -3.0627052783966064,
1306
+ "logits/rejected": -3.0874767303466797,
1307
+ "logps/chosen": -384.46795654296875,
1308
+ "logps/rejected": -338.3650817871094,
1309
+ "loss": 0.1731,
1310
+ "rewards/accuracies": 0.887499988079071,
1311
+ "rewards/chosen": 1.2397066354751587,
1312
+ "rewards/margins": 6.408433437347412,
1313
+ "rewards/rejected": -5.168726921081543,
1314
+ "step": 860
1315
+ },
1316
+ {
1317
+ "epoch": 0.7894736842105263,
1318
+ "grad_norm": 44.77250114819241,
1319
+ "learning_rate": 6.462091042537576e-08,
1320
+ "logits/chosen": -3.144641876220703,
1321
+ "logits/rejected": -3.09205961227417,
1322
+ "logps/chosen": -467.0062561035156,
1323
+ "logps/rejected": -395.1363830566406,
1324
+ "loss": 0.2021,
1325
+ "rewards/accuracies": 0.862500011920929,
1326
+ "rewards/chosen": 1.8845285177230835,
1327
+ "rewards/margins": 7.703507423400879,
1328
+ "rewards/rejected": -5.818978786468506,
1329
+ "step": 870
1330
+ },
1331
+ {
1332
+ "epoch": 0.7985480943738656,
1333
+ "grad_norm": 118.93536447485907,
1334
+ "learning_rate": 5.9397577482259043e-08,
1335
+ "logits/chosen": -3.0359034538269043,
1336
+ "logits/rejected": -3.035188674926758,
1337
+ "logps/chosen": -356.08587646484375,
1338
+ "logps/rejected": -359.81884765625,
1339
+ "loss": 0.1989,
1340
+ "rewards/accuracies": 0.8999999761581421,
1341
+ "rewards/chosen": 0.9912107586860657,
1342
+ "rewards/margins": 6.407823085784912,
1343
+ "rewards/rejected": -5.41661262512207,
1344
+ "step": 880
1345
+ },
1346
+ {
1347
+ "epoch": 0.8076225045372051,
1348
+ "grad_norm": 44.83410379465266,
1349
+ "learning_rate": 5.436577792287841e-08,
1350
+ "logits/chosen": -3.1080565452575684,
1351
+ "logits/rejected": -3.024658679962158,
1352
+ "logps/chosen": -350.00421142578125,
1353
+ "logps/rejected": -313.59234619140625,
1354
+ "loss": 0.184,
1355
+ "rewards/accuracies": 0.9125000238418579,
1356
+ "rewards/chosen": 1.5829899311065674,
1357
+ "rewards/margins": 7.088255882263184,
1358
+ "rewards/rejected": -5.505266189575195,
1359
+ "step": 890
1360
+ },
1361
+ {
1362
+ "epoch": 0.8166969147005445,
1363
+ "grad_norm": 38.85789898863309,
1364
+ "learning_rate": 4.953056812365958e-08,
1365
+ "logits/chosen": -3.08073353767395,
1366
+ "logits/rejected": -3.0578064918518066,
1367
+ "logps/chosen": -371.7347717285156,
1368
+ "logps/rejected": -363.08514404296875,
1369
+ "loss": 0.1924,
1370
+ "rewards/accuracies": 0.875,
1371
+ "rewards/chosen": 1.1432292461395264,
1372
+ "rewards/margins": 6.080376148223877,
1373
+ "rewards/rejected": -4.9371466636657715,
1374
+ "step": 900
1375
+ },
1376
+ {
1377
+ "epoch": 0.8257713248638838,
1378
+ "grad_norm": 30.66043802418784,
1379
+ "learning_rate": 4.489680691106279e-08,
1380
+ "logits/chosen": -3.1917724609375,
1381
+ "logits/rejected": -3.0947258472442627,
1382
+ "logps/chosen": -456.81982421875,
1383
+ "logps/rejected": -366.9622497558594,
1384
+ "loss": 0.1746,
1385
+ "rewards/accuracies": 0.9375,
1386
+ "rewards/chosen": 1.2509504556655884,
1387
+ "rewards/margins": 6.275618076324463,
1388
+ "rewards/rejected": -5.024667739868164,
1389
+ "step": 910
1390
+ },
1391
+ {
1392
+ "epoch": 0.8348457350272233,
1393
+ "grad_norm": 31.60875396882119,
1394
+ "learning_rate": 4.046915067902443e-08,
1395
+ "logits/chosen": -3.182232141494751,
1396
+ "logits/rejected": -3.031777858734131,
1397
+ "logps/chosen": -385.55963134765625,
1398
+ "logps/rejected": -349.5807800292969,
1399
+ "loss": 0.1717,
1400
+ "rewards/accuracies": 0.949999988079071,
1401
+ "rewards/chosen": 1.7857484817504883,
1402
+ "rewards/margins": 7.652297019958496,
1403
+ "rewards/rejected": -5.866549491882324,
1404
+ "step": 920
1405
+ },
1406
+ {
1407
+ "epoch": 0.8439201451905626,
1408
+ "grad_norm": 44.88851150727403,
1409
+ "learning_rate": 3.625204870981974e-08,
1410
+ "logits/chosen": -3.136620044708252,
1411
+ "logits/rejected": -3.059222459793091,
1412
+ "logps/chosen": -377.5750427246094,
1413
+ "logps/rejected": -349.54327392578125,
1414
+ "loss": 0.1835,
1415
+ "rewards/accuracies": 0.9125000238418579,
1416
+ "rewards/chosen": 1.178045630455017,
1417
+ "rewards/margins": 6.5388946533203125,
1418
+ "rewards/rejected": -5.360849380493164,
1419
+ "step": 930
1420
+ },
1421
+ {
1422
+ "epoch": 0.852994555353902,
1423
+ "grad_norm": 45.78958197714914,
1424
+ "learning_rate": 3.2249738703049175e-08,
1425
+ "logits/chosen": -3.097872257232666,
1426
+ "logits/rejected": -3.0364792346954346,
1427
+ "logps/chosen": -413.3067321777344,
1428
+ "logps/rejected": -384.83538818359375,
1429
+ "loss": 0.1759,
1430
+ "rewards/accuracies": 0.925000011920929,
1431
+ "rewards/chosen": 1.3575388193130493,
1432
+ "rewards/margins": 6.335871696472168,
1433
+ "rewards/rejected": -4.978333473205566,
1434
+ "step": 940
1435
+ },
1436
+ {
1437
+ "epoch": 0.8620689655172413,
1438
+ "grad_norm": 35.84076730560574,
1439
+ "learning_rate": 2.8466242517240142e-08,
1440
+ "logits/chosen": -3.125441551208496,
1441
+ "logits/rejected": -3.012545108795166,
1442
+ "logps/chosen": -380.4359436035156,
1443
+ "logps/rejected": -364.7444152832031,
1444
+ "loss": 0.1642,
1445
+ "rewards/accuracies": 0.949999988079071,
1446
+ "rewards/chosen": 1.327865719795227,
1447
+ "rewards/margins": 6.545771598815918,
1448
+ "rewards/rejected": -5.2179059982299805,
1449
+ "step": 950
1450
+ },
1451
+ {
1452
+ "epoch": 0.8711433756805808,
1453
+ "grad_norm": 44.73068485554119,
1454
+ "learning_rate": 2.4905362128344652e-08,
1455
+ "logits/chosen": -3.1192257404327393,
1456
+ "logits/rejected": -3.086743116378784,
1457
+ "logps/chosen": -386.00775146484375,
1458
+ "logps/rejected": -357.411865234375,
1459
+ "loss": 0.1977,
1460
+ "rewards/accuracies": 0.875,
1461
+ "rewards/chosen": 1.0490394830703735,
1462
+ "rewards/margins": 5.79413366317749,
1463
+ "rewards/rejected": -4.745095252990723,
1464
+ "step": 960
1465
+ },
1466
+ {
1467
+ "epoch": 0.8802177858439202,
1468
+ "grad_norm": 46.24126513458063,
1469
+ "learning_rate": 2.1570675809193554e-08,
1470
+ "logits/chosen": -3.0832366943359375,
1471
+ "logits/rejected": -3.063671350479126,
1472
+ "logps/chosen": -352.98785400390625,
1473
+ "logps/rejected": -328.38775634765625,
1474
+ "loss": 0.1693,
1475
+ "rewards/accuracies": 0.8999999761581421,
1476
+ "rewards/chosen": 1.1795412302017212,
1477
+ "rewards/margins": 6.687762260437012,
1478
+ "rewards/rejected": -5.508220672607422,
1479
+ "step": 970
1480
+ },
1481
+ {
1482
+ "epoch": 0.8892921960072595,
1483
+ "grad_norm": 40.90013103941717,
1484
+ "learning_rate": 1.846553453374586e-08,
1485
+ "logits/chosen": -3.10916805267334,
1486
+ "logits/rejected": -3.096108913421631,
1487
+ "logps/chosen": -332.032958984375,
1488
+ "logps/rejected": -333.1717529296875,
1489
+ "loss": 0.19,
1490
+ "rewards/accuracies": 0.862500011920929,
1491
+ "rewards/chosen": 0.7762575149536133,
1492
+ "rewards/margins": 5.550572395324707,
1493
+ "rewards/rejected": -4.774314880371094,
1494
+ "step": 980
1495
+ },
1496
+ {
1497
+ "epoch": 0.8983666061705989,
1498
+ "grad_norm": 34.87323084074699,
1499
+ "learning_rate": 1.559305860974805e-08,
1500
+ "logits/chosen": -3.114400625228882,
1501
+ "logits/rejected": -3.0474765300750732,
1502
+ "logps/chosen": -375.05072021484375,
1503
+ "logps/rejected": -326.6473693847656,
1504
+ "loss": 0.1674,
1505
+ "rewards/accuracies": 0.9624999761581421,
1506
+ "rewards/chosen": 1.794228196144104,
1507
+ "rewards/margins": 7.251187324523926,
1508
+ "rewards/rejected": -5.456958293914795,
1509
+ "step": 990
1510
+ },
1511
+ {
1512
+ "epoch": 0.9074410163339383,
1513
+ "grad_norm": 68.79219384818164,
1514
+ "learning_rate": 1.2956134543185449e-08,
1515
+ "logits/chosen": -3.1376454830169678,
1516
+ "logits/rejected": -3.045968532562256,
1517
+ "logps/chosen": -385.92266845703125,
1518
+ "logps/rejected": -303.8971252441406,
1519
+ "loss": 0.1882,
1520
+ "rewards/accuracies": 0.8999999761581421,
1521
+ "rewards/chosen": 1.1436764001846313,
1522
+ "rewards/margins": 6.106300354003906,
1523
+ "rewards/rejected": -4.962623596191406,
1524
+ "step": 1000
1525
+ },
1526
+ {
1527
+ "epoch": 0.9165154264972777,
1528
+ "grad_norm": 41.53283307570751,
1529
+ "learning_rate": 1.0557412137677884e-08,
1530
+ "logits/chosen": -3.101411819458008,
1531
+ "logits/rejected": -3.023592472076416,
1532
+ "logps/chosen": -395.31597900390625,
1533
+ "logps/rejected": -361.7346496582031,
1534
+ "loss": 0.1771,
1535
+ "rewards/accuracies": 0.8999999761581421,
1536
+ "rewards/chosen": 0.987866222858429,
1537
+ "rewards/margins": 5.9940056800842285,
1538
+ "rewards/rejected": -5.006139755249023,
1539
+ "step": 1010
1540
+ },
1541
+ {
1542
+ "epoch": 0.925589836660617,
1543
+ "grad_norm": 38.24151872414285,
1544
+ "learning_rate": 8.399301831733403e-09,
1545
+ "logits/chosen": -3.124439239501953,
1546
+ "logits/rejected": -3.0538604259490967,
1547
+ "logps/chosen": -365.45355224609375,
1548
+ "logps/rejected": -359.1190490722656,
1549
+ "loss": 0.1784,
1550
+ "rewards/accuracies": 0.887499988079071,
1551
+ "rewards/chosen": 1.254525899887085,
1552
+ "rewards/margins": 6.763253688812256,
1553
+ "rewards/rejected": -5.508728981018066,
1554
+ "step": 1020
1555
+ },
1556
+ {
1557
+ "epoch": 0.9346642468239564,
1558
+ "grad_norm": 42.11264132797008,
1559
+ "learning_rate": 6.483972276536576e-09,
1560
+ "logits/chosen": -3.1174588203430176,
1561
+ "logits/rejected": -3.0175864696502686,
1562
+ "logps/chosen": -428.16937255859375,
1563
+ "logps/rejected": -335.56707763671875,
1564
+ "loss": 0.1856,
1565
+ "rewards/accuracies": 0.949999988079071,
1566
+ "rewards/chosen": 1.0685935020446777,
1567
+ "rewards/margins": 6.026675701141357,
1568
+ "rewards/rejected": -4.95808219909668,
1569
+ "step": 1030
1570
+ },
1571
+ {
1572
+ "epoch": 0.9437386569872959,
1573
+ "grad_norm": 48.30353514815893,
1574
+ "learning_rate": 4.813348156704866e-09,
1575
+ "logits/chosen": -3.0982871055603027,
1576
+ "logits/rejected": -3.007462978363037,
1577
+ "logps/chosen": -365.95343017578125,
1578
+ "logps/rejected": -376.6304626464844,
1579
+ "loss": 0.1849,
1580
+ "rewards/accuracies": 0.925000011920929,
1581
+ "rewards/chosen": 1.2602510452270508,
1582
+ "rewards/margins": 7.303647041320801,
1583
+ "rewards/rejected": -6.043395042419434,
1584
+ "step": 1040
1585
+ },
1586
+ {
1587
+ "epoch": 0.9528130671506352,
1588
+ "grad_norm": 42.00692162522527,
1589
+ "learning_rate": 3.389108256203338e-09,
1590
+ "logits/chosen": -3.146289348602295,
1591
+ "logits/rejected": -2.9762940406799316,
1592
+ "logps/chosen": -395.01666259765625,
1593
+ "logps/rejected": -327.5099792480469,
1594
+ "loss": 0.1715,
1595
+ "rewards/accuracies": 0.8999999761581421,
1596
+ "rewards/chosen": 1.2526695728302002,
1597
+ "rewards/margins": 6.668323516845703,
1598
+ "rewards/rejected": -5.415653705596924,
1599
+ "step": 1050
1600
+ },
1601
+ {
1602
+ "epoch": 0.9618874773139746,
1603
+ "grad_norm": 65.97199815970751,
1604
+ "learning_rate": 2.2126837713609403e-09,
1605
+ "logits/chosen": -3.0963053703308105,
1606
+ "logits/rejected": -3.0015339851379395,
1607
+ "logps/chosen": -363.32708740234375,
1608
+ "logps/rejected": -339.29852294921875,
1609
+ "loss": 0.1755,
1610
+ "rewards/accuracies": 0.925000011920929,
1611
+ "rewards/chosen": 1.2676588296890259,
1612
+ "rewards/margins": 6.559351444244385,
1613
+ "rewards/rejected": -5.29169225692749,
1614
+ "step": 1060
1615
+ },
1616
+ {
1617
+ "epoch": 0.9709618874773139,
1618
+ "grad_norm": 46.66363739025269,
1619
+ "learning_rate": 1.2852568726837987e-09,
1620
+ "logits/chosen": -3.1610124111175537,
1621
+ "logits/rejected": -3.092536449432373,
1622
+ "logps/chosen": -431.3716735839844,
1623
+ "logps/rejected": -377.6732177734375,
1624
+ "loss": 0.2216,
1625
+ "rewards/accuracies": 0.9125000238418579,
1626
+ "rewards/chosen": 1.3350967168807983,
1627
+ "rewards/margins": 6.358585834503174,
1628
+ "rewards/rejected": -5.023488998413086,
1629
+ "step": 1070
1630
+ },
1631
+ {
1632
+ "epoch": 0.9800362976406534,
1633
+ "grad_norm": 30.34640117074649,
1634
+ "learning_rate": 6.077595169105277e-10,
1635
+ "logits/chosen": -3.076800584793091,
1636
+ "logits/rejected": -3.0786032676696777,
1637
+ "logps/chosen": -358.31878662109375,
1638
+ "logps/rejected": -334.080810546875,
1639
+ "loss": 0.1916,
1640
+ "rewards/accuracies": 0.9375,
1641
+ "rewards/chosen": 1.380605936050415,
1642
+ "rewards/margins": 6.937164306640625,
1643
+ "rewards/rejected": -5.556557655334473,
1644
+ "step": 1080
1645
+ },
1646
+ {
1647
+ "epoch": 0.9891107078039928,
1648
+ "grad_norm": 32.7478445581115,
1649
+ "learning_rate": 1.8087251050369344e-10,
1650
+ "logits/chosen": -3.1240782737731934,
1651
+ "logits/rejected": -3.0358004570007324,
1652
+ "logps/chosen": -377.9797668457031,
1653
+ "logps/rejected": -362.23028564453125,
1654
+ "loss": 0.1972,
1655
+ "rewards/accuracies": 0.9375,
1656
+ "rewards/chosen": 1.7212746143341064,
1657
+ "rewards/margins": 7.479613304138184,
1658
+ "rewards/rejected": -5.758338451385498,
1659
+ "step": 1090
1660
+ },
1661
+ {
1662
+ "epoch": 0.9981851179673321,
1663
+ "grad_norm": 39.30875311875232,
1664
+ "learning_rate": 5.024825517951914e-12,
1665
+ "logits/chosen": -3.1050829887390137,
1666
+ "logits/rejected": -3.1044745445251465,
1667
+ "logps/chosen": -381.04168701171875,
1668
+ "logps/rejected": -363.96197509765625,
1669
+ "loss": 0.1954,
1670
+ "rewards/accuracies": 0.9125000238418579,
1671
+ "rewards/chosen": 1.0158889293670654,
1672
+ "rewards/margins": 5.7684783935546875,
1673
+ "rewards/rejected": -4.752589225769043,
1674
+ "step": 1100
1675
+ },
1676
+ {
1677
+ "epoch": 1.0,
1678
+ "step": 1102,
1679
+ "total_flos": 0.0,
1680
+ "train_loss": 0.22562925693344074,
1681
+ "train_runtime": 6258.663,
1682
+ "train_samples_per_second": 11.267,
1683
+ "train_steps_per_second": 0.176
1684
+ }
1685
+ ],
1686
+ "logging_steps": 10,
1687
+ "max_steps": 1102,
1688
+ "num_input_tokens_seen": 0,
1689
+ "num_train_epochs": 1,
1690
+ "save_steps": 500,
1691
+ "stateful_callbacks": {
1692
+ "TrainerControl": {
1693
+ "args": {
1694
+ "should_epoch_stop": false,
1695
+ "should_evaluate": false,
1696
+ "should_log": false,
1697
+ "should_save": true,
1698
+ "should_training_stop": true
1699
+ },
1700
+ "attributes": {}
1701
+ }
1702
+ },
1703
+ "total_flos": 0.0,
1704
+ "train_batch_size": 4,
1705
+ "trial_name": null,
1706
+ "trial_params": null
1707
+ }