lole25 commited on
Commit
1d0ea21
1 Parent(s): 3dd6f42

Model save

Browse files
README.md ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: peft
4
+ tags:
5
+ - trl
6
+ - dpo
7
+ - generated_from_trainer
8
+ base_model: DUAL-GPO/phi-2-irepo-chatml-merged-i1
9
+ model-index:
10
+ - name: phi-2-irepo-chatml-v4-i2
11
+ results: []
12
+ ---
13
+
14
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
+ should probably proofread and complete it, then remove this comment. -->
16
+
17
+ # phi-2-irepo-chatml-v4-i2
18
+
19
+ This model is a fine-tuned version of [DUAL-GPO/phi-2-irepo-chatml-merged-i1](https://huggingface.co/DUAL-GPO/phi-2-irepo-chatml-merged-i1) on the None dataset.
20
+
21
+ ## Model description
22
+
23
+ More information needed
24
+
25
+ ## Intended uses & limitations
26
+
27
+ More information needed
28
+
29
+ ## Training and evaluation data
30
+
31
+ More information needed
32
+
33
+ ## Training procedure
34
+
35
+ ### Training hyperparameters
36
+
37
+ The following hyperparameters were used during training:
38
+ - learning_rate: 5e-06
39
+ - train_batch_size: 4
40
+ - eval_batch_size: 4
41
+ - seed: 42
42
+ - distributed_type: multi-GPU
43
+ - gradient_accumulation_steps: 4
44
+ - total_train_batch_size: 16
45
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
46
+ - lr_scheduler_type: cosine
47
+ - lr_scheduler_warmup_ratio: 0.1
48
+ - num_epochs: 1
49
+
50
+ ### Training results
51
+
52
+
53
+
54
+ ### Framework versions
55
+
56
+ - PEFT 0.7.1
57
+ - Transformers 4.36.2
58
+ - Pytorch 2.1.2+cu121
59
+ - Datasets 2.14.6
60
+ - Tokenizers 0.15.2
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7abdfe181d0d18a6df74036bfdf162de668e1aa99b7578cc56ef7714bd4958fb
3
  size 335579632
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df1533bfce11a58af9ae234bf668afdaf0ac258c8d62f4f9ff386525b7424fee
3
  size 335579632
all_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.19300631308555602,
4
+ "train_runtime": 11612.5974,
5
+ "train_samples": 30000,
6
+ "train_samples_per_second": 2.583,
7
+ "train_steps_per_second": 0.161
8
+ }
runs/May22_09-27-31_gpu4-119-5/events.out.tfevents.1716334211.gpu4-119-5.2627167.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cb35923953465771357b4a417bb0274e5cacf7883f74d25d43eb3b91d2287bf8
3
- size 119520
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fba1c654a7e49b8d28d8c79eef29eeb8bdb0aa448d48ba27ca4bc410f9d2529f
3
+ size 124312
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.19300631308555602,
4
+ "train_runtime": 11612.5974,
5
+ "train_samples": 30000,
6
+ "train_samples_per_second": 2.583,
7
+ "train_steps_per_second": 0.161
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,2662 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 500,
6
+ "global_step": 1875,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0,
13
+ "learning_rate": 2.6595744680851065e-08,
14
+ "logits/chosen": -0.1435682475566864,
15
+ "logits/rejected": -0.2822338938713074,
16
+ "logps/chosen": -552.8109130859375,
17
+ "logps/rejected": -630.5926513671875,
18
+ "loss": 0.247,
19
+ "rewards/accuracies": 0.0,
20
+ "rewards/chosen": 0.0,
21
+ "rewards/margins": 0.0,
22
+ "rewards/rejected": 0.0,
23
+ "step": 1
24
+ },
25
+ {
26
+ "epoch": 0.01,
27
+ "learning_rate": 2.6595744680851066e-07,
28
+ "logits/chosen": -0.16537699103355408,
29
+ "logits/rejected": -0.1148439273238182,
30
+ "logps/chosen": -533.032470703125,
31
+ "logps/rejected": -578.028564453125,
32
+ "loss": 0.2072,
33
+ "rewards/accuracies": 0.375,
34
+ "rewards/chosen": 8.621174492873251e-05,
35
+ "rewards/margins": 2.2910655388841406e-05,
36
+ "rewards/rejected": 6.330109317786992e-05,
37
+ "step": 10
38
+ },
39
+ {
40
+ "epoch": 0.01,
41
+ "learning_rate": 5.319148936170213e-07,
42
+ "logits/chosen": -0.17788179218769073,
43
+ "logits/rejected": -0.07320378720760345,
44
+ "logps/chosen": -506.20989990234375,
45
+ "logps/rejected": -561.3789672851562,
46
+ "loss": 0.2124,
47
+ "rewards/accuracies": 0.38749998807907104,
48
+ "rewards/chosen": -0.00013401822070591152,
49
+ "rewards/margins": 1.3708593542105518e-05,
50
+ "rewards/rejected": -0.00014772682334296405,
51
+ "step": 20
52
+ },
53
+ {
54
+ "epoch": 0.02,
55
+ "learning_rate": 7.97872340425532e-07,
56
+ "logits/chosen": -0.19918851554393768,
57
+ "logits/rejected": -0.13439549505710602,
58
+ "logps/chosen": -486.73468017578125,
59
+ "logps/rejected": -483.3734436035156,
60
+ "loss": 0.2112,
61
+ "rewards/accuracies": 0.4000000059604645,
62
+ "rewards/chosen": -0.0002551859070081264,
63
+ "rewards/margins": -5.1160714065190405e-05,
64
+ "rewards/rejected": -0.00020402518566697836,
65
+ "step": 30
66
+ },
67
+ {
68
+ "epoch": 0.02,
69
+ "learning_rate": 1.0638297872340427e-06,
70
+ "logits/chosen": -0.12800553441047668,
71
+ "logits/rejected": -0.045729465782642365,
72
+ "logps/chosen": -474.02825927734375,
73
+ "logps/rejected": -471.7972717285156,
74
+ "loss": 0.1915,
75
+ "rewards/accuracies": 0.39375001192092896,
76
+ "rewards/chosen": -0.0005546818720176816,
77
+ "rewards/margins": -1.8026912584900856e-05,
78
+ "rewards/rejected": -0.0005366549594327807,
79
+ "step": 40
80
+ },
81
+ {
82
+ "epoch": 0.03,
83
+ "learning_rate": 1.3297872340425533e-06,
84
+ "logits/chosen": -0.16574732959270477,
85
+ "logits/rejected": -0.09961490333080292,
86
+ "logps/chosen": -512.2724609375,
87
+ "logps/rejected": -547.2150268554688,
88
+ "loss": 0.2085,
89
+ "rewards/accuracies": 0.4749999940395355,
90
+ "rewards/chosen": -0.00024178791500162333,
91
+ "rewards/margins": 0.00024111787206493318,
92
+ "rewards/rejected": -0.00048290580161847174,
93
+ "step": 50
94
+ },
95
+ {
96
+ "epoch": 0.03,
97
+ "learning_rate": 1.595744680851064e-06,
98
+ "logits/chosen": -0.10868887603282928,
99
+ "logits/rejected": -0.14005155861377716,
100
+ "logps/chosen": -513.1686401367188,
101
+ "logps/rejected": -552.2679443359375,
102
+ "loss": 0.1949,
103
+ "rewards/accuracies": 0.44999998807907104,
104
+ "rewards/chosen": -0.00046477210707962513,
105
+ "rewards/margins": 0.00030607733060605824,
106
+ "rewards/rejected": -0.0007708495249971747,
107
+ "step": 60
108
+ },
109
+ {
110
+ "epoch": 0.04,
111
+ "learning_rate": 1.8617021276595745e-06,
112
+ "logits/chosen": -0.1557120382785797,
113
+ "logits/rejected": -0.14936389029026031,
114
+ "logps/chosen": -495.12396240234375,
115
+ "logps/rejected": -488.8611755371094,
116
+ "loss": 0.2186,
117
+ "rewards/accuracies": 0.4124999940395355,
118
+ "rewards/chosen": -0.0014903920236974955,
119
+ "rewards/margins": 6.922415923327208e-05,
120
+ "rewards/rejected": -0.0015596161829307675,
121
+ "step": 70
122
+ },
123
+ {
124
+ "epoch": 0.04,
125
+ "learning_rate": 2.1276595744680853e-06,
126
+ "logits/chosen": -0.09320048987865448,
127
+ "logits/rejected": -0.19148024916648865,
128
+ "logps/chosen": -500.7992248535156,
129
+ "logps/rejected": -555.9862060546875,
130
+ "loss": 0.2,
131
+ "rewards/accuracies": 0.45625001192092896,
132
+ "rewards/chosen": -0.0017530046170577407,
133
+ "rewards/margins": 0.0014083541464060545,
134
+ "rewards/rejected": -0.003161358879879117,
135
+ "step": 80
136
+ },
137
+ {
138
+ "epoch": 0.05,
139
+ "learning_rate": 2.393617021276596e-06,
140
+ "logits/chosen": -0.1890203058719635,
141
+ "logits/rejected": -0.10575833171606064,
142
+ "logps/chosen": -448.5530700683594,
143
+ "logps/rejected": -470.9744567871094,
144
+ "loss": 0.209,
145
+ "rewards/accuracies": 0.40625,
146
+ "rewards/chosen": -0.003437761915847659,
147
+ "rewards/margins": 0.0008986371685750782,
148
+ "rewards/rejected": -0.0043363990262150764,
149
+ "step": 90
150
+ },
151
+ {
152
+ "epoch": 0.05,
153
+ "learning_rate": 2.6595744680851065e-06,
154
+ "logits/chosen": -0.1906287968158722,
155
+ "logits/rejected": -0.15528397262096405,
156
+ "logps/chosen": -502.1371154785156,
157
+ "logps/rejected": -580.352294921875,
158
+ "loss": 0.1987,
159
+ "rewards/accuracies": 0.44999998807907104,
160
+ "rewards/chosen": -0.005366797558963299,
161
+ "rewards/margins": 0.003352800849825144,
162
+ "rewards/rejected": -0.00871959887444973,
163
+ "step": 100
164
+ },
165
+ {
166
+ "epoch": 0.06,
167
+ "learning_rate": 2.9255319148936174e-06,
168
+ "logits/chosen": -0.1939711570739746,
169
+ "logits/rejected": -0.13397586345672607,
170
+ "logps/chosen": -510.639404296875,
171
+ "logps/rejected": -555.0497436523438,
172
+ "loss": 0.2059,
173
+ "rewards/accuracies": 0.44999998807907104,
174
+ "rewards/chosen": -0.01002412661910057,
175
+ "rewards/margins": 0.003309743944555521,
176
+ "rewards/rejected": -0.013333870097994804,
177
+ "step": 110
178
+ },
179
+ {
180
+ "epoch": 0.06,
181
+ "learning_rate": 3.191489361702128e-06,
182
+ "logits/chosen": -0.13544224202632904,
183
+ "logits/rejected": -0.1556098461151123,
184
+ "logps/chosen": -558.7181396484375,
185
+ "logps/rejected": -572.3422241210938,
186
+ "loss": 0.1984,
187
+ "rewards/accuracies": 0.4000000059604645,
188
+ "rewards/chosen": -0.016881179064512253,
189
+ "rewards/margins": 0.001452689990401268,
190
+ "rewards/rejected": -0.01833386905491352,
191
+ "step": 120
192
+ },
193
+ {
194
+ "epoch": 0.07,
195
+ "learning_rate": 3.457446808510639e-06,
196
+ "logits/chosen": -0.23009908199310303,
197
+ "logits/rejected": -0.06849310547113419,
198
+ "logps/chosen": -488.73760986328125,
199
+ "logps/rejected": -536.1317138671875,
200
+ "loss": 0.2037,
201
+ "rewards/accuracies": 0.512499988079071,
202
+ "rewards/chosen": -0.013895799405872822,
203
+ "rewards/margins": 0.0098473085090518,
204
+ "rewards/rejected": -0.02374310791492462,
205
+ "step": 130
206
+ },
207
+ {
208
+ "epoch": 0.07,
209
+ "learning_rate": 3.723404255319149e-06,
210
+ "logits/chosen": -0.22787579894065857,
211
+ "logits/rejected": -0.22026856243610382,
212
+ "logps/chosen": -506.5633850097656,
213
+ "logps/rejected": -527.6223754882812,
214
+ "loss": 0.2149,
215
+ "rewards/accuracies": 0.41874998807907104,
216
+ "rewards/chosen": -0.023789843544363976,
217
+ "rewards/margins": 0.007460168097168207,
218
+ "rewards/rejected": -0.031250011175870895,
219
+ "step": 140
220
+ },
221
+ {
222
+ "epoch": 0.08,
223
+ "learning_rate": 3.98936170212766e-06,
224
+ "logits/chosen": -0.19232454895973206,
225
+ "logits/rejected": -0.20156431198120117,
226
+ "logps/chosen": -561.6283569335938,
227
+ "logps/rejected": -535.9696655273438,
228
+ "loss": 0.2204,
229
+ "rewards/accuracies": 0.40625,
230
+ "rewards/chosen": -0.03427774831652641,
231
+ "rewards/margins": 0.0032845172099769115,
232
+ "rewards/rejected": -0.03756226599216461,
233
+ "step": 150
234
+ },
235
+ {
236
+ "epoch": 0.09,
237
+ "learning_rate": 4.255319148936171e-06,
238
+ "logits/chosen": -0.12956355512142181,
239
+ "logits/rejected": -0.2759852707386017,
240
+ "logps/chosen": -593.6724243164062,
241
+ "logps/rejected": -623.8087158203125,
242
+ "loss": 0.204,
243
+ "rewards/accuracies": 0.48750001192092896,
244
+ "rewards/chosen": -0.03722642734646797,
245
+ "rewards/margins": 0.012259968556463718,
246
+ "rewards/rejected": -0.04948639124631882,
247
+ "step": 160
248
+ },
249
+ {
250
+ "epoch": 0.09,
251
+ "learning_rate": 4.521276595744681e-06,
252
+ "logits/chosen": -0.29936036467552185,
253
+ "logits/rejected": -0.23110389709472656,
254
+ "logps/chosen": -537.4310913085938,
255
+ "logps/rejected": -559.8766479492188,
256
+ "loss": 0.2115,
257
+ "rewards/accuracies": 0.4312500059604645,
258
+ "rewards/chosen": -0.04294046759605408,
259
+ "rewards/margins": 0.013050531037151814,
260
+ "rewards/rejected": -0.05599099397659302,
261
+ "step": 170
262
+ },
263
+ {
264
+ "epoch": 0.1,
265
+ "learning_rate": 4.787234042553192e-06,
266
+ "logits/chosen": -0.22940175235271454,
267
+ "logits/rejected": -0.15090861916542053,
268
+ "logps/chosen": -498.94219970703125,
269
+ "logps/rejected": -544.394287109375,
270
+ "loss": 0.1869,
271
+ "rewards/accuracies": 0.46875,
272
+ "rewards/chosen": -0.03456363454461098,
273
+ "rewards/margins": 0.021605905145406723,
274
+ "rewards/rejected": -0.0561695322394371,
275
+ "step": 180
276
+ },
277
+ {
278
+ "epoch": 0.1,
279
+ "learning_rate": 4.999982660399688e-06,
280
+ "logits/chosen": -0.25657087564468384,
281
+ "logits/rejected": -0.2293904572725296,
282
+ "logps/chosen": -565.2684936523438,
283
+ "logps/rejected": -628.2935791015625,
284
+ "loss": 0.1942,
285
+ "rewards/accuracies": 0.5,
286
+ "rewards/chosen": -0.05312346667051315,
287
+ "rewards/margins": 0.03432563692331314,
288
+ "rewards/rejected": -0.0874490961432457,
289
+ "step": 190
290
+ },
291
+ {
292
+ "epoch": 0.11,
293
+ "learning_rate": 4.99937579964398e-06,
294
+ "logits/chosen": -0.27247872948646545,
295
+ "logits/rejected": -0.16829442977905273,
296
+ "logps/chosen": -588.1021728515625,
297
+ "logps/rejected": -624.6795654296875,
298
+ "loss": 0.2013,
299
+ "rewards/accuracies": 0.4625000059604645,
300
+ "rewards/chosen": -0.07817868888378143,
301
+ "rewards/margins": 0.012142921797931194,
302
+ "rewards/rejected": -0.0903216004371643,
303
+ "step": 200
304
+ },
305
+ {
306
+ "epoch": 0.11,
307
+ "learning_rate": 4.9979021993870645e-06,
308
+ "logits/chosen": -0.1175961121916771,
309
+ "logits/rejected": -0.19183820486068726,
310
+ "logps/chosen": -587.1693115234375,
311
+ "logps/rejected": -589.8121337890625,
312
+ "loss": 0.1982,
313
+ "rewards/accuracies": 0.44999998807907104,
314
+ "rewards/chosen": -0.05276166647672653,
315
+ "rewards/margins": 0.009214815683662891,
316
+ "rewards/rejected": -0.06197648122906685,
317
+ "step": 210
318
+ },
319
+ {
320
+ "epoch": 0.12,
321
+ "learning_rate": 4.995562370647553e-06,
322
+ "logits/chosen": -0.1543715000152588,
323
+ "logits/rejected": -0.12788452208042145,
324
+ "logps/chosen": -565.2959594726562,
325
+ "logps/rejected": -637.3449096679688,
326
+ "loss": 0.2083,
327
+ "rewards/accuracies": 0.4937500059604645,
328
+ "rewards/chosen": -0.036317385733127594,
329
+ "rewards/margins": 0.019043080508708954,
330
+ "rewards/rejected": -0.05536046624183655,
331
+ "step": 220
332
+ },
333
+ {
334
+ "epoch": 0.12,
335
+ "learning_rate": 4.992357124836838e-06,
336
+ "logits/chosen": -0.24705199897289276,
337
+ "logits/rejected": -0.26284027099609375,
338
+ "logps/chosen": -516.439453125,
339
+ "logps/rejected": -574.214111328125,
340
+ "loss": 0.1917,
341
+ "rewards/accuracies": 0.44999998807907104,
342
+ "rewards/chosen": -0.028448155149817467,
343
+ "rewards/margins": 0.02016386389732361,
344
+ "rewards/rejected": -0.048612017184495926,
345
+ "step": 230
346
+ },
347
+ {
348
+ "epoch": 0.13,
349
+ "learning_rate": 4.9882875734777044e-06,
350
+ "logits/chosen": -0.2014235258102417,
351
+ "logits/rejected": -0.05569050833582878,
352
+ "logps/chosen": -596.5665893554688,
353
+ "logps/rejected": -604.5208129882812,
354
+ "loss": 0.2009,
355
+ "rewards/accuracies": 0.48124998807907104,
356
+ "rewards/chosen": -0.032715924084186554,
357
+ "rewards/margins": 0.013567534275352955,
358
+ "rewards/rejected": -0.046283453702926636,
359
+ "step": 240
360
+ },
361
+ {
362
+ "epoch": 0.13,
363
+ "learning_rate": 4.983355127818882e-06,
364
+ "logits/chosen": -0.15398232638835907,
365
+ "logits/rejected": -0.1951122283935547,
366
+ "logps/chosen": -564.0726318359375,
367
+ "logps/rejected": -575.0238037109375,
368
+ "loss": 0.2,
369
+ "rewards/accuracies": 0.4124999940395355,
370
+ "rewards/chosen": -0.045795269310474396,
371
+ "rewards/margins": 0.016760127618908882,
372
+ "rewards/rejected": -0.06255538761615753,
373
+ "step": 250
374
+ },
375
+ {
376
+ "epoch": 0.14,
377
+ "learning_rate": 4.977561498345639e-06,
378
+ "logits/chosen": -0.22048966586589813,
379
+ "logits/rejected": -0.14720526337623596,
380
+ "logps/chosen": -541.268310546875,
381
+ "logps/rejected": -575.94482421875,
382
+ "loss": 0.1918,
383
+ "rewards/accuracies": 0.5375000238418579,
384
+ "rewards/chosen": -0.039269499480724335,
385
+ "rewards/margins": 0.02872983179986477,
386
+ "rewards/rejected": -0.06799933314323425,
387
+ "step": 260
388
+ },
389
+ {
390
+ "epoch": 0.14,
391
+ "learning_rate": 4.970908694186624e-06,
392
+ "logits/chosen": -0.12767961621284485,
393
+ "logits/rejected": -0.17970682680606842,
394
+ "logps/chosen": -586.6306762695312,
395
+ "logps/rejected": -654.3707275390625,
396
+ "loss": 0.1793,
397
+ "rewards/accuracies": 0.5249999761581421,
398
+ "rewards/chosen": -0.050308506935834885,
399
+ "rewards/margins": 0.04710409790277481,
400
+ "rewards/rejected": -0.0974126011133194,
401
+ "step": 270
402
+ },
403
+ {
404
+ "epoch": 0.15,
405
+ "learning_rate": 4.9633990224171305e-06,
406
+ "logits/chosen": -0.246888667345047,
407
+ "logits/rejected": -0.23898451030254364,
408
+ "logps/chosen": -540.099609375,
409
+ "logps/rejected": -619.6549072265625,
410
+ "loss": 0.1998,
411
+ "rewards/accuracies": 0.42500001192092896,
412
+ "rewards/chosen": -0.05168341472744942,
413
+ "rewards/margins": 0.023662500083446503,
414
+ "rewards/rejected": -0.07534591108560562,
415
+ "step": 280
416
+ },
417
+ {
418
+ "epoch": 0.15,
419
+ "learning_rate": 4.955035087259046e-06,
420
+ "logits/chosen": -0.17880979180335999,
421
+ "logits/rejected": -0.17022277414798737,
422
+ "logps/chosen": -519.5186767578125,
423
+ "logps/rejected": -577.2052001953125,
424
+ "loss": 0.1936,
425
+ "rewards/accuracies": 0.4937500059604645,
426
+ "rewards/chosen": -0.036568932235240936,
427
+ "rewards/margins": 0.0271470807492733,
428
+ "rewards/rejected": -0.06371601670980453,
429
+ "step": 290
430
+ },
431
+ {
432
+ "epoch": 0.16,
433
+ "learning_rate": 4.945819789177756e-06,
434
+ "logits/chosen": -0.22478008270263672,
435
+ "logits/rejected": -0.12748220562934875,
436
+ "logps/chosen": -538.5833129882812,
437
+ "logps/rejected": -555.4600830078125,
438
+ "loss": 0.2204,
439
+ "rewards/accuracies": 0.41874998807907104,
440
+ "rewards/chosen": -0.0454036220908165,
441
+ "rewards/margins": 0.00936979427933693,
442
+ "rewards/rejected": -0.05477341264486313,
443
+ "step": 300
444
+ },
445
+ {
446
+ "epoch": 0.17,
447
+ "learning_rate": 4.935756323876306e-06,
448
+ "logits/chosen": -0.16740605235099792,
449
+ "logits/rejected": -0.1558527946472168,
450
+ "logps/chosen": -514.8778076171875,
451
+ "logps/rejected": -554.3466796875,
452
+ "loss": 0.2064,
453
+ "rewards/accuracies": 0.4437499940395355,
454
+ "rewards/chosen": -0.03959917649626732,
455
+ "rewards/margins": 0.016423584893345833,
456
+ "rewards/rejected": -0.056022755801677704,
457
+ "step": 310
458
+ },
459
+ {
460
+ "epoch": 0.17,
461
+ "learning_rate": 4.924848181187199e-06,
462
+ "logits/chosen": -0.21729227900505066,
463
+ "logits/rejected": -0.13360168039798737,
464
+ "logps/chosen": -550.1074829101562,
465
+ "logps/rejected": -548.8150634765625,
466
+ "loss": 0.2088,
467
+ "rewards/accuracies": 0.41874998807907104,
468
+ "rewards/chosen": -0.03942955285310745,
469
+ "rewards/margins": 0.015313336625695229,
470
+ "rewards/rejected": -0.05474288389086723,
471
+ "step": 320
472
+ },
473
+ {
474
+ "epoch": 0.18,
475
+ "learning_rate": 4.913099143862173e-06,
476
+ "logits/chosen": -0.1903410106897354,
477
+ "logits/rejected": -0.047067493200302124,
478
+ "logps/chosen": -524.9461669921875,
479
+ "logps/rejected": -586.0892333984375,
480
+ "loss": 0.1892,
481
+ "rewards/accuracies": 0.4625000059604645,
482
+ "rewards/chosen": -0.039661504328250885,
483
+ "rewards/margins": 0.02599067986011505,
484
+ "rewards/rejected": -0.06565217673778534,
485
+ "step": 330
486
+ },
487
+ {
488
+ "epoch": 0.18,
489
+ "learning_rate": 4.900513286260416e-06,
490
+ "logits/chosen": -0.13064715266227722,
491
+ "logits/rejected": -0.11442460119724274,
492
+ "logps/chosen": -598.4253540039062,
493
+ "logps/rejected": -655.3058471679688,
494
+ "loss": 0.2005,
495
+ "rewards/accuracies": 0.46875,
496
+ "rewards/chosen": -0.06107966974377632,
497
+ "rewards/margins": 0.02944565750658512,
498
+ "rewards/rejected": -0.09052532911300659,
499
+ "step": 340
500
+ },
501
+ {
502
+ "epoch": 0.19,
503
+ "learning_rate": 4.887094972935645e-06,
504
+ "logits/chosen": -0.19566695392131805,
505
+ "logits/rejected": -0.12640057504177094,
506
+ "logps/chosen": -542.46533203125,
507
+ "logps/rejected": -590.942626953125,
508
+ "loss": 0.1869,
509
+ "rewards/accuracies": 0.4749999940395355,
510
+ "rewards/chosen": -0.046198152005672455,
511
+ "rewards/margins": 0.034528639167547226,
512
+ "rewards/rejected": -0.08072678744792938,
513
+ "step": 350
514
+ },
515
+ {
516
+ "epoch": 0.19,
517
+ "learning_rate": 4.87284885712256e-06,
518
+ "logits/chosen": -0.15978778898715973,
519
+ "logits/rejected": -0.08649591356515884,
520
+ "logps/chosen": -535.330810546875,
521
+ "logps/rejected": -571.8146362304688,
522
+ "loss": 0.2087,
523
+ "rewards/accuracies": 0.46875,
524
+ "rewards/chosen": -0.05062000826001167,
525
+ "rewards/margins": 0.016931967809796333,
526
+ "rewards/rejected": -0.06755197793245316,
527
+ "step": 360
528
+ },
529
+ {
530
+ "epoch": 0.2,
531
+ "learning_rate": 4.857779879123181e-06,
532
+ "logits/chosen": -0.14306055009365082,
533
+ "logits/rejected": -0.18863961100578308,
534
+ "logps/chosen": -550.8839111328125,
535
+ "logps/rejected": -664.7866821289062,
536
+ "loss": 0.1922,
537
+ "rewards/accuracies": 0.48750001192092896,
538
+ "rewards/chosen": -0.04616117477416992,
539
+ "rewards/margins": 0.050246696919202805,
540
+ "rewards/rejected": -0.09640787541866302,
541
+ "step": 370
542
+ },
543
+ {
544
+ "epoch": 0.2,
545
+ "learning_rate": 4.841893264593643e-06,
546
+ "logits/chosen": -0.14398136734962463,
547
+ "logits/rejected": -0.12858401238918304,
548
+ "logps/chosen": -536.4591674804688,
549
+ "logps/rejected": -554.4386596679688,
550
+ "loss": 0.1966,
551
+ "rewards/accuracies": 0.4437499940395355,
552
+ "rewards/chosen": -0.046029798686504364,
553
+ "rewards/margins": 0.020984839648008347,
554
+ "rewards/rejected": -0.0670146495103836,
555
+ "step": 380
556
+ },
557
+ {
558
+ "epoch": 0.21,
559
+ "learning_rate": 4.825194522732023e-06,
560
+ "logits/chosen": -0.16872592270374298,
561
+ "logits/rejected": -0.04804924130439758,
562
+ "logps/chosen": -523.2348022460938,
563
+ "logps/rejected": -565.3974609375,
564
+ "loss": 0.1916,
565
+ "rewards/accuracies": 0.48750001192092896,
566
+ "rewards/chosen": -0.04487336426973343,
567
+ "rewards/margins": 0.028753217309713364,
568
+ "rewards/rejected": -0.07362658530473709,
569
+ "step": 390
570
+ },
571
+ {
572
+ "epoch": 0.21,
573
+ "learning_rate": 4.807689444367853e-06,
574
+ "logits/chosen": -0.16394059360027313,
575
+ "logits/rejected": -0.07825560867786407,
576
+ "logps/chosen": -569.9908447265625,
577
+ "logps/rejected": -620.1237182617188,
578
+ "loss": 0.1977,
579
+ "rewards/accuracies": 0.46875,
580
+ "rewards/chosen": -0.05317618325352669,
581
+ "rewards/margins": 0.03387502580881119,
582
+ "rewards/rejected": -0.08705119788646698,
583
+ "step": 400
584
+ },
585
+ {
586
+ "epoch": 0.22,
587
+ "learning_rate": 4.78938409995396e-06,
588
+ "logits/chosen": -0.15660491585731506,
589
+ "logits/rejected": -0.05097333714365959,
590
+ "logps/chosen": -575.72607421875,
591
+ "logps/rejected": -647.863037109375,
592
+ "loss": 0.1964,
593
+ "rewards/accuracies": 0.512499988079071,
594
+ "rewards/chosen": -0.04157186672091484,
595
+ "rewards/margins": 0.042187802493572235,
596
+ "rewards/rejected": -0.08375968039035797,
597
+ "step": 410
598
+ },
599
+ {
600
+ "epoch": 0.22,
601
+ "learning_rate": 4.770284837461342e-06,
602
+ "logits/chosen": -0.1467301845550537,
603
+ "logits/rejected": -0.041234057396650314,
604
+ "logps/chosen": -506.21075439453125,
605
+ "logps/rejected": -573.1678466796875,
606
+ "loss": 0.1813,
607
+ "rewards/accuracies": 0.48124998807907104,
608
+ "rewards/chosen": -0.03669622167944908,
609
+ "rewards/margins": 0.04058472067117691,
610
+ "rewards/rejected": -0.07728095352649689,
611
+ "step": 420
612
+ },
613
+ {
614
+ "epoch": 0.23,
615
+ "learning_rate": 4.7503982801778015e-06,
616
+ "logits/chosen": -0.049983613193035126,
617
+ "logits/rejected": -0.09682069718837738,
618
+ "logps/chosen": -542.2359619140625,
619
+ "logps/rejected": -583.8301391601562,
620
+ "loss": 0.1926,
621
+ "rewards/accuracies": 0.44999998807907104,
622
+ "rewards/chosen": -0.04950016364455223,
623
+ "rewards/margins": 0.027737829834222794,
624
+ "rewards/rejected": -0.07723800092935562,
625
+ "step": 430
626
+ },
627
+ {
628
+ "epoch": 0.23,
629
+ "learning_rate": 4.729731324411104e-06,
630
+ "logits/chosen": -0.170378640294075,
631
+ "logits/rejected": -0.12245140224695206,
632
+ "logps/chosen": -491.5496520996094,
633
+ "logps/rejected": -573.3743896484375,
634
+ "loss": 0.1994,
635
+ "rewards/accuracies": 0.4937500059604645,
636
+ "rewards/chosen": -0.0331207700073719,
637
+ "rewards/margins": 0.03893864154815674,
638
+ "rewards/rejected": -0.07205941528081894,
639
+ "step": 440
640
+ },
641
+ {
642
+ "epoch": 0.24,
643
+ "learning_rate": 4.7082911370974645e-06,
644
+ "logits/chosen": -0.21018514037132263,
645
+ "logits/rejected": -0.09919361770153046,
646
+ "logps/chosen": -500.1617736816406,
647
+ "logps/rejected": -528.3983154296875,
648
+ "loss": 0.1829,
649
+ "rewards/accuracies": 0.44999998807907104,
650
+ "rewards/chosen": -0.03072548285126686,
651
+ "rewards/margins": 0.03223457187414169,
652
+ "rewards/rejected": -0.06296005100011826,
653
+ "step": 450
654
+ },
655
+ {
656
+ "epoch": 0.25,
657
+ "learning_rate": 4.68608515331618e-06,
658
+ "logits/chosen": -0.103789784014225,
659
+ "logits/rejected": -0.09898993372917175,
660
+ "logps/chosen": -548.4000854492188,
661
+ "logps/rejected": -562.94482421875,
662
+ "loss": 0.2153,
663
+ "rewards/accuracies": 0.46875,
664
+ "rewards/chosen": -0.03969540819525719,
665
+ "rewards/margins": 0.025323593989014626,
666
+ "rewards/rejected": -0.06501900404691696,
667
+ "step": 460
668
+ },
669
+ {
670
+ "epoch": 0.25,
671
+ "learning_rate": 4.663121073711269e-06,
672
+ "logits/chosen": -0.1721930205821991,
673
+ "logits/rejected": -0.1523117572069168,
674
+ "logps/chosen": -503.42657470703125,
675
+ "logps/rejected": -561.2427368164062,
676
+ "loss": 0.2054,
677
+ "rewards/accuracies": 0.45625001192092896,
678
+ "rewards/chosen": -0.0317218080163002,
679
+ "rewards/margins": 0.039346493780612946,
680
+ "rewards/rejected": -0.07106830924749374,
681
+ "step": 470
682
+ },
683
+ {
684
+ "epoch": 0.26,
685
+ "learning_rate": 4.63940686182103e-06,
686
+ "logits/chosen": -0.10965099185705185,
687
+ "logits/rejected": -0.08603724092245102,
688
+ "logps/chosen": -558.77490234375,
689
+ "logps/rejected": -615.6389770507812,
690
+ "loss": 0.1821,
691
+ "rewards/accuracies": 0.543749988079071,
692
+ "rewards/chosen": -0.03726140409708023,
693
+ "rewards/margins": 0.03986964374780655,
694
+ "rewards/rejected": -0.07713104784488678,
695
+ "step": 480
696
+ },
697
+ {
698
+ "epoch": 0.26,
699
+ "learning_rate": 4.614950741316425e-06,
700
+ "logits/chosen": -0.16542614996433258,
701
+ "logits/rejected": -0.024035543203353882,
702
+ "logps/chosen": -553.4063720703125,
703
+ "logps/rejected": -628.834716796875,
704
+ "loss": 0.1962,
705
+ "rewards/accuracies": 0.48124998807907104,
706
+ "rewards/chosen": -0.033900871872901917,
707
+ "rewards/margins": 0.026126693934202194,
708
+ "rewards/rejected": -0.06002756953239441,
709
+ "step": 490
710
+ },
711
+ {
712
+ "epoch": 0.27,
713
+ "learning_rate": 4.589761193149254e-06,
714
+ "logits/chosen": -0.13893479108810425,
715
+ "logits/rejected": -0.03659617155790329,
716
+ "logps/chosen": -565.3289794921875,
717
+ "logps/rejected": -582.9500732421875,
718
+ "loss": 0.2002,
719
+ "rewards/accuracies": 0.5062500238418579,
720
+ "rewards/chosen": -0.034871987998485565,
721
+ "rewards/margins": 0.018284570425748825,
722
+ "rewards/rejected": -0.05315656214952469,
723
+ "step": 500
724
+ },
725
+ {
726
+ "epoch": 0.27,
727
+ "learning_rate": 4.563846952611112e-06,
728
+ "logits/chosen": -0.03639233857393265,
729
+ "logits/rejected": -0.13991451263427734,
730
+ "logps/chosen": -574.9199829101562,
731
+ "logps/rejected": -610.3167114257812,
732
+ "loss": 0.2025,
733
+ "rewards/accuracies": 0.48124998807907104,
734
+ "rewards/chosen": -0.028040558099746704,
735
+ "rewards/margins": 0.02244061976671219,
736
+ "rewards/rejected": -0.05048117786645889,
737
+ "step": 510
738
+ },
739
+ {
740
+ "epoch": 0.28,
741
+ "learning_rate": 4.537217006304141e-06,
742
+ "logits/chosen": -0.1792822778224945,
743
+ "logits/rejected": -0.08993463218212128,
744
+ "logps/chosen": -525.2308349609375,
745
+ "logps/rejected": -558.35791015625,
746
+ "loss": 0.2095,
747
+ "rewards/accuracies": 0.5062500238418579,
748
+ "rewards/chosen": -0.022352229803800583,
749
+ "rewards/margins": 0.015106442384421825,
750
+ "rewards/rejected": -0.03745867311954498,
751
+ "step": 520
752
+ },
753
+ {
754
+ "epoch": 0.28,
755
+ "learning_rate": 4.50988058902464e-06,
756
+ "logits/chosen": -0.10850954055786133,
757
+ "logits/rejected": -0.0964195728302002,
758
+ "logps/chosen": -534.2761840820312,
759
+ "logps/rejected": -559.3724975585938,
760
+ "loss": 0.2077,
761
+ "rewards/accuracies": 0.4375,
762
+ "rewards/chosen": -0.020403413102030754,
763
+ "rewards/margins": 0.021993646398186684,
764
+ "rewards/rejected": -0.04239705950021744,
765
+ "step": 530
766
+ },
767
+ {
768
+ "epoch": 0.29,
769
+ "learning_rate": 4.481847180560593e-06,
770
+ "logits/chosen": -0.05597566440701485,
771
+ "logits/rejected": -0.1112387552857399,
772
+ "logps/chosen": -582.2784423828125,
773
+ "logps/rejected": -618.896728515625,
774
+ "loss": 0.197,
775
+ "rewards/accuracies": 0.46875,
776
+ "rewards/chosen": -0.032762907445430756,
777
+ "rewards/margins": 0.018652746453881264,
778
+ "rewards/rejected": -0.05141565203666687,
779
+ "step": 540
780
+ },
781
+ {
782
+ "epoch": 0.29,
783
+ "learning_rate": 4.453126502404253e-06,
784
+ "logits/chosen": -0.06729540228843689,
785
+ "logits/rejected": -0.045423537492752075,
786
+ "logps/chosen": -526.5831298828125,
787
+ "logps/rejected": -574.4923706054688,
788
+ "loss": 0.1946,
789
+ "rewards/accuracies": 0.550000011920929,
790
+ "rewards/chosen": -0.027647733688354492,
791
+ "rewards/margins": 0.02961956337094307,
792
+ "rewards/rejected": -0.05726729705929756,
793
+ "step": 550
794
+ },
795
+ {
796
+ "epoch": 0.3,
797
+ "learning_rate": 4.423728514380892e-06,
798
+ "logits/chosen": -0.09008444845676422,
799
+ "logits/rejected": 0.0276274923235178,
800
+ "logps/chosen": -544.626708984375,
801
+ "logps/rejected": -580.9148559570312,
802
+ "loss": 0.2048,
803
+ "rewards/accuracies": 0.4749999940395355,
804
+ "rewards/chosen": -0.036577023565769196,
805
+ "rewards/margins": 0.01596749946475029,
806
+ "rewards/rejected": -0.052544523030519485,
807
+ "step": 560
808
+ },
809
+ {
810
+ "epoch": 0.3,
811
+ "learning_rate": 4.393663411194918e-06,
812
+ "logits/chosen": -0.05868130922317505,
813
+ "logits/rejected": -0.04863614961504936,
814
+ "logps/chosen": -531.1651000976562,
815
+ "logps/rejected": -553.3944702148438,
816
+ "loss": 0.2092,
817
+ "rewards/accuracies": 0.42500001192092896,
818
+ "rewards/chosen": -0.03158984333276749,
819
+ "rewards/margins": 0.02088005281984806,
820
+ "rewards/rejected": -0.052469898015260696,
821
+ "step": 570
822
+ },
823
+ {
824
+ "epoch": 0.31,
825
+ "learning_rate": 4.362941618894523e-06,
826
+ "logits/chosen": -0.08449320495128632,
827
+ "logits/rejected": -0.012650961056351662,
828
+ "logps/chosen": -541.2258911132812,
829
+ "logps/rejected": -583.4708862304688,
830
+ "loss": 0.2071,
831
+ "rewards/accuracies": 0.44999998807907104,
832
+ "rewards/chosen": -0.03278597444295883,
833
+ "rewards/margins": 0.016275811940431595,
834
+ "rewards/rejected": -0.04906178265810013,
835
+ "step": 580
836
+ },
837
+ {
838
+ "epoch": 0.31,
839
+ "learning_rate": 4.331573791256116e-06,
840
+ "logits/chosen": -0.05021784454584122,
841
+ "logits/rejected": -0.0996134951710701,
842
+ "logps/chosen": -490.427978515625,
843
+ "logps/rejected": -564.3570556640625,
844
+ "loss": 0.1908,
845
+ "rewards/accuracies": 0.53125,
846
+ "rewards/chosen": -0.025160646066069603,
847
+ "rewards/margins": 0.028088271617889404,
848
+ "rewards/rejected": -0.053248919546604156,
849
+ "step": 590
850
+ },
851
+ {
852
+ "epoch": 0.32,
853
+ "learning_rate": 4.299570806089786e-06,
854
+ "logits/chosen": 0.0037119381595402956,
855
+ "logits/rejected": -0.03845081478357315,
856
+ "logps/chosen": -517.3943481445312,
857
+ "logps/rejected": -546.4298095703125,
858
+ "loss": 0.1878,
859
+ "rewards/accuracies": 0.46875,
860
+ "rewards/chosen": -0.03542790561914444,
861
+ "rewards/margins": 0.03082282282412052,
862
+ "rewards/rejected": -0.06625071913003922,
863
+ "step": 600
864
+ },
865
+ {
866
+ "epoch": 0.33,
867
+ "learning_rate": 4.266943761467057e-06,
868
+ "logits/chosen": -0.11214573681354523,
869
+ "logits/rejected": -0.048879437148571014,
870
+ "logps/chosen": -585.6790771484375,
871
+ "logps/rejected": -597.4764404296875,
872
+ "loss": 0.1957,
873
+ "rewards/accuracies": 0.48750001192092896,
874
+ "rewards/chosen": -0.04125026613473892,
875
+ "rewards/margins": 0.022660288959741592,
876
+ "rewards/rejected": -0.06391056627035141,
877
+ "step": 610
878
+ },
879
+ {
880
+ "epoch": 0.33,
881
+ "learning_rate": 4.233703971872287e-06,
882
+ "logits/chosen": -0.03887901455163956,
883
+ "logits/rejected": -0.041454676538705826,
884
+ "logps/chosen": -554.0848999023438,
885
+ "logps/rejected": -593.8275756835938,
886
+ "loss": 0.1925,
887
+ "rewards/accuracies": 0.5,
888
+ "rewards/chosen": -0.04528883099555969,
889
+ "rewards/margins": 0.03674833104014397,
890
+ "rewards/rejected": -0.08203716576099396,
891
+ "step": 620
892
+ },
893
+ {
894
+ "epoch": 0.34,
895
+ "learning_rate": 4.1998629642789925e-06,
896
+ "logits/chosen": -0.14514538645744324,
897
+ "logits/rejected": -0.12874510884284973,
898
+ "logps/chosen": -580.5797729492188,
899
+ "logps/rejected": -627.352294921875,
900
+ "loss": 0.1972,
901
+ "rewards/accuracies": 0.48124998807907104,
902
+ "rewards/chosen": -0.053490232676267624,
903
+ "rewards/margins": 0.03829261288046837,
904
+ "rewards/rejected": -0.09178284555673599,
905
+ "step": 630
906
+ },
907
+ {
908
+ "epoch": 0.34,
909
+ "learning_rate": 4.165432474152505e-06,
910
+ "logits/chosen": -0.10119237750768661,
911
+ "logits/rejected": -0.12084090709686279,
912
+ "logps/chosen": -556.23828125,
913
+ "logps/rejected": -601.0460205078125,
914
+ "loss": 0.1939,
915
+ "rewards/accuracies": 0.46875,
916
+ "rewards/chosen": -0.05898448824882507,
917
+ "rewards/margins": 0.03952840715646744,
918
+ "rewards/rejected": -0.09851288795471191,
919
+ "step": 640
920
+ },
921
+ {
922
+ "epoch": 0.35,
923
+ "learning_rate": 4.130424441380308e-06,
924
+ "logits/chosen": -0.051937490701675415,
925
+ "logits/rejected": -0.11804668605327606,
926
+ "logps/chosen": -618.0116577148438,
927
+ "logps/rejected": -681.9761962890625,
928
+ "loss": 0.1944,
929
+ "rewards/accuracies": 0.46875,
930
+ "rewards/chosen": -0.08069413900375366,
931
+ "rewards/margins": 0.053174007683992386,
932
+ "rewards/rejected": -0.13386814296245575,
933
+ "step": 650
934
+ },
935
+ {
936
+ "epoch": 0.35,
937
+ "learning_rate": 4.09485100613151e-06,
938
+ "logits/chosen": 0.03249294310808182,
939
+ "logits/rejected": -0.03189364820718765,
940
+ "logps/chosen": -587.2728881835938,
941
+ "logps/rejected": -678.32568359375,
942
+ "loss": 0.2106,
943
+ "rewards/accuracies": 0.5,
944
+ "rewards/chosen": -0.06832827627658844,
945
+ "rewards/margins": 0.0339791476726532,
946
+ "rewards/rejected": -0.10230742394924164,
947
+ "step": 660
948
+ },
949
+ {
950
+ "epoch": 0.36,
951
+ "learning_rate": 4.058724504646834e-06,
952
+ "logits/chosen": 0.003849644912406802,
953
+ "logits/rejected": -0.07050515711307526,
954
+ "logps/chosen": -485.66375732421875,
955
+ "logps/rejected": -613.0921020507812,
956
+ "loss": 0.1938,
957
+ "rewards/accuracies": 0.48750001192092896,
958
+ "rewards/chosen": -0.0477924570441246,
959
+ "rewards/margins": 0.05937989428639412,
960
+ "rewards/rejected": -0.10717234760522842,
961
+ "step": 670
962
+ },
963
+ {
964
+ "epoch": 0.36,
965
+ "learning_rate": 4.022057464960632e-06,
966
+ "logits/chosen": -0.034047119319438934,
967
+ "logits/rejected": 0.01398265827447176,
968
+ "logps/chosen": -568.3038940429688,
969
+ "logps/rejected": -603.4803466796875,
970
+ "loss": 0.2027,
971
+ "rewards/accuracies": 0.48124998807907104,
972
+ "rewards/chosen": -0.04010879248380661,
973
+ "rewards/margins": 0.03294765576720238,
974
+ "rewards/rejected": -0.07305644452571869,
975
+ "step": 680
976
+ },
977
+ {
978
+ "epoch": 0.37,
979
+ "learning_rate": 3.984862602556383e-06,
980
+ "logits/chosen": -0.0736241266131401,
981
+ "logits/rejected": -0.03072969987988472,
982
+ "logps/chosen": -584.6604614257812,
983
+ "logps/rejected": -646.9960327148438,
984
+ "loss": 0.1893,
985
+ "rewards/accuracies": 0.581250011920929,
986
+ "rewards/chosen": -0.022528892382979393,
987
+ "rewards/margins": 0.03268754482269287,
988
+ "rewards/rejected": -0.055216431617736816,
989
+ "step": 690
990
+ },
991
+ {
992
+ "epoch": 0.37,
993
+ "learning_rate": 3.947152815957187e-06,
994
+ "logits/chosen": -0.07144252955913544,
995
+ "logits/rejected": -0.04082520306110382,
996
+ "logps/chosen": -536.7813720703125,
997
+ "logps/rejected": -604.7335205078125,
998
+ "loss": 0.185,
999
+ "rewards/accuracies": 0.5687500238418579,
1000
+ "rewards/chosen": -0.04123537242412567,
1001
+ "rewards/margins": 0.04173678532242775,
1002
+ "rewards/rejected": -0.08297216892242432,
1003
+ "step": 700
1004
+ },
1005
+ {
1006
+ "epoch": 0.38,
1007
+ "learning_rate": 3.908941182252785e-06,
1008
+ "logits/chosen": -0.04457836225628853,
1009
+ "logits/rejected": -0.019069496542215347,
1010
+ "logps/chosen": -504.98260498046875,
1011
+ "logps/rejected": -545.35107421875,
1012
+ "loss": 0.201,
1013
+ "rewards/accuracies": 0.4937500059604645,
1014
+ "rewards/chosen": -0.03742678835988045,
1015
+ "rewards/margins": 0.03181907907128334,
1016
+ "rewards/rejected": -0.06924585998058319,
1017
+ "step": 710
1018
+ },
1019
+ {
1020
+ "epoch": 0.38,
1021
+ "learning_rate": 3.8702409525646535e-06,
1022
+ "logits/chosen": -0.08643130213022232,
1023
+ "logits/rejected": -0.00629496481269598,
1024
+ "logps/chosen": -506.75128173828125,
1025
+ "logps/rejected": -579.1588745117188,
1026
+ "loss": 0.2058,
1027
+ "rewards/accuracies": 0.48124998807907104,
1028
+ "rewards/chosen": -0.03903960436582565,
1029
+ "rewards/margins": 0.03844233229756355,
1030
+ "rewards/rejected": -0.0774819403886795,
1031
+ "step": 720
1032
+ },
1033
+ {
1034
+ "epoch": 0.39,
1035
+ "learning_rate": 3.8310655474507495e-06,
1036
+ "logits/chosen": -0.10445193201303482,
1037
+ "logits/rejected": -0.05900994688272476,
1038
+ "logps/chosen": -561.2862548828125,
1039
+ "logps/rejected": -646.6036987304688,
1040
+ "loss": 0.1828,
1041
+ "rewards/accuracies": 0.5562499761581421,
1042
+ "rewards/chosen": -0.03991767764091492,
1043
+ "rewards/margins": 0.04778400436043739,
1044
+ "rewards/rejected": -0.08770167827606201,
1045
+ "step": 730
1046
+ },
1047
+ {
1048
+ "epoch": 0.39,
1049
+ "learning_rate": 3.7914285522515002e-06,
1050
+ "logits/chosen": -0.10235663503408432,
1051
+ "logits/rejected": -0.06297244131565094,
1052
+ "logps/chosen": -533.1697387695312,
1053
+ "logps/rejected": -621.258056640625,
1054
+ "loss": 0.1897,
1055
+ "rewards/accuracies": 0.512499988079071,
1056
+ "rewards/chosen": -0.035120584070682526,
1057
+ "rewards/margins": 0.046875499188899994,
1058
+ "rewards/rejected": -0.08199608325958252,
1059
+ "step": 740
1060
+ },
1061
+ {
1062
+ "epoch": 0.4,
1063
+ "learning_rate": 3.751343712378639e-06,
1064
+ "logits/chosen": 0.04809337109327316,
1065
+ "logits/rejected": -0.11011433601379395,
1066
+ "logps/chosen": -501.5302734375,
1067
+ "logps/rejected": -548.520263671875,
1068
+ "loss": 0.198,
1069
+ "rewards/accuracies": 0.44999998807907104,
1070
+ "rewards/chosen": -0.040851037949323654,
1071
+ "rewards/margins": 0.02717413939535618,
1072
+ "rewards/rejected": -0.06802517920732498,
1073
+ "step": 750
1074
+ },
1075
+ {
1076
+ "epoch": 0.41,
1077
+ "learning_rate": 3.710824928548546e-06,
1078
+ "logits/chosen": -0.01540103554725647,
1079
+ "logits/rejected": 0.029544491320848465,
1080
+ "logps/chosen": -581.104736328125,
1081
+ "logps/rejected": -622.2809448242188,
1082
+ "loss": 0.1955,
1083
+ "rewards/accuracies": 0.4749999940395355,
1084
+ "rewards/chosen": -0.03365876525640488,
1085
+ "rewards/margins": 0.01963328942656517,
1086
+ "rewards/rejected": -0.05329205468297005,
1087
+ "step": 760
1088
+ },
1089
+ {
1090
+ "epoch": 0.41,
1091
+ "learning_rate": 3.6698862519617225e-06,
1092
+ "logits/chosen": 0.05280674248933792,
1093
+ "logits/rejected": -0.0824475884437561,
1094
+ "logps/chosen": -551.7106323242188,
1095
+ "logps/rejected": -647.3377685546875,
1096
+ "loss": 0.1755,
1097
+ "rewards/accuracies": 0.581250011920929,
1098
+ "rewards/chosen": -0.03773096203804016,
1099
+ "rewards/margins": 0.051321662962436676,
1100
+ "rewards/rejected": -0.08905263245105743,
1101
+ "step": 770
1102
+ },
1103
+ {
1104
+ "epoch": 0.42,
1105
+ "learning_rate": 3.6285418794300793e-06,
1106
+ "logits/chosen": 0.035738877952098846,
1107
+ "logits/rejected": -0.04512631893157959,
1108
+ "logps/chosen": -562.4360961914062,
1109
+ "logps/rejected": -655.2886962890625,
1110
+ "loss": 0.1919,
1111
+ "rewards/accuracies": 0.4937500059604645,
1112
+ "rewards/chosen": -0.041907649487257004,
1113
+ "rewards/margins": 0.0431944876909256,
1114
+ "rewards/rejected": -0.0851021409034729,
1115
+ "step": 780
1116
+ },
1117
+ {
1118
+ "epoch": 0.42,
1119
+ "learning_rate": 3.5868061484537365e-06,
1120
+ "logits/chosen": -0.029743824154138565,
1121
+ "logits/rejected": 0.053073786199092865,
1122
+ "logps/chosen": -578.3715209960938,
1123
+ "logps/rejected": -652.7194213867188,
1124
+ "loss": 0.1883,
1125
+ "rewards/accuracies": 0.4937500059604645,
1126
+ "rewards/chosen": -0.043207161128520966,
1127
+ "rewards/margins": 0.047172196209430695,
1128
+ "rewards/rejected": -0.09037934988737106,
1129
+ "step": 790
1130
+ },
1131
+ {
1132
+ "epoch": 0.43,
1133
+ "learning_rate": 3.5446935322490285e-06,
1134
+ "logits/chosen": -0.006876747123897076,
1135
+ "logits/rejected": 0.015042210929095745,
1136
+ "logps/chosen": -525.9793090820312,
1137
+ "logps/rejected": -600.8638916015625,
1138
+ "loss": 0.1998,
1139
+ "rewards/accuracies": 0.46875,
1140
+ "rewards/chosen": -0.03883466124534607,
1141
+ "rewards/margins": 0.05057210475206375,
1142
+ "rewards/rejected": -0.08940677344799042,
1143
+ "step": 800
1144
+ },
1145
+ {
1146
+ "epoch": 0.43,
1147
+ "learning_rate": 3.502218634729447e-06,
1148
+ "logits/chosen": -0.06274811178445816,
1149
+ "logits/rejected": 0.020462125539779663,
1150
+ "logps/chosen": -544.5744018554688,
1151
+ "logps/rejected": -636.985107421875,
1152
+ "loss": 0.199,
1153
+ "rewards/accuracies": 0.5,
1154
+ "rewards/chosen": -0.02988346479833126,
1155
+ "rewards/margins": 0.046063631772994995,
1156
+ "rewards/rejected": -0.07594709098339081,
1157
+ "step": 810
1158
+ },
1159
+ {
1160
+ "epoch": 0.44,
1161
+ "learning_rate": 3.459396185441265e-06,
1162
+ "logits/chosen": -0.034183721989393234,
1163
+ "logits/rejected": 0.007283804006874561,
1164
+ "logps/chosen": -590.5462646484375,
1165
+ "logps/rejected": -617.2237548828125,
1166
+ "loss": 0.1814,
1167
+ "rewards/accuracies": 0.5,
1168
+ "rewards/chosen": -0.03479190170764923,
1169
+ "rewards/margins": 0.03565431758761406,
1170
+ "rewards/rejected": -0.07044622302055359,
1171
+ "step": 820
1172
+ },
1173
+ {
1174
+ "epoch": 0.44,
1175
+ "learning_rate": 3.4162410344555834e-06,
1176
+ "logits/chosen": -0.031631968915462494,
1177
+ "logits/rejected": -0.03749103471636772,
1178
+ "logps/chosen": -553.4362182617188,
1179
+ "logps/rejected": -585.8868408203125,
1180
+ "loss": 0.1936,
1181
+ "rewards/accuracies": 0.543749988079071,
1182
+ "rewards/chosen": -0.025158634409308434,
1183
+ "rewards/margins": 0.02824149653315544,
1184
+ "rewards/rejected": -0.053400129079818726,
1185
+ "step": 830
1186
+ },
1187
+ {
1188
+ "epoch": 0.45,
1189
+ "learning_rate": 3.3727681472185937e-06,
1190
+ "logits/chosen": -0.017966533079743385,
1191
+ "logits/rejected": 0.07461138069629669,
1192
+ "logps/chosen": -504.60809326171875,
1193
+ "logps/rejected": -541.7296142578125,
1194
+ "loss": 0.2005,
1195
+ "rewards/accuracies": 0.46875,
1196
+ "rewards/chosen": -0.018680742010474205,
1197
+ "rewards/margins": 0.02315039001405239,
1198
+ "rewards/rejected": -0.0418311282992363,
1199
+ "step": 840
1200
+ },
1201
+ {
1202
+ "epoch": 0.45,
1203
+ "learning_rate": 3.3289925993618217e-06,
1204
+ "logits/chosen": -0.05572665482759476,
1205
+ "logits/rejected": 0.0033084123861044645,
1206
+ "logps/chosen": -538.4072875976562,
1207
+ "logps/rejected": -589.1475219726562,
1208
+ "loss": 0.1734,
1209
+ "rewards/accuracies": 0.5687500238418579,
1210
+ "rewards/chosen": -0.017685720697045326,
1211
+ "rewards/margins": 0.02718517556786537,
1212
+ "rewards/rejected": -0.04487089440226555,
1213
+ "step": 850
1214
+ },
1215
+ {
1216
+ "epoch": 0.46,
1217
+ "learning_rate": 3.2849295714741643e-06,
1218
+ "logits/chosen": -0.019068999215960503,
1219
+ "logits/rejected": -0.08031287044286728,
1220
+ "logps/chosen": -477.96832275390625,
1221
+ "logps/rejected": -538.3767700195312,
1222
+ "loss": 0.2003,
1223
+ "rewards/accuracies": 0.48750001192092896,
1224
+ "rewards/chosen": -0.017096657305955887,
1225
+ "rewards/margins": 0.022515540942549706,
1226
+ "rewards/rejected": -0.03961219638586044,
1227
+ "step": 860
1228
+ },
1229
+ {
1230
+ "epoch": 0.46,
1231
+ "learning_rate": 3.2405943438375287e-06,
1232
+ "logits/chosen": -0.10866532474756241,
1233
+ "logits/rejected": -0.014922107569873333,
1234
+ "logps/chosen": -573.4918212890625,
1235
+ "logps/rejected": -601.1939086914062,
1236
+ "loss": 0.1981,
1237
+ "rewards/accuracies": 0.5375000238418579,
1238
+ "rewards/chosen": -0.030659830197691917,
1239
+ "rewards/margins": 0.022852713242173195,
1240
+ "rewards/rejected": -0.05351255089044571,
1241
+ "step": 870
1242
+ },
1243
+ {
1244
+ "epoch": 0.47,
1245
+ "learning_rate": 3.1960022911279036e-06,
1246
+ "logits/chosen": -0.005284407641738653,
1247
+ "logits/rejected": -0.043508388102054596,
1248
+ "logps/chosen": -556.1735229492188,
1249
+ "logps/rejected": -627.1209106445312,
1250
+ "loss": 0.2072,
1251
+ "rewards/accuracies": 0.6312500238418579,
1252
+ "rewards/chosen": -0.0352107435464859,
1253
+ "rewards/margins": 0.038544684648513794,
1254
+ "rewards/rejected": -0.0737554281949997,
1255
+ "step": 880
1256
+ },
1257
+ {
1258
+ "epoch": 0.47,
1259
+ "learning_rate": 3.1511688770836844e-06,
1260
+ "logits/chosen": -0.06900519132614136,
1261
+ "logits/rejected": -0.008347516879439354,
1262
+ "logps/chosen": -518.9674072265625,
1263
+ "logps/rejected": -563.190185546875,
1264
+ "loss": 0.1873,
1265
+ "rewards/accuracies": 0.46875,
1266
+ "rewards/chosen": -0.0382576659321785,
1267
+ "rewards/margins": 0.03923247009515762,
1268
+ "rewards/rejected": -0.07749013602733612,
1269
+ "step": 890
1270
+ },
1271
+ {
1272
+ "epoch": 0.48,
1273
+ "learning_rate": 3.1061096491431307e-06,
1274
+ "logits/chosen": -0.004668307490646839,
1275
+ "logits/rejected": -0.05605720728635788,
1276
+ "logps/chosen": -560.2337646484375,
1277
+ "logps/rejected": -637.04931640625,
1278
+ "loss": 0.1747,
1279
+ "rewards/accuracies": 0.512499988079071,
1280
+ "rewards/chosen": -0.06013607978820801,
1281
+ "rewards/margins": 0.052308034151792526,
1282
+ "rewards/rejected": -0.11244411766529083,
1283
+ "step": 900
1284
+ },
1285
+ {
1286
+ "epoch": 0.49,
1287
+ "learning_rate": 3.0608402330527796e-06,
1288
+ "logits/chosen": -0.06335186958312988,
1289
+ "logits/rejected": -0.027245525270700455,
1290
+ "logps/chosen": -516.86865234375,
1291
+ "logps/rejected": -573.42724609375,
1292
+ "loss": 0.1955,
1293
+ "rewards/accuracies": 0.46875,
1294
+ "rewards/chosen": -0.06095459312200546,
1295
+ "rewards/margins": 0.03399496152997017,
1296
+ "rewards/rejected": -0.09494956582784653,
1297
+ "step": 910
1298
+ },
1299
+ {
1300
+ "epoch": 0.49,
1301
+ "learning_rate": 3.0153763274487176e-06,
1302
+ "logits/chosen": -0.016411593183875084,
1303
+ "logits/rejected": 0.005086073186248541,
1304
+ "logps/chosen": -552.3294677734375,
1305
+ "logps/rejected": -607.4609375,
1306
+ "loss": 0.1793,
1307
+ "rewards/accuracies": 0.4625000059604645,
1308
+ "rewards/chosen": -0.06465917080640793,
1309
+ "rewards/margins": 0.038704242557287216,
1310
+ "rewards/rejected": -0.10336340963840485,
1311
+ "step": 920
1312
+ },
1313
+ {
1314
+ "epoch": 0.5,
1315
+ "learning_rate": 2.9697336984125683e-06,
1316
+ "logits/chosen": -0.0311849657446146,
1317
+ "logits/rejected": -0.05022461339831352,
1318
+ "logps/chosen": -507.9979553222656,
1319
+ "logps/rejected": -572.2631225585938,
1320
+ "loss": 0.1788,
1321
+ "rewards/accuracies": 0.518750011920929,
1322
+ "rewards/chosen": -0.03872525319457054,
1323
+ "rewards/margins": 0.03657948225736618,
1324
+ "rewards/rejected": -0.07530473172664642,
1325
+ "step": 930
1326
+ },
1327
+ {
1328
+ "epoch": 0.5,
1329
+ "learning_rate": 2.923928174004094e-06,
1330
+ "logits/chosen": -0.1109534353017807,
1331
+ "logits/rejected": -0.057219941169023514,
1332
+ "logps/chosen": -550.4554443359375,
1333
+ "logps/rejected": -599.104736328125,
1334
+ "loss": 0.1928,
1335
+ "rewards/accuracies": 0.5375000238418579,
1336
+ "rewards/chosen": -0.05719093605875969,
1337
+ "rewards/margins": 0.02921503223478794,
1338
+ "rewards/rejected": -0.08640597015619278,
1339
+ "step": 940
1340
+ },
1341
+ {
1342
+ "epoch": 0.51,
1343
+ "learning_rate": 2.8779756387723036e-06,
1344
+ "logits/chosen": -0.009409350343048573,
1345
+ "logits/rejected": 0.0022618831135332584,
1346
+ "logps/chosen": -604.61181640625,
1347
+ "logps/rejected": -644.1531982421875,
1348
+ "loss": 0.1886,
1349
+ "rewards/accuracies": 0.5562499761581421,
1350
+ "rewards/chosen": -0.05070986598730087,
1351
+ "rewards/margins": 0.03501267358660698,
1352
+ "rewards/rejected": -0.08572253584861755,
1353
+ "step": 950
1354
+ },
1355
+ {
1356
+ "epoch": 0.51,
1357
+ "learning_rate": 2.831892028246968e-06,
1358
+ "logits/chosen": -0.06726431846618652,
1359
+ "logits/rejected": -0.08842380344867706,
1360
+ "logps/chosen": -606.5565185546875,
1361
+ "logps/rejected": -661.2351684570312,
1362
+ "loss": 0.2006,
1363
+ "rewards/accuracies": 0.4937500059604645,
1364
+ "rewards/chosen": -0.05755617097020149,
1365
+ "rewards/margins": 0.024521058425307274,
1366
+ "rewards/rejected": -0.08207723498344421,
1367
+ "step": 960
1368
+ },
1369
+ {
1370
+ "epoch": 0.52,
1371
+ "learning_rate": 2.7856933234124617e-06,
1372
+ "logits/chosen": 0.021266190335154533,
1373
+ "logits/rejected": -0.03655446320772171,
1374
+ "logps/chosen": -579.4450073242188,
1375
+ "logps/rejected": -625.4588623046875,
1376
+ "loss": 0.1998,
1377
+ "rewards/accuracies": 0.4937500059604645,
1378
+ "rewards/chosen": -0.0589175745844841,
1379
+ "rewards/margins": 0.03393319249153137,
1380
+ "rewards/rejected": -0.09285076707601547,
1381
+ "step": 970
1382
+ },
1383
+ {
1384
+ "epoch": 0.52,
1385
+ "learning_rate": 2.7393955451658387e-06,
1386
+ "logits/chosen": -0.001839430071413517,
1387
+ "logits/rejected": 0.02721802517771721,
1388
+ "logps/chosen": -569.8568115234375,
1389
+ "logps/rejected": -593.9893798828125,
1390
+ "loss": 0.1865,
1391
+ "rewards/accuracies": 0.512499988079071,
1392
+ "rewards/chosen": -0.058799803256988525,
1393
+ "rewards/margins": 0.0315038226544857,
1394
+ "rewards/rejected": -0.09030362218618393,
1395
+ "step": 980
1396
+ },
1397
+ {
1398
+ "epoch": 0.53,
1399
+ "learning_rate": 2.6930147487610667e-06,
1400
+ "logits/chosen": -0.014354798011481762,
1401
+ "logits/rejected": -0.015522453002631664,
1402
+ "logps/chosen": -546.56884765625,
1403
+ "logps/rejected": -624.5516967773438,
1404
+ "loss": 0.2,
1405
+ "rewards/accuracies": 0.4937500059604645,
1406
+ "rewards/chosen": -0.06721802055835724,
1407
+ "rewards/margins": 0.030059367418289185,
1408
+ "rewards/rejected": -0.09727738052606583,
1409
+ "step": 990
1410
+ },
1411
+ {
1412
+ "epoch": 0.53,
1413
+ "learning_rate": 2.6465670182413487e-06,
1414
+ "logits/chosen": 0.009198218584060669,
1415
+ "logits/rejected": -0.031388141214847565,
1416
+ "logps/chosen": -546.1246337890625,
1417
+ "logps/rejected": -607.7100219726562,
1418
+ "loss": 0.1942,
1419
+ "rewards/accuracies": 0.512499988079071,
1420
+ "rewards/chosen": -0.050740886479616165,
1421
+ "rewards/margins": 0.04155651852488518,
1422
+ "rewards/rejected": -0.09229740500450134,
1423
+ "step": 1000
1424
+ },
1425
+ {
1426
+ "epoch": 0.54,
1427
+ "learning_rate": 2.6000684608614594e-06,
1428
+ "logits/chosen": -0.1307494342327118,
1429
+ "logits/rejected": 0.043267279863357544,
1430
+ "logps/chosen": -562.8900146484375,
1431
+ "logps/rejected": -619.1513671875,
1432
+ "loss": 0.1872,
1433
+ "rewards/accuracies": 0.48124998807907104,
1434
+ "rewards/chosen": -0.06052188202738762,
1435
+ "rewards/margins": 0.034762654453516006,
1436
+ "rewards/rejected": -0.09528453648090363,
1437
+ "step": 1010
1438
+ },
1439
+ {
1440
+ "epoch": 0.54,
1441
+ "learning_rate": 2.5535352015020338e-06,
1442
+ "logits/chosen": -0.09118489176034927,
1443
+ "logits/rejected": -0.04373834654688835,
1444
+ "logps/chosen": -546.9132690429688,
1445
+ "logps/rejected": -610.7948608398438,
1446
+ "loss": 0.1832,
1447
+ "rewards/accuracies": 0.5,
1448
+ "rewards/chosen": -0.06443742662668228,
1449
+ "rewards/margins": 0.0450962632894516,
1450
+ "rewards/rejected": -0.10953368991613388,
1451
+ "step": 1020
1452
+ },
1453
+ {
1454
+ "epoch": 0.55,
1455
+ "learning_rate": 2.506983377077741e-06,
1456
+ "logits/chosen": -0.10389532893896103,
1457
+ "logits/rejected": 0.04294447600841522,
1458
+ "logps/chosen": -597.6129760742188,
1459
+ "logps/rejected": -630.375,
1460
+ "loss": 0.1955,
1461
+ "rewards/accuracies": 0.4749999940395355,
1462
+ "rewards/chosen": -0.050885915756225586,
1463
+ "rewards/margins": 0.04200722649693489,
1464
+ "rewards/rejected": -0.09289313852787018,
1465
+ "step": 1030
1466
+ },
1467
+ {
1468
+ "epoch": 0.55,
1469
+ "learning_rate": 2.460429130941289e-06,
1470
+ "logits/chosen": -0.05054370313882828,
1471
+ "logits/rejected": -0.041653625667095184,
1472
+ "logps/chosen": -562.7481689453125,
1473
+ "logps/rejected": -637.655029296875,
1474
+ "loss": 0.1907,
1475
+ "rewards/accuracies": 0.5,
1476
+ "rewards/chosen": -0.06117384508252144,
1477
+ "rewards/margins": 0.04911119490861893,
1478
+ "rewards/rejected": -0.11028504371643066,
1479
+ "step": 1040
1480
+ },
1481
+ {
1482
+ "epoch": 0.56,
1483
+ "learning_rate": 2.413888607285192e-06,
1484
+ "logits/chosen": -0.13274100422859192,
1485
+ "logits/rejected": -0.03395792096853256,
1486
+ "logps/chosen": -592.2869262695312,
1487
+ "logps/rejected": -656.2505493164062,
1488
+ "loss": 0.1916,
1489
+ "rewards/accuracies": 0.4625000059604645,
1490
+ "rewards/chosen": -0.06654617935419083,
1491
+ "rewards/margins": 0.048628997057676315,
1492
+ "rewards/rejected": -0.11517517268657684,
1493
+ "step": 1050
1494
+ },
1495
+ {
1496
+ "epoch": 0.57,
1497
+ "learning_rate": 2.367377945543249e-06,
1498
+ "logits/chosen": -0.04836379736661911,
1499
+ "logits/rejected": 0.0366315022110939,
1500
+ "logps/chosen": -561.9847412109375,
1501
+ "logps/rejected": -609.58935546875,
1502
+ "loss": 0.1841,
1503
+ "rewards/accuracies": 0.48124998807907104,
1504
+ "rewards/chosen": -0.05483894422650337,
1505
+ "rewards/margins": 0.03531566262245178,
1506
+ "rewards/rejected": -0.09015461057424545,
1507
+ "step": 1060
1508
+ },
1509
+ {
1510
+ "epoch": 0.57,
1511
+ "learning_rate": 2.320913274793676e-06,
1512
+ "logits/chosen": -0.02548578381538391,
1513
+ "logits/rejected": -0.011551322415471077,
1514
+ "logps/chosen": -517.5578002929688,
1515
+ "logps/rejected": -587.1166381835938,
1516
+ "loss": 0.1911,
1517
+ "rewards/accuracies": 0.53125,
1518
+ "rewards/chosen": -0.03753050044178963,
1519
+ "rewards/margins": 0.05013161897659302,
1520
+ "rewards/rejected": -0.08766211569309235,
1521
+ "step": 1070
1522
+ },
1523
+ {
1524
+ "epoch": 0.58,
1525
+ "learning_rate": 2.27451070816582e-06,
1526
+ "logits/chosen": -0.07835321873426437,
1527
+ "logits/rejected": -0.008995351381599903,
1528
+ "logps/chosen": -603.7330932617188,
1529
+ "logps/rejected": -606.139404296875,
1530
+ "loss": 0.1808,
1531
+ "rewards/accuracies": 0.550000011920929,
1532
+ "rewards/chosen": -0.05653967335820198,
1533
+ "rewards/margins": 0.02898581698536873,
1534
+ "rewards/rejected": -0.08552548289299011,
1535
+ "step": 1080
1536
+ },
1537
+ {
1538
+ "epoch": 0.58,
1539
+ "learning_rate": 2.228186337252414e-06,
1540
+ "logits/chosen": -0.05141784995794296,
1541
+ "logits/rejected": -0.023404929786920547,
1542
+ "logps/chosen": -539.2532958984375,
1543
+ "logps/rejected": -597.3132934570312,
1544
+ "loss": 0.1901,
1545
+ "rewards/accuracies": 0.5687500238418579,
1546
+ "rewards/chosen": -0.04464349150657654,
1547
+ "rewards/margins": 0.027696922421455383,
1548
+ "rewards/rejected": -0.07234041392803192,
1549
+ "step": 1090
1550
+ },
1551
+ {
1552
+ "epoch": 0.59,
1553
+ "learning_rate": 2.1819562265292946e-06,
1554
+ "logits/chosen": -0.024759747087955475,
1555
+ "logits/rejected": -0.13666841387748718,
1556
+ "logps/chosen": -521.7346801757812,
1557
+ "logps/rejected": -593.8477783203125,
1558
+ "loss": 0.1896,
1559
+ "rewards/accuracies": 0.4625000059604645,
1560
+ "rewards/chosen": -0.04558413475751877,
1561
+ "rewards/margins": 0.028859147801995277,
1562
+ "rewards/rejected": -0.0744432732462883,
1563
+ "step": 1100
1564
+ },
1565
+ {
1566
+ "epoch": 0.59,
1567
+ "learning_rate": 2.1358364077845236e-06,
1568
+ "logits/chosen": -0.1430703103542328,
1569
+ "logits/rejected": -0.04963989928364754,
1570
+ "logps/chosen": -554.0672607421875,
1571
+ "logps/rejected": -595.87939453125,
1572
+ "loss": 0.1941,
1573
+ "rewards/accuracies": 0.4937500059604645,
1574
+ "rewards/chosen": -0.04022881016135216,
1575
+ "rewards/margins": 0.03246025741100311,
1576
+ "rewards/rejected": -0.07268907129764557,
1577
+ "step": 1110
1578
+ },
1579
+ {
1580
+ "epoch": 0.6,
1581
+ "learning_rate": 2.089842874558849e-06,
1582
+ "logits/chosen": -0.010374871082603931,
1583
+ "logits/rejected": -0.07016284763813019,
1584
+ "logps/chosen": -510.901611328125,
1585
+ "logps/rejected": -578.7679443359375,
1586
+ "loss": 0.2011,
1587
+ "rewards/accuracies": 0.5687500238418579,
1588
+ "rewards/chosen": -0.03890245407819748,
1589
+ "rewards/margins": 0.032768718898296356,
1590
+ "rewards/rejected": -0.07167118042707443,
1591
+ "step": 1120
1592
+ },
1593
+ {
1594
+ "epoch": 0.6,
1595
+ "learning_rate": 2.0439915765994242e-06,
1596
+ "logits/chosen": -0.06316359341144562,
1597
+ "logits/rejected": 0.0023398040793836117,
1598
+ "logps/chosen": -512.5825805664062,
1599
+ "logps/rejected": -597.9288330078125,
1600
+ "loss": 0.1817,
1601
+ "rewards/accuracies": 0.5249999761581421,
1602
+ "rewards/chosen": -0.04139110818505287,
1603
+ "rewards/margins": 0.03969164192676544,
1604
+ "rewards/rejected": -0.08108274638652802,
1605
+ "step": 1130
1606
+ },
1607
+ {
1608
+ "epoch": 0.61,
1609
+ "learning_rate": 1.9982984143287186e-06,
1610
+ "logits/chosen": -0.0061457292176783085,
1611
+ "logits/rejected": -0.014603073708713055,
1612
+ "logps/chosen": -573.0762939453125,
1613
+ "logps/rejected": -628.219970703125,
1614
+ "loss": 0.1784,
1615
+ "rewards/accuracies": 0.581250011920929,
1616
+ "rewards/chosen": -0.0488489493727684,
1617
+ "rewards/margins": 0.04788987711071968,
1618
+ "rewards/rejected": -0.09673883020877838,
1619
+ "step": 1140
1620
+ },
1621
+ {
1622
+ "epoch": 0.61,
1623
+ "learning_rate": 1.95277923333053e-06,
1624
+ "logits/chosen": -0.03414573520421982,
1625
+ "logits/rejected": -0.014127236790955067,
1626
+ "logps/chosen": -551.8556518554688,
1627
+ "logps/rejected": -587.9727783203125,
1628
+ "loss": 0.184,
1629
+ "rewards/accuracies": 0.5062500238418579,
1630
+ "rewards/chosen": -0.04027246683835983,
1631
+ "rewards/margins": 0.03260934352874756,
1632
+ "rewards/rejected": -0.07288181036710739,
1633
+ "step": 1150
1634
+ },
1635
+ {
1636
+ "epoch": 0.62,
1637
+ "learning_rate": 1.9074498188550156e-06,
1638
+ "logits/chosen": -0.040990687906742096,
1639
+ "logits/rejected": 0.027432560920715332,
1640
+ "logps/chosen": -519.9183349609375,
1641
+ "logps/rejected": -585.5509033203125,
1642
+ "loss": 0.1709,
1643
+ "rewards/accuracies": 0.581250011920929,
1644
+ "rewards/chosen": -0.028263911604881287,
1645
+ "rewards/margins": 0.038620587438344955,
1646
+ "rewards/rejected": -0.06688449531793594,
1647
+ "step": 1160
1648
+ },
1649
+ {
1650
+ "epoch": 0.62,
1651
+ "learning_rate": 1.862325890344643e-06,
1652
+ "logits/chosen": -0.09657453745603561,
1653
+ "logits/rejected": -0.028930921107530594,
1654
+ "logps/chosen": -509.68731689453125,
1655
+ "logps/rejected": -545.7371215820312,
1656
+ "loss": 0.1907,
1657
+ "rewards/accuracies": 0.4749999940395355,
1658
+ "rewards/chosen": -0.03510227054357529,
1659
+ "rewards/margins": 0.028696129098534584,
1660
+ "rewards/rejected": -0.06379840523004532,
1661
+ "step": 1170
1662
+ },
1663
+ {
1664
+ "epoch": 0.63,
1665
+ "learning_rate": 1.817423095982972e-06,
1666
+ "logits/chosen": -0.001943819923326373,
1667
+ "logits/rejected": -0.10663602501153946,
1668
+ "logps/chosen": -541.8197021484375,
1669
+ "logps/rejected": -631.3278198242188,
1670
+ "loss": 0.183,
1671
+ "rewards/accuracies": 0.550000011920929,
1672
+ "rewards/chosen": -0.05557183176279068,
1673
+ "rewards/margins": 0.04637579992413521,
1674
+ "rewards/rejected": -0.10194762796163559,
1675
+ "step": 1180
1676
+ },
1677
+ {
1678
+ "epoch": 0.63,
1679
+ "learning_rate": 1.7727570072681293e-06,
1680
+ "logits/chosen": -0.03551124408841133,
1681
+ "logits/rejected": -0.03228951245546341,
1682
+ "logps/chosen": -513.2832641601562,
1683
+ "logps/rejected": -647.9851684570312,
1684
+ "loss": 0.1855,
1685
+ "rewards/accuracies": 0.53125,
1686
+ "rewards/chosen": -0.03109642304480076,
1687
+ "rewards/margins": 0.06740632653236389,
1688
+ "rewards/rejected": -0.0985027551651001,
1689
+ "step": 1190
1690
+ },
1691
+ {
1692
+ "epoch": 0.64,
1693
+ "learning_rate": 1.7283431136128961e-06,
1694
+ "logits/chosen": -0.006384107284247875,
1695
+ "logits/rejected": -0.0014162927400320768,
1696
+ "logps/chosen": -545.4998168945312,
1697
+ "logps/rejected": -616.1009521484375,
1698
+ "loss": 0.1799,
1699
+ "rewards/accuracies": 0.550000011920929,
1700
+ "rewards/chosen": -0.057301461696624756,
1701
+ "rewards/margins": 0.0508766695857048,
1702
+ "rewards/rejected": -0.10817812383174896,
1703
+ "step": 1200
1704
+ },
1705
+ {
1706
+ "epoch": 0.65,
1707
+ "learning_rate": 1.6841968169732478e-06,
1708
+ "logits/chosen": 0.031141120940446854,
1709
+ "logits/rejected": -0.0957186371088028,
1710
+ "logps/chosen": -535.7457275390625,
1711
+ "logps/rejected": -653.5030517578125,
1712
+ "loss": 0.1861,
1713
+ "rewards/accuracies": 0.5062500238418579,
1714
+ "rewards/chosen": -0.06120613217353821,
1715
+ "rewards/margins": 0.046815842390060425,
1716
+ "rewards/rejected": -0.10802197456359863,
1717
+ "step": 1210
1718
+ },
1719
+ {
1720
+ "epoch": 0.65,
1721
+ "learning_rate": 1.6403334265072284e-06,
1722
+ "logits/chosen": -0.06112520769238472,
1723
+ "logits/rejected": 0.02096955105662346,
1724
+ "logps/chosen": -497.9486389160156,
1725
+ "logps/rejected": -615.8877563476562,
1726
+ "loss": 0.1872,
1727
+ "rewards/accuracies": 0.5062500238418579,
1728
+ "rewards/chosen": -0.03988587483763695,
1729
+ "rewards/margins": 0.04784765467047691,
1730
+ "rewards/rejected": -0.08773352205753326,
1731
+ "step": 1220
1732
+ },
1733
+ {
1734
+ "epoch": 0.66,
1735
+ "learning_rate": 1.5967681532660066e-06,
1736
+ "logits/chosen": -0.07220903038978577,
1737
+ "logits/rejected": 0.042104389518499374,
1738
+ "logps/chosen": -510.8187561035156,
1739
+ "logps/rejected": -635.9632568359375,
1740
+ "loss": 0.1726,
1741
+ "rewards/accuracies": 0.53125,
1742
+ "rewards/chosen": -0.04127117991447449,
1743
+ "rewards/margins": 0.06924723088741302,
1744
+ "rewards/rejected": -0.11051841825246811,
1745
+ "step": 1230
1746
+ },
1747
+ {
1748
+ "epoch": 0.66,
1749
+ "learning_rate": 1.5535161049189463e-06,
1750
+ "logits/chosen": -0.14815333485603333,
1751
+ "logits/rejected": -0.03901112079620361,
1752
+ "logps/chosen": -537.5811157226562,
1753
+ "logps/rejected": -621.5980834960938,
1754
+ "loss": 0.1831,
1755
+ "rewards/accuracies": 0.5562499761581421,
1756
+ "rewards/chosen": -0.058847375214099884,
1757
+ "rewards/margins": 0.05733171105384827,
1758
+ "rewards/rejected": -0.11617908626794815,
1759
+ "step": 1240
1760
+ },
1761
+ {
1762
+ "epoch": 0.67,
1763
+ "learning_rate": 1.5105922805145356e-06,
1764
+ "logits/chosen": -0.029178712517023087,
1765
+ "logits/rejected": -0.11420316994190216,
1766
+ "logps/chosen": -573.003662109375,
1767
+ "logps/rejected": -637.9074096679688,
1768
+ "loss": 0.1971,
1769
+ "rewards/accuracies": 0.45625001192092896,
1770
+ "rewards/chosen": -0.07729945331811905,
1771
+ "rewards/margins": 0.04123348742723465,
1772
+ "rewards/rejected": -0.1185329332947731,
1773
+ "step": 1250
1774
+ },
1775
+ {
1776
+ "epoch": 0.67,
1777
+ "learning_rate": 1.4680115652789823e-06,
1778
+ "logits/chosen": -0.07775846868753433,
1779
+ "logits/rejected": -0.07050256431102753,
1780
+ "logps/chosen": -551.1271362304688,
1781
+ "logps/rejected": -654.5675659179688,
1782
+ "loss": 0.1784,
1783
+ "rewards/accuracies": 0.581250011920929,
1784
+ "rewards/chosen": -0.06159217283129692,
1785
+ "rewards/margins": 0.07197807729244232,
1786
+ "rewards/rejected": -0.13357023894786835,
1787
+ "step": 1260
1788
+ },
1789
+ {
1790
+ "epoch": 0.68,
1791
+ "learning_rate": 1.4257887254542767e-06,
1792
+ "logits/chosen": -0.052437614649534225,
1793
+ "logits/rejected": -0.10926377773284912,
1794
+ "logps/chosen": -546.0693969726562,
1795
+ "logps/rejected": -625.9730834960938,
1796
+ "loss": 0.1806,
1797
+ "rewards/accuracies": 0.5062500238418579,
1798
+ "rewards/chosen": -0.053220171481370926,
1799
+ "rewards/margins": 0.04990285634994507,
1800
+ "rewards/rejected": -0.1031230241060257,
1801
+ "step": 1270
1802
+ },
1803
+ {
1804
+ "epoch": 0.68,
1805
+ "learning_rate": 1.3839384031775227e-06,
1806
+ "logits/chosen": 0.014421490021049976,
1807
+ "logits/rejected": -0.14018869400024414,
1808
+ "logps/chosen": -590.1350708007812,
1809
+ "logps/rejected": -659.4343872070312,
1810
+ "loss": 0.2018,
1811
+ "rewards/accuracies": 0.5375000238418579,
1812
+ "rewards/chosen": -0.051880426704883575,
1813
+ "rewards/margins": 0.05124548077583313,
1814
+ "rewards/rejected": -0.1031259074807167,
1815
+ "step": 1280
1816
+ },
1817
+ {
1818
+ "epoch": 0.69,
1819
+ "learning_rate": 1.342475111403298e-06,
1820
+ "logits/chosen": -0.0805274024605751,
1821
+ "logits/rejected": -0.01422274298965931,
1822
+ "logps/chosen": -526.9544067382812,
1823
+ "logps/rejected": -622.9330444335938,
1824
+ "loss": 0.1859,
1825
+ "rewards/accuracies": 0.46875,
1826
+ "rewards/chosen": -0.06152445822954178,
1827
+ "rewards/margins": 0.049809303134679794,
1828
+ "rewards/rejected": -0.11133376508951187,
1829
+ "step": 1290
1830
+ },
1831
+ {
1832
+ "epoch": 0.69,
1833
+ "learning_rate": 1.3014132288708209e-06,
1834
+ "logits/chosen": -0.04959375411272049,
1835
+ "logits/rejected": -0.12484890222549438,
1836
+ "logps/chosen": -564.7056274414062,
1837
+ "logps/rejected": -604.1170654296875,
1838
+ "loss": 0.2058,
1839
+ "rewards/accuracies": 0.44999998807907104,
1840
+ "rewards/chosen": -0.05899285152554512,
1841
+ "rewards/margins": 0.032194919884204865,
1842
+ "rewards/rejected": -0.09118776023387909,
1843
+ "step": 1300
1844
+ },
1845
+ {
1846
+ "epoch": 0.7,
1847
+ "learning_rate": 1.2607669951176549e-06,
1848
+ "logits/chosen": -0.046819113194942474,
1849
+ "logits/rejected": -0.02196955308318138,
1850
+ "logps/chosen": -544.0050048828125,
1851
+ "logps/rejected": -577.3683471679688,
1852
+ "loss": 0.1872,
1853
+ "rewards/accuracies": 0.518750011920929,
1854
+ "rewards/chosen": -0.0536922812461853,
1855
+ "rewards/margins": 0.038987286388874054,
1856
+ "rewards/rejected": -0.09267956763505936,
1857
+ "step": 1310
1858
+ },
1859
+ {
1860
+ "epoch": 0.7,
1861
+ "learning_rate": 1.2205505055416891e-06,
1862
+ "logits/chosen": -0.07024160772562027,
1863
+ "logits/rejected": -0.004563197493553162,
1864
+ "logps/chosen": -593.686279296875,
1865
+ "logps/rejected": -663.3975830078125,
1866
+ "loss": 0.2018,
1867
+ "rewards/accuracies": 0.512499988079071,
1868
+ "rewards/chosen": -0.05577061325311661,
1869
+ "rewards/margins": 0.04607420414686203,
1870
+ "rewards/rejected": -0.10184482485055923,
1871
+ "step": 1320
1872
+ },
1873
+ {
1874
+ "epoch": 0.71,
1875
+ "learning_rate": 1.1807777065131002e-06,
1876
+ "logits/chosen": -0.07550742477178574,
1877
+ "logits/rejected": -0.007421135902404785,
1878
+ "logps/chosen": -610.5904541015625,
1879
+ "logps/rejected": -628.4864501953125,
1880
+ "loss": 0.2139,
1881
+ "rewards/accuracies": 0.42500001192092896,
1882
+ "rewards/chosen": -0.06365049630403519,
1883
+ "rewards/margins": 0.018766289576888084,
1884
+ "rewards/rejected": -0.08241678774356842,
1885
+ "step": 1330
1886
+ },
1887
+ {
1888
+ "epoch": 0.71,
1889
+ "learning_rate": 1.1414623905380012e-06,
1890
+ "logits/chosen": -0.020526980981230736,
1891
+ "logits/rejected": -0.038831211626529694,
1892
+ "logps/chosen": -613.1051635742188,
1893
+ "logps/rejected": -662.9939575195312,
1894
+ "loss": 0.1844,
1895
+ "rewards/accuracies": 0.518750011920929,
1896
+ "rewards/chosen": -0.057425010949373245,
1897
+ "rewards/margins": 0.03942105546593666,
1898
+ "rewards/rejected": -0.09684605151414871,
1899
+ "step": 1340
1900
+ },
1901
+ {
1902
+ "epoch": 0.72,
1903
+ "learning_rate": 1.1026181914754388e-06,
1904
+ "logits/chosen": -0.05669023469090462,
1905
+ "logits/rejected": -0.005404030904173851,
1906
+ "logps/chosen": -522.0977783203125,
1907
+ "logps/rejected": -629.5926513671875,
1908
+ "loss": 0.1752,
1909
+ "rewards/accuracies": 0.5874999761581421,
1910
+ "rewards/chosen": -0.0492081418633461,
1911
+ "rewards/margins": 0.0713469535112381,
1912
+ "rewards/rejected": -0.1205551028251648,
1913
+ "step": 1350
1914
+ },
1915
+ {
1916
+ "epoch": 0.73,
1917
+ "learning_rate": 1.0642585798094136e-06,
1918
+ "logits/chosen": -0.03868585079908371,
1919
+ "logits/rejected": -0.15266025066375732,
1920
+ "logps/chosen": -537.9027099609375,
1921
+ "logps/rejected": -588.400146484375,
1922
+ "loss": 0.1912,
1923
+ "rewards/accuracies": 0.4625000059604645,
1924
+ "rewards/chosen": -0.04704970866441727,
1925
+ "rewards/margins": 0.03578130528330803,
1926
+ "rewards/rejected": -0.0828310176730156,
1927
+ "step": 1360
1928
+ },
1929
+ {
1930
+ "epoch": 0.73,
1931
+ "learning_rate": 1.0263968579775522e-06,
1932
+ "logits/chosen": -0.039743565022945404,
1933
+ "logits/rejected": -0.05752366781234741,
1934
+ "logps/chosen": -541.8310546875,
1935
+ "logps/rejected": -601.9227294921875,
1936
+ "loss": 0.2017,
1937
+ "rewards/accuracies": 0.5062500238418579,
1938
+ "rewards/chosen": -0.04554866999387741,
1939
+ "rewards/margins": 0.043122462928295135,
1940
+ "rewards/rejected": -0.08867113292217255,
1941
+ "step": 1370
1942
+ },
1943
+ {
1944
+ "epoch": 0.74,
1945
+ "learning_rate": 9.89046155758058e-07,
1946
+ "logits/chosen": -0.01734338328242302,
1947
+ "logits/rejected": -0.03730212524533272,
1948
+ "logps/chosen": -536.6981201171875,
1949
+ "logps/rejected": -599.0525512695312,
1950
+ "loss": 0.1948,
1951
+ "rewards/accuracies": 0.48124998807907104,
1952
+ "rewards/chosen": -0.05502275377511978,
1953
+ "rewards/margins": 0.03183166682720184,
1954
+ "rewards/rejected": -0.08685442060232162,
1955
+ "step": 1380
1956
+ },
1957
+ {
1958
+ "epoch": 0.74,
1959
+ "learning_rate": 9.52219425716534e-07,
1960
+ "logits/chosen": -0.0649690106511116,
1961
+ "logits/rejected": -0.030281294137239456,
1962
+ "logps/chosen": -510.13037109375,
1963
+ "logps/rejected": -563.4937133789062,
1964
+ "loss": 0.1828,
1965
+ "rewards/accuracies": 0.518750011920929,
1966
+ "rewards/chosen": -0.04460709169507027,
1967
+ "rewards/margins": 0.03472636267542839,
1968
+ "rewards/rejected": -0.07933344691991806,
1969
+ "step": 1390
1970
+ },
1971
+ {
1972
+ "epoch": 0.75,
1973
+ "learning_rate": 9.15929438714262e-07,
1974
+ "logits/chosen": -0.039043981581926346,
1975
+ "logits/rejected": -0.025227446109056473,
1976
+ "logps/chosen": -558.3739013671875,
1977
+ "logps/rejected": -604.859375,
1978
+ "loss": 0.1936,
1979
+ "rewards/accuracies": 0.550000011920929,
1980
+ "rewards/chosen": -0.042985010892152786,
1981
+ "rewards/margins": 0.040365200489759445,
1982
+ "rewards/rejected": -0.08335021138191223,
1983
+ "step": 1400
1984
+ },
1985
+ {
1986
+ "epoch": 0.75,
1987
+ "learning_rate": 8.801887794794911e-07,
1988
+ "logits/chosen": -0.044082947075366974,
1989
+ "logits/rejected": 0.006080737803131342,
1990
+ "logps/chosen": -583.2000122070312,
1991
+ "logps/rejected": -629.2266845703125,
1992
+ "loss": 0.1878,
1993
+ "rewards/accuracies": 0.53125,
1994
+ "rewards/chosen": -0.042018450796604156,
1995
+ "rewards/margins": 0.03190966323018074,
1996
+ "rewards/rejected": -0.0739281103014946,
1997
+ "step": 1410
1998
+ },
1999
+ {
2000
+ "epoch": 0.76,
2001
+ "learning_rate": 8.450098422432787e-07,
2002
+ "logits/chosen": -0.014850256033241749,
2003
+ "logits/rejected": -0.14079946279525757,
2004
+ "logps/chosen": -530.8626708984375,
2005
+ "logps/rejected": -583.5850830078125,
2006
+ "loss": 0.1933,
2007
+ "rewards/accuracies": 0.46875,
2008
+ "rewards/chosen": -0.0459921732544899,
2009
+ "rewards/margins": 0.03213239461183548,
2010
+ "rewards/rejected": -0.07812456786632538,
2011
+ "step": 1420
2012
+ },
2013
+ {
2014
+ "epoch": 0.76,
2015
+ "learning_rate": 8.104048264413858e-07,
2016
+ "logits/chosen": -0.10342959314584732,
2017
+ "logits/rejected": 0.01154373586177826,
2018
+ "logps/chosen": -565.1043701171875,
2019
+ "logps/rejected": -615.1676025390625,
2020
+ "loss": 0.1802,
2021
+ "rewards/accuracies": 0.543749988079071,
2022
+ "rewards/chosen": -0.0521150641143322,
2023
+ "rewards/margins": 0.03459666296839714,
2024
+ "rewards/rejected": -0.08671173453330994,
2025
+ "step": 1430
2026
+ },
2027
+ {
2028
+ "epoch": 0.77,
2029
+ "learning_rate": 7.763857324837321e-07,
2030
+ "logits/chosen": -0.025015175342559814,
2031
+ "logits/rejected": -0.002726233098655939,
2032
+ "logps/chosen": -575.7015380859375,
2033
+ "logps/rejected": -641.2514038085938,
2034
+ "loss": 0.1856,
2035
+ "rewards/accuracies": 0.5375000238418579,
2036
+ "rewards/chosen": -0.04756324738264084,
2037
+ "rewards/margins": 0.04533480852842331,
2038
+ "rewards/rejected": -0.09289805591106415,
2039
+ "step": 1440
2040
+ },
2041
+ {
2042
+ "epoch": 0.77,
2043
+ "learning_rate": 7.429643575928605e-07,
2044
+ "logits/chosen": 0.006048756651580334,
2045
+ "logits/rejected": -0.0436415895819664,
2046
+ "logps/chosen": -544.4678955078125,
2047
+ "logps/rejected": -597.9765625,
2048
+ "loss": 0.1921,
2049
+ "rewards/accuracies": 0.4937500059604645,
2050
+ "rewards/chosen": -0.0511322021484375,
2051
+ "rewards/margins": 0.02459198795258999,
2052
+ "rewards/rejected": -0.07572419196367264,
2053
+ "step": 1450
2054
+ },
2055
+ {
2056
+ "epoch": 0.78,
2057
+ "learning_rate": 7.101522917128709e-07,
2058
+ "logits/chosen": -0.06234214827418327,
2059
+ "logits/rejected": 0.00918728206306696,
2060
+ "logps/chosen": -555.4013061523438,
2061
+ "logps/rejected": -603.6314697265625,
2062
+ "loss": 0.195,
2063
+ "rewards/accuracies": 0.5062500238418579,
2064
+ "rewards/chosen": -0.05011584237217903,
2065
+ "rewards/margins": 0.03172696754336357,
2066
+ "rewards/rejected": -0.0818428173661232,
2067
+ "step": 1460
2068
+ },
2069
+ {
2070
+ "epoch": 0.78,
2071
+ "learning_rate": 6.779609134902312e-07,
2072
+ "logits/chosen": -0.043599605560302734,
2073
+ "logits/rejected": -0.10574928671121597,
2074
+ "logps/chosen": -516.0931396484375,
2075
+ "logps/rejected": -598.8382568359375,
2076
+ "loss": 0.1858,
2077
+ "rewards/accuracies": 0.5,
2078
+ "rewards/chosen": -0.047776833176612854,
2079
+ "rewards/margins": 0.048279162496328354,
2080
+ "rewards/rejected": -0.09605599194765091,
2081
+ "step": 1470
2082
+ },
2083
+ {
2084
+ "epoch": 0.79,
2085
+ "learning_rate": 6.464013863278629e-07,
2086
+ "logits/chosen": -0.001982520567253232,
2087
+ "logits/rejected": -0.0532495453953743,
2088
+ "logps/chosen": -522.6332397460938,
2089
+ "logps/rejected": -572.3375244140625,
2090
+ "loss": 0.1889,
2091
+ "rewards/accuracies": 0.5062500238418579,
2092
+ "rewards/chosen": -0.04353585094213486,
2093
+ "rewards/margins": 0.04176926985383034,
2094
+ "rewards/rejected": -0.0853051170706749,
2095
+ "step": 1480
2096
+ },
2097
+ {
2098
+ "epoch": 0.79,
2099
+ "learning_rate": 6.154846545138696e-07,
2100
+ "logits/chosen": -0.05790393799543381,
2101
+ "logits/rejected": -0.00911637395620346,
2102
+ "logps/chosen": -572.7697143554688,
2103
+ "logps/rejected": -623.4581909179688,
2104
+ "loss": 0.1777,
2105
+ "rewards/accuracies": 0.53125,
2106
+ "rewards/chosen": -0.049880485981702805,
2107
+ "rewards/margins": 0.043653905391693115,
2108
+ "rewards/rejected": -0.09353439509868622,
2109
+ "step": 1490
2110
+ },
2111
+ {
2112
+ "epoch": 0.8,
2113
+ "learning_rate": 5.852214394262515e-07,
2114
+ "logits/chosen": -0.012275640852749348,
2115
+ "logits/rejected": -0.04724960774183273,
2116
+ "logps/chosen": -537.2650756835938,
2117
+ "logps/rejected": -643.6207275390625,
2118
+ "loss": 0.1696,
2119
+ "rewards/accuracies": 0.5249999761581421,
2120
+ "rewards/chosen": -0.03831138834357262,
2121
+ "rewards/margins": 0.057165808975696564,
2122
+ "rewards/rejected": -0.09547720104455948,
2123
+ "step": 1500
2124
+ },
2125
+ {
2126
+ "epoch": 0.81,
2127
+ "learning_rate": 5.556222358149191e-07,
2128
+ "logits/chosen": -0.03172417730093002,
2129
+ "logits/rejected": -0.1095399409532547,
2130
+ "logps/chosen": -506.0896911621094,
2131
+ "logps/rejected": -564.1954345703125,
2132
+ "loss": 0.1888,
2133
+ "rewards/accuracies": 0.4937500059604645,
2134
+ "rewards/chosen": -0.047778304666280746,
2135
+ "rewards/margins": 0.02731451950967312,
2136
+ "rewards/rejected": -0.07509282976388931,
2137
+ "step": 1510
2138
+ },
2139
+ {
2140
+ "epoch": 0.81,
2141
+ "learning_rate": 5.266973081622992e-07,
2142
+ "logits/chosen": -0.079510398209095,
2143
+ "logits/rejected": -0.025600697845220566,
2144
+ "logps/chosen": -502.23529052734375,
2145
+ "logps/rejected": -601.0836181640625,
2146
+ "loss": 0.1855,
2147
+ "rewards/accuracies": 0.53125,
2148
+ "rewards/chosen": -0.05057615786790848,
2149
+ "rewards/margins": 0.03538522124290466,
2150
+ "rewards/rejected": -0.08596137166023254,
2151
+ "step": 1520
2152
+ },
2153
+ {
2154
+ "epoch": 0.82,
2155
+ "learning_rate": 4.984566871237942e-07,
2156
+ "logits/chosen": -0.05354876443743706,
2157
+ "logits/rejected": -0.003774791955947876,
2158
+ "logps/chosen": -595.66357421875,
2159
+ "logps/rejected": -653.7220458984375,
2160
+ "loss": 0.1824,
2161
+ "rewards/accuracies": 0.574999988079071,
2162
+ "rewards/chosen": -0.0476485937833786,
2163
+ "rewards/margins": 0.04878585413098335,
2164
+ "rewards/rejected": -0.09643445163965225,
2165
+ "step": 1530
2166
+ },
2167
+ {
2168
+ "epoch": 0.82,
2169
+ "learning_rate": 4.709101660493251e-07,
2170
+ "logits/chosen": -0.051042985171079636,
2171
+ "logits/rejected": -0.04658942669630051,
2172
+ "logps/chosen": -561.7081298828125,
2173
+ "logps/rejected": -605.6749877929688,
2174
+ "loss": 0.1711,
2175
+ "rewards/accuracies": 0.5687500238418579,
2176
+ "rewards/chosen": -0.04670659080147743,
2177
+ "rewards/margins": 0.05042141675949097,
2178
+ "rewards/rejected": -0.0971280112862587,
2179
+ "step": 1540
2180
+ },
2181
+ {
2182
+ "epoch": 0.83,
2183
+ "learning_rate": 4.440672975871743e-07,
2184
+ "logits/chosen": -0.05640207976102829,
2185
+ "logits/rejected": -0.034627966582775116,
2186
+ "logps/chosen": -542.857421875,
2187
+ "logps/rejected": -607.437744140625,
2188
+ "loss": 0.1882,
2189
+ "rewards/accuracies": 0.543749988079071,
2190
+ "rewards/chosen": -0.04426370561122894,
2191
+ "rewards/margins": 0.04681876674294472,
2192
+ "rewards/rejected": -0.09108246862888336,
2193
+ "step": 1550
2194
+ },
2195
+ {
2196
+ "epoch": 0.83,
2197
+ "learning_rate": 4.1793739037129134e-07,
2198
+ "logits/chosen": -0.02437787875533104,
2199
+ "logits/rejected": -0.011664156801998615,
2200
+ "logps/chosen": -626.6923828125,
2201
+ "logps/rejected": -659.685546875,
2202
+ "loss": 0.1894,
2203
+ "rewards/accuracies": 0.550000011920929,
2204
+ "rewards/chosen": -0.052075475454330444,
2205
+ "rewards/margins": 0.04108622670173645,
2206
+ "rewards/rejected": -0.0931617021560669,
2207
+ "step": 1560
2208
+ },
2209
+ {
2210
+ "epoch": 0.84,
2211
+ "learning_rate": 3.9252950579322405e-07,
2212
+ "logits/chosen": -0.03457965329289436,
2213
+ "logits/rejected": -0.04087120667099953,
2214
+ "logps/chosen": -503.4874572753906,
2215
+ "logps/rejected": -566.9150390625,
2216
+ "loss": 0.186,
2217
+ "rewards/accuracies": 0.4749999940395355,
2218
+ "rewards/chosen": -0.0481523722410202,
2219
+ "rewards/margins": 0.04138932749629021,
2220
+ "rewards/rejected": -0.08954168856143951,
2221
+ "step": 1570
2222
+ },
2223
+ {
2224
+ "epoch": 0.84,
2225
+ "learning_rate": 3.6785245485978864e-07,
2226
+ "logits/chosen": -0.12483050674200058,
2227
+ "logits/rejected": -0.037600237876176834,
2228
+ "logps/chosen": -563.97314453125,
2229
+ "logps/rejected": -627.5960693359375,
2230
+ "loss": 0.2002,
2231
+ "rewards/accuracies": 0.550000011920929,
2232
+ "rewards/chosen": -0.06187894195318222,
2233
+ "rewards/margins": 0.03723733127117157,
2234
+ "rewards/rejected": -0.0991162657737732,
2235
+ "step": 1580
2236
+ },
2237
+ {
2238
+ "epoch": 0.85,
2239
+ "learning_rate": 3.43914795137566e-07,
2240
+ "logits/chosen": -0.05473123863339424,
2241
+ "logits/rejected": -0.07926475256681442,
2242
+ "logps/chosen": -551.5614624023438,
2243
+ "logps/rejected": -604.369384765625,
2244
+ "loss": 0.184,
2245
+ "rewards/accuracies": 0.512499988079071,
2246
+ "rewards/chosen": -0.046952735632658005,
2247
+ "rewards/margins": 0.046221982687711716,
2248
+ "rewards/rejected": -0.09317471832036972,
2249
+ "step": 1590
2250
+ },
2251
+ {
2252
+ "epoch": 0.85,
2253
+ "learning_rate": 3.207248277852901e-07,
2254
+ "logits/chosen": -0.1253429353237152,
2255
+ "logits/rejected": -0.08013667166233063,
2256
+ "logps/chosen": -565.1073608398438,
2257
+ "logps/rejected": -605.0519409179688,
2258
+ "loss": 0.1981,
2259
+ "rewards/accuracies": 0.4375,
2260
+ "rewards/chosen": -0.05863552540540695,
2261
+ "rewards/margins": 0.026683971285820007,
2262
+ "rewards/rejected": -0.08531949669122696,
2263
+ "step": 1600
2264
+ },
2265
+ {
2266
+ "epoch": 0.86,
2267
+ "learning_rate": 2.9829059467515074e-07,
2268
+ "logits/chosen": -0.05738731101155281,
2269
+ "logits/rejected": -0.054315079003572464,
2270
+ "logps/chosen": -526.6010131835938,
2271
+ "logps/rejected": -576.3739013671875,
2272
+ "loss": 0.1928,
2273
+ "rewards/accuracies": 0.5062500238418579,
2274
+ "rewards/chosen": -0.04895392805337906,
2275
+ "rewards/margins": 0.04191981628537178,
2276
+ "rewards/rejected": -0.09087374061346054,
2277
+ "step": 1610
2278
+ },
2279
+ {
2280
+ "epoch": 0.86,
2281
+ "learning_rate": 2.766198756040153e-07,
2282
+ "logits/chosen": -0.02228875830769539,
2283
+ "logits/rejected": -0.031645432114601135,
2284
+ "logps/chosen": -580.3348388671875,
2285
+ "logps/rejected": -626.4335327148438,
2286
+ "loss": 0.1859,
2287
+ "rewards/accuracies": 0.5874999761581421,
2288
+ "rewards/chosen": -0.05218944698572159,
2289
+ "rewards/margins": 0.04899830371141434,
2290
+ "rewards/rejected": -0.10118775069713593,
2291
+ "step": 1620
2292
+ },
2293
+ {
2294
+ "epoch": 0.87,
2295
+ "learning_rate": 2.5572018559553155e-07,
2296
+ "logits/chosen": 0.026733970269560814,
2297
+ "logits/rejected": -0.07179022580385208,
2298
+ "logps/chosen": -546.0059814453125,
2299
+ "logps/rejected": -609.9427490234375,
2300
+ "loss": 0.1785,
2301
+ "rewards/accuracies": 0.5625,
2302
+ "rewards/chosen": -0.05384701490402222,
2303
+ "rewards/margins": 0.04153439775109291,
2304
+ "rewards/rejected": -0.09538142383098602,
2305
+ "step": 1630
2306
+ },
2307
+ {
2308
+ "epoch": 0.87,
2309
+ "learning_rate": 2.3559877229404864e-07,
2310
+ "logits/chosen": -0.10137800127267838,
2311
+ "logits/rejected": -0.06986267119646072,
2312
+ "logps/chosen": -515.6650390625,
2313
+ "logps/rejected": -605.7598876953125,
2314
+ "loss": 0.182,
2315
+ "rewards/accuracies": 0.550000011920929,
2316
+ "rewards/chosen": -0.050880689173936844,
2317
+ "rewards/margins": 0.054004967212677,
2318
+ "rewards/rejected": -0.10488566011190414,
2319
+ "step": 1640
2320
+ },
2321
+ {
2322
+ "epoch": 0.88,
2323
+ "learning_rate": 2.1626261345126576e-07,
2324
+ "logits/chosen": -0.05389503762125969,
2325
+ "logits/rejected": -0.05351116135716438,
2326
+ "logps/chosen": -596.9784545898438,
2327
+ "logps/rejected": -662.634521484375,
2328
+ "loss": 0.185,
2329
+ "rewards/accuracies": 0.5687500238418579,
2330
+ "rewards/chosen": -0.06371568143367767,
2331
+ "rewards/margins": 0.05009468272328377,
2332
+ "rewards/rejected": -0.11381037533283234,
2333
+ "step": 1650
2334
+ },
2335
+ {
2336
+ "epoch": 0.89,
2337
+ "learning_rate": 1.9771841450646505e-07,
2338
+ "logits/chosen": -0.12013034522533417,
2339
+ "logits/rejected": 0.005744548514485359,
2340
+ "logps/chosen": -562.4629516601562,
2341
+ "logps/rejected": -641.6666259765625,
2342
+ "loss": 0.1958,
2343
+ "rewards/accuracies": 0.5562499761581421,
2344
+ "rewards/chosen": -0.047561004757881165,
2345
+ "rewards/margins": 0.061145197600126266,
2346
+ "rewards/rejected": -0.10870621353387833,
2347
+ "step": 1660
2348
+ },
2349
+ {
2350
+ "epoch": 0.89,
2351
+ "learning_rate": 1.7997260626118758e-07,
2352
+ "logits/chosen": -0.09952473640441895,
2353
+ "logits/rejected": -0.06873885542154312,
2354
+ "logps/chosen": -495.61199951171875,
2355
+ "logps/rejected": -587.5067138671875,
2356
+ "loss": 0.1836,
2357
+ "rewards/accuracies": 0.4937500059604645,
2358
+ "rewards/chosen": -0.05300045758485794,
2359
+ "rewards/margins": 0.050713229924440384,
2360
+ "rewards/rejected": -0.10371369123458862,
2361
+ "step": 1670
2362
+ },
2363
+ {
2364
+ "epoch": 0.9,
2365
+ "learning_rate": 1.6303134264914365e-07,
2366
+ "logits/chosen": -0.025821620598435402,
2367
+ "logits/rejected": -0.15098163485527039,
2368
+ "logps/chosen": -559.591064453125,
2369
+ "logps/rejected": -673.056640625,
2370
+ "loss": 0.1711,
2371
+ "rewards/accuracies": 0.5562499761581421,
2372
+ "rewards/chosen": -0.06007053703069687,
2373
+ "rewards/margins": 0.05461307615041733,
2374
+ "rewards/rejected": -0.1146836131811142,
2375
+ "step": 1680
2376
+ },
2377
+ {
2378
+ "epoch": 0.9,
2379
+ "learning_rate": 1.469004986021355e-07,
2380
+ "logits/chosen": 0.037708066403865814,
2381
+ "logits/rejected": -0.11987550556659698,
2382
+ "logps/chosen": -585.4603881835938,
2383
+ "logps/rejected": -654.9968872070312,
2384
+ "loss": 0.1686,
2385
+ "rewards/accuracies": 0.550000011920929,
2386
+ "rewards/chosen": -0.05478737875819206,
2387
+ "rewards/margins": 0.05192810297012329,
2388
+ "rewards/rejected": -0.10671547800302505,
2389
+ "step": 1690
2390
+ },
2391
+ {
2392
+ "epoch": 0.91,
2393
+ "learning_rate": 1.315856680127367e-07,
2394
+ "logits/chosen": -0.023753302171826363,
2395
+ "logits/rejected": -0.022690903395414352,
2396
+ "logps/chosen": -563.29345703125,
2397
+ "logps/rejected": -643.427490234375,
2398
+ "loss": 0.1791,
2399
+ "rewards/accuracies": 0.53125,
2400
+ "rewards/chosen": -0.05222300440073013,
2401
+ "rewards/margins": 0.05298025533556938,
2402
+ "rewards/rejected": -0.10520325601100922,
2403
+ "step": 1700
2404
+ },
2405
+ {
2406
+ "epoch": 0.91,
2407
+ "learning_rate": 1.1709216179442817e-07,
2408
+ "logits/chosen": -0.07842310518026352,
2409
+ "logits/rejected": -0.06540755927562714,
2410
+ "logps/chosen": -601.8659057617188,
2411
+ "logps/rejected": -662.879150390625,
2412
+ "loss": 0.1923,
2413
+ "rewards/accuracies": 0.53125,
2414
+ "rewards/chosen": -0.06227841228246689,
2415
+ "rewards/margins": 0.040394991636276245,
2416
+ "rewards/rejected": -0.10267341136932373,
2417
+ "step": 1710
2418
+ },
2419
+ {
2420
+ "epoch": 0.92,
2421
+ "learning_rate": 1.0342500603986421e-07,
2422
+ "logits/chosen": -0.05102648586034775,
2423
+ "logits/rejected": -0.06519290059804916,
2424
+ "logps/chosen": -558.7926025390625,
2425
+ "logps/rejected": -606.8863525390625,
2426
+ "loss": 0.1921,
2427
+ "rewards/accuracies": 0.4749999940395355,
2428
+ "rewards/chosen": -0.0554363913834095,
2429
+ "rewards/margins": 0.03473091870546341,
2430
+ "rewards/rejected": -0.09016730636358261,
2431
+ "step": 1720
2432
+ },
2433
+ {
2434
+ "epoch": 0.92,
2435
+ "learning_rate": 9.058894027791643e-08,
2436
+ "logits/chosen": -0.005893229506909847,
2437
+ "logits/rejected": -0.06061660498380661,
2438
+ "logps/chosen": -520.56201171875,
2439
+ "logps/rejected": -629.1943969726562,
2440
+ "loss": 0.1805,
2441
+ "rewards/accuracies": 0.59375,
2442
+ "rewards/chosen": -0.0561298243701458,
2443
+ "rewards/margins": 0.04909069091081619,
2444
+ "rewards/rejected": -0.10522051900625229,
2445
+ "step": 1730
2446
+ },
2447
+ {
2448
+ "epoch": 0.93,
2449
+ "learning_rate": 7.858841583008592e-08,
2450
+ "logits/chosen": -0.02195947803556919,
2451
+ "logits/rejected": -0.0024458705447614193,
2452
+ "logps/chosen": -613.8434448242188,
2453
+ "logps/rejected": -640.0768432617188,
2454
+ "loss": 0.2113,
2455
+ "rewards/accuracies": 0.5,
2456
+ "rewards/chosen": -0.06373412162065506,
2457
+ "rewards/margins": 0.03569970279932022,
2458
+ "rewards/rejected": -0.09943382441997528,
2459
+ "step": 1740
2460
+ },
2461
+ {
2462
+ "epoch": 0.93,
2463
+ "learning_rate": 6.742759426686313e-08,
2464
+ "logits/chosen": -0.05320558696985245,
2465
+ "logits/rejected": -0.048701487481594086,
2466
+ "logps/chosen": -581.4453125,
2467
+ "logps/rejected": -626.544677734375,
2468
+ "loss": 0.1786,
2469
+ "rewards/accuracies": 0.5062500238418579,
2470
+ "rewards/chosen": -0.058798860758543015,
2471
+ "rewards/margins": 0.04494379833340645,
2472
+ "rewards/rejected": -0.10374267399311066,
2473
+ "step": 1750
2474
+ },
2475
+ {
2476
+ "epoch": 0.94,
2477
+ "learning_rate": 5.7110345964571104e-08,
2478
+ "logits/chosen": -0.06332091242074966,
2479
+ "logits/rejected": -0.0014598838752135634,
2480
+ "logps/chosen": -652.0838623046875,
2481
+ "logps/rejected": -733.8502807617188,
2482
+ "loss": 0.1778,
2483
+ "rewards/accuracies": 0.59375,
2484
+ "rewards/chosen": -0.057877421379089355,
2485
+ "rewards/margins": 0.06819779425859451,
2486
+ "rewards/rejected": -0.12607520818710327,
2487
+ "step": 1760
2488
+ },
2489
+ {
2490
+ "epoch": 0.94,
2491
+ "learning_rate": 4.764024876318357e-08,
2492
+ "logits/chosen": -0.10024017095565796,
2493
+ "logits/rejected": 0.01046714186668396,
2494
+ "logps/chosen": -522.90185546875,
2495
+ "logps/rejected": -571.7759399414062,
2496
+ "loss": 0.2054,
2497
+ "rewards/accuracies": 0.5062500238418579,
2498
+ "rewards/chosen": -0.06100086122751236,
2499
+ "rewards/margins": 0.03126870468258858,
2500
+ "rewards/rejected": -0.09226956218481064,
2501
+ "step": 1770
2502
+ },
2503
+ {
2504
+ "epoch": 0.95,
2505
+ "learning_rate": 3.902058672559633e-08,
2506
+ "logits/chosen": 0.03307167440652847,
2507
+ "logits/rejected": -0.036007530987262726,
2508
+ "logps/chosen": -591.3616943359375,
2509
+ "logps/rejected": -665.3886108398438,
2510
+ "loss": 0.1901,
2511
+ "rewards/accuracies": 0.48750001192092896,
2512
+ "rewards/chosen": -0.05448399856686592,
2513
+ "rewards/margins": 0.04371560364961624,
2514
+ "rewards/rejected": -0.09819960594177246,
2515
+ "step": 1780
2516
+ },
2517
+ {
2518
+ "epoch": 0.95,
2519
+ "learning_rate": 3.125434899876933e-08,
2520
+ "logits/chosen": 0.0261197779327631,
2521
+ "logits/rejected": -0.10165154933929443,
2522
+ "logps/chosen": -611.7322387695312,
2523
+ "logps/rejected": -695.073974609375,
2524
+ "loss": 0.197,
2525
+ "rewards/accuracies": 0.5375000238418579,
2526
+ "rewards/chosen": -0.060929931700229645,
2527
+ "rewards/margins": 0.049871496856212616,
2528
+ "rewards/rejected": -0.11080142110586166,
2529
+ "step": 1790
2530
+ },
2531
+ {
2532
+ "epoch": 0.96,
2533
+ "learning_rate": 2.4344228777145873e-08,
2534
+ "logits/chosen": -0.10922287404537201,
2535
+ "logits/rejected": -0.036579955369234085,
2536
+ "logps/chosen": -538.944091796875,
2537
+ "logps/rejected": -600.6095581054688,
2538
+ "loss": 0.1917,
2539
+ "rewards/accuracies": 0.512499988079071,
2540
+ "rewards/chosen": -0.05488158017396927,
2541
+ "rewards/margins": 0.04075679928064346,
2542
+ "rewards/rejected": -0.09563836455345154,
2543
+ "step": 1800
2544
+ },
2545
+ {
2546
+ "epoch": 0.97,
2547
+ "learning_rate": 1.829262236869772e-08,
2548
+ "logits/chosen": 0.006532920990139246,
2549
+ "logits/rejected": -0.0661952868103981,
2550
+ "logps/chosen": -537.886962890625,
2551
+ "logps/rejected": -634.1477661132812,
2552
+ "loss": 0.2016,
2553
+ "rewards/accuracies": 0.543749988079071,
2554
+ "rewards/chosen": -0.05655653029680252,
2555
+ "rewards/margins": 0.06689898669719696,
2556
+ "rewards/rejected": -0.12345552444458008,
2557
+ "step": 1810
2558
+ },
2559
+ {
2560
+ "epoch": 0.97,
2561
+ "learning_rate": 1.3101628363929586e-08,
2562
+ "logits/chosen": -0.0753311961889267,
2563
+ "logits/rejected": -0.013547460548579693,
2564
+ "logps/chosen": -561.2381591796875,
2565
+ "logps/rejected": -576.2022094726562,
2566
+ "loss": 0.1924,
2567
+ "rewards/accuracies": 0.518750011920929,
2568
+ "rewards/chosen": -0.058451224118471146,
2569
+ "rewards/margins": 0.025429999455809593,
2570
+ "rewards/rejected": -0.08388122171163559,
2571
+ "step": 1820
2572
+ },
2573
+ {
2574
+ "epoch": 0.98,
2575
+ "learning_rate": 8.773046908123195e-09,
2576
+ "logits/chosen": -0.022709984332323074,
2577
+ "logits/rejected": -0.028875475749373436,
2578
+ "logps/chosen": -606.6187744140625,
2579
+ "logps/rejected": -637.9049072265625,
2580
+ "loss": 0.198,
2581
+ "rewards/accuracies": 0.5249999761581421,
2582
+ "rewards/chosen": -0.072281613945961,
2583
+ "rewards/margins": 0.030539941042661667,
2584
+ "rewards/rejected": -0.10282155126333237,
2585
+ "step": 1830
2586
+ },
2587
+ {
2588
+ "epoch": 0.98,
2589
+ "learning_rate": 5.308379077080817e-09,
2590
+ "logits/chosen": 0.0019970417488366365,
2591
+ "logits/rejected": -0.02689496800303459,
2592
+ "logps/chosen": -602.9913940429688,
2593
+ "logps/rejected": -666.972412109375,
2594
+ "loss": 0.196,
2595
+ "rewards/accuracies": 0.53125,
2596
+ "rewards/chosen": -0.057968251407146454,
2597
+ "rewards/margins": 0.04878325015306473,
2598
+ "rewards/rejected": -0.10675151646137238,
2599
+ "step": 1840
2600
+ },
2601
+ {
2602
+ "epoch": 0.99,
2603
+ "learning_rate": 2.7088263565760996e-09,
2604
+ "logits/chosen": -0.10940905660390854,
2605
+ "logits/rejected": -0.03552763909101486,
2606
+ "logps/chosen": -565.7769775390625,
2607
+ "logps/rejected": -600.9927368164062,
2608
+ "loss": 0.1964,
2609
+ "rewards/accuracies": 0.48750001192092896,
2610
+ "rewards/chosen": -0.04456068575382233,
2611
+ "rewards/margins": 0.03636551275849342,
2612
+ "rewards/rejected": -0.08092619478702545,
2613
+ "step": 1850
2614
+ },
2615
+ {
2616
+ "epoch": 0.99,
2617
+ "learning_rate": 9.752902257023633e-10,
2618
+ "logits/chosen": -0.02857157588005066,
2619
+ "logits/rejected": 0.05758611112833023,
2620
+ "logps/chosen": -594.5465087890625,
2621
+ "logps/rejected": -621.6343994140625,
2622
+ "loss": 0.1832,
2623
+ "rewards/accuracies": 0.5,
2624
+ "rewards/chosen": -0.05167108029127121,
2625
+ "rewards/margins": 0.037888895720243454,
2626
+ "rewards/rejected": -0.08955998718738556,
2627
+ "step": 1860
2628
+ },
2629
+ {
2630
+ "epoch": 1.0,
2631
+ "learning_rate": 1.083718442532189e-10,
2632
+ "logits/chosen": -0.07677461951971054,
2633
+ "logits/rejected": 0.008063477464020252,
2634
+ "logps/chosen": -562.1439208984375,
2635
+ "logps/rejected": -642.2802734375,
2636
+ "loss": 0.1852,
2637
+ "rewards/accuracies": 0.5375000238418579,
2638
+ "rewards/chosen": -0.0575174018740654,
2639
+ "rewards/margins": 0.049335263669490814,
2640
+ "rewards/rejected": -0.10685265064239502,
2641
+ "step": 1870
2642
+ },
2643
+ {
2644
+ "epoch": 1.0,
2645
+ "step": 1875,
2646
+ "total_flos": 0.0,
2647
+ "train_loss": 0.19300631308555602,
2648
+ "train_runtime": 11612.5974,
2649
+ "train_samples_per_second": 2.583,
2650
+ "train_steps_per_second": 0.161
2651
+ }
2652
+ ],
2653
+ "logging_steps": 10,
2654
+ "max_steps": 1875,
2655
+ "num_input_tokens_seen": 0,
2656
+ "num_train_epochs": 1,
2657
+ "save_steps": 100,
2658
+ "total_flos": 0.0,
2659
+ "train_batch_size": 4,
2660
+ "trial_name": null,
2661
+ "trial_params": null
2662
+ }