Model save
Browse files- README.md +84 -0
- adapter_model.safetensors +1 -1
- all_results.json +21 -0
- eval_results.json +16 -0
- train_results.json +8 -0
- trainer_state.json +1972 -0
README.md
ADDED
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
library_name: peft
|
4 |
+
tags:
|
5 |
+
- trl
|
6 |
+
- dpo
|
7 |
+
- generated_from_trainer
|
8 |
+
base_model: mistralai/Mistral-7B-Instruct-v0.2
|
9 |
+
model-index:
|
10 |
+
- name: zephyr-7b-dpo-qlora-pairrm
|
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 |
+
# zephyr-7b-dpo-qlora-pairrm
|
18 |
+
|
19 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) on the None dataset.
|
20 |
+
It achieves the following results on the evaluation set:
|
21 |
+
- Loss: 0.6773
|
22 |
+
- Rewards/chosen: -1.5442
|
23 |
+
- Rewards/rejected: -1.6837
|
24 |
+
- Rewards/accuracies: 0.5687
|
25 |
+
- Rewards/margins: 0.1395
|
26 |
+
- Logps/rejected: -394.7031
|
27 |
+
- Logps/chosen: -375.1367
|
28 |
+
- Logits/rejected: -4.4436
|
29 |
+
- Logits/chosen: -4.4568
|
30 |
+
|
31 |
+
## Model description
|
32 |
+
|
33 |
+
More information needed
|
34 |
+
|
35 |
+
## Intended uses & limitations
|
36 |
+
|
37 |
+
More information needed
|
38 |
+
|
39 |
+
## Training and evaluation data
|
40 |
+
|
41 |
+
More information needed
|
42 |
+
|
43 |
+
## Training procedure
|
44 |
+
|
45 |
+
### Training hyperparameters
|
46 |
+
|
47 |
+
The following hyperparameters were used during training:
|
48 |
+
- learning_rate: 5e-06
|
49 |
+
- train_batch_size: 4
|
50 |
+
- eval_batch_size: 8
|
51 |
+
- seed: 42
|
52 |
+
- distributed_type: multi-GPU
|
53 |
+
- gradient_accumulation_steps: 4
|
54 |
+
- total_train_batch_size: 16
|
55 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
56 |
+
- lr_scheduler_type: cosine
|
57 |
+
- lr_scheduler_warmup_ratio: 0.1
|
58 |
+
- num_epochs: 1
|
59 |
+
|
60 |
+
### Training results
|
61 |
+
|
62 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
63 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
64 |
+
| 0.6909 | 0.08 | 100 | 0.6921 | -0.0169 | -0.0192 | 0.5427 | 0.0023 | -228.2522 | -222.4038 | -2.6219 | -2.6247 |
|
65 |
+
| 0.684 | 0.16 | 200 | 0.6873 | -0.0803 | -0.0944 | 0.5567 | 0.0141 | -235.7721 | -228.7468 | -2.7599 | -2.7629 |
|
66 |
+
| 0.6795 | 0.24 | 300 | 0.6839 | -0.5138 | -0.5516 | 0.5460 | 0.0378 | -281.4856 | -272.0929 | -3.5141 | -3.5199 |
|
67 |
+
| 0.6561 | 0.32 | 400 | 0.6812 | -0.8158 | -0.8788 | 0.5573 | 0.0630 | -314.2105 | -302.2954 | -3.7484 | -3.7580 |
|
68 |
+
| 0.633 | 0.4 | 500 | 0.6787 | -0.9027 | -0.9810 | 0.5597 | 0.0782 | -324.4269 | -310.9858 | -4.0978 | -4.1077 |
|
69 |
+
| 0.6302 | 0.48 | 600 | 0.6785 | -1.1692 | -1.2692 | 0.5597 | 0.1000 | -353.2493 | -337.6355 | -4.4318 | -4.4435 |
|
70 |
+
| 0.5743 | 0.56 | 700 | 0.6835 | -1.5435 | -1.6640 | 0.5630 | 0.1205 | -392.7273 | -375.0575 | -4.5047 | -4.5182 |
|
71 |
+
| 0.6443 | 0.64 | 800 | 0.6779 | -1.3860 | -1.5069 | 0.5667 | 0.1209 | -377.0208 | -359.3108 | -4.2453 | -4.2572 |
|
72 |
+
| 0.6651 | 0.72 | 900 | 0.6819 | -1.6633 | -1.8040 | 0.5693 | 0.1408 | -406.7332 | -387.0414 | -4.6039 | -4.6178 |
|
73 |
+
| 0.5993 | 0.8 | 1000 | 0.6785 | -1.5776 | -1.7191 | 0.5683 | 0.1415 | -398.2364 | -378.4713 | -4.5356 | -4.5491 |
|
74 |
+
| 0.6759 | 0.88 | 1100 | 0.6778 | -1.5515 | -1.6923 | 0.5687 | 0.1408 | -395.5604 | -375.8654 | -4.4722 | -4.4855 |
|
75 |
+
| 0.6402 | 0.96 | 1200 | 0.6773 | -1.5439 | -1.6837 | 0.5690 | 0.1398 | -394.7002 | -375.1028 | -4.4444 | -4.4577 |
|
76 |
+
|
77 |
+
|
78 |
+
### Framework versions
|
79 |
+
|
80 |
+
- PEFT 0.7.1
|
81 |
+
- Transformers 4.36.2
|
82 |
+
- Pytorch 2.1.2
|
83 |
+
- Datasets 2.14.6
|
84 |
+
- Tokenizers 0.15.0
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1342238560
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f66dedcfa579dcdfb9f1d78c397d7645e70f57c173df9dd20bfc82911ce23de6
|
3 |
size 1342238560
|
all_results.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -4.456838607788086,
|
4 |
+
"eval_logits/rejected": -4.443554401397705,
|
5 |
+
"eval_logps/chosen": -375.13671875,
|
6 |
+
"eval_logps/rejected": -394.703125,
|
7 |
+
"eval_loss": 0.6773372888565063,
|
8 |
+
"eval_rewards/accuracies": 0.5686666369438171,
|
9 |
+
"eval_rewards/chosen": -1.5442434549331665,
|
10 |
+
"eval_rewards/margins": 0.13949742913246155,
|
11 |
+
"eval_rewards/rejected": -1.6837408542633057,
|
12 |
+
"eval_runtime": 1623.421,
|
13 |
+
"eval_samples": 2994,
|
14 |
+
"eval_samples_per_second": 1.844,
|
15 |
+
"eval_steps_per_second": 0.231,
|
16 |
+
"train_loss": 0.6475976420174225,
|
17 |
+
"train_runtime": 42677.4758,
|
18 |
+
"train_samples": 19996,
|
19 |
+
"train_samples_per_second": 0.469,
|
20 |
+
"train_steps_per_second": 0.029
|
21 |
+
}
|
eval_results.json
ADDED
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -4.456838607788086,
|
4 |
+
"eval_logits/rejected": -4.443554401397705,
|
5 |
+
"eval_logps/chosen": -375.13671875,
|
6 |
+
"eval_logps/rejected": -394.703125,
|
7 |
+
"eval_loss": 0.6773372888565063,
|
8 |
+
"eval_rewards/accuracies": 0.5686666369438171,
|
9 |
+
"eval_rewards/chosen": -1.5442434549331665,
|
10 |
+
"eval_rewards/margins": 0.13949742913246155,
|
11 |
+
"eval_rewards/rejected": -1.6837408542633057,
|
12 |
+
"eval_runtime": 1623.421,
|
13 |
+
"eval_samples": 2994,
|
14 |
+
"eval_samples_per_second": 1.844,
|
15 |
+
"eval_steps_per_second": 0.231
|
16 |
+
}
|
train_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.6475976420174225,
|
4 |
+
"train_runtime": 42677.4758,
|
5 |
+
"train_samples": 19996,
|
6 |
+
"train_samples_per_second": 0.469,
|
7 |
+
"train_steps_per_second": 0.029
|
8 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,1972 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.9993998799759952,
|
5 |
+
"eval_steps": 100,
|
6 |
+
"global_step": 1249,
|
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": 4e-08,
|
14 |
+
"logits/chosen": -2.682399272918701,
|
15 |
+
"logits/rejected": -2.7047135829925537,
|
16 |
+
"logps/chosen": -275.10638427734375,
|
17 |
+
"logps/rejected": -271.8466491699219,
|
18 |
+
"loss": 0.6931,
|
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": 4.0000000000000003e-07,
|
28 |
+
"logits/chosen": -2.606243848800659,
|
29 |
+
"logits/rejected": -2.633491277694702,
|
30 |
+
"logps/chosen": -301.7389831542969,
|
31 |
+
"logps/rejected": -324.2469787597656,
|
32 |
+
"loss": 0.6931,
|
33 |
+
"rewards/accuracies": 0.4166666567325592,
|
34 |
+
"rewards/chosen": 0.00018933300452772528,
|
35 |
+
"rewards/margins": 9.413135558133945e-06,
|
36 |
+
"rewards/rejected": 0.00017991992353927344,
|
37 |
+
"step": 10
|
38 |
+
},
|
39 |
+
{
|
40 |
+
"epoch": 0.02,
|
41 |
+
"learning_rate": 8.000000000000001e-07,
|
42 |
+
"logits/chosen": -2.5866377353668213,
|
43 |
+
"logits/rejected": -2.5900259017944336,
|
44 |
+
"logps/chosen": -269.0643615722656,
|
45 |
+
"logps/rejected": -289.1509094238281,
|
46 |
+
"loss": 0.6931,
|
47 |
+
"rewards/accuracies": 0.53125,
|
48 |
+
"rewards/chosen": -0.0011652575340121984,
|
49 |
+
"rewards/margins": 6.114605639595538e-05,
|
50 |
+
"rewards/rejected": -0.0012264035176485777,
|
51 |
+
"step": 20
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"epoch": 0.02,
|
55 |
+
"learning_rate": 1.2000000000000002e-06,
|
56 |
+
"logits/chosen": -2.57534122467041,
|
57 |
+
"logits/rejected": -2.5880730152130127,
|
58 |
+
"logps/chosen": -291.5533752441406,
|
59 |
+
"logps/rejected": -311.4080505371094,
|
60 |
+
"loss": 0.693,
|
61 |
+
"rewards/accuracies": 0.46875,
|
62 |
+
"rewards/chosen": -0.0030827566515654325,
|
63 |
+
"rewards/margins": 0.0002471129409968853,
|
64 |
+
"rewards/rejected": -0.0033298698253929615,
|
65 |
+
"step": 30
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"epoch": 0.03,
|
69 |
+
"learning_rate": 1.6000000000000001e-06,
|
70 |
+
"logits/chosen": -2.612988233566284,
|
71 |
+
"logits/rejected": -2.619412899017334,
|
72 |
+
"logps/chosen": -264.4569396972656,
|
73 |
+
"logps/rejected": -273.54107666015625,
|
74 |
+
"loss": 0.6927,
|
75 |
+
"rewards/accuracies": 0.5375000238418579,
|
76 |
+
"rewards/chosen": -0.005209661088883877,
|
77 |
+
"rewards/margins": 0.0008221397292800248,
|
78 |
+
"rewards/rejected": -0.006031800992786884,
|
79 |
+
"step": 40
|
80 |
+
},
|
81 |
+
{
|
82 |
+
"epoch": 0.04,
|
83 |
+
"learning_rate": 2.0000000000000003e-06,
|
84 |
+
"logits/chosen": -2.563633680343628,
|
85 |
+
"logits/rejected": -2.5437724590301514,
|
86 |
+
"logps/chosen": -264.8982849121094,
|
87 |
+
"logps/rejected": -269.68804931640625,
|
88 |
+
"loss": 0.6932,
|
89 |
+
"rewards/accuracies": 0.44999998807907104,
|
90 |
+
"rewards/chosen": -0.008133028633892536,
|
91 |
+
"rewards/margins": -1.2002186849713326e-05,
|
92 |
+
"rewards/rejected": -0.008121026679873466,
|
93 |
+
"step": 50
|
94 |
+
},
|
95 |
+
{
|
96 |
+
"epoch": 0.05,
|
97 |
+
"learning_rate": 2.4000000000000003e-06,
|
98 |
+
"logits/chosen": -2.6302762031555176,
|
99 |
+
"logits/rejected": -2.633596420288086,
|
100 |
+
"logps/chosen": -277.2352600097656,
|
101 |
+
"logps/rejected": -296.19476318359375,
|
102 |
+
"loss": 0.6921,
|
103 |
+
"rewards/accuracies": 0.550000011920929,
|
104 |
+
"rewards/chosen": -0.013791452161967754,
|
105 |
+
"rewards/margins": 0.0020633486565202475,
|
106 |
+
"rewards/rejected": -0.01585480198264122,
|
107 |
+
"step": 60
|
108 |
+
},
|
109 |
+
{
|
110 |
+
"epoch": 0.06,
|
111 |
+
"learning_rate": 2.8000000000000003e-06,
|
112 |
+
"logits/chosen": -2.6230902671813965,
|
113 |
+
"logits/rejected": -2.6135356426239014,
|
114 |
+
"logps/chosen": -280.782958984375,
|
115 |
+
"logps/rejected": -286.6590881347656,
|
116 |
+
"loss": 0.6931,
|
117 |
+
"rewards/accuracies": 0.4937500059604645,
|
118 |
+
"rewards/chosen": -0.01885022595524788,
|
119 |
+
"rewards/margins": 0.00015832395001780242,
|
120 |
+
"rewards/rejected": -0.01900855079293251,
|
121 |
+
"step": 70
|
122 |
+
},
|
123 |
+
{
|
124 |
+
"epoch": 0.06,
|
125 |
+
"learning_rate": 3.2000000000000003e-06,
|
126 |
+
"logits/chosen": -2.6521055698394775,
|
127 |
+
"logits/rejected": -2.650635004043579,
|
128 |
+
"logps/chosen": -276.92156982421875,
|
129 |
+
"logps/rejected": -297.58477783203125,
|
130 |
+
"loss": 0.6924,
|
131 |
+
"rewards/accuracies": 0.5249999761581421,
|
132 |
+
"rewards/chosen": -0.017211003229022026,
|
133 |
+
"rewards/margins": 0.0015935760457068682,
|
134 |
+
"rewards/rejected": -0.018804579973220825,
|
135 |
+
"step": 80
|
136 |
+
},
|
137 |
+
{
|
138 |
+
"epoch": 0.07,
|
139 |
+
"learning_rate": 3.6000000000000003e-06,
|
140 |
+
"logits/chosen": -2.650332450866699,
|
141 |
+
"logits/rejected": -2.639291286468506,
|
142 |
+
"logps/chosen": -311.17218017578125,
|
143 |
+
"logps/rejected": -316.7694091796875,
|
144 |
+
"loss": 0.6891,
|
145 |
+
"rewards/accuracies": 0.59375,
|
146 |
+
"rewards/chosen": -0.017588406801223755,
|
147 |
+
"rewards/margins": 0.008481341414153576,
|
148 |
+
"rewards/rejected": -0.026069749146699905,
|
149 |
+
"step": 90
|
150 |
+
},
|
151 |
+
{
|
152 |
+
"epoch": 0.08,
|
153 |
+
"learning_rate": 4.000000000000001e-06,
|
154 |
+
"logits/chosen": -2.6498260498046875,
|
155 |
+
"logits/rejected": -2.6482200622558594,
|
156 |
+
"logps/chosen": -276.51043701171875,
|
157 |
+
"logps/rejected": -290.5121154785156,
|
158 |
+
"loss": 0.6909,
|
159 |
+
"rewards/accuracies": 0.550000011920929,
|
160 |
+
"rewards/chosen": -0.02430056408047676,
|
161 |
+
"rewards/margins": 0.004733243025839329,
|
162 |
+
"rewards/rejected": -0.029033806174993515,
|
163 |
+
"step": 100
|
164 |
+
},
|
165 |
+
{
|
166 |
+
"epoch": 0.08,
|
167 |
+
"eval_logits/chosen": -2.6246910095214844,
|
168 |
+
"eval_logits/rejected": -2.6219358444213867,
|
169 |
+
"eval_logps/chosen": -222.40379333496094,
|
170 |
+
"eval_logps/rejected": -228.25218200683594,
|
171 |
+
"eval_loss": 0.6920604109764099,
|
172 |
+
"eval_rewards/accuracies": 0.5426666736602783,
|
173 |
+
"eval_rewards/chosen": -0.01691427268087864,
|
174 |
+
"eval_rewards/margins": 0.0023173687513917685,
|
175 |
+
"eval_rewards/rejected": -0.01923164166510105,
|
176 |
+
"eval_runtime": 1621.9667,
|
177 |
+
"eval_samples_per_second": 1.846,
|
178 |
+
"eval_steps_per_second": 0.231,
|
179 |
+
"step": 100
|
180 |
+
},
|
181 |
+
{
|
182 |
+
"epoch": 0.09,
|
183 |
+
"learning_rate": 4.4e-06,
|
184 |
+
"logits/chosen": -2.6501355171203613,
|
185 |
+
"logits/rejected": -2.652731418609619,
|
186 |
+
"logps/chosen": -298.9729919433594,
|
187 |
+
"logps/rejected": -309.8643493652344,
|
188 |
+
"loss": 0.6891,
|
189 |
+
"rewards/accuracies": 0.6000000238418579,
|
190 |
+
"rewards/chosen": -0.025512468069791794,
|
191 |
+
"rewards/margins": 0.008652618154883385,
|
192 |
+
"rewards/rejected": -0.03416508436203003,
|
193 |
+
"step": 110
|
194 |
+
},
|
195 |
+
{
|
196 |
+
"epoch": 0.1,
|
197 |
+
"learning_rate": 4.800000000000001e-06,
|
198 |
+
"logits/chosen": -2.676758289337158,
|
199 |
+
"logits/rejected": -2.6891021728515625,
|
200 |
+
"logps/chosen": -278.479736328125,
|
201 |
+
"logps/rejected": -296.43212890625,
|
202 |
+
"loss": 0.6903,
|
203 |
+
"rewards/accuracies": 0.4937500059604645,
|
204 |
+
"rewards/chosen": -0.031918395310640335,
|
205 |
+
"rewards/margins": 0.006463131867349148,
|
206 |
+
"rewards/rejected": -0.03838152438402176,
|
207 |
+
"step": 120
|
208 |
+
},
|
209 |
+
{
|
210 |
+
"epoch": 0.1,
|
211 |
+
"learning_rate": 4.999755876225375e-06,
|
212 |
+
"logits/chosen": -2.645005702972412,
|
213 |
+
"logits/rejected": -2.62728214263916,
|
214 |
+
"logps/chosen": -294.44366455078125,
|
215 |
+
"logps/rejected": -315.0718994140625,
|
216 |
+
"loss": 0.6878,
|
217 |
+
"rewards/accuracies": 0.574999988079071,
|
218 |
+
"rewards/chosen": -0.035983841866254807,
|
219 |
+
"rewards/margins": 0.011375428177416325,
|
220 |
+
"rewards/rejected": -0.047359269112348557,
|
221 |
+
"step": 130
|
222 |
+
},
|
223 |
+
{
|
224 |
+
"epoch": 0.11,
|
225 |
+
"learning_rate": 4.997803172081864e-06,
|
226 |
+
"logits/chosen": -2.6803853511810303,
|
227 |
+
"logits/rejected": -2.680997371673584,
|
228 |
+
"logps/chosen": -289.1062927246094,
|
229 |
+
"logps/rejected": -302.7191467285156,
|
230 |
+
"loss": 0.6855,
|
231 |
+
"rewards/accuracies": 0.5625,
|
232 |
+
"rewards/chosen": -0.041079964488744736,
|
233 |
+
"rewards/margins": 0.016574053093791008,
|
234 |
+
"rewards/rejected": -0.057654011994600296,
|
235 |
+
"step": 140
|
236 |
+
},
|
237 |
+
{
|
238 |
+
"epoch": 0.12,
|
239 |
+
"learning_rate": 4.9938992891651825e-06,
|
240 |
+
"logits/chosen": -2.6616640090942383,
|
241 |
+
"logits/rejected": -2.6513876914978027,
|
242 |
+
"logps/chosen": -277.707763671875,
|
243 |
+
"logps/rejected": -300.9044494628906,
|
244 |
+
"loss": 0.683,
|
245 |
+
"rewards/accuracies": 0.6000000238418579,
|
246 |
+
"rewards/chosen": -0.05293840169906616,
|
247 |
+
"rewards/margins": 0.0216156505048275,
|
248 |
+
"rewards/rejected": -0.07455406337976456,
|
249 |
+
"step": 150
|
250 |
+
},
|
251 |
+
{
|
252 |
+
"epoch": 0.13,
|
253 |
+
"learning_rate": 4.988047277024456e-06,
|
254 |
+
"logits/chosen": -2.7210304737091064,
|
255 |
+
"logits/rejected": -2.7316393852233887,
|
256 |
+
"logps/chosen": -288.5920715332031,
|
257 |
+
"logps/rejected": -304.19964599609375,
|
258 |
+
"loss": 0.6804,
|
259 |
+
"rewards/accuracies": 0.643750011920929,
|
260 |
+
"rewards/chosen": -0.07934443652629852,
|
261 |
+
"rewards/margins": 0.026776760816574097,
|
262 |
+
"rewards/rejected": -0.10612119734287262,
|
263 |
+
"step": 160
|
264 |
+
},
|
265 |
+
{
|
266 |
+
"epoch": 0.14,
|
267 |
+
"learning_rate": 4.980251707005417e-06,
|
268 |
+
"logits/chosen": -2.71783185005188,
|
269 |
+
"logits/rejected": -2.690868854522705,
|
270 |
+
"logps/chosen": -307.91888427734375,
|
271 |
+
"logps/rejected": -315.943359375,
|
272 |
+
"loss": 0.689,
|
273 |
+
"rewards/accuracies": 0.53125,
|
274 |
+
"rewards/chosen": -0.10906956344842911,
|
275 |
+
"rewards/margins": 0.009946177713572979,
|
276 |
+
"rewards/rejected": -0.11901573836803436,
|
277 |
+
"step": 170
|
278 |
+
},
|
279 |
+
{
|
280 |
+
"epoch": 0.14,
|
281 |
+
"learning_rate": 4.970518668679459e-06,
|
282 |
+
"logits/chosen": -2.729719638824463,
|
283 |
+
"logits/rejected": -2.714111804962158,
|
284 |
+
"logps/chosen": -304.322998046875,
|
285 |
+
"logps/rejected": -311.78167724609375,
|
286 |
+
"loss": 0.6816,
|
287 |
+
"rewards/accuracies": 0.574999988079071,
|
288 |
+
"rewards/chosen": -0.09336166828870773,
|
289 |
+
"rewards/margins": 0.0258515365421772,
|
290 |
+
"rewards/rejected": -0.11921320110559464,
|
291 |
+
"step": 180
|
292 |
+
},
|
293 |
+
{
|
294 |
+
"epoch": 0.15,
|
295 |
+
"learning_rate": 4.958855765086722e-06,
|
296 |
+
"logits/chosen": -2.757159471511841,
|
297 |
+
"logits/rejected": -2.7543435096740723,
|
298 |
+
"logps/chosen": -284.23687744140625,
|
299 |
+
"logps/rejected": -293.60595703125,
|
300 |
+
"loss": 0.6848,
|
301 |
+
"rewards/accuracies": 0.5562499761581421,
|
302 |
+
"rewards/chosen": -0.06431926041841507,
|
303 |
+
"rewards/margins": 0.019170444458723068,
|
304 |
+
"rewards/rejected": -0.08348970115184784,
|
305 |
+
"step": 190
|
306 |
+
},
|
307 |
+
{
|
308 |
+
"epoch": 0.16,
|
309 |
+
"learning_rate": 4.945272106796919e-06,
|
310 |
+
"logits/chosen": -2.770078420639038,
|
311 |
+
"logits/rejected": -2.7745845317840576,
|
312 |
+
"logps/chosen": -285.0936584472656,
|
313 |
+
"logps/rejected": -300.80975341796875,
|
314 |
+
"loss": 0.684,
|
315 |
+
"rewards/accuracies": 0.5874999761581421,
|
316 |
+
"rewards/chosen": -0.0751299113035202,
|
317 |
+
"rewards/margins": 0.02216259017586708,
|
318 |
+
"rewards/rejected": -0.09729250520467758,
|
319 |
+
"step": 200
|
320 |
+
},
|
321 |
+
{
|
322 |
+
"epoch": 0.16,
|
323 |
+
"eval_logits/chosen": -2.762911558151245,
|
324 |
+
"eval_logits/rejected": -2.759880542755127,
|
325 |
+
"eval_logps/chosen": -228.746826171875,
|
326 |
+
"eval_logps/rejected": -235.77212524414062,
|
327 |
+
"eval_loss": 0.6872997283935547,
|
328 |
+
"eval_rewards/accuracies": 0.5566666722297668,
|
329 |
+
"eval_rewards/chosen": -0.08034466207027435,
|
330 |
+
"eval_rewards/margins": 0.014086335897445679,
|
331 |
+
"eval_rewards/rejected": -0.09443099796772003,
|
332 |
+
"eval_runtime": 1664.3062,
|
333 |
+
"eval_samples_per_second": 1.799,
|
334 |
+
"eval_steps_per_second": 0.225,
|
335 |
+
"step": 200
|
336 |
+
},
|
337 |
+
{
|
338 |
+
"epoch": 0.17,
|
339 |
+
"learning_rate": 4.929778304792537e-06,
|
340 |
+
"logits/chosen": -2.7531464099884033,
|
341 |
+
"logits/rejected": -2.7568936347961426,
|
342 |
+
"logps/chosen": -310.95513916015625,
|
343 |
+
"logps/rejected": -315.43353271484375,
|
344 |
+
"loss": 0.6749,
|
345 |
+
"rewards/accuracies": 0.612500011920929,
|
346 |
+
"rewards/chosen": -0.08977816253900528,
|
347 |
+
"rewards/margins": 0.044371671974658966,
|
348 |
+
"rewards/rejected": -0.13414981961250305,
|
349 |
+
"step": 210
|
350 |
+
},
|
351 |
+
{
|
352 |
+
"epoch": 0.18,
|
353 |
+
"learning_rate": 4.912386462179987e-06,
|
354 |
+
"logits/chosen": -2.7818262577056885,
|
355 |
+
"logits/rejected": -2.777902126312256,
|
356 |
+
"logps/chosen": -298.80780029296875,
|
357 |
+
"logps/rejected": -325.99853515625,
|
358 |
+
"loss": 0.6741,
|
359 |
+
"rewards/accuracies": 0.606249988079071,
|
360 |
+
"rewards/chosen": -0.10158131271600723,
|
361 |
+
"rewards/margins": 0.044485487043857574,
|
362 |
+
"rewards/rejected": -0.1460667848587036,
|
363 |
+
"step": 220
|
364 |
+
},
|
365 |
+
{
|
366 |
+
"epoch": 0.18,
|
367 |
+
"learning_rate": 4.893110164735167e-06,
|
368 |
+
"logits/chosen": -2.8827590942382812,
|
369 |
+
"logits/rejected": -2.8796629905700684,
|
370 |
+
"logps/chosen": -305.14068603515625,
|
371 |
+
"logps/rejected": -319.39471435546875,
|
372 |
+
"loss": 0.6796,
|
373 |
+
"rewards/accuracies": 0.543749988079071,
|
374 |
+
"rewards/chosen": -0.1311652809381485,
|
375 |
+
"rewards/margins": 0.0319681391119957,
|
376 |
+
"rewards/rejected": -0.1631334125995636,
|
377 |
+
"step": 230
|
378 |
+
},
|
379 |
+
{
|
380 |
+
"epoch": 0.19,
|
381 |
+
"learning_rate": 4.871964470290823e-06,
|
382 |
+
"logits/chosen": -2.90864634513855,
|
383 |
+
"logits/rejected": -2.9260551929473877,
|
384 |
+
"logps/chosen": -309.19970703125,
|
385 |
+
"logps/rejected": -332.04119873046875,
|
386 |
+
"loss": 0.6654,
|
387 |
+
"rewards/accuracies": 0.643750011920929,
|
388 |
+
"rewards/chosen": -0.1595773994922638,
|
389 |
+
"rewards/margins": 0.06538807600736618,
|
390 |
+
"rewards/rejected": -0.22496548295021057,
|
391 |
+
"step": 240
|
392 |
+
},
|
393 |
+
{
|
394 |
+
"epoch": 0.2,
|
395 |
+
"learning_rate": 4.848965896974006e-06,
|
396 |
+
"logits/chosen": -2.947906494140625,
|
397 |
+
"logits/rejected": -2.940717935562134,
|
398 |
+
"logps/chosen": -302.98651123046875,
|
399 |
+
"logps/rejected": -325.380859375,
|
400 |
+
"loss": 0.6756,
|
401 |
+
"rewards/accuracies": 0.5562499761581421,
|
402 |
+
"rewards/chosen": -0.19427716732025146,
|
403 |
+
"rewards/margins": 0.043551910668611526,
|
404 |
+
"rewards/rejected": -0.2378290891647339,
|
405 |
+
"step": 250
|
406 |
+
},
|
407 |
+
{
|
408 |
+
"epoch": 0.21,
|
409 |
+
"learning_rate": 4.8241324103028055e-06,
|
410 |
+
"logits/chosen": -3.116504430770874,
|
411 |
+
"logits/rejected": -3.088792324066162,
|
412 |
+
"logps/chosen": -312.82159423828125,
|
413 |
+
"logps/rejected": -328.6378479003906,
|
414 |
+
"loss": 0.6614,
|
415 |
+
"rewards/accuracies": 0.606249988079071,
|
416 |
+
"rewards/chosen": -0.24597156047821045,
|
417 |
+
"rewards/margins": 0.07650710642337799,
|
418 |
+
"rewards/rejected": -0.322478711605072,
|
419 |
+
"step": 260
|
420 |
+
},
|
421 |
+
{
|
422 |
+
"epoch": 0.22,
|
423 |
+
"learning_rate": 4.797483409152438e-06,
|
424 |
+
"logits/chosen": -3.2219741344451904,
|
425 |
+
"logits/rejected": -3.211695432662964,
|
426 |
+
"logps/chosen": -308.4544372558594,
|
427 |
+
"logps/rejected": -333.4085388183594,
|
428 |
+
"loss": 0.6625,
|
429 |
+
"rewards/accuracies": 0.581250011920929,
|
430 |
+
"rewards/chosen": -0.3319942355155945,
|
431 |
+
"rewards/margins": 0.08054188638925552,
|
432 |
+
"rewards/rejected": -0.412536084651947,
|
433 |
+
"step": 270
|
434 |
+
},
|
435 |
+
{
|
436 |
+
"epoch": 0.22,
|
437 |
+
"learning_rate": 4.769039710601669e-06,
|
438 |
+
"logits/chosen": -3.368110179901123,
|
439 |
+
"logits/rejected": -3.3736987113952637,
|
440 |
+
"logps/chosen": -316.3270263671875,
|
441 |
+
"logps/rejected": -338.3800048828125,
|
442 |
+
"loss": 0.6608,
|
443 |
+
"rewards/accuracies": 0.6499999761581421,
|
444 |
+
"rewards/chosen": -0.42326441407203674,
|
445 |
+
"rewards/margins": 0.08332939445972443,
|
446 |
+
"rewards/rejected": -0.5065938234329224,
|
447 |
+
"step": 280
|
448 |
+
},
|
449 |
+
{
|
450 |
+
"epoch": 0.23,
|
451 |
+
"learning_rate": 4.738823533671383e-06,
|
452 |
+
"logits/chosen": -3.503385543823242,
|
453 |
+
"logits/rejected": -3.490826368331909,
|
454 |
+
"logps/chosen": -351.4194030761719,
|
455 |
+
"logps/rejected": -368.7189025878906,
|
456 |
+
"loss": 0.6789,
|
457 |
+
"rewards/accuracies": 0.59375,
|
458 |
+
"rewards/chosen": -0.5290869474411011,
|
459 |
+
"rewards/margins": 0.048080917447805405,
|
460 |
+
"rewards/rejected": -0.5771678686141968,
|
461 |
+
"step": 290
|
462 |
+
},
|
463 |
+
{
|
464 |
+
"epoch": 0.24,
|
465 |
+
"learning_rate": 4.706858481968017e-06,
|
466 |
+
"logits/chosen": -3.464003801345825,
|
467 |
+
"logits/rejected": -3.469198226928711,
|
468 |
+
"logps/chosen": -340.75830078125,
|
469 |
+
"logps/rejected": -352.869384765625,
|
470 |
+
"loss": 0.6795,
|
471 |
+
"rewards/accuracies": 0.550000011920929,
|
472 |
+
"rewards/chosen": -0.6002467274665833,
|
473 |
+
"rewards/margins": 0.04778647795319557,
|
474 |
+
"rewards/rejected": -0.6480332612991333,
|
475 |
+
"step": 300
|
476 |
+
},
|
477 |
+
{
|
478 |
+
"epoch": 0.24,
|
479 |
+
"eval_logits/chosen": -3.5198659896850586,
|
480 |
+
"eval_logits/rejected": -3.5140511989593506,
|
481 |
+
"eval_logps/chosen": -272.0929260253906,
|
482 |
+
"eval_logps/rejected": -281.4856262207031,
|
483 |
+
"eval_loss": 0.683901846408844,
|
484 |
+
"eval_rewards/accuracies": 0.5460000038146973,
|
485 |
+
"eval_rewards/chosen": -0.5138051509857178,
|
486 |
+
"eval_rewards/margins": 0.037760715931653976,
|
487 |
+
"eval_rewards/rejected": -0.551565945148468,
|
488 |
+
"eval_runtime": 1688.0776,
|
489 |
+
"eval_samples_per_second": 1.774,
|
490 |
+
"eval_steps_per_second": 0.222,
|
491 |
+
"step": 300
|
492 |
+
},
|
493 |
+
{
|
494 |
+
"epoch": 0.25,
|
495 |
+
"learning_rate": 4.673169525245416e-06,
|
496 |
+
"logits/chosen": -3.4468257427215576,
|
497 |
+
"logits/rejected": -3.422842502593994,
|
498 |
+
"logps/chosen": -337.1869812011719,
|
499 |
+
"logps/rejected": -369.9530029296875,
|
500 |
+
"loss": 0.6606,
|
501 |
+
"rewards/accuracies": 0.6000000238418579,
|
502 |
+
"rewards/chosen": -0.5638504028320312,
|
503 |
+
"rewards/margins": 0.09270543605089188,
|
504 |
+
"rewards/rejected": -0.6565557718276978,
|
505 |
+
"step": 310
|
506 |
+
},
|
507 |
+
{
|
508 |
+
"epoch": 0.26,
|
509 |
+
"learning_rate": 4.63778297989952e-06,
|
510 |
+
"logits/chosen": -3.598461866378784,
|
511 |
+
"logits/rejected": -3.5812110900878906,
|
512 |
+
"logps/chosen": -344.33819580078125,
|
513 |
+
"logps/rejected": -364.0965270996094,
|
514 |
+
"loss": 0.6658,
|
515 |
+
"rewards/accuracies": 0.612500011920929,
|
516 |
+
"rewards/chosen": -0.5991551280021667,
|
517 |
+
"rewards/margins": 0.08249818533658981,
|
518 |
+
"rewards/rejected": -0.6816532015800476,
|
519 |
+
"step": 320
|
520 |
+
},
|
521 |
+
{
|
522 |
+
"epoch": 0.26,
|
523 |
+
"learning_rate": 4.60072648841109e-06,
|
524 |
+
"logits/chosen": -3.7547969818115234,
|
525 |
+
"logits/rejected": -3.74609637260437,
|
526 |
+
"logps/chosen": -364.9189758300781,
|
527 |
+
"logps/rejected": -394.4504089355469,
|
528 |
+
"loss": 0.6271,
|
529 |
+
"rewards/accuracies": 0.699999988079071,
|
530 |
+
"rewards/chosen": -0.7271077632904053,
|
531 |
+
"rewards/margins": 0.17919641733169556,
|
532 |
+
"rewards/rejected": -0.9063041806221008,
|
533 |
+
"step": 330
|
534 |
+
},
|
535 |
+
{
|
536 |
+
"epoch": 0.27,
|
537 |
+
"learning_rate": 4.562028997752574e-06,
|
538 |
+
"logits/chosen": -3.9404075145721436,
|
539 |
+
"logits/rejected": -3.9303627014160156,
|
540 |
+
"logps/chosen": -380.1604919433594,
|
541 |
+
"logps/rejected": -408.8302307128906,
|
542 |
+
"loss": 0.676,
|
543 |
+
"rewards/accuracies": 0.5874999761581421,
|
544 |
+
"rewards/chosen": -0.8844378590583801,
|
545 |
+
"rewards/margins": 0.08313676714897156,
|
546 |
+
"rewards/rejected": -0.9675747156143188,
|
547 |
+
"step": 340
|
548 |
+
},
|
549 |
+
{
|
550 |
+
"epoch": 0.28,
|
551 |
+
"learning_rate": 4.521720736775947e-06,
|
552 |
+
"logits/chosen": -3.974989414215088,
|
553 |
+
"logits/rejected": -3.998753786087036,
|
554 |
+
"logps/chosen": -397.6668701171875,
|
555 |
+
"logps/rejected": -411.4002990722656,
|
556 |
+
"loss": 0.6559,
|
557 |
+
"rewards/accuracies": 0.625,
|
558 |
+
"rewards/chosen": -1.0197855234146118,
|
559 |
+
"rewards/margins": 0.12257959693670273,
|
560 |
+
"rewards/rejected": -1.1423652172088623,
|
561 |
+
"step": 350
|
562 |
+
},
|
563 |
+
{
|
564 |
+
"epoch": 0.29,
|
565 |
+
"learning_rate": 4.479833192599198e-06,
|
566 |
+
"logits/chosen": -3.942905902862549,
|
567 |
+
"logits/rejected": -3.9193332195281982,
|
568 |
+
"logps/chosen": -387.99420166015625,
|
569 |
+
"logps/rejected": -410.8706970214844,
|
570 |
+
"loss": 0.6613,
|
571 |
+
"rewards/accuracies": 0.581250011920929,
|
572 |
+
"rewards/chosen": -0.8740938305854797,
|
573 |
+
"rewards/margins": 0.10989212989807129,
|
574 |
+
"rewards/rejected": -0.9839859008789062,
|
575 |
+
"step": 360
|
576 |
+
},
|
577 |
+
{
|
578 |
+
"epoch": 0.3,
|
579 |
+
"learning_rate": 4.436399086009928e-06,
|
580 |
+
"logits/chosen": -3.781745195388794,
|
581 |
+
"logits/rejected": -3.746504306793213,
|
582 |
+
"logps/chosen": -363.17657470703125,
|
583 |
+
"logps/rejected": -384.13323974609375,
|
584 |
+
"loss": 0.6487,
|
585 |
+
"rewards/accuracies": 0.6625000238418579,
|
586 |
+
"rewards/chosen": -0.7560319304466248,
|
587 |
+
"rewards/margins": 0.12066509574651718,
|
588 |
+
"rewards/rejected": -0.8766969442367554,
|
589 |
+
"step": 370
|
590 |
+
},
|
591 |
+
{
|
592 |
+
"epoch": 0.3,
|
593 |
+
"learning_rate": 4.391452345905239e-06,
|
594 |
+
"logits/chosen": -3.672318696975708,
|
595 |
+
"logits/rejected": -3.6834559440612793,
|
596 |
+
"logps/chosen": -373.22344970703125,
|
597 |
+
"logps/rejected": -389.52777099609375,
|
598 |
+
"loss": 0.6611,
|
599 |
+
"rewards/accuracies": 0.5625,
|
600 |
+
"rewards/chosen": -0.8263392448425293,
|
601 |
+
"rewards/margins": 0.11394073814153671,
|
602 |
+
"rewards/rejected": -0.9402799606323242,
|
603 |
+
"step": 380
|
604 |
+
},
|
605 |
+
{
|
606 |
+
"epoch": 0.31,
|
607 |
+
"learning_rate": 4.3450280827879125e-06,
|
608 |
+
"logits/chosen": -3.7310726642608643,
|
609 |
+
"logits/rejected": -3.7564334869384766,
|
610 |
+
"logps/chosen": -374.36260986328125,
|
611 |
+
"logps/rejected": -394.80010986328125,
|
612 |
+
"loss": 0.6607,
|
613 |
+
"rewards/accuracies": 0.6187499761581421,
|
614 |
+
"rewards/chosen": -0.8552727699279785,
|
615 |
+
"rewards/margins": 0.11295287311077118,
|
616 |
+
"rewards/rejected": -0.9682257771492004,
|
617 |
+
"step": 390
|
618 |
+
},
|
619 |
+
{
|
620 |
+
"epoch": 0.32,
|
621 |
+
"learning_rate": 4.297162561339554e-06,
|
622 |
+
"logits/chosen": -3.6382896900177,
|
623 |
+
"logits/rejected": -3.6042380332946777,
|
624 |
+
"logps/chosen": -380.48590087890625,
|
625 |
+
"logps/rejected": -407.5166015625,
|
626 |
+
"loss": 0.6561,
|
627 |
+
"rewards/accuracies": 0.6187499761581421,
|
628 |
+
"rewards/chosen": -0.8749423027038574,
|
629 |
+
"rewards/margins": 0.1411461979150772,
|
630 |
+
"rewards/rejected": -1.016088604927063,
|
631 |
+
"step": 400
|
632 |
+
},
|
633 |
+
{
|
634 |
+
"epoch": 0.32,
|
635 |
+
"eval_logits/chosen": -3.7580296993255615,
|
636 |
+
"eval_logits/rejected": -3.7484097480773926,
|
637 |
+
"eval_logps/chosen": -302.29541015625,
|
638 |
+
"eval_logps/rejected": -314.21051025390625,
|
639 |
+
"eval_loss": 0.6812021136283875,
|
640 |
+
"eval_rewards/accuracies": 0.5573333501815796,
|
641 |
+
"eval_rewards/chosen": -0.815830409526825,
|
642 |
+
"eval_rewards/margins": 0.06298430263996124,
|
643 |
+
"eval_rewards/rejected": -0.8788146376609802,
|
644 |
+
"eval_runtime": 1618.6439,
|
645 |
+
"eval_samples_per_second": 1.85,
|
646 |
+
"eval_steps_per_second": 0.232,
|
647 |
+
"step": 400
|
648 |
+
},
|
649 |
+
{
|
650 |
+
"epoch": 0.33,
|
651 |
+
"learning_rate": 4.247893172092157e-06,
|
652 |
+
"logits/chosen": -3.615405559539795,
|
653 |
+
"logits/rejected": -3.612015962600708,
|
654 |
+
"logps/chosen": -370.3289794921875,
|
655 |
+
"logps/rejected": -405.1226806640625,
|
656 |
+
"loss": 0.6581,
|
657 |
+
"rewards/accuracies": 0.6312500238418579,
|
658 |
+
"rewards/chosen": -0.9003459215164185,
|
659 |
+
"rewards/margins": 0.13107234239578247,
|
660 |
+
"rewards/rejected": -1.0314182043075562,
|
661 |
+
"step": 410
|
662 |
+
},
|
663 |
+
{
|
664 |
+
"epoch": 0.34,
|
665 |
+
"learning_rate": 4.197258402220187e-06,
|
666 |
+
"logits/chosen": -3.65478515625,
|
667 |
+
"logits/rejected": -3.6666762828826904,
|
668 |
+
"logps/chosen": -379.541748046875,
|
669 |
+
"logps/rejected": -418.70880126953125,
|
670 |
+
"loss": 0.657,
|
671 |
+
"rewards/accuracies": 0.6000000238418579,
|
672 |
+
"rewards/chosen": -0.9110058546066284,
|
673 |
+
"rewards/margins": 0.3442539870738983,
|
674 |
+
"rewards/rejected": -1.2552598714828491,
|
675 |
+
"step": 420
|
676 |
+
},
|
677 |
+
{
|
678 |
+
"epoch": 0.34,
|
679 |
+
"learning_rate": 4.145297805476023e-06,
|
680 |
+
"logits/chosen": -3.6817328929901123,
|
681 |
+
"logits/rejected": -3.688814640045166,
|
682 |
+
"logps/chosen": -376.09783935546875,
|
683 |
+
"logps/rejected": -406.4654846191406,
|
684 |
+
"loss": 0.6439,
|
685 |
+
"rewards/accuracies": 0.637499988079071,
|
686 |
+
"rewards/chosen": -0.8991987109184265,
|
687 |
+
"rewards/margins": 0.1862173080444336,
|
688 |
+
"rewards/rejected": -1.0854160785675049,
|
689 |
+
"step": 430
|
690 |
+
},
|
691 |
+
{
|
692 |
+
"epoch": 0.35,
|
693 |
+
"learning_rate": 4.092051971292228e-06,
|
694 |
+
"logits/chosen": -3.716754198074341,
|
695 |
+
"logits/rejected": -3.7106194496154785,
|
696 |
+
"logps/chosen": -376.1932067871094,
|
697 |
+
"logps/rejected": -401.6561584472656,
|
698 |
+
"loss": 0.6554,
|
699 |
+
"rewards/accuracies": 0.643750011920929,
|
700 |
+
"rewards/chosen": -0.954433798789978,
|
701 |
+
"rewards/margins": 0.13505356013774872,
|
702 |
+
"rewards/rejected": -1.0894873142242432,
|
703 |
+
"step": 440
|
704 |
+
},
|
705 |
+
{
|
706 |
+
"epoch": 0.36,
|
707 |
+
"learning_rate": 4.037562493074792e-06,
|
708 |
+
"logits/chosen": -3.8129425048828125,
|
709 |
+
"logits/rejected": -3.833683729171753,
|
710 |
+
"logps/chosen": -403.4329528808594,
|
711 |
+
"logps/rejected": -423.4547424316406,
|
712 |
+
"loss": 0.6365,
|
713 |
+
"rewards/accuracies": 0.6812499761581421,
|
714 |
+
"rewards/chosen": -1.0463995933532715,
|
715 |
+
"rewards/margins": 0.19165393710136414,
|
716 |
+
"rewards/rejected": -1.2380534410476685,
|
717 |
+
"step": 450
|
718 |
+
},
|
719 |
+
{
|
720 |
+
"epoch": 0.37,
|
721 |
+
"learning_rate": 3.981871935712112e-06,
|
722 |
+
"logits/chosen": -3.983973741531372,
|
723 |
+
"logits/rejected": -3.940070629119873,
|
724 |
+
"logps/chosen": -385.3892517089844,
|
725 |
+
"logps/rejected": -416.1813049316406,
|
726 |
+
"loss": 0.6533,
|
727 |
+
"rewards/accuracies": 0.5687500238418579,
|
728 |
+
"rewards/chosen": -1.0447807312011719,
|
729 |
+
"rewards/margins": 0.14543746411800385,
|
730 |
+
"rewards/rejected": -1.190218210220337,
|
731 |
+
"step": 460
|
732 |
+
},
|
733 |
+
{
|
734 |
+
"epoch": 0.38,
|
735 |
+
"learning_rate": 3.925023802325094e-06,
|
736 |
+
"logits/chosen": -4.042995929718018,
|
737 |
+
"logits/rejected": -4.0136399269104,
|
738 |
+
"logps/chosen": -401.97381591796875,
|
739 |
+
"logps/rejected": -437.3184509277344,
|
740 |
+
"loss": 0.6455,
|
741 |
+
"rewards/accuracies": 0.6187499761581421,
|
742 |
+
"rewards/chosen": -1.0883817672729492,
|
743 |
+
"rewards/margins": 0.2549799978733063,
|
744 |
+
"rewards/rejected": -1.343361735343933,
|
745 |
+
"step": 470
|
746 |
+
},
|
747 |
+
{
|
748 |
+
"epoch": 0.38,
|
749 |
+
"learning_rate": 3.867062500284342e-06,
|
750 |
+
"logits/chosen": -4.073556423187256,
|
751 |
+
"logits/rejected": -4.043025493621826,
|
752 |
+
"logps/chosen": -388.99090576171875,
|
753 |
+
"logps/rejected": -425.56787109375,
|
754 |
+
"loss": 0.6433,
|
755 |
+
"rewards/accuracies": 0.612500011920929,
|
756 |
+
"rewards/chosen": -1.1277819871902466,
|
757 |
+
"rewards/margins": 0.1785212755203247,
|
758 |
+
"rewards/rejected": -1.3063032627105713,
|
759 |
+
"step": 480
|
760 |
+
},
|
761 |
+
{
|
762 |
+
"epoch": 0.39,
|
763 |
+
"learning_rate": 3.8080333065209885e-06,
|
764 |
+
"logits/chosen": -4.076624393463135,
|
765 |
+
"logits/rejected": -4.084932327270508,
|
766 |
+
"logps/chosen": -391.6429138183594,
|
767 |
+
"logps/rejected": -392.52362060546875,
|
768 |
+
"loss": 0.7064,
|
769 |
+
"rewards/accuracies": 0.53125,
|
770 |
+
"rewards/chosen": -1.1167463064193726,
|
771 |
+
"rewards/margins": 0.06357622146606445,
|
772 |
+
"rewards/rejected": -1.180322527885437,
|
773 |
+
"step": 490
|
774 |
+
},
|
775 |
+
{
|
776 |
+
"epoch": 0.4,
|
777 |
+
"learning_rate": 3.7479823321582624e-06,
|
778 |
+
"logits/chosen": -3.93993878364563,
|
779 |
+
"logits/rejected": -3.9026870727539062,
|
780 |
+
"logps/chosen": -378.84918212890625,
|
781 |
+
"logps/rejected": -416.9828186035156,
|
782 |
+
"loss": 0.633,
|
783 |
+
"rewards/accuracies": 0.6187499761581421,
|
784 |
+
"rewards/chosen": -0.9214572906494141,
|
785 |
+
"rewards/margins": 0.19020086526870728,
|
786 |
+
"rewards/rejected": -1.1116580963134766,
|
787 |
+
"step": 500
|
788 |
+
},
|
789 |
+
{
|
790 |
+
"epoch": 0.4,
|
791 |
+
"eval_logits/chosen": -4.107741355895996,
|
792 |
+
"eval_logits/rejected": -4.097755432128906,
|
793 |
+
"eval_logps/chosen": -310.98577880859375,
|
794 |
+
"eval_logps/rejected": -324.42694091796875,
|
795 |
+
"eval_loss": 0.678744912147522,
|
796 |
+
"eval_rewards/accuracies": 0.5596666932106018,
|
797 |
+
"eval_rewards/chosen": -0.9027342796325684,
|
798 |
+
"eval_rewards/margins": 0.0782446414232254,
|
799 |
+
"eval_rewards/rejected": -0.9809789061546326,
|
800 |
+
"eval_runtime": 1619.4632,
|
801 |
+
"eval_samples_per_second": 1.849,
|
802 |
+
"eval_steps_per_second": 0.232,
|
803 |
+
"step": 500
|
804 |
+
},
|
805 |
+
{
|
806 |
+
"epoch": 0.41,
|
807 |
+
"learning_rate": 3.686956486491419e-06,
|
808 |
+
"logits/chosen": -3.9462807178497314,
|
809 |
+
"logits/rejected": -3.941249132156372,
|
810 |
+
"logps/chosen": -386.48590087890625,
|
811 |
+
"logps/rejected": -424.64508056640625,
|
812 |
+
"loss": 0.6313,
|
813 |
+
"rewards/accuracies": 0.637499988079071,
|
814 |
+
"rewards/chosen": -0.9227803349494934,
|
815 |
+
"rewards/margins": 0.29515519738197327,
|
816 |
+
"rewards/rejected": -1.217935562133789,
|
817 |
+
"step": 510
|
818 |
+
},
|
819 |
+
{
|
820 |
+
"epoch": 0.42,
|
821 |
+
"learning_rate": 3.625003440344166e-06,
|
822 |
+
"logits/chosen": -4.044493198394775,
|
823 |
+
"logits/rejected": -4.073317527770996,
|
824 |
+
"logps/chosen": -369.5896301269531,
|
825 |
+
"logps/rejected": -382.105224609375,
|
826 |
+
"loss": 0.664,
|
827 |
+
"rewards/accuracies": 0.6000000238418579,
|
828 |
+
"rewards/chosen": -0.9153301119804382,
|
829 |
+
"rewards/margins": 0.08651997148990631,
|
830 |
+
"rewards/rejected": -1.0018501281738281,
|
831 |
+
"step": 520
|
832 |
+
},
|
833 |
+
{
|
834 |
+
"epoch": 0.42,
|
835 |
+
"learning_rate": 3.562171588830231e-06,
|
836 |
+
"logits/chosen": -4.016448497772217,
|
837 |
+
"logits/rejected": -3.9958884716033936,
|
838 |
+
"logps/chosen": -377.91912841796875,
|
839 |
+
"logps/rejected": -404.9534606933594,
|
840 |
+
"loss": 0.6943,
|
841 |
+
"rewards/accuracies": 0.5625,
|
842 |
+
"rewards/chosen": -0.8825477361679077,
|
843 |
+
"rewards/margins": 0.0769728347659111,
|
844 |
+
"rewards/rejected": -0.9595205187797546,
|
845 |
+
"step": 530
|
846 |
+
},
|
847 |
+
{
|
848 |
+
"epoch": 0.43,
|
849 |
+
"learning_rate": 3.4985100135491245e-06,
|
850 |
+
"logits/chosen": -4.008540630340576,
|
851 |
+
"logits/rejected": -3.9679737091064453,
|
852 |
+
"logps/chosen": -382.65924072265625,
|
853 |
+
"logps/rejected": -425.71954345703125,
|
854 |
+
"loss": 0.629,
|
855 |
+
"rewards/accuracies": 0.6499999761581421,
|
856 |
+
"rewards/chosen": -0.8818261027336121,
|
857 |
+
"rewards/margins": 0.2214728146791458,
|
858 |
+
"rewards/rejected": -1.1032989025115967,
|
859 |
+
"step": 540
|
860 |
+
},
|
861 |
+
{
|
862 |
+
"epoch": 0.44,
|
863 |
+
"learning_rate": 3.4340684442456673e-06,
|
864 |
+
"logits/chosen": -4.049837589263916,
|
865 |
+
"logits/rejected": -4.043179035186768,
|
866 |
+
"logps/chosen": -384.42938232421875,
|
867 |
+
"logps/rejected": -410.5345153808594,
|
868 |
+
"loss": 0.6434,
|
869 |
+
"rewards/accuracies": 0.6187499761581421,
|
870 |
+
"rewards/chosen": -0.9648186564445496,
|
871 |
+
"rewards/margins": 0.1496874988079071,
|
872 |
+
"rewards/rejected": -1.1145063638687134,
|
873 |
+
"step": 550
|
874 |
+
},
|
875 |
+
{
|
876 |
+
"epoch": 0.45,
|
877 |
+
"learning_rate": 3.3688972199631974e-06,
|
878 |
+
"logits/chosen": -4.042217254638672,
|
879 |
+
"logits/rejected": -4.064842224121094,
|
880 |
+
"logps/chosen": -387.6910705566406,
|
881 |
+
"logps/rejected": -413.78662109375,
|
882 |
+
"loss": 0.6329,
|
883 |
+
"rewards/accuracies": 0.6812499761581421,
|
884 |
+
"rewards/chosen": -0.9935464859008789,
|
885 |
+
"rewards/margins": 0.24879872798919678,
|
886 |
+
"rewards/rejected": -1.2423454523086548,
|
887 |
+
"step": 560
|
888 |
+
},
|
889 |
+
{
|
890 |
+
"epoch": 0.46,
|
891 |
+
"learning_rate": 3.3030472497208354e-06,
|
892 |
+
"logits/chosen": -4.0646257400512695,
|
893 |
+
"logits/rejected": -4.018919944763184,
|
894 |
+
"logps/chosen": -385.509765625,
|
895 |
+
"logps/rejected": -450.21429443359375,
|
896 |
+
"loss": 0.6219,
|
897 |
+
"rewards/accuracies": 0.625,
|
898 |
+
"rewards/chosen": -1.0814400911331177,
|
899 |
+
"rewards/margins": 0.26284489035606384,
|
900 |
+
"rewards/rejected": -1.3442847728729248,
|
901 |
+
"step": 570
|
902 |
+
},
|
903 |
+
{
|
904 |
+
"epoch": 0.46,
|
905 |
+
"learning_rate": 3.236569972745492e-06,
|
906 |
+
"logits/chosen": -4.145129203796387,
|
907 |
+
"logits/rejected": -4.1337690353393555,
|
908 |
+
"logps/chosen": -380.5148620605469,
|
909 |
+
"logps/rejected": -406.1288146972656,
|
910 |
+
"loss": 0.6628,
|
911 |
+
"rewards/accuracies": 0.65625,
|
912 |
+
"rewards/chosen": -1.0649586915969849,
|
913 |
+
"rewards/margins": 0.15909543633460999,
|
914 |
+
"rewards/rejected": -1.2240540981292725,
|
915 |
+
"step": 580
|
916 |
+
},
|
917 |
+
{
|
918 |
+
"epoch": 0.47,
|
919 |
+
"learning_rate": 3.1695173182897126e-06,
|
920 |
+
"logits/chosen": -4.133418083190918,
|
921 |
+
"logits/rejected": -4.11216402053833,
|
922 |
+
"logps/chosen": -398.23626708984375,
|
923 |
+
"logps/rejected": -435.56646728515625,
|
924 |
+
"loss": 0.6554,
|
925 |
+
"rewards/accuracies": 0.6312500238418579,
|
926 |
+
"rewards/chosen": -1.1820625066757202,
|
927 |
+
"rewards/margins": 0.16709741950035095,
|
928 |
+
"rewards/rejected": -1.349160075187683,
|
929 |
+
"step": 590
|
930 |
+
},
|
931 |
+
{
|
932 |
+
"epoch": 0.48,
|
933 |
+
"learning_rate": 3.10194166506673e-06,
|
934 |
+
"logits/chosen": -4.175902366638184,
|
935 |
+
"logits/rejected": -4.1359357833862305,
|
936 |
+
"logps/chosen": -390.8631286621094,
|
937 |
+
"logps/rejected": -439.28436279296875,
|
938 |
+
"loss": 0.6302,
|
939 |
+
"rewards/accuracies": 0.6312500238418579,
|
940 |
+
"rewards/chosen": -1.061226487159729,
|
941 |
+
"rewards/margins": 0.2546694874763489,
|
942 |
+
"rewards/rejected": -1.3158957958221436,
|
943 |
+
"step": 600
|
944 |
+
},
|
945 |
+
{
|
946 |
+
"epoch": 0.48,
|
947 |
+
"eval_logits/chosen": -4.443523406982422,
|
948 |
+
"eval_logits/rejected": -4.431849002838135,
|
949 |
+
"eval_logps/chosen": -337.6354675292969,
|
950 |
+
"eval_logps/rejected": -353.24932861328125,
|
951 |
+
"eval_loss": 0.6784851551055908,
|
952 |
+
"eval_rewards/accuracies": 0.5596666932106018,
|
953 |
+
"eval_rewards/chosen": -1.1692306995391846,
|
954 |
+
"eval_rewards/margins": 0.09997232258319855,
|
955 |
+
"eval_rewards/rejected": -1.2692030668258667,
|
956 |
+
"eval_runtime": 1619.3984,
|
957 |
+
"eval_samples_per_second": 1.849,
|
958 |
+
"eval_steps_per_second": 0.232,
|
959 |
+
"step": 600
|
960 |
+
},
|
961 |
+
{
|
962 |
+
"epoch": 0.49,
|
963 |
+
"learning_rate": 3.0338958003344115e-06,
|
964 |
+
"logits/chosen": -4.3245649337768555,
|
965 |
+
"logits/rejected": -4.272718906402588,
|
966 |
+
"logps/chosen": -396.521240234375,
|
967 |
+
"logps/rejected": -432.1895446777344,
|
968 |
+
"loss": 0.6883,
|
969 |
+
"rewards/accuracies": 0.675000011920929,
|
970 |
+
"rewards/chosen": -1.2242915630340576,
|
971 |
+
"rewards/margins": 0.24001073837280273,
|
972 |
+
"rewards/rejected": -1.4643023014068604,
|
973 |
+
"step": 610
|
974 |
+
},
|
975 |
+
{
|
976 |
+
"epoch": 0.5,
|
977 |
+
"learning_rate": 2.9654328786600823e-06,
|
978 |
+
"logits/chosen": -4.306203365325928,
|
979 |
+
"logits/rejected": -4.252989768981934,
|
980 |
+
"logps/chosen": -397.3540344238281,
|
981 |
+
"logps/rejected": -442.0118103027344,
|
982 |
+
"loss": 0.6197,
|
983 |
+
"rewards/accuracies": 0.6937500238418579,
|
984 |
+
"rewards/chosen": -1.2108550071716309,
|
985 |
+
"rewards/margins": 0.21409356594085693,
|
986 |
+
"rewards/rejected": -1.4249485731124878,
|
987 |
+
"step": 620
|
988 |
+
},
|
989 |
+
{
|
990 |
+
"epoch": 0.5,
|
991 |
+
"learning_rate": 2.896606380398402e-06,
|
992 |
+
"logits/chosen": -4.365767478942871,
|
993 |
+
"logits/rejected": -4.406495094299316,
|
994 |
+
"logps/chosen": -417.7538146972656,
|
995 |
+
"logps/rejected": -443.05548095703125,
|
996 |
+
"loss": 0.6623,
|
997 |
+
"rewards/accuracies": 0.643750011920929,
|
998 |
+
"rewards/chosen": -1.3476839065551758,
|
999 |
+
"rewards/margins": 0.18296115100383759,
|
1000 |
+
"rewards/rejected": -1.5306451320648193,
|
1001 |
+
"step": 630
|
1002 |
+
},
|
1003 |
+
{
|
1004 |
+
"epoch": 0.51,
|
1005 |
+
"learning_rate": 2.827470069914772e-06,
|
1006 |
+
"logits/chosen": -4.2744035720825195,
|
1007 |
+
"logits/rejected": -4.236593723297119,
|
1008 |
+
"logps/chosen": -425.37908935546875,
|
1009 |
+
"logps/rejected": -453.02862548828125,
|
1010 |
+
"loss": 0.6756,
|
1011 |
+
"rewards/accuracies": 0.574999988079071,
|
1012 |
+
"rewards/chosen": -1.3940837383270264,
|
1013 |
+
"rewards/margins": 0.14192768931388855,
|
1014 |
+
"rewards/rejected": -1.5360115766525269,
|
1015 |
+
"step": 640
|
1016 |
+
},
|
1017 |
+
{
|
1018 |
+
"epoch": 0.52,
|
1019 |
+
"learning_rate": 2.7580779535868675e-06,
|
1020 |
+
"logits/chosen": -4.252664089202881,
|
1021 |
+
"logits/rejected": -4.254392147064209,
|
1022 |
+
"logps/chosen": -409.19378662109375,
|
1023 |
+
"logps/rejected": -438.70556640625,
|
1024 |
+
"loss": 0.6575,
|
1025 |
+
"rewards/accuracies": 0.675000011920929,
|
1026 |
+
"rewards/chosen": -1.230445384979248,
|
1027 |
+
"rewards/margins": 0.18368306756019592,
|
1028 |
+
"rewards/rejected": -1.4141284227371216,
|
1029 |
+
"step": 650
|
1030 |
+
},
|
1031 |
+
{
|
1032 |
+
"epoch": 0.53,
|
1033 |
+
"learning_rate": 2.688484237617129e-06,
|
1034 |
+
"logits/chosen": -4.151357650756836,
|
1035 |
+
"logits/rejected": -4.122767925262451,
|
1036 |
+
"logps/chosen": -400.3338928222656,
|
1037 |
+
"logps/rejected": -436.16571044921875,
|
1038 |
+
"loss": 0.6222,
|
1039 |
+
"rewards/accuracies": 0.65625,
|
1040 |
+
"rewards/chosen": -1.108682632446289,
|
1041 |
+
"rewards/margins": 0.24583642184734344,
|
1042 |
+
"rewards/rejected": -1.3545191287994385,
|
1043 |
+
"step": 660
|
1044 |
+
},
|
1045 |
+
{
|
1046 |
+
"epoch": 0.54,
|
1047 |
+
"learning_rate": 2.6187432856891585e-06,
|
1048 |
+
"logits/chosen": -4.1051225662231445,
|
1049 |
+
"logits/rejected": -4.0978288650512695,
|
1050 |
+
"logps/chosen": -407.8655700683594,
|
1051 |
+
"logps/rejected": -446.1006774902344,
|
1052 |
+
"loss": 0.6432,
|
1053 |
+
"rewards/accuracies": 0.574999988079071,
|
1054 |
+
"rewards/chosen": -1.1575034856796265,
|
1055 |
+
"rewards/margins": 0.18804897367954254,
|
1056 |
+
"rewards/rejected": -1.3455523252487183,
|
1057 |
+
"step": 670
|
1058 |
+
},
|
1059 |
+
{
|
1060 |
+
"epoch": 0.54,
|
1061 |
+
"learning_rate": 2.548909576501096e-06,
|
1062 |
+
"logits/chosen": -4.150703430175781,
|
1063 |
+
"logits/rejected": -4.143389701843262,
|
1064 |
+
"logps/chosen": -413.3582458496094,
|
1065 |
+
"logps/rejected": -443.1758728027344,
|
1066 |
+
"loss": 0.6353,
|
1067 |
+
"rewards/accuracies": 0.581250011920929,
|
1068 |
+
"rewards/chosen": -1.2182948589324951,
|
1069 |
+
"rewards/margins": 0.19429844617843628,
|
1070 |
+
"rewards/rejected": -1.4125932455062866,
|
1071 |
+
"step": 680
|
1072 |
+
},
|
1073 |
+
{
|
1074 |
+
"epoch": 0.55,
|
1075 |
+
"learning_rate": 2.4790376612091503e-06,
|
1076 |
+
"logits/chosen": -4.271695613861084,
|
1077 |
+
"logits/rejected": -4.229399681091309,
|
1078 |
+
"logps/chosen": -443.7752990722656,
|
1079 |
+
"logps/rejected": -475.551025390625,
|
1080 |
+
"loss": 0.6236,
|
1081 |
+
"rewards/accuracies": 0.6812499761581421,
|
1082 |
+
"rewards/chosen": -1.456383228302002,
|
1083 |
+
"rewards/margins": 0.23586714267730713,
|
1084 |
+
"rewards/rejected": -1.6922504901885986,
|
1085 |
+
"step": 690
|
1086 |
+
},
|
1087 |
+
{
|
1088 |
+
"epoch": 0.56,
|
1089 |
+
"learning_rate": 2.40918212081453e-06,
|
1090 |
+
"logits/chosen": -4.358768939971924,
|
1091 |
+
"logits/rejected": -4.3066534996032715,
|
1092 |
+
"logps/chosen": -407.3147888183594,
|
1093 |
+
"logps/rejected": -466.919921875,
|
1094 |
+
"loss": 0.5743,
|
1095 |
+
"rewards/accuracies": 0.6937500238418579,
|
1096 |
+
"rewards/chosen": -1.3695669174194336,
|
1097 |
+
"rewards/margins": 0.4060749411582947,
|
1098 |
+
"rewards/rejected": -1.7756417989730835,
|
1099 |
+
"step": 700
|
1100 |
+
},
|
1101 |
+
{
|
1102 |
+
"epoch": 0.56,
|
1103 |
+
"eval_logits/chosen": -4.518208026885986,
|
1104 |
+
"eval_logits/rejected": -4.504719257354736,
|
1105 |
+
"eval_logps/chosen": -375.0574645996094,
|
1106 |
+
"eval_logps/rejected": -392.7273254394531,
|
1107 |
+
"eval_loss": 0.6835331916809082,
|
1108 |
+
"eval_rewards/accuracies": 0.5630000233650208,
|
1109 |
+
"eval_rewards/chosen": -1.5434508323669434,
|
1110 |
+
"eval_rewards/margins": 0.12053229659795761,
|
1111 |
+
"eval_rewards/rejected": -1.6639831066131592,
|
1112 |
+
"eval_runtime": 1619.3262,
|
1113 |
+
"eval_samples_per_second": 1.849,
|
1114 |
+
"eval_steps_per_second": 0.232,
|
1115 |
+
"step": 700
|
1116 |
+
},
|
1117 |
+
{
|
1118 |
+
"epoch": 0.57,
|
1119 |
+
"learning_rate": 2.3393975235270654e-06,
|
1120 |
+
"logits/chosen": -4.303341865539551,
|
1121 |
+
"logits/rejected": -4.286491394042969,
|
1122 |
+
"logps/chosen": -452.05718994140625,
|
1123 |
+
"logps/rejected": -493.17144775390625,
|
1124 |
+
"loss": 0.6602,
|
1125 |
+
"rewards/accuracies": 0.5874999761581421,
|
1126 |
+
"rewards/chosen": -1.5600488185882568,
|
1127 |
+
"rewards/margins": 0.23919770121574402,
|
1128 |
+
"rewards/rejected": -1.7992465496063232,
|
1129 |
+
"step": 710
|
1130 |
+
},
|
1131 |
+
{
|
1132 |
+
"epoch": 0.58,
|
1133 |
+
"learning_rate": 2.2697383821388153e-06,
|
1134 |
+
"logits/chosen": -4.293368816375732,
|
1135 |
+
"logits/rejected": -4.3109025955200195,
|
1136 |
+
"logps/chosen": -435.1492614746094,
|
1137 |
+
"logps/rejected": -460.28033447265625,
|
1138 |
+
"loss": 0.6504,
|
1139 |
+
"rewards/accuracies": 0.5874999761581421,
|
1140 |
+
"rewards/chosen": -1.4816340208053589,
|
1141 |
+
"rewards/margins": 0.2186344861984253,
|
1142 |
+
"rewards/rejected": -1.7002685070037842,
|
1143 |
+
"step": 720
|
1144 |
+
},
|
1145 |
+
{
|
1146 |
+
"epoch": 0.58,
|
1147 |
+
"learning_rate": 2.2002591114409657e-06,
|
1148 |
+
"logits/chosen": -4.212637424468994,
|
1149 |
+
"logits/rejected": -4.208783149719238,
|
1150 |
+
"logps/chosen": -432.06805419921875,
|
1151 |
+
"logps/rejected": -468.85601806640625,
|
1152 |
+
"loss": 0.6325,
|
1153 |
+
"rewards/accuracies": 0.637499988079071,
|
1154 |
+
"rewards/chosen": -1.403395652770996,
|
1155 |
+
"rewards/margins": 0.24927303194999695,
|
1156 |
+
"rewards/rejected": -1.6526685953140259,
|
1157 |
+
"step": 730
|
1158 |
+
},
|
1159 |
+
{
|
1160 |
+
"epoch": 0.59,
|
1161 |
+
"learning_rate": 2.131013985717285e-06,
|
1162 |
+
"logits/chosen": -4.271391868591309,
|
1163 |
+
"logits/rejected": -4.220091819763184,
|
1164 |
+
"logps/chosen": -442.1729431152344,
|
1165 |
+
"logps/rejected": -491.44384765625,
|
1166 |
+
"loss": 0.6362,
|
1167 |
+
"rewards/accuracies": 0.675000011920929,
|
1168 |
+
"rewards/chosen": -1.4394041299819946,
|
1169 |
+
"rewards/margins": 0.2614460587501526,
|
1170 |
+
"rewards/rejected": -1.7008501291275024,
|
1171 |
+
"step": 740
|
1172 |
+
},
|
1173 |
+
{
|
1174 |
+
"epoch": 0.6,
|
1175 |
+
"learning_rate": 2.062057096347338e-06,
|
1176 |
+
"logits/chosen": -4.25800895690918,
|
1177 |
+
"logits/rejected": -4.223499774932861,
|
1178 |
+
"logps/chosen": -419.89495849609375,
|
1179 |
+
"logps/rejected": -435.30389404296875,
|
1180 |
+
"loss": 0.6593,
|
1181 |
+
"rewards/accuracies": 0.643750011920929,
|
1182 |
+
"rewards/chosen": -1.2841438055038452,
|
1183 |
+
"rewards/margins": 0.15097984671592712,
|
1184 |
+
"rewards/rejected": -1.4351234436035156,
|
1185 |
+
"step": 750
|
1186 |
+
},
|
1187 |
+
{
|
1188 |
+
"epoch": 0.61,
|
1189 |
+
"learning_rate": 1.9934423095525733e-06,
|
1190 |
+
"logits/chosen": -4.121432304382324,
|
1191 |
+
"logits/rejected": -4.1321306228637695,
|
1192 |
+
"logps/chosen": -416.6729431152344,
|
1193 |
+
"logps/rejected": -442.0414123535156,
|
1194 |
+
"loss": 0.6534,
|
1195 |
+
"rewards/accuracies": 0.6000000238418579,
|
1196 |
+
"rewards/chosen": -1.1921896934509277,
|
1197 |
+
"rewards/margins": 0.2546849846839905,
|
1198 |
+
"rewards/rejected": -1.4468748569488525,
|
1199 |
+
"step": 760
|
1200 |
+
},
|
1201 |
+
{
|
1202 |
+
"epoch": 0.62,
|
1203 |
+
"learning_rate": 1.9252232243182986e-06,
|
1204 |
+
"logits/chosen": -4.221813678741455,
|
1205 |
+
"logits/rejected": -4.169572830200195,
|
1206 |
+
"logps/chosen": -377.6047668457031,
|
1207 |
+
"logps/rejected": -425.6900939941406,
|
1208 |
+
"loss": 0.6026,
|
1209 |
+
"rewards/accuracies": 0.6875,
|
1210 |
+
"rewards/chosen": -1.071094036102295,
|
1211 |
+
"rewards/margins": 0.307754784822464,
|
1212 |
+
"rewards/rejected": -1.3788487911224365,
|
1213 |
+
"step": 770
|
1214 |
+
},
|
1215 |
+
{
|
1216 |
+
"epoch": 0.62,
|
1217 |
+
"learning_rate": 1.8574531305244043e-06,
|
1218 |
+
"logits/chosen": -4.010577201843262,
|
1219 |
+
"logits/rejected": -3.9968719482421875,
|
1220 |
+
"logps/chosen": -418.9974060058594,
|
1221 |
+
"logps/rejected": -465.41650390625,
|
1222 |
+
"loss": 0.5956,
|
1223 |
+
"rewards/accuracies": 0.6875,
|
1224 |
+
"rewards/chosen": -1.2276278734207153,
|
1225 |
+
"rewards/margins": 0.3166094124317169,
|
1226 |
+
"rewards/rejected": -1.5442373752593994,
|
1227 |
+
"step": 780
|
1228 |
+
},
|
1229 |
+
{
|
1230 |
+
"epoch": 0.63,
|
1231 |
+
"learning_rate": 1.7901849673175559e-06,
|
1232 |
+
"logits/chosen": -4.077489376068115,
|
1233 |
+
"logits/rejected": -4.033568382263184,
|
1234 |
+
"logps/chosen": -420.52984619140625,
|
1235 |
+
"logps/rejected": -455.3899841308594,
|
1236 |
+
"loss": 0.6434,
|
1237 |
+
"rewards/accuracies": 0.6187499761581421,
|
1238 |
+
"rewards/chosen": -1.2396165132522583,
|
1239 |
+
"rewards/margins": 0.20500020682811737,
|
1240 |
+
"rewards/rejected": -1.4446165561676025,
|
1241 |
+
"step": 790
|
1242 |
+
},
|
1243 |
+
{
|
1244 |
+
"epoch": 0.64,
|
1245 |
+
"learning_rate": 1.7234712817573555e-06,
|
1246 |
+
"logits/chosen": -4.063477516174316,
|
1247 |
+
"logits/rejected": -4.062304496765137,
|
1248 |
+
"logps/chosen": -456.8814392089844,
|
1249 |
+
"logps/rejected": -480.8173828125,
|
1250 |
+
"loss": 0.6443,
|
1251 |
+
"rewards/accuracies": 0.65625,
|
1252 |
+
"rewards/chosen": -1.4422929286956787,
|
1253 |
+
"rewards/margins": 0.22365431487560272,
|
1254 |
+
"rewards/rejected": -1.6659473180770874,
|
1255 |
+
"step": 800
|
1256 |
+
},
|
1257 |
+
{
|
1258 |
+
"epoch": 0.64,
|
1259 |
+
"eval_logits/chosen": -4.257167816162109,
|
1260 |
+
"eval_logits/rejected": -4.245348930358887,
|
1261 |
+
"eval_logps/chosen": -359.3107604980469,
|
1262 |
+
"eval_logps/rejected": -377.020751953125,
|
1263 |
+
"eval_loss": 0.6778839230537415,
|
1264 |
+
"eval_rewards/accuracies": 0.5666666626930237,
|
1265 |
+
"eval_rewards/chosen": -1.385983943939209,
|
1266 |
+
"eval_rewards/margins": 0.12093351036310196,
|
1267 |
+
"eval_rewards/rejected": -1.5069174766540527,
|
1268 |
+
"eval_runtime": 1620.4754,
|
1269 |
+
"eval_samples_per_second": 1.848,
|
1270 |
+
"eval_steps_per_second": 0.231,
|
1271 |
+
"step": 800
|
1272 |
+
},
|
1273 |
+
{
|
1274 |
+
"epoch": 0.65,
|
1275 |
+
"learning_rate": 1.6573641877687936e-06,
|
1276 |
+
"logits/chosen": -4.076521873474121,
|
1277 |
+
"logits/rejected": -4.057218074798584,
|
1278 |
+
"logps/chosen": -422.85699462890625,
|
1279 |
+
"logps/rejected": -470.95465087890625,
|
1280 |
+
"loss": 0.6161,
|
1281 |
+
"rewards/accuracies": 0.6499999761581421,
|
1282 |
+
"rewards/chosen": -1.3746501207351685,
|
1283 |
+
"rewards/margins": 0.2610599100589752,
|
1284 |
+
"rewards/rejected": -1.6357100009918213,
|
1285 |
+
"step": 810
|
1286 |
+
},
|
1287 |
+
{
|
1288 |
+
"epoch": 0.66,
|
1289 |
+
"learning_rate": 1.591915325433034e-06,
|
1290 |
+
"logits/chosen": -4.133788108825684,
|
1291 |
+
"logits/rejected": -4.142486572265625,
|
1292 |
+
"logps/chosen": -414.0416564941406,
|
1293 |
+
"logps/rejected": -449.82373046875,
|
1294 |
+
"loss": 0.6194,
|
1295 |
+
"rewards/accuracies": 0.6499999761581421,
|
1296 |
+
"rewards/chosen": -1.315294861793518,
|
1297 |
+
"rewards/margins": 0.30571961402893066,
|
1298 |
+
"rewards/rejected": -1.6210145950317383,
|
1299 |
+
"step": 820
|
1300 |
+
},
|
1301 |
+
{
|
1302 |
+
"epoch": 0.66,
|
1303 |
+
"learning_rate": 1.5271758206483664e-06,
|
1304 |
+
"logits/chosen": -4.143270015716553,
|
1305 |
+
"logits/rejected": -4.132315158843994,
|
1306 |
+
"logps/chosen": -438.13037109375,
|
1307 |
+
"logps/rejected": -471.50030517578125,
|
1308 |
+
"loss": 0.6481,
|
1309 |
+
"rewards/accuracies": 0.6937500238418579,
|
1310 |
+
"rewards/chosen": -1.4561512470245361,
|
1311 |
+
"rewards/margins": 0.22672787308692932,
|
1312 |
+
"rewards/rejected": -1.682879090309143,
|
1313 |
+
"step": 830
|
1314 |
+
},
|
1315 |
+
{
|
1316 |
+
"epoch": 0.67,
|
1317 |
+
"learning_rate": 1.4631962451927966e-06,
|
1318 |
+
"logits/chosen": -4.0487775802612305,
|
1319 |
+
"logits/rejected": -4.032698631286621,
|
1320 |
+
"logps/chosen": -431.3470153808594,
|
1321 |
+
"logps/rejected": -473.2076110839844,
|
1322 |
+
"loss": 0.6076,
|
1323 |
+
"rewards/accuracies": 0.6812499761581421,
|
1324 |
+
"rewards/chosen": -1.3846697807312012,
|
1325 |
+
"rewards/margins": 0.27571359276771545,
|
1326 |
+
"rewards/rejected": -1.6603834629058838,
|
1327 |
+
"step": 840
|
1328 |
+
},
|
1329 |
+
{
|
1330 |
+
"epoch": 0.68,
|
1331 |
+
"learning_rate": 1.4000265772195032e-06,
|
1332 |
+
"logits/chosen": -4.225982666015625,
|
1333 |
+
"logits/rejected": -4.171608924865723,
|
1334 |
+
"logps/chosen": -430.31201171875,
|
1335 |
+
"logps/rejected": -475.3255920410156,
|
1336 |
+
"loss": 0.6196,
|
1337 |
+
"rewards/accuracies": 0.6625000238418579,
|
1338 |
+
"rewards/chosen": -1.3736594915390015,
|
1339 |
+
"rewards/margins": 0.2828761339187622,
|
1340 |
+
"rewards/rejected": -1.6565355062484741,
|
1341 |
+
"step": 850
|
1342 |
+
},
|
1343 |
+
{
|
1344 |
+
"epoch": 0.69,
|
1345 |
+
"learning_rate": 1.3377161622160077e-06,
|
1346 |
+
"logits/chosen": -4.169137001037598,
|
1347 |
+
"logits/rejected": -4.160987854003906,
|
1348 |
+
"logps/chosen": -430.1625061035156,
|
1349 |
+
"logps/rejected": -470.7685546875,
|
1350 |
+
"loss": 0.6046,
|
1351 |
+
"rewards/accuracies": 0.7250000238418579,
|
1352 |
+
"rewards/chosen": -1.3943125009536743,
|
1353 |
+
"rewards/margins": 0.2797546982765198,
|
1354 |
+
"rewards/rejected": -1.6740672588348389,
|
1355 |
+
"step": 860
|
1356 |
+
},
|
1357 |
+
{
|
1358 |
+
"epoch": 0.7,
|
1359 |
+
"learning_rate": 1.276313674457553e-06,
|
1360 |
+
"logits/chosen": -4.306554794311523,
|
1361 |
+
"logits/rejected": -4.296151161193848,
|
1362 |
+
"logps/chosen": -415.35101318359375,
|
1363 |
+
"logps/rejected": -470.18145751953125,
|
1364 |
+
"loss": 0.5782,
|
1365 |
+
"rewards/accuracies": 0.71875,
|
1366 |
+
"rewards/chosen": -1.413987636566162,
|
1367 |
+
"rewards/margins": 0.35374554991722107,
|
1368 |
+
"rewards/rejected": -1.7677332162857056,
|
1369 |
+
"step": 870
|
1370 |
+
},
|
1371 |
+
{
|
1372 |
+
"epoch": 0.7,
|
1373 |
+
"learning_rate": 1.2158670789848095e-06,
|
1374 |
+
"logits/chosen": -4.3886284828186035,
|
1375 |
+
"logits/rejected": -4.385241508483887,
|
1376 |
+
"logps/chosen": -460.49005126953125,
|
1377 |
+
"logps/rejected": -502.52130126953125,
|
1378 |
+
"loss": 0.608,
|
1379 |
+
"rewards/accuracies": 0.675000011920929,
|
1380 |
+
"rewards/chosen": -1.6725581884384155,
|
1381 |
+
"rewards/margins": 0.34351325035095215,
|
1382 |
+
"rewards/rejected": -2.016071319580078,
|
1383 |
+
"step": 880
|
1384 |
+
},
|
1385 |
+
{
|
1386 |
+
"epoch": 0.71,
|
1387 |
+
"learning_rate": 1.1564235941356016e-06,
|
1388 |
+
"logits/chosen": -4.544154167175293,
|
1389 |
+
"logits/rejected": -4.480313301086426,
|
1390 |
+
"logps/chosen": -452.08245849609375,
|
1391 |
+
"logps/rejected": -509.763427734375,
|
1392 |
+
"loss": 0.6428,
|
1393 |
+
"rewards/accuracies": 0.5874999761581421,
|
1394 |
+
"rewards/chosen": -1.7576968669891357,
|
1395 |
+
"rewards/margins": 0.3635411858558655,
|
1396 |
+
"rewards/rejected": -2.1212382316589355,
|
1397 |
+
"step": 890
|
1398 |
+
},
|
1399 |
+
{
|
1400 |
+
"epoch": 0.72,
|
1401 |
+
"learning_rate": 1.0980296546599254e-06,
|
1402 |
+
"logits/chosen": -4.359221935272217,
|
1403 |
+
"logits/rejected": -4.357415199279785,
|
1404 |
+
"logps/chosen": -476.0204162597656,
|
1405 |
+
"logps/rejected": -513.69287109375,
|
1406 |
+
"loss": 0.6651,
|
1407 |
+
"rewards/accuracies": 0.6000000238418579,
|
1408 |
+
"rewards/chosen": -1.769547462463379,
|
1409 |
+
"rewards/margins": 0.4242987036705017,
|
1410 |
+
"rewards/rejected": -2.1938462257385254,
|
1411 |
+
"step": 900
|
1412 |
+
},
|
1413 |
+
{
|
1414 |
+
"epoch": 0.72,
|
1415 |
+
"eval_logits/chosen": -4.617808818817139,
|
1416 |
+
"eval_logits/rejected": -4.603901386260986,
|
1417 |
+
"eval_logps/chosen": -387.0414123535156,
|
1418 |
+
"eval_logps/rejected": -406.733154296875,
|
1419 |
+
"eval_loss": 0.6818779706954956,
|
1420 |
+
"eval_rewards/accuracies": 0.5693333148956299,
|
1421 |
+
"eval_rewards/chosen": -1.6632905006408691,
|
1422 |
+
"eval_rewards/margins": 0.14075076580047607,
|
1423 |
+
"eval_rewards/rejected": -1.8040413856506348,
|
1424 |
+
"eval_runtime": 1619.8038,
|
1425 |
+
"eval_samples_per_second": 1.848,
|
1426 |
+
"eval_steps_per_second": 0.232,
|
1427 |
+
"step": 900
|
1428 |
+
},
|
1429 |
+
{
|
1430 |
+
"epoch": 0.73,
|
1431 |
+
"learning_rate": 1.040730875447083e-06,
|
1432 |
+
"logits/chosen": -4.3218793869018555,
|
1433 |
+
"logits/rejected": -4.3270487785339355,
|
1434 |
+
"logps/chosen": -451.809814453125,
|
1435 |
+
"logps/rejected": -478.60491943359375,
|
1436 |
+
"loss": 0.6503,
|
1437 |
+
"rewards/accuracies": 0.668749988079071,
|
1438 |
+
"rewards/chosen": -1.565606713294983,
|
1439 |
+
"rewards/margins": 0.20372018218040466,
|
1440 |
+
"rewards/rejected": -1.7693268060684204,
|
1441 |
+
"step": 910
|
1442 |
+
},
|
1443 |
+
{
|
1444 |
+
"epoch": 0.74,
|
1445 |
+
"learning_rate": 9.845720158932414e-07,
|
1446 |
+
"logits/chosen": -4.289103031158447,
|
1447 |
+
"logits/rejected": -4.304252624511719,
|
1448 |
+
"logps/chosen": -420.05926513671875,
|
1449 |
+
"logps/rejected": -442.1366271972656,
|
1450 |
+
"loss": 0.6505,
|
1451 |
+
"rewards/accuracies": 0.625,
|
1452 |
+
"rewards/chosen": -1.4516870975494385,
|
1453 |
+
"rewards/margins": 0.20668797194957733,
|
1454 |
+
"rewards/rejected": -1.6583747863769531,
|
1455 |
+
"step": 920
|
1456 |
+
},
|
1457 |
+
{
|
1458 |
+
"epoch": 0.74,
|
1459 |
+
"learning_rate": 9.295969449372796e-07,
|
1460 |
+
"logits/chosen": -4.317067623138428,
|
1461 |
+
"logits/rejected": -4.289021968841553,
|
1462 |
+
"logps/chosen": -417.4593811035156,
|
1463 |
+
"logps/rejected": -454.826904296875,
|
1464 |
+
"loss": 0.6219,
|
1465 |
+
"rewards/accuracies": 0.6625000238418579,
|
1466 |
+
"rewards/chosen": -1.3394687175750732,
|
1467 |
+
"rewards/margins": 0.25516074895858765,
|
1468 |
+
"rewards/rejected": -1.5946292877197266,
|
1469 |
+
"step": 930
|
1470 |
+
},
|
1471 |
+
{
|
1472 |
+
"epoch": 0.75,
|
1473 |
+
"learning_rate": 8.758486067922176e-07,
|
1474 |
+
"logits/chosen": -4.306538105010986,
|
1475 |
+
"logits/rejected": -4.262487411499023,
|
1476 |
+
"logps/chosen": -414.806884765625,
|
1477 |
+
"logps/rejected": -465.29931640625,
|
1478 |
+
"loss": 0.5998,
|
1479 |
+
"rewards/accuracies": 0.699999988079071,
|
1480 |
+
"rewards/chosen": -1.3654025793075562,
|
1481 |
+
"rewards/margins": 0.3171504735946655,
|
1482 |
+
"rewards/rejected": -1.6825529336929321,
|
1483 |
+
"step": 940
|
1484 |
+
},
|
1485 |
+
{
|
1486 |
+
"epoch": 0.76,
|
1487 |
+
"learning_rate": 8.233689873990006e-07,
|
1488 |
+
"logits/chosen": -4.317531585693359,
|
1489 |
+
"logits/rejected": -4.282795429229736,
|
1490 |
+
"logps/chosen": -420.382568359375,
|
1491 |
+
"logps/rejected": -493.355712890625,
|
1492 |
+
"loss": 0.5953,
|
1493 |
+
"rewards/accuracies": 0.6312500238418579,
|
1494 |
+
"rewards/chosen": -1.419424295425415,
|
1495 |
+
"rewards/margins": 0.5507811903953552,
|
1496 |
+
"rewards/rejected": -1.970205545425415,
|
1497 |
+
"step": 950
|
1498 |
+
},
|
1499 |
+
{
|
1500 |
+
"epoch": 0.77,
|
1501 |
+
"learning_rate": 7.721990816288555e-07,
|
1502 |
+
"logits/chosen": -4.283775329589844,
|
1503 |
+
"logits/rejected": -4.2459330558776855,
|
1504 |
+
"logps/chosen": -395.4678955078125,
|
1505 |
+
"logps/rejected": -434.7349548339844,
|
1506 |
+
"loss": 0.6193,
|
1507 |
+
"rewards/accuracies": 0.6875,
|
1508 |
+
"rewards/chosen": -1.3240183591842651,
|
1509 |
+
"rewards/margins": 0.2636045813560486,
|
1510 |
+
"rewards/rejected": -1.587622880935669,
|
1511 |
+
"step": 960
|
1512 |
+
},
|
1513 |
+
{
|
1514 |
+
"epoch": 0.78,
|
1515 |
+
"learning_rate": 7.223788612598148e-07,
|
1516 |
+
"logits/chosen": -4.310162544250488,
|
1517 |
+
"logits/rejected": -4.287927627563477,
|
1518 |
+
"logps/chosen": -436.2978515625,
|
1519 |
+
"logps/rejected": -470.63079833984375,
|
1520 |
+
"loss": 0.6316,
|
1521 |
+
"rewards/accuracies": 0.643750011920929,
|
1522 |
+
"rewards/chosen": -1.4355452060699463,
|
1523 |
+
"rewards/margins": 0.18634586036205292,
|
1524 |
+
"rewards/rejected": -1.6218910217285156,
|
1525 |
+
"step": 970
|
1526 |
+
},
|
1527 |
+
{
|
1528 |
+
"epoch": 0.78,
|
1529 |
+
"learning_rate": 6.73947243752448e-07,
|
1530 |
+
"logits/chosen": -4.377969264984131,
|
1531 |
+
"logits/rejected": -4.389736652374268,
|
1532 |
+
"logps/chosen": -458.318359375,
|
1533 |
+
"logps/rejected": -487.77874755859375,
|
1534 |
+
"loss": 0.6632,
|
1535 |
+
"rewards/accuracies": 0.612500011920929,
|
1536 |
+
"rewards/chosen": -1.6293509006500244,
|
1537 |
+
"rewards/margins": 0.36782675981521606,
|
1538 |
+
"rewards/rejected": -1.9971777200698853,
|
1539 |
+
"step": 980
|
1540 |
+
},
|
1541 |
+
{
|
1542 |
+
"epoch": 0.79,
|
1543 |
+
"learning_rate": 6.269420618491759e-07,
|
1544 |
+
"logits/chosen": -4.321467876434326,
|
1545 |
+
"logits/rejected": -4.286978721618652,
|
1546 |
+
"logps/chosen": -420.07220458984375,
|
1547 |
+
"logps/rejected": -458.8563537597656,
|
1548 |
+
"loss": 0.6383,
|
1549 |
+
"rewards/accuracies": 0.625,
|
1550 |
+
"rewards/chosen": -1.4905959367752075,
|
1551 |
+
"rewards/margins": 0.22982291877269745,
|
1552 |
+
"rewards/rejected": -1.720418930053711,
|
1553 |
+
"step": 990
|
1554 |
+
},
|
1555 |
+
{
|
1556 |
+
"epoch": 0.8,
|
1557 |
+
"learning_rate": 5.814000340209267e-07,
|
1558 |
+
"logits/chosen": -4.299461364746094,
|
1559 |
+
"logits/rejected": -4.237879276275635,
|
1560 |
+
"logps/chosen": -435.0042419433594,
|
1561 |
+
"logps/rejected": -491.1070861816406,
|
1562 |
+
"loss": 0.5993,
|
1563 |
+
"rewards/accuracies": 0.7250000238418579,
|
1564 |
+
"rewards/chosen": -1.4882529973983765,
|
1565 |
+
"rewards/margins": 0.3519567549228668,
|
1566 |
+
"rewards/rejected": -1.8402099609375,
|
1567 |
+
"step": 1000
|
1568 |
+
},
|
1569 |
+
{
|
1570 |
+
"epoch": 0.8,
|
1571 |
+
"eval_logits/chosen": -4.549124717712402,
|
1572 |
+
"eval_logits/rejected": -4.535641193389893,
|
1573 |
+
"eval_logps/chosen": -378.4713439941406,
|
1574 |
+
"eval_logps/rejected": -398.2364196777344,
|
1575 |
+
"eval_loss": 0.6785325407981873,
|
1576 |
+
"eval_rewards/accuracies": 0.5683333277702332,
|
1577 |
+
"eval_rewards/chosen": -1.577589511871338,
|
1578 |
+
"eval_rewards/margins": 0.1414840668439865,
|
1579 |
+
"eval_rewards/rejected": -1.7190735340118408,
|
1580 |
+
"eval_runtime": 1620.6034,
|
1581 |
+
"eval_samples_per_second": 1.847,
|
1582 |
+
"eval_steps_per_second": 0.231,
|
1583 |
+
"step": 1000
|
1584 |
+
},
|
1585 |
+
{
|
1586 |
+
"epoch": 0.81,
|
1587 |
+
"learning_rate": 5.373567357842111e-07,
|
1588 |
+
"logits/chosen": -4.278590202331543,
|
1589 |
+
"logits/rejected": -4.246352195739746,
|
1590 |
+
"logps/chosen": -429.3436584472656,
|
1591 |
+
"logps/rejected": -471.84722900390625,
|
1592 |
+
"loss": 0.5957,
|
1593 |
+
"rewards/accuracies": 0.643750011920929,
|
1594 |
+
"rewards/chosen": -1.422374963760376,
|
1595 |
+
"rewards/margins": 0.33030739426612854,
|
1596 |
+
"rewards/rejected": -1.7526824474334717,
|
1597 |
+
"step": 1010
|
1598 |
+
},
|
1599 |
+
{
|
1600 |
+
"epoch": 0.82,
|
1601 |
+
"learning_rate": 4.948465719110226e-07,
|
1602 |
+
"logits/chosen": -4.373248100280762,
|
1603 |
+
"logits/rejected": -4.3430304527282715,
|
1604 |
+
"logps/chosen": -415.3047790527344,
|
1605 |
+
"logps/rejected": -445.019287109375,
|
1606 |
+
"loss": 0.6443,
|
1607 |
+
"rewards/accuracies": 0.625,
|
1608 |
+
"rewards/chosen": -1.40928053855896,
|
1609 |
+
"rewards/margins": 0.228702574968338,
|
1610 |
+
"rewards/rejected": -1.6379830837249756,
|
1611 |
+
"step": 1020
|
1612 |
+
},
|
1613 |
+
{
|
1614 |
+
"epoch": 0.82,
|
1615 |
+
"learning_rate": 4.539027495532766e-07,
|
1616 |
+
"logits/chosen": -4.33120059967041,
|
1617 |
+
"logits/rejected": -4.347836494445801,
|
1618 |
+
"logps/chosen": -415.25555419921875,
|
1619 |
+
"logps/rejected": -453.518798828125,
|
1620 |
+
"loss": 0.6177,
|
1621 |
+
"rewards/accuracies": 0.65625,
|
1622 |
+
"rewards/chosen": -1.4035775661468506,
|
1623 |
+
"rewards/margins": 0.28108319640159607,
|
1624 |
+
"rewards/rejected": -1.6846606731414795,
|
1625 |
+
"step": 1030
|
1626 |
+
},
|
1627 |
+
{
|
1628 |
+
"epoch": 0.83,
|
1629 |
+
"learning_rate": 4.14557252302783e-07,
|
1630 |
+
"logits/chosen": -4.341530799865723,
|
1631 |
+
"logits/rejected": -4.305572032928467,
|
1632 |
+
"logps/chosen": -433.0341796875,
|
1633 |
+
"logps/rejected": -472.1036682128906,
|
1634 |
+
"loss": 0.6713,
|
1635 |
+
"rewards/accuracies": 0.637499988079071,
|
1636 |
+
"rewards/chosen": -1.5271722078323364,
|
1637 |
+
"rewards/margins": 0.23339995741844177,
|
1638 |
+
"rewards/rejected": -1.760572075843811,
|
1639 |
+
"step": 1040
|
1640 |
+
},
|
1641 |
+
{
|
1642 |
+
"epoch": 0.84,
|
1643 |
+
"learning_rate": 3.7684081520700884e-07,
|
1644 |
+
"logits/chosen": -4.2357988357543945,
|
1645 |
+
"logits/rejected": -4.237625598907471,
|
1646 |
+
"logps/chosen": -450.54827880859375,
|
1647 |
+
"logps/rejected": -471.02423095703125,
|
1648 |
+
"loss": 0.6429,
|
1649 |
+
"rewards/accuracies": 0.643750011920929,
|
1650 |
+
"rewards/chosen": -1.5025393962860107,
|
1651 |
+
"rewards/margins": 0.2623196542263031,
|
1652 |
+
"rewards/rejected": -1.7648589611053467,
|
1653 |
+
"step": 1050
|
1654 |
+
},
|
1655 |
+
{
|
1656 |
+
"epoch": 0.85,
|
1657 |
+
"learning_rate": 3.407829007601507e-07,
|
1658 |
+
"logits/chosen": -4.270508766174316,
|
1659 |
+
"logits/rejected": -4.225382328033447,
|
1660 |
+
"logps/chosen": -428.3861389160156,
|
1661 |
+
"logps/rejected": -479.0707092285156,
|
1662 |
+
"loss": 0.6119,
|
1663 |
+
"rewards/accuracies": 0.668749988079071,
|
1664 |
+
"rewards/chosen": -1.4203639030456543,
|
1665 |
+
"rewards/margins": 0.3360464870929718,
|
1666 |
+
"rewards/rejected": -1.7564103603363037,
|
1667 |
+
"step": 1060
|
1668 |
+
},
|
1669 |
+
{
|
1670 |
+
"epoch": 0.86,
|
1671 |
+
"learning_rate": 3.064116758882724e-07,
|
1672 |
+
"logits/chosen": -4.24053955078125,
|
1673 |
+
"logits/rejected": -4.1828999519348145,
|
1674 |
+
"logps/chosen": -443.78485107421875,
|
1675 |
+
"logps/rejected": -504.28070068359375,
|
1676 |
+
"loss": 0.6037,
|
1677 |
+
"rewards/accuracies": 0.6875,
|
1678 |
+
"rewards/chosen": -1.5132863521575928,
|
1679 |
+
"rewards/margins": 0.4130435883998871,
|
1680 |
+
"rewards/rejected": -1.9263302087783813,
|
1681 |
+
"step": 1070
|
1682 |
+
},
|
1683 |
+
{
|
1684 |
+
"epoch": 0.86,
|
1685 |
+
"learning_rate": 2.737539899464908e-07,
|
1686 |
+
"logits/chosen": -4.2971696853637695,
|
1687 |
+
"logits/rejected": -4.288466453552246,
|
1688 |
+
"logps/chosen": -403.1819763183594,
|
1689 |
+
"logps/rejected": -454.8402404785156,
|
1690 |
+
"loss": 0.6141,
|
1691 |
+
"rewards/accuracies": 0.643750011920929,
|
1692 |
+
"rewards/chosen": -1.3817451000213623,
|
1693 |
+
"rewards/margins": 0.379900187253952,
|
1694 |
+
"rewards/rejected": -1.7616455554962158,
|
1695 |
+
"step": 1080
|
1696 |
+
},
|
1697 |
+
{
|
1698 |
+
"epoch": 0.87,
|
1699 |
+
"learning_rate": 2.4283535374538645e-07,
|
1700 |
+
"logits/chosen": -4.242150783538818,
|
1701 |
+
"logits/rejected": -4.240577220916748,
|
1702 |
+
"logps/chosen": -434.7925720214844,
|
1703 |
+
"logps/rejected": -473.67803955078125,
|
1704 |
+
"loss": 0.6122,
|
1705 |
+
"rewards/accuracies": 0.637499988079071,
|
1706 |
+
"rewards/chosen": -1.4003090858459473,
|
1707 |
+
"rewards/margins": 0.31833842396736145,
|
1708 |
+
"rewards/rejected": -1.7186473608016968,
|
1709 |
+
"step": 1090
|
1710 |
+
},
|
1711 |
+
{
|
1712 |
+
"epoch": 0.88,
|
1713 |
+
"learning_rate": 2.1367991962303298e-07,
|
1714 |
+
"logits/chosen": -4.232297420501709,
|
1715 |
+
"logits/rejected": -4.2173943519592285,
|
1716 |
+
"logps/chosen": -418.9490661621094,
|
1717 |
+
"logps/rejected": -446.4789123535156,
|
1718 |
+
"loss": 0.6759,
|
1719 |
+
"rewards/accuracies": 0.5874999761581421,
|
1720 |
+
"rewards/chosen": -1.5112884044647217,
|
1721 |
+
"rewards/margins": 0.1443391889333725,
|
1722 |
+
"rewards/rejected": -1.6556276082992554,
|
1723 |
+
"step": 1100
|
1724 |
+
},
|
1725 |
+
{
|
1726 |
+
"epoch": 0.88,
|
1727 |
+
"eval_logits/chosen": -4.485485553741455,
|
1728 |
+
"eval_logits/rejected": -4.472198009490967,
|
1729 |
+
"eval_logps/chosen": -375.86541748046875,
|
1730 |
+
"eval_logps/rejected": -395.56036376953125,
|
1731 |
+
"eval_loss": 0.6777821779251099,
|
1732 |
+
"eval_rewards/accuracies": 0.5686666369438171,
|
1733 |
+
"eval_rewards/chosen": -1.5515305995941162,
|
1734 |
+
"eval_rewards/margins": 0.14078289270401,
|
1735 |
+
"eval_rewards/rejected": -1.6923134326934814,
|
1736 |
+
"eval_runtime": 1618.2899,
|
1737 |
+
"eval_samples_per_second": 1.85,
|
1738 |
+
"eval_steps_per_second": 0.232,
|
1739 |
+
"step": 1100
|
1740 |
+
},
|
1741 |
+
{
|
1742 |
+
"epoch": 0.89,
|
1743 |
+
"learning_rate": 1.8631046257820278e-07,
|
1744 |
+
"logits/chosen": -4.233702659606934,
|
1745 |
+
"logits/rejected": -4.249814033508301,
|
1746 |
+
"logps/chosen": -432.20458984375,
|
1747 |
+
"logps/rejected": -461.7394104003906,
|
1748 |
+
"loss": 0.6275,
|
1749 |
+
"rewards/accuracies": 0.6499999761581421,
|
1750 |
+
"rewards/chosen": -1.41855788230896,
|
1751 |
+
"rewards/margins": 0.2600322365760803,
|
1752 |
+
"rewards/rejected": -1.678590178489685,
|
1753 |
+
"step": 1110
|
1754 |
+
},
|
1755 |
+
{
|
1756 |
+
"epoch": 0.9,
|
1757 |
+
"learning_rate": 1.6074836247950143e-07,
|
1758 |
+
"logits/chosen": -4.206725597381592,
|
1759 |
+
"logits/rejected": -4.209759712219238,
|
1760 |
+
"logps/chosen": -443.52459716796875,
|
1761 |
+
"logps/rejected": -470.96588134765625,
|
1762 |
+
"loss": 0.6415,
|
1763 |
+
"rewards/accuracies": 0.6312500238418579,
|
1764 |
+
"rewards/chosen": -1.4318974018096924,
|
1765 |
+
"rewards/margins": 0.21563585102558136,
|
1766 |
+
"rewards/rejected": -1.6475334167480469,
|
1767 |
+
"step": 1120
|
1768 |
+
},
|
1769 |
+
{
|
1770 |
+
"epoch": 0.9,
|
1771 |
+
"learning_rate": 1.370135873643097e-07,
|
1772 |
+
"logits/chosen": -4.226916313171387,
|
1773 |
+
"logits/rejected": -4.266509532928467,
|
1774 |
+
"logps/chosen": -436.6336975097656,
|
1775 |
+
"logps/rejected": -456.3214416503906,
|
1776 |
+
"loss": 0.6262,
|
1777 |
+
"rewards/accuracies": 0.6499999761581421,
|
1778 |
+
"rewards/chosen": -1.4590485095977783,
|
1779 |
+
"rewards/margins": 0.2844700217247009,
|
1780 |
+
"rewards/rejected": -1.7435184717178345,
|
1781 |
+
"step": 1130
|
1782 |
+
},
|
1783 |
+
{
|
1784 |
+
"epoch": 0.91,
|
1785 |
+
"learning_rate": 1.1512467784059372e-07,
|
1786 |
+
"logits/chosen": -4.302299499511719,
|
1787 |
+
"logits/rejected": -4.261002540588379,
|
1788 |
+
"logps/chosen": -398.5301208496094,
|
1789 |
+
"logps/rejected": -437.83673095703125,
|
1790 |
+
"loss": 0.6221,
|
1791 |
+
"rewards/accuracies": 0.6312500238418579,
|
1792 |
+
"rewards/chosen": -1.3698493242263794,
|
1793 |
+
"rewards/margins": 0.2959776520729065,
|
1794 |
+
"rewards/rejected": -1.6658270359039307,
|
1795 |
+
"step": 1140
|
1796 |
+
},
|
1797 |
+
{
|
1798 |
+
"epoch": 0.92,
|
1799 |
+
"learning_rate": 9.509873260376251e-08,
|
1800 |
+
"logits/chosen": -4.228875160217285,
|
1801 |
+
"logits/rejected": -4.17572021484375,
|
1802 |
+
"logps/chosen": -429.0230407714844,
|
1803 |
+
"logps/rejected": -501.6388244628906,
|
1804 |
+
"loss": 0.5814,
|
1805 |
+
"rewards/accuracies": 0.699999988079071,
|
1806 |
+
"rewards/chosen": -1.4344505071640015,
|
1807 |
+
"rewards/margins": 0.4450379014015198,
|
1808 |
+
"rewards/rejected": -1.8794885873794556,
|
1809 |
+
"step": 1150
|
1810 |
+
},
|
1811 |
+
{
|
1812 |
+
"epoch": 0.93,
|
1813 |
+
"learning_rate": 7.695139507988559e-08,
|
1814 |
+
"logits/chosen": -4.261081218719482,
|
1815 |
+
"logits/rejected": -4.282492160797119,
|
1816 |
+
"logps/chosen": -444.05096435546875,
|
1817 |
+
"logps/rejected": -479.0032653808594,
|
1818 |
+
"loss": 0.6376,
|
1819 |
+
"rewards/accuracies": 0.6000000238418579,
|
1820 |
+
"rewards/chosen": -1.457777500152588,
|
1821 |
+
"rewards/margins": 0.2464013397693634,
|
1822 |
+
"rewards/rejected": -1.704178810119629,
|
1823 |
+
"step": 1160
|
1824 |
+
},
|
1825 |
+
{
|
1826 |
+
"epoch": 0.94,
|
1827 |
+
"learning_rate": 6.069684120570684e-08,
|
1828 |
+
"logits/chosen": -4.286696434020996,
|
1829 |
+
"logits/rejected": -4.1916022300720215,
|
1830 |
+
"logps/chosen": -431.74053955078125,
|
1831 |
+
"logps/rejected": -482.0506896972656,
|
1832 |
+
"loss": 0.6355,
|
1833 |
+
"rewards/accuracies": 0.625,
|
1834 |
+
"rewards/chosen": -1.5193744897842407,
|
1835 |
+
"rewards/margins": 0.3015151619911194,
|
1836 |
+
"rewards/rejected": -1.8208894729614258,
|
1837 |
+
"step": 1170
|
1838 |
+
},
|
1839 |
+
{
|
1840 |
+
"epoch": 0.94,
|
1841 |
+
"learning_rate": 4.634776835499871e-08,
|
1842 |
+
"logits/chosen": -4.216092109680176,
|
1843 |
+
"logits/rejected": -4.185781955718994,
|
1844 |
+
"logps/chosen": -413.5948791503906,
|
1845 |
+
"logps/rejected": -455.0482482910156,
|
1846 |
+
"loss": 0.6371,
|
1847 |
+
"rewards/accuracies": 0.6625000238418579,
|
1848 |
+
"rewards/chosen": -1.4376914501190186,
|
1849 |
+
"rewards/margins": 0.27367502450942993,
|
1850 |
+
"rewards/rejected": -1.7113662958145142,
|
1851 |
+
"step": 1180
|
1852 |
+
},
|
1853 |
+
{
|
1854 |
+
"epoch": 0.95,
|
1855 |
+
"learning_rate": 3.3915385419908964e-08,
|
1856 |
+
"logits/chosen": -4.179436206817627,
|
1857 |
+
"logits/rejected": -4.222240447998047,
|
1858 |
+
"logps/chosen": -430.4244079589844,
|
1859 |
+
"logps/rejected": -464.51629638671875,
|
1860 |
+
"loss": 0.6297,
|
1861 |
+
"rewards/accuracies": 0.6000000238418579,
|
1862 |
+
"rewards/chosen": -1.4367833137512207,
|
1863 |
+
"rewards/margins": 0.29136672616004944,
|
1864 |
+
"rewards/rejected": -1.7281500101089478,
|
1865 |
+
"step": 1190
|
1866 |
+
},
|
1867 |
+
{
|
1868 |
+
"epoch": 0.96,
|
1869 |
+
"learning_rate": 2.3409404055043938e-08,
|
1870 |
+
"logits/chosen": -4.306519508361816,
|
1871 |
+
"logits/rejected": -4.272242546081543,
|
1872 |
+
"logps/chosen": -437.791259765625,
|
1873 |
+
"logps/rejected": -471.58575439453125,
|
1874 |
+
"loss": 0.6402,
|
1875 |
+
"rewards/accuracies": 0.5874999761581421,
|
1876 |
+
"rewards/chosen": -1.5211617946624756,
|
1877 |
+
"rewards/margins": 0.2325367033481598,
|
1878 |
+
"rewards/rejected": -1.7536985874176025,
|
1879 |
+
"step": 1200
|
1880 |
+
},
|
1881 |
+
{
|
1882 |
+
"epoch": 0.96,
|
1883 |
+
"eval_logits/chosen": -4.457742691040039,
|
1884 |
+
"eval_logits/rejected": -4.444427967071533,
|
1885 |
+
"eval_logps/chosen": -375.10284423828125,
|
1886 |
+
"eval_logps/rejected": -394.7001647949219,
|
1887 |
+
"eval_loss": 0.6772644519805908,
|
1888 |
+
"eval_rewards/accuracies": 0.5690000057220459,
|
1889 |
+
"eval_rewards/chosen": -1.5439047813415527,
|
1890 |
+
"eval_rewards/margins": 0.13980673253536224,
|
1891 |
+
"eval_rewards/rejected": -1.6837116479873657,
|
1892 |
+
"eval_runtime": 1624.6813,
|
1893 |
+
"eval_samples_per_second": 1.843,
|
1894 |
+
"eval_steps_per_second": 0.231,
|
1895 |
+
"step": 1200
|
1896 |
+
},
|
1897 |
+
{
|
1898 |
+
"epoch": 0.97,
|
1899 |
+
"learning_rate": 1.4838031091134186e-08,
|
1900 |
+
"logits/chosen": -4.243393898010254,
|
1901 |
+
"logits/rejected": -4.175347805023193,
|
1902 |
+
"logps/chosen": -404.5989990234375,
|
1903 |
+
"logps/rejected": -469.9576110839844,
|
1904 |
+
"loss": 0.5982,
|
1905 |
+
"rewards/accuracies": 0.6937500238418579,
|
1906 |
+
"rewards/chosen": -1.3969910144805908,
|
1907 |
+
"rewards/margins": 0.3869909644126892,
|
1908 |
+
"rewards/rejected": -1.7839819192886353,
|
1909 |
+
"step": 1210
|
1910 |
+
},
|
1911 |
+
{
|
1912 |
+
"epoch": 0.98,
|
1913 |
+
"learning_rate": 8.207962124201774e-09,
|
1914 |
+
"logits/chosen": -4.263760566711426,
|
1915 |
+
"logits/rejected": -4.225130081176758,
|
1916 |
+
"logps/chosen": -435.9114685058594,
|
1917 |
+
"logps/rejected": -472.66839599609375,
|
1918 |
+
"loss": 0.6017,
|
1919 |
+
"rewards/accuracies": 0.675000011920929,
|
1920 |
+
"rewards/chosen": -1.37319016456604,
|
1921 |
+
"rewards/margins": 0.3209651708602905,
|
1922 |
+
"rewards/rejected": -1.6941554546356201,
|
1923 |
+
"step": 1220
|
1924 |
+
},
|
1925 |
+
{
|
1926 |
+
"epoch": 0.98,
|
1927 |
+
"learning_rate": 3.5243762852441023e-09,
|
1928 |
+
"logits/chosen": -4.210858345031738,
|
1929 |
+
"logits/rejected": -4.163503170013428,
|
1930 |
+
"logps/chosen": -427.44342041015625,
|
1931 |
+
"logps/rejected": -469.5209045410156,
|
1932 |
+
"loss": 0.6376,
|
1933 |
+
"rewards/accuracies": 0.6000000238418579,
|
1934 |
+
"rewards/chosen": -1.4386152029037476,
|
1935 |
+
"rewards/margins": 0.2547362446784973,
|
1936 |
+
"rewards/rejected": -1.6933513879776,
|
1937 |
+
"step": 1230
|
1938 |
+
},
|
1939 |
+
{
|
1940 |
+
"epoch": 0.99,
|
1941 |
+
"learning_rate": 7.909321945129278e-10,
|
1942 |
+
"logits/chosen": -4.18636417388916,
|
1943 |
+
"logits/rejected": -4.141125679016113,
|
1944 |
+
"logps/chosen": -443.1625061035156,
|
1945 |
+
"logps/rejected": -493.43975830078125,
|
1946 |
+
"loss": 0.6016,
|
1947 |
+
"rewards/accuracies": 0.6499999761581421,
|
1948 |
+
"rewards/chosen": -1.4015429019927979,
|
1949 |
+
"rewards/margins": 0.35573890805244446,
|
1950 |
+
"rewards/rejected": -1.7572818994522095,
|
1951 |
+
"step": 1240
|
1952 |
+
},
|
1953 |
+
{
|
1954 |
+
"epoch": 1.0,
|
1955 |
+
"step": 1249,
|
1956 |
+
"total_flos": 0.0,
|
1957 |
+
"train_loss": 0.6475976420174225,
|
1958 |
+
"train_runtime": 42677.4758,
|
1959 |
+
"train_samples_per_second": 0.469,
|
1960 |
+
"train_steps_per_second": 0.029
|
1961 |
+
}
|
1962 |
+
],
|
1963 |
+
"logging_steps": 10,
|
1964 |
+
"max_steps": 1249,
|
1965 |
+
"num_input_tokens_seen": 0,
|
1966 |
+
"num_train_epochs": 1,
|
1967 |
+
"save_steps": 100,
|
1968 |
+
"total_flos": 0.0,
|
1969 |
+
"train_batch_size": 4,
|
1970 |
+
"trial_name": null,
|
1971 |
+
"trial_params": null
|
1972 |
+
}
|