Commit
•
415f17b
1
Parent(s):
146c680
Model save
Browse files- README.md +84 -0
- adapter_model.safetensors +1 -1
- all_results.json +21 -0
- eval_results.json +16 -0
- runs/Mar30_15-30-11_Braylon/events.out.tfevents.1711773088.Braylon.21497.0 +2 -2
- runs/Mar30_15-30-11_Braylon/events.out.tfevents.1711787753.Braylon.21497.1 +3 -0
- train_results.json +8 -0
- trainer_state.json +1972 -0
README.md
ADDED
|
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
base_model: microsoft/phi-2
|
| 9 |
+
model-index:
|
| 10 |
+
- name: phi-2-ipo-test-iter-0
|
| 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-ipo-test-iter-0
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [microsoft/phi-2](https://huggingface.co/microsoft/phi-2) on the None dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 2546.4375
|
| 22 |
+
- Rewards/chosen: -0.1591
|
| 23 |
+
- Rewards/rejected: -0.1612
|
| 24 |
+
- Rewards/accuracies: 0.5220
|
| 25 |
+
- Rewards/margins: 0.0021
|
| 26 |
+
- Logps/rejected: -249.6534
|
| 27 |
+
- Logps/chosen: -272.5227
|
| 28 |
+
- Logits/rejected: 0.4171
|
| 29 |
+
- Logits/chosen: 0.3526
|
| 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: 4
|
| 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: 4
|
| 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 |
+
| 2477.3281 | 0.32 | 100 | 2500.7156 | -0.0018 | -0.0018 | 0.4930 | -0.0000 | -233.7207 | -256.7978 | 0.8796 | 0.8221 |
|
| 65 |
+
| 2224.3488 | 0.64 | 200 | 2499.8904 | -0.0195 | -0.0198 | 0.5015 | 0.0003 | -235.5204 | -258.5673 | 0.8051 | 0.7462 |
|
| 66 |
+
| 1898.0719 | 0.96 | 300 | 2505.6912 | -0.0563 | -0.0571 | 0.5140 | 0.0008 | -239.2530 | -262.2491 | 0.6844 | 0.6233 |
|
| 67 |
+
| 1879.8852 | 1.28 | 400 | 2516.0835 | -0.0944 | -0.0957 | 0.5200 | 0.0013 | -243.1053 | -266.0533 | 0.5839 | 0.5215 |
|
| 68 |
+
| 1917.2811 | 1.6 | 500 | 2527.1995 | -0.1156 | -0.1170 | 0.5115 | 0.0014 | -245.2343 | -268.1747 | 0.5244 | 0.4611 |
|
| 69 |
+
| 1799.3824 | 1.92 | 600 | 2534.4292 | -0.1363 | -0.1381 | 0.5210 | 0.0018 | -247.3504 | -270.2482 | 0.4714 | 0.4075 |
|
| 70 |
+
| 1751.5762 | 2.24 | 700 | 2531.3550 | -0.1448 | -0.1474 | 0.5180 | 0.0026 | -248.2780 | -271.0988 | 0.4545 | 0.3906 |
|
| 71 |
+
| 1711.1711 | 2.56 | 800 | 2536.2451 | -0.1487 | -0.1511 | 0.5145 | 0.0024 | -248.6440 | -271.4834 | 0.4402 | 0.3759 |
|
| 72 |
+
| 1894.4447 | 2.88 | 900 | 2542.6299 | -0.1549 | -0.1570 | 0.5235 | 0.0022 | -249.2417 | -272.1000 | 0.4262 | 0.3618 |
|
| 73 |
+
| 1798.5389 | 3.2 | 1000 | 2542.7288 | -0.1581 | -0.1604 | 0.5205 | 0.0023 | -249.5780 | -272.4200 | 0.4202 | 0.3559 |
|
| 74 |
+
| 1834.9711 | 3.52 | 1100 | 2542.2373 | -0.1586 | -0.1610 | 0.5205 | 0.0024 | -249.6345 | -272.4703 | 0.4177 | 0.3532 |
|
| 75 |
+
| 1765.5148 | 3.84 | 1200 | 2546.1714 | -0.1589 | -0.1610 | 0.5220 | 0.0021 | -249.6357 | -272.5010 | 0.4160 | 0.3515 |
|
| 76 |
+
|
| 77 |
+
|
| 78 |
+
### Framework versions
|
| 79 |
+
|
| 80 |
+
- PEFT 0.7.1
|
| 81 |
+
- Transformers 4.36.2
|
| 82 |
+
- Pytorch 2.2.1+cu121
|
| 83 |
+
- Datasets 2.14.6
|
| 84 |
+
- Tokenizers 0.15.2
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 41977616
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5c1e67946b196556debdce64a2a65e344ae1ad85cba635e5b08b0a06b1003e4
|
| 3 |
size 41977616
|
all_results.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.99,
|
| 3 |
+
"eval_logits/chosen": 0.3525713086128235,
|
| 4 |
+
"eval_logits/rejected": 0.4171118438243866,
|
| 5 |
+
"eval_logps/chosen": -272.5227355957031,
|
| 6 |
+
"eval_logps/rejected": -249.6534423828125,
|
| 7 |
+
"eval_loss": 2546.4375,
|
| 8 |
+
"eval_rewards/accuracies": 0.5220000147819519,
|
| 9 |
+
"eval_rewards/chosen": -0.15908558666706085,
|
| 10 |
+
"eval_rewards/margins": 0.0020662047900259495,
|
| 11 |
+
"eval_rewards/rejected": -0.16115178167819977,
|
| 12 |
+
"eval_runtime": 411.7057,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 4.858,
|
| 15 |
+
"eval_steps_per_second": 1.214,
|
| 16 |
+
"train_loss": 1918.4646684695513,
|
| 17 |
+
"train_runtime": 14252.7073,
|
| 18 |
+
"train_samples": 30567,
|
| 19 |
+
"train_samples_per_second": 1.403,
|
| 20 |
+
"train_steps_per_second": 0.088
|
| 21 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.99,
|
| 3 |
+
"eval_logits/chosen": 0.3525713086128235,
|
| 4 |
+
"eval_logits/rejected": 0.4171118438243866,
|
| 5 |
+
"eval_logps/chosen": -272.5227355957031,
|
| 6 |
+
"eval_logps/rejected": -249.6534423828125,
|
| 7 |
+
"eval_loss": 2546.4375,
|
| 8 |
+
"eval_rewards/accuracies": 0.5220000147819519,
|
| 9 |
+
"eval_rewards/chosen": -0.15908558666706085,
|
| 10 |
+
"eval_rewards/margins": 0.0020662047900259495,
|
| 11 |
+
"eval_rewards/rejected": -0.16115178167819977,
|
| 12 |
+
"eval_runtime": 411.7057,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 4.858,
|
| 15 |
+
"eval_steps_per_second": 1.214
|
| 16 |
+
}
|
runs/Mar30_15-30-11_Braylon/events.out.tfevents.1711773088.Braylon.21497.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fcefbd3d4afe4692920068afa1be5a0b979d01eb00227742305862b7f1dfd8ae
|
| 3 |
+
size 93140
|
runs/Mar30_15-30-11_Braylon/events.out.tfevents.1711787753.Braylon.21497.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15ca0c62afb04058b08608cbc96562a5e226f7f6122a4dd3668d55f4f0f0cd9f
|
| 3 |
+
size 828
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.99,
|
| 3 |
+
"train_loss": 1918.4646684695513,
|
| 4 |
+
"train_runtime": 14252.7073,
|
| 5 |
+
"train_samples": 30567,
|
| 6 |
+
"train_samples_per_second": 1.403,
|
| 7 |
+
"train_steps_per_second": 0.088
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1972 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.9936,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 1248,
|
| 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": 0.7221256494522095,
|
| 15 |
+
"logits/rejected": 0.8745549917221069,
|
| 16 |
+
"logps/chosen": -277.6833801269531,
|
| 17 |
+
"logps/rejected": -189.9869384765625,
|
| 18 |
+
"loss": 2500.0,
|
| 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": 0.03,
|
| 27 |
+
"learning_rate": 4.0000000000000003e-07,
|
| 28 |
+
"logits/chosen": 0.6776503324508667,
|
| 29 |
+
"logits/rejected": 0.7872164845466614,
|
| 30 |
+
"logps/chosen": -273.4142761230469,
|
| 31 |
+
"logps/rejected": -216.78836059570312,
|
| 32 |
+
"loss": 2505.8785,
|
| 33 |
+
"rewards/accuracies": 0.3888888955116272,
|
| 34 |
+
"rewards/chosen": 0.00015669333515688777,
|
| 35 |
+
"rewards/margins": -0.0005680065951310098,
|
| 36 |
+
"rewards/rejected": 0.0007246998138725758,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.06,
|
| 41 |
+
"learning_rate": 8.000000000000001e-07,
|
| 42 |
+
"logits/chosen": 0.6071363687515259,
|
| 43 |
+
"logits/rejected": 0.8961409330368042,
|
| 44 |
+
"logps/chosen": -252.6411590576172,
|
| 45 |
+
"logps/rejected": -188.6440887451172,
|
| 46 |
+
"loss": 2498.5875,
|
| 47 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 48 |
+
"rewards/chosen": 0.0007678332040086389,
|
| 49 |
+
"rewards/margins": 0.00017280881002079695,
|
| 50 |
+
"rewards/rejected": 0.0005950243212282658,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.1,
|
| 55 |
+
"learning_rate": 1.2000000000000002e-06,
|
| 56 |
+
"logits/chosen": 0.5971012711524963,
|
| 57 |
+
"logits/rejected": 0.7454315423965454,
|
| 58 |
+
"logps/chosen": -272.0901184082031,
|
| 59 |
+
"logps/rejected": -205.0573272705078,
|
| 60 |
+
"loss": 2504.0305,
|
| 61 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 62 |
+
"rewards/chosen": -0.00030031849746592343,
|
| 63 |
+
"rewards/margins": -0.0003763613640330732,
|
| 64 |
+
"rewards/rejected": 7.604288111906499e-05,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.13,
|
| 69 |
+
"learning_rate": 1.6000000000000001e-06,
|
| 70 |
+
"logits/chosen": 0.5639680027961731,
|
| 71 |
+
"logits/rejected": 0.8156954646110535,
|
| 72 |
+
"logps/chosen": -254.08987426757812,
|
| 73 |
+
"logps/rejected": -199.48912048339844,
|
| 74 |
+
"loss": 2499.0203,
|
| 75 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 76 |
+
"rewards/chosen": -0.00016066078387666494,
|
| 77 |
+
"rewards/margins": 0.00013786301133222878,
|
| 78 |
+
"rewards/rejected": -0.00029852380976080894,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.16,
|
| 83 |
+
"learning_rate": 2.0000000000000003e-06,
|
| 84 |
+
"logits/chosen": 0.6404772996902466,
|
| 85 |
+
"logits/rejected": 0.8446690440177917,
|
| 86 |
+
"logps/chosen": -264.58636474609375,
|
| 87 |
+
"logps/rejected": -211.51986694335938,
|
| 88 |
+
"loss": 2499.9363,
|
| 89 |
+
"rewards/accuracies": 0.46875,
|
| 90 |
+
"rewards/chosen": 0.0005515815573744476,
|
| 91 |
+
"rewards/margins": 3.0463817893178202e-05,
|
| 92 |
+
"rewards/rejected": 0.0005211178213357925,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.19,
|
| 97 |
+
"learning_rate": 2.4000000000000003e-06,
|
| 98 |
+
"logits/chosen": 0.693733274936676,
|
| 99 |
+
"logits/rejected": 0.8246955871582031,
|
| 100 |
+
"logps/chosen": -266.50872802734375,
|
| 101 |
+
"logps/rejected": -209.1096649169922,
|
| 102 |
+
"loss": 2498.2361,
|
| 103 |
+
"rewards/accuracies": 0.4375,
|
| 104 |
+
"rewards/chosen": 0.0005181363085284829,
|
| 105 |
+
"rewards/margins": 0.00020206482440698892,
|
| 106 |
+
"rewards/rejected": 0.0003160713822580874,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.22,
|
| 111 |
+
"learning_rate": 2.8000000000000003e-06,
|
| 112 |
+
"logits/chosen": 0.6724094152450562,
|
| 113 |
+
"logits/rejected": 0.8244352340698242,
|
| 114 |
+
"logps/chosen": -256.1933898925781,
|
| 115 |
+
"logps/rejected": -198.73048400878906,
|
| 116 |
+
"loss": 2492.3055,
|
| 117 |
+
"rewards/accuracies": 0.518750011920929,
|
| 118 |
+
"rewards/chosen": 0.00043690926395356655,
|
| 119 |
+
"rewards/margins": 0.0007959330687299371,
|
| 120 |
+
"rewards/rejected": -0.0003590236883610487,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.26,
|
| 125 |
+
"learning_rate": 3.2000000000000003e-06,
|
| 126 |
+
"logits/chosen": 0.6599764227867126,
|
| 127 |
+
"logits/rejected": 0.8170899152755737,
|
| 128 |
+
"logps/chosen": -273.2989807128906,
|
| 129 |
+
"logps/rejected": -216.25827026367188,
|
| 130 |
+
"loss": 2488.535,
|
| 131 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 132 |
+
"rewards/chosen": 0.0022243678104132414,
|
| 133 |
+
"rewards/margins": 0.0011816672049462795,
|
| 134 |
+
"rewards/rejected": 0.0010427006054669619,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.29,
|
| 139 |
+
"learning_rate": 3.6000000000000003e-06,
|
| 140 |
+
"logits/chosen": 0.6013220548629761,
|
| 141 |
+
"logits/rejected": 0.8273889422416687,
|
| 142 |
+
"logps/chosen": -268.2584533691406,
|
| 143 |
+
"logps/rejected": -213.9058380126953,
|
| 144 |
+
"loss": 2482.6732,
|
| 145 |
+
"rewards/accuracies": 0.574999988079071,
|
| 146 |
+
"rewards/chosen": 0.003236656542867422,
|
| 147 |
+
"rewards/margins": 0.0017823975067585707,
|
| 148 |
+
"rewards/rejected": 0.001454259268939495,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.32,
|
| 153 |
+
"learning_rate": 4.000000000000001e-06,
|
| 154 |
+
"logits/chosen": 0.6121121644973755,
|
| 155 |
+
"logits/rejected": 0.8102381825447083,
|
| 156 |
+
"logps/chosen": -258.34576416015625,
|
| 157 |
+
"logps/rejected": -212.81478881835938,
|
| 158 |
+
"loss": 2477.3281,
|
| 159 |
+
"rewards/accuracies": 0.543749988079071,
|
| 160 |
+
"rewards/chosen": 0.0035489716101437807,
|
| 161 |
+
"rewards/margins": 0.0023237697314471006,
|
| 162 |
+
"rewards/rejected": 0.001225201180204749,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.32,
|
| 167 |
+
"eval_logits/chosen": 0.8220780491828918,
|
| 168 |
+
"eval_logits/rejected": 0.8795672655105591,
|
| 169 |
+
"eval_logps/chosen": -256.7978210449219,
|
| 170 |
+
"eval_logps/rejected": -233.720703125,
|
| 171 |
+
"eval_loss": 2500.715576171875,
|
| 172 |
+
"eval_rewards/accuracies": 0.49300000071525574,
|
| 173 |
+
"eval_rewards/chosen": -0.0018364518182352185,
|
| 174 |
+
"eval_rewards/margins": -1.2206258361402433e-05,
|
| 175 |
+
"eval_rewards/rejected": -0.0018242454389110208,
|
| 176 |
+
"eval_runtime": 442.8545,
|
| 177 |
+
"eval_samples_per_second": 4.516,
|
| 178 |
+
"eval_steps_per_second": 1.129,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.35,
|
| 183 |
+
"learning_rate": 4.4e-06,
|
| 184 |
+
"logits/chosen": 0.6344770193099976,
|
| 185 |
+
"logits/rejected": 0.8470233082771301,
|
| 186 |
+
"logps/chosen": -257.6260986328125,
|
| 187 |
+
"logps/rejected": -205.66622924804688,
|
| 188 |
+
"loss": 2466.1889,
|
| 189 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 190 |
+
"rewards/chosen": 0.005115890875458717,
|
| 191 |
+
"rewards/margins": 0.003451352473348379,
|
| 192 |
+
"rewards/rejected": 0.0016645386349409819,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.38,
|
| 197 |
+
"learning_rate": 4.800000000000001e-06,
|
| 198 |
+
"logits/chosen": 0.6225894689559937,
|
| 199 |
+
"logits/rejected": 0.7519047856330872,
|
| 200 |
+
"logps/chosen": -273.39776611328125,
|
| 201 |
+
"logps/rejected": -211.7797088623047,
|
| 202 |
+
"loss": 2446.4322,
|
| 203 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 204 |
+
"rewards/chosen": 0.007990755140781403,
|
| 205 |
+
"rewards/margins": 0.005494705401360989,
|
| 206 |
+
"rewards/rejected": 0.0024960506707429886,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.42,
|
| 211 |
+
"learning_rate": 4.999755441268144e-06,
|
| 212 |
+
"logits/chosen": 0.5374079942703247,
|
| 213 |
+
"logits/rejected": 0.8167151212692261,
|
| 214 |
+
"logps/chosen": -283.385498046875,
|
| 215 |
+
"logps/rejected": -228.09716796875,
|
| 216 |
+
"loss": 2434.1764,
|
| 217 |
+
"rewards/accuracies": 0.606249988079071,
|
| 218 |
+
"rewards/chosen": 0.010817321017384529,
|
| 219 |
+
"rewards/margins": 0.006822638213634491,
|
| 220 |
+
"rewards/rejected": 0.003994682338088751,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.45,
|
| 225 |
+
"learning_rate": 4.997799258487003e-06,
|
| 226 |
+
"logits/chosen": 0.6246897578239441,
|
| 227 |
+
"logits/rejected": 0.8092406988143921,
|
| 228 |
+
"logps/chosen": -273.5729064941406,
|
| 229 |
+
"logps/rejected": -220.0849609375,
|
| 230 |
+
"loss": 2406.7561,
|
| 231 |
+
"rewards/accuracies": 0.606249988079071,
|
| 232 |
+
"rewards/chosen": 0.013732816092669964,
|
| 233 |
+
"rewards/margins": 0.009700324386358261,
|
| 234 |
+
"rewards/rejected": 0.004032492637634277,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.48,
|
| 239 |
+
"learning_rate": 4.993888423734898e-06,
|
| 240 |
+
"logits/chosen": 0.6373761892318726,
|
| 241 |
+
"logits/rejected": 0.7924760580062866,
|
| 242 |
+
"logps/chosen": -259.81890869140625,
|
| 243 |
+
"logps/rejected": -198.0768585205078,
|
| 244 |
+
"loss": 2378.9869,
|
| 245 |
+
"rewards/accuracies": 0.668749988079071,
|
| 246 |
+
"rewards/chosen": 0.017864007502794266,
|
| 247 |
+
"rewards/margins": 0.01282494980841875,
|
| 248 |
+
"rewards/rejected": 0.00503905676305294,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.51,
|
| 253 |
+
"learning_rate": 4.988025997434253e-06,
|
| 254 |
+
"logits/chosen": 0.6211603283882141,
|
| 255 |
+
"logits/rejected": 0.7996471524238586,
|
| 256 |
+
"logps/chosen": -262.85723876953125,
|
| 257 |
+
"logps/rejected": -209.2670135498047,
|
| 258 |
+
"loss": 2347.3615,
|
| 259 |
+
"rewards/accuracies": 0.65625,
|
| 260 |
+
"rewards/chosen": 0.02218160405755043,
|
| 261 |
+
"rewards/margins": 0.016304031014442444,
|
| 262 |
+
"rewards/rejected": 0.005877572111785412,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.54,
|
| 267 |
+
"learning_rate": 4.980216567224801e-06,
|
| 268 |
+
"logits/chosen": 0.5590766668319702,
|
| 269 |
+
"logits/rejected": 0.8358405232429504,
|
| 270 |
+
"logps/chosen": -268.93450927734375,
|
| 271 |
+
"logps/rejected": -217.07223510742188,
|
| 272 |
+
"loss": 2314.8727,
|
| 273 |
+
"rewards/accuracies": 0.668749988079071,
|
| 274 |
+
"rewards/chosen": 0.027453165501356125,
|
| 275 |
+
"rewards/margins": 0.02034853585064411,
|
| 276 |
+
"rewards/rejected": 0.0071046315133571625,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.58,
|
| 281 |
+
"learning_rate": 4.970466244373527e-06,
|
| 282 |
+
"logits/chosen": 0.5857344269752502,
|
| 283 |
+
"logits/rejected": 0.8025603294372559,
|
| 284 |
+
"logps/chosen": -269.49420166015625,
|
| 285 |
+
"logps/rejected": -224.668701171875,
|
| 286 |
+
"loss": 2323.1209,
|
| 287 |
+
"rewards/accuracies": 0.675000011920929,
|
| 288 |
+
"rewards/chosen": 0.026119906455278397,
|
| 289 |
+
"rewards/margins": 0.019484227523207664,
|
| 290 |
+
"rewards/rejected": 0.006635676138103008,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.61,
|
| 295 |
+
"learning_rate": 4.958782658992307e-06,
|
| 296 |
+
"logits/chosen": 0.6378930807113647,
|
| 297 |
+
"logits/rejected": 0.7625473737716675,
|
| 298 |
+
"logps/chosen": -245.82064819335938,
|
| 299 |
+
"logps/rejected": -212.6953582763672,
|
| 300 |
+
"loss": 2353.0727,
|
| 301 |
+
"rewards/accuracies": 0.59375,
|
| 302 |
+
"rewards/chosen": 0.02194877900183201,
|
| 303 |
+
"rewards/margins": 0.01612677052617073,
|
| 304 |
+
"rewards/rejected": 0.0058220066130161285,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.64,
|
| 309 |
+
"learning_rate": 4.945174954066957e-06,
|
| 310 |
+
"logits/chosen": 0.5794962048530579,
|
| 311 |
+
"logits/rejected": 0.7216407656669617,
|
| 312 |
+
"logps/chosen": -262.74481201171875,
|
| 313 |
+
"logps/rejected": -205.56753540039062,
|
| 314 |
+
"loss": 2224.3488,
|
| 315 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 316 |
+
"rewards/chosen": 0.040737103670835495,
|
| 317 |
+
"rewards/margins": 0.03179619461297989,
|
| 318 |
+
"rewards/rejected": 0.008940907195210457,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.64,
|
| 323 |
+
"eval_logits/chosen": 0.7462335824966431,
|
| 324 |
+
"eval_logits/rejected": 0.8050766587257385,
|
| 325 |
+
"eval_logps/chosen": -258.5672912597656,
|
| 326 |
+
"eval_logps/rejected": -235.52040100097656,
|
| 327 |
+
"eval_loss": 2499.890380859375,
|
| 328 |
+
"eval_rewards/accuracies": 0.5015000104904175,
|
| 329 |
+
"eval_rewards/chosen": -0.019531190395355225,
|
| 330 |
+
"eval_rewards/margins": 0.0002903530257754028,
|
| 331 |
+
"eval_rewards/rejected": -0.019821541383862495,
|
| 332 |
+
"eval_runtime": 412.3797,
|
| 333 |
+
"eval_samples_per_second": 4.85,
|
| 334 |
+
"eval_steps_per_second": 1.212,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.67,
|
| 339 |
+
"learning_rate": 4.929653778302397e-06,
|
| 340 |
+
"logits/chosen": 0.6170880198478699,
|
| 341 |
+
"logits/rejected": 0.7665790915489197,
|
| 342 |
+
"logps/chosen": -260.6783447265625,
|
| 343 |
+
"logps/rejected": -205.8716278076172,
|
| 344 |
+
"loss": 2165.3709,
|
| 345 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 346 |
+
"rewards/chosen": 0.04661153629422188,
|
| 347 |
+
"rewards/margins": 0.039808522909879684,
|
| 348 |
+
"rewards/rejected": 0.006803011987358332,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.7,
|
| 353 |
+
"learning_rate": 4.912231277789509e-06,
|
| 354 |
+
"logits/chosen": 0.5444914102554321,
|
| 355 |
+
"logits/rejected": 0.7560594081878662,
|
| 356 |
+
"logps/chosen": -277.1501770019531,
|
| 357 |
+
"logps/rejected": -227.8493194580078,
|
| 358 |
+
"loss": 2199.4646,
|
| 359 |
+
"rewards/accuracies": 0.606249988079071,
|
| 360 |
+
"rewards/chosen": 0.05624958872795105,
|
| 361 |
+
"rewards/margins": 0.037255048751831055,
|
| 362 |
+
"rewards/rejected": 0.018994538113474846,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.74,
|
| 367 |
+
"learning_rate": 4.892921086500219e-06,
|
| 368 |
+
"logits/chosen": 0.5499123930931091,
|
| 369 |
+
"logits/rejected": 0.6719511151313782,
|
| 370 |
+
"logps/chosen": -243.00390625,
|
| 371 |
+
"logps/rejected": -205.0264434814453,
|
| 372 |
+
"loss": 2218.1998,
|
| 373 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 374 |
+
"rewards/chosen": 0.04481152817606926,
|
| 375 |
+
"rewards/margins": 0.03403898701071739,
|
| 376 |
+
"rewards/rejected": 0.010772541165351868,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.77,
|
| 381 |
+
"learning_rate": 4.871738315618236e-06,
|
| 382 |
+
"logits/chosen": 0.5359088182449341,
|
| 383 |
+
"logits/rejected": 0.7013985514640808,
|
| 384 |
+
"logps/chosen": -263.36322021484375,
|
| 385 |
+
"logps/rejected": -217.5428924560547,
|
| 386 |
+
"loss": 2121.8713,
|
| 387 |
+
"rewards/accuracies": 0.612500011920929,
|
| 388 |
+
"rewards/chosen": 0.06463086605072021,
|
| 389 |
+
"rewards/margins": 0.04686326906085014,
|
| 390 |
+
"rewards/rejected": 0.017767589539289474,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.8,
|
| 395 |
+
"learning_rate": 4.848699541713801e-06,
|
| 396 |
+
"logits/chosen": 0.5078493356704712,
|
| 397 |
+
"logits/rejected": 0.6534050703048706,
|
| 398 |
+
"logps/chosen": -255.99612426757812,
|
| 399 |
+
"logps/rejected": -209.0691680908203,
|
| 400 |
+
"loss": 2179.4311,
|
| 401 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 402 |
+
"rewards/chosen": 0.05505412817001343,
|
| 403 |
+
"rewards/margins": 0.04023104906082153,
|
| 404 |
+
"rewards/rejected": 0.014823079109191895,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.83,
|
| 409 |
+
"learning_rate": 4.823822793771696e-06,
|
| 410 |
+
"logits/chosen": 0.5251200795173645,
|
| 411 |
+
"logits/rejected": 0.6907744407653809,
|
| 412 |
+
"logps/chosen": -244.97616577148438,
|
| 413 |
+
"logps/rejected": -193.0167694091797,
|
| 414 |
+
"loss": 2039.2832,
|
| 415 |
+
"rewards/accuracies": 0.699999988079071,
|
| 416 |
+
"rewards/chosen": 0.06581679731607437,
|
| 417 |
+
"rewards/margins": 0.05999577045440674,
|
| 418 |
+
"rewards/rejected": 0.005821021273732185,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.86,
|
| 423 |
+
"learning_rate": 4.797127539082669e-06,
|
| 424 |
+
"logits/chosen": 0.5102941393852234,
|
| 425 |
+
"logits/rejected": 0.6884879469871521,
|
| 426 |
+
"logps/chosen": -267.79193115234375,
|
| 427 |
+
"logps/rejected": -216.37808227539062,
|
| 428 |
+
"loss": 2084.2711,
|
| 429 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 430 |
+
"rewards/chosen": 0.06373900920152664,
|
| 431 |
+
"rewards/margins": 0.05440465360879898,
|
| 432 |
+
"rewards/rejected": 0.009334356524050236,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.9,
|
| 437 |
+
"learning_rate": 4.7686346680093135e-06,
|
| 438 |
+
"logits/chosen": 0.46306952834129333,
|
| 439 |
+
"logits/rejected": 0.6997040510177612,
|
| 440 |
+
"logps/chosen": -248.92501831054688,
|
| 441 |
+
"logps/rejected": -192.95266723632812,
|
| 442 |
+
"loss": 1956.5348,
|
| 443 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 444 |
+
"rewards/chosen": 0.08462236821651459,
|
| 445 |
+
"rewards/margins": 0.07454703748226166,
|
| 446 |
+
"rewards/rejected": 0.01007532887160778,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.93,
|
| 451 |
+
"learning_rate": 4.738366477638319e-06,
|
| 452 |
+
"logits/chosen": 0.49755996465682983,
|
| 453 |
+
"logits/rejected": 0.6823209524154663,
|
| 454 |
+
"logps/chosen": -252.31265258789062,
|
| 455 |
+
"logps/rejected": -203.94357299804688,
|
| 456 |
+
"loss": 2082.6836,
|
| 457 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 458 |
+
"rewards/chosen": 0.060582734644412994,
|
| 459 |
+
"rewards/margins": 0.0560123547911644,
|
| 460 |
+
"rewards/rejected": 0.004570374730974436,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.96,
|
| 465 |
+
"learning_rate": 4.7063466543318965e-06,
|
| 466 |
+
"logits/chosen": 0.40442362427711487,
|
| 467 |
+
"logits/rejected": 0.643377423286438,
|
| 468 |
+
"logps/chosen": -253.7049102783203,
|
| 469 |
+
"logps/rejected": -207.00146484375,
|
| 470 |
+
"loss": 1898.0719,
|
| 471 |
+
"rewards/accuracies": 0.6875,
|
| 472 |
+
"rewards/chosen": 0.09280969202518463,
|
| 473 |
+
"rewards/margins": 0.08169680833816528,
|
| 474 |
+
"rewards/rejected": 0.011112888343632221,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.96,
|
| 479 |
+
"eval_logits/chosen": 0.6232779622077942,
|
| 480 |
+
"eval_logits/rejected": 0.6844168305397034,
|
| 481 |
+
"eval_logps/chosen": -262.24908447265625,
|
| 482 |
+
"eval_logps/rejected": -239.25296020507812,
|
| 483 |
+
"eval_loss": 2505.691162109375,
|
| 484 |
+
"eval_rewards/accuracies": 0.5139999985694885,
|
| 485 |
+
"eval_rewards/chosen": -0.05634931102395058,
|
| 486 |
+
"eval_rewards/margins": 0.0007975505432114005,
|
| 487 |
+
"eval_rewards/rejected": -0.057146865874528885,
|
| 488 |
+
"eval_runtime": 411.7351,
|
| 489 |
+
"eval_samples_per_second": 4.857,
|
| 490 |
+
"eval_steps_per_second": 1.214,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.99,
|
| 495 |
+
"learning_rate": 4.672600255192022e-06,
|
| 496 |
+
"logits/chosen": 0.4478415846824646,
|
| 497 |
+
"logits/rejected": 0.6094005703926086,
|
| 498 |
+
"logps/chosen": -280.0593566894531,
|
| 499 |
+
"logps/rejected": -230.06820678710938,
|
| 500 |
+
"loss": 1902.6814,
|
| 501 |
+
"rewards/accuracies": 0.699999988079071,
|
| 502 |
+
"rewards/chosen": 0.08015590906143188,
|
| 503 |
+
"rewards/margins": 0.08636869490146637,
|
| 504 |
+
"rewards/rejected": -0.006212792359292507,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 1.02,
|
| 509 |
+
"learning_rate": 4.6371536884520115e-06,
|
| 510 |
+
"logits/chosen": 0.4466761648654938,
|
| 511 |
+
"logits/rejected": 0.5766214728355408,
|
| 512 |
+
"logps/chosen": -260.2146911621094,
|
| 513 |
+
"logps/rejected": -202.6782989501953,
|
| 514 |
+
"loss": 1965.6586,
|
| 515 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 516 |
+
"rewards/chosen": 0.08317091315984726,
|
| 517 |
+
"rewards/margins": 0.07663293927907944,
|
| 518 |
+
"rewards/rejected": 0.006537970155477524,
|
| 519 |
+
"step": 320
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 1.06,
|
| 523 |
+
"learning_rate": 4.600034692810764e-06,
|
| 524 |
+
"logits/chosen": 0.4064536690711975,
|
| 525 |
+
"logits/rejected": 0.6527116894721985,
|
| 526 |
+
"logps/chosen": -285.71063232421875,
|
| 527 |
+
"logps/rejected": -232.9168243408203,
|
| 528 |
+
"loss": 1807.0604,
|
| 529 |
+
"rewards/accuracies": 0.675000011920929,
|
| 530 |
+
"rewards/chosen": 0.1063646450638771,
|
| 531 |
+
"rewards/margins": 0.10833654552698135,
|
| 532 |
+
"rewards/rejected": -0.001971897669136524,
|
| 533 |
+
"step": 330
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 1.09,
|
| 537 |
+
"learning_rate": 4.561272315725852e-06,
|
| 538 |
+
"logits/chosen": 0.39116889238357544,
|
| 539 |
+
"logits/rejected": 0.5679996609687805,
|
| 540 |
+
"logps/chosen": -269.14727783203125,
|
| 541 |
+
"logps/rejected": -226.08987426757812,
|
| 542 |
+
"loss": 1983.8736,
|
| 543 |
+
"rewards/accuracies": 0.675000011920929,
|
| 544 |
+
"rewards/chosen": 0.09996851533651352,
|
| 545 |
+
"rewards/margins": 0.08038529753684998,
|
| 546 |
+
"rewards/rejected": 0.019583214074373245,
|
| 547 |
+
"step": 340
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 1.12,
|
| 551 |
+
"learning_rate": 4.520896890682449e-06,
|
| 552 |
+
"logits/chosen": 0.3980609178543091,
|
| 553 |
+
"logits/rejected": 0.5835639834403992,
|
| 554 |
+
"logps/chosen": -261.222412109375,
|
| 555 |
+
"logps/rejected": -211.2742156982422,
|
| 556 |
+
"loss": 1903.2615,
|
| 557 |
+
"rewards/accuracies": 0.625,
|
| 558 |
+
"rewards/chosen": 0.11096767336130142,
|
| 559 |
+
"rewards/margins": 0.10369744151830673,
|
| 560 |
+
"rewards/rejected": 0.007270221598446369,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 1.15,
|
| 565 |
+
"learning_rate": 4.478940013455864e-06,
|
| 566 |
+
"logits/chosen": 0.32668036222457886,
|
| 567 |
+
"logits/rejected": 0.642693042755127,
|
| 568 |
+
"logps/chosen": -248.02194213867188,
|
| 569 |
+
"logps/rejected": -195.903564453125,
|
| 570 |
+
"loss": 1819.0176,
|
| 571 |
+
"rewards/accuracies": 0.675000011920929,
|
| 572 |
+
"rewards/chosen": 0.11030755937099457,
|
| 573 |
+
"rewards/margins": 0.10598069429397583,
|
| 574 |
+
"rewards/rejected": 0.004326865542680025,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 1.18,
|
| 579 |
+
"learning_rate": 4.435434517386281e-06,
|
| 580 |
+
"logits/chosen": 0.40880435705184937,
|
| 581 |
+
"logits/rejected": 0.6121729612350464,
|
| 582 |
+
"logps/chosen": -256.9469909667969,
|
| 583 |
+
"logps/rejected": -210.01284790039062,
|
| 584 |
+
"loss": 1875.7564,
|
| 585 |
+
"rewards/accuracies": 0.65625,
|
| 586 |
+
"rewards/chosen": 0.10248272120952606,
|
| 587 |
+
"rewards/margins": 0.09737477451562881,
|
| 588 |
+
"rewards/rejected": 0.005107959732413292,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 1.22,
|
| 593 |
+
"learning_rate": 4.39041444768504e-06,
|
| 594 |
+
"logits/chosen": 0.35375842452049255,
|
| 595 |
+
"logits/rejected": 0.5443102121353149,
|
| 596 |
+
"logps/chosen": -271.12091064453125,
|
| 597 |
+
"logps/rejected": -223.96395874023438,
|
| 598 |
+
"loss": 1840.3432,
|
| 599 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 600 |
+
"rewards/chosen": 0.10606244951486588,
|
| 601 |
+
"rewards/margins": 0.11150339990854263,
|
| 602 |
+
"rewards/rejected": -0.005440945271402597,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 1.25,
|
| 607 |
+
"learning_rate": 4.343915034792569e-06,
|
| 608 |
+
"logits/chosen": 0.33456122875213623,
|
| 609 |
+
"logits/rejected": 0.6237267255783081,
|
| 610 |
+
"logps/chosen": -248.20458984375,
|
| 611 |
+
"logps/rejected": -200.05455017089844,
|
| 612 |
+
"loss": 1773.2205,
|
| 613 |
+
"rewards/accuracies": 0.6875,
|
| 614 |
+
"rewards/chosen": 0.11527778208255768,
|
| 615 |
+
"rewards/margins": 0.11671481281518936,
|
| 616 |
+
"rewards/rejected": -0.0014370165299624205,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.28,
|
| 621 |
+
"learning_rate": 4.295972666808811e-06,
|
| 622 |
+
"logits/chosen": 0.3166826665401459,
|
| 623 |
+
"logits/rejected": 0.5174378156661987,
|
| 624 |
+
"logps/chosen": -285.5495910644531,
|
| 625 |
+
"logps/rejected": -236.46630859375,
|
| 626 |
+
"loss": 1879.8852,
|
| 627 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 628 |
+
"rewards/chosen": 0.10617715120315552,
|
| 629 |
+
"rewards/margins": 0.1061965599656105,
|
| 630 |
+
"rewards/rejected": -1.941286063811276e-05,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.28,
|
| 635 |
+
"eval_logits/chosen": 0.5215258598327637,
|
| 636 |
+
"eval_logits/rejected": 0.5839402079582214,
|
| 637 |
+
"eval_logps/chosen": -266.0533447265625,
|
| 638 |
+
"eval_logps/rejected": -243.10528564453125,
|
| 639 |
+
"eval_loss": 2516.08349609375,
|
| 640 |
+
"eval_rewards/accuracies": 0.5199999809265137,
|
| 641 |
+
"eval_rewards/chosen": -0.09439140558242798,
|
| 642 |
+
"eval_rewards/margins": 0.001278651412576437,
|
| 643 |
+
"eval_rewards/rejected": -0.09567005187273026,
|
| 644 |
+
"eval_runtime": 412.4154,
|
| 645 |
+
"eval_samples_per_second": 4.849,
|
| 646 |
+
"eval_steps_per_second": 1.212,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 1.31,
|
| 651 |
+
"learning_rate": 4.246624861017732e-06,
|
| 652 |
+
"logits/chosen": 0.31735214591026306,
|
| 653 |
+
"logits/rejected": 0.45617708563804626,
|
| 654 |
+
"logps/chosen": -242.1129608154297,
|
| 655 |
+
"logps/rejected": -200.98440551757812,
|
| 656 |
+
"loss": 1814.2686,
|
| 657 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 658 |
+
"rewards/chosen": 0.13390564918518066,
|
| 659 |
+
"rewards/margins": 0.11606212705373764,
|
| 660 |
+
"rewards/rejected": 0.01784351095557213,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.34,
|
| 665 |
+
"learning_rate": 4.195910234528186e-06,
|
| 666 |
+
"logits/chosen": 0.34283334016799927,
|
| 667 |
+
"logits/rejected": 0.532746434211731,
|
| 668 |
+
"logps/chosen": -250.3861541748047,
|
| 669 |
+
"logps/rejected": -209.1632843017578,
|
| 670 |
+
"loss": 1950.3166,
|
| 671 |
+
"rewards/accuracies": 0.550000011920929,
|
| 672 |
+
"rewards/chosen": 0.08673225343227386,
|
| 673 |
+
"rewards/margins": 0.09020966291427612,
|
| 674 |
+
"rewards/rejected": -0.003477415069937706,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.38,
|
| 679 |
+
"learning_rate": 4.143868474054098e-06,
|
| 680 |
+
"logits/chosen": 0.2972554564476013,
|
| 681 |
+
"logits/rejected": 0.4938802123069763,
|
| 682 |
+
"logps/chosen": -245.431640625,
|
| 683 |
+
"logps/rejected": -195.8451385498047,
|
| 684 |
+
"loss": 1783.8299,
|
| 685 |
+
"rewards/accuracies": 0.675000011920929,
|
| 686 |
+
"rewards/chosen": 0.1252884864807129,
|
| 687 |
+
"rewards/margins": 0.11878180503845215,
|
| 688 |
+
"rewards/rejected": 0.006506689824163914,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.41,
|
| 693 |
+
"learning_rate": 4.0905403048576545e-06,
|
| 694 |
+
"logits/chosen": 0.3040269911289215,
|
| 695 |
+
"logits/rejected": 0.5599544048309326,
|
| 696 |
+
"logps/chosen": -254.08749389648438,
|
| 697 |
+
"logps/rejected": -213.42941284179688,
|
| 698 |
+
"loss": 1878.9361,
|
| 699 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 700 |
+
"rewards/chosen": 0.1011713370680809,
|
| 701 |
+
"rewards/margins": 0.10365436226129532,
|
| 702 |
+
"rewards/rejected": -0.002483018906787038,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.44,
|
| 707 |
+
"learning_rate": 4.035967458879751e-06,
|
| 708 |
+
"logits/chosen": 0.29073578119277954,
|
| 709 |
+
"logits/rejected": 0.4535519480705261,
|
| 710 |
+
"logps/chosen": -226.4980926513672,
|
| 711 |
+
"logps/rejected": -191.19338989257812,
|
| 712 |
+
"loss": 1882.7074,
|
| 713 |
+
"rewards/accuracies": 0.706250011920929,
|
| 714 |
+
"rewards/chosen": 0.11003535985946655,
|
| 715 |
+
"rewards/margins": 0.10973072052001953,
|
| 716 |
+
"rewards/rejected": 0.000304625544231385,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.47,
|
| 721 |
+
"learning_rate": 3.980192642082682e-06,
|
| 722 |
+
"logits/chosen": 0.27826622128486633,
|
| 723 |
+
"logits/rejected": 0.4509051740169525,
|
| 724 |
+
"logps/chosen": -246.6437225341797,
|
| 725 |
+
"logps/rejected": -199.22531127929688,
|
| 726 |
+
"loss": 1963.9541,
|
| 727 |
+
"rewards/accuracies": 0.637499988079071,
|
| 728 |
+
"rewards/chosen": 0.10869649797677994,
|
| 729 |
+
"rewards/margins": 0.10766948759555817,
|
| 730 |
+
"rewards/rejected": 0.0010270171333104372,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.5,
|
| 735 |
+
"learning_rate": 3.923259501030604e-06,
|
| 736 |
+
"logits/chosen": 0.24556589126586914,
|
| 737 |
+
"logits/rejected": 0.4926172196865082,
|
| 738 |
+
"logps/chosen": -247.1698455810547,
|
| 739 |
+
"logps/rejected": -208.8766326904297,
|
| 740 |
+
"loss": 1971.4818,
|
| 741 |
+
"rewards/accuracies": 0.625,
|
| 742 |
+
"rewards/chosen": 0.0876002386212349,
|
| 743 |
+
"rewards/margins": 0.08931747823953629,
|
| 744 |
+
"rewards/rejected": -0.0017172384541481733,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.54,
|
| 749 |
+
"learning_rate": 3.865212588733927e-06,
|
| 750 |
+
"logits/chosen": 0.3068960905075073,
|
| 751 |
+
"logits/rejected": 0.515642523765564,
|
| 752 |
+
"logps/chosen": -253.49880981445312,
|
| 753 |
+
"logps/rejected": -201.22650146484375,
|
| 754 |
+
"loss": 1692.4139,
|
| 755 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 756 |
+
"rewards/chosen": 0.12761953473091125,
|
| 757 |
+
"rewards/margins": 0.15515872836112976,
|
| 758 |
+
"rewards/rejected": -0.027539223432540894,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.57,
|
| 763 |
+
"learning_rate": 3.8060973297843773e-06,
|
| 764 |
+
"logits/chosen": 0.2675209045410156,
|
| 765 |
+
"logits/rejected": 0.4657977521419525,
|
| 766 |
+
"logps/chosen": -260.93353271484375,
|
| 767 |
+
"logps/rejected": -221.60214233398438,
|
| 768 |
+
"loss": 1822.4361,
|
| 769 |
+
"rewards/accuracies": 0.6875,
|
| 770 |
+
"rewards/chosen": 0.10896619409322739,
|
| 771 |
+
"rewards/margins": 0.10793409496545792,
|
| 772 |
+
"rewards/rejected": 0.0010320901637896895,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.6,
|
| 777 |
+
"learning_rate": 3.7459599848079965e-06,
|
| 778 |
+
"logits/chosen": 0.32330405712127686,
|
| 779 |
+
"logits/rejected": 0.49660053849220276,
|
| 780 |
+
"logps/chosen": -270.75567626953125,
|
| 781 |
+
"logps/rejected": -223.2246856689453,
|
| 782 |
+
"loss": 1917.2811,
|
| 783 |
+
"rewards/accuracies": 0.606249988079071,
|
| 784 |
+
"rewards/chosen": 0.1301083266735077,
|
| 785 |
+
"rewards/margins": 0.10672901570796967,
|
| 786 |
+
"rewards/rejected": 0.023379310965538025,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.6,
|
| 791 |
+
"eval_logits/chosen": 0.4610958397388458,
|
| 792 |
+
"eval_logits/rejected": 0.5243986248970032,
|
| 793 |
+
"eval_logps/chosen": -268.1746826171875,
|
| 794 |
+
"eval_logps/rejected": -245.2342987060547,
|
| 795 |
+
"eval_loss": 2527.199462890625,
|
| 796 |
+
"eval_rewards/accuracies": 0.5115000009536743,
|
| 797 |
+
"eval_rewards/chosen": -0.11560481786727905,
|
| 798 |
+
"eval_rewards/margins": 0.0013554621255025268,
|
| 799 |
+
"eval_rewards/rejected": -0.11696028709411621,
|
| 800 |
+
"eval_runtime": 411.7771,
|
| 801 |
+
"eval_samples_per_second": 4.857,
|
| 802 |
+
"eval_steps_per_second": 1.214,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 1.63,
|
| 807 |
+
"learning_rate": 3.684847614263898e-06,
|
| 808 |
+
"logits/chosen": 0.2653389871120453,
|
| 809 |
+
"logits/rejected": 0.5690582990646362,
|
| 810 |
+
"logps/chosen": -266.63861083984375,
|
| 811 |
+
"logps/rejected": -214.3343963623047,
|
| 812 |
+
"loss": 1769.2539,
|
| 813 |
+
"rewards/accuracies": 0.65625,
|
| 814 |
+
"rewards/chosen": 0.1135379895567894,
|
| 815 |
+
"rewards/margins": 0.13335202634334564,
|
| 816 |
+
"rewards/rejected": -0.01981404609978199,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.66,
|
| 821 |
+
"learning_rate": 3.622808041617133e-06,
|
| 822 |
+
"logits/chosen": 0.23678083717823029,
|
| 823 |
+
"logits/rejected": 0.3923517167568207,
|
| 824 |
+
"logps/chosen": -251.62295532226562,
|
| 825 |
+
"logps/rejected": -213.0308074951172,
|
| 826 |
+
"loss": 1775.6238,
|
| 827 |
+
"rewards/accuracies": 0.699999988079071,
|
| 828 |
+
"rewards/chosen": 0.12413735687732697,
|
| 829 |
+
"rewards/margins": 0.11967913061380386,
|
| 830 |
+
"rewards/rejected": 0.0044582299888134,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.7,
|
| 835 |
+
"learning_rate": 3.559889815914441e-06,
|
| 836 |
+
"logits/chosen": 0.28451576828956604,
|
| 837 |
+
"logits/rejected": 0.5491675138473511,
|
| 838 |
+
"logps/chosen": -242.1103057861328,
|
| 839 |
+
"logps/rejected": -196.47967529296875,
|
| 840 |
+
"loss": 1708.515,
|
| 841 |
+
"rewards/accuracies": 0.699999988079071,
|
| 842 |
+
"rewards/chosen": 0.13650254905223846,
|
| 843 |
+
"rewards/margins": 0.1416713297367096,
|
| 844 |
+
"rewards/rejected": -0.005168789066374302,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.73,
|
| 849 |
+
"learning_rate": 3.496142173792219e-06,
|
| 850 |
+
"logits/chosen": 0.23486502468585968,
|
| 851 |
+
"logits/rejected": 0.4870077073574066,
|
| 852 |
+
"logps/chosen": -244.1999969482422,
|
| 853 |
+
"logps/rejected": -209.3772430419922,
|
| 854 |
+
"loss": 1659.4633,
|
| 855 |
+
"rewards/accuracies": 0.737500011920929,
|
| 856 |
+
"rewards/chosen": 0.13094991445541382,
|
| 857 |
+
"rewards/margins": 0.14943310618400574,
|
| 858 |
+
"rewards/rejected": -0.018483208492398262,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.76,
|
| 863 |
+
"learning_rate": 3.4316150009464023e-06,
|
| 864 |
+
"logits/chosen": 0.2991487979888916,
|
| 865 |
+
"logits/rejected": 0.5229448080062866,
|
| 866 |
+
"logps/chosen": -264.29534912109375,
|
| 867 |
+
"logps/rejected": -222.90609741210938,
|
| 868 |
+
"loss": 1999.6244,
|
| 869 |
+
"rewards/accuracies": 0.643750011920929,
|
| 870 |
+
"rewards/chosen": 0.1066334992647171,
|
| 871 |
+
"rewards/margins": 0.0911371111869812,
|
| 872 |
+
"rewards/rejected": 0.015496388077735901,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.79,
|
| 877 |
+
"learning_rate": 3.366358793094433e-06,
|
| 878 |
+
"logits/chosen": 0.18731743097305298,
|
| 879 |
+
"logits/rejected": 0.40507373213768005,
|
| 880 |
+
"logps/chosen": -245.1287078857422,
|
| 881 |
+
"logps/rejected": -194.5263214111328,
|
| 882 |
+
"loss": 1803.6146,
|
| 883 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 884 |
+
"rewards/chosen": 0.15052266418933868,
|
| 885 |
+
"rewards/margins": 0.13785871863365173,
|
| 886 |
+
"rewards/rejected": 0.012663939967751503,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.82,
|
| 891 |
+
"learning_rate": 3.3004246164598535e-06,
|
| 892 |
+
"logits/chosen": 0.20718684792518616,
|
| 893 |
+
"logits/rejected": 0.46155864000320435,
|
| 894 |
+
"logps/chosen": -270.30792236328125,
|
| 895 |
+
"logps/rejected": -233.5759735107422,
|
| 896 |
+
"loss": 1710.7586,
|
| 897 |
+
"rewards/accuracies": 0.668749988079071,
|
| 898 |
+
"rewards/chosen": 0.14311200380325317,
|
| 899 |
+
"rewards/margins": 0.14155760407447815,
|
| 900 |
+
"rewards/rejected": 0.0015544015914201736,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.86,
|
| 905 |
+
"learning_rate": 3.233864067810446e-06,
|
| 906 |
+
"logits/chosen": 0.3184027075767517,
|
| 907 |
+
"logits/rejected": 0.4623110890388489,
|
| 908 |
+
"logps/chosen": -252.27920532226562,
|
| 909 |
+
"logps/rejected": -208.2698974609375,
|
| 910 |
+
"loss": 1681.7773,
|
| 911 |
+
"rewards/accuracies": 0.699999988079071,
|
| 912 |
+
"rewards/chosen": 0.137790247797966,
|
| 913 |
+
"rewards/margins": 0.15522876381874084,
|
| 914 |
+
"rewards/rejected": -0.017438489943742752,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.89,
|
| 919 |
+
"learning_rate": 3.1667292340812077e-06,
|
| 920 |
+
"logits/chosen": 0.1437700092792511,
|
| 921 |
+
"logits/rejected": 0.3409765958786011,
|
| 922 |
+
"logps/chosen": -247.97341918945312,
|
| 923 |
+
"logps/rejected": -218.58163452148438,
|
| 924 |
+
"loss": 1856.1729,
|
| 925 |
+
"rewards/accuracies": 0.668749988079071,
|
| 926 |
+
"rewards/chosen": 0.13051538169384003,
|
| 927 |
+
"rewards/margins": 0.12071319669485092,
|
| 928 |
+
"rewards/rejected": 0.009802192449569702,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.92,
|
| 933 |
+
"learning_rate": 3.099072651613728e-06,
|
| 934 |
+
"logits/chosen": 0.24014055728912354,
|
| 935 |
+
"logits/rejected": 0.3376144766807556,
|
| 936 |
+
"logps/chosen": -236.4192352294922,
|
| 937 |
+
"logps/rejected": -202.04861450195312,
|
| 938 |
+
"loss": 1799.3824,
|
| 939 |
+
"rewards/accuracies": 0.65625,
|
| 940 |
+
"rewards/chosen": 0.11552337557077408,
|
| 941 |
+
"rewards/margins": 0.12494631111621857,
|
| 942 |
+
"rewards/rejected": -0.009422937408089638,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.92,
|
| 947 |
+
"eval_logits/chosen": 0.40748730301856995,
|
| 948 |
+
"eval_logits/rejected": 0.4713863432407379,
|
| 949 |
+
"eval_logps/chosen": -270.2481994628906,
|
| 950 |
+
"eval_logps/rejected": -247.35043334960938,
|
| 951 |
+
"eval_loss": 2534.42919921875,
|
| 952 |
+
"eval_rewards/accuracies": 0.5210000276565552,
|
| 953 |
+
"eval_rewards/chosen": -0.13634006679058075,
|
| 954 |
+
"eval_rewards/margins": 0.001781497965566814,
|
| 955 |
+
"eval_rewards/rejected": -0.13812156021595,
|
| 956 |
+
"eval_runtime": 412.3385,
|
| 957 |
+
"eval_samples_per_second": 4.85,
|
| 958 |
+
"eval_steps_per_second": 1.213,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 1.95,
|
| 963 |
+
"learning_rate": 3.0309472650438982e-06,
|
| 964 |
+
"logits/chosen": 0.2844320833683014,
|
| 965 |
+
"logits/rejected": 0.44563204050064087,
|
| 966 |
+
"logps/chosen": -256.5829772949219,
|
| 967 |
+
"logps/rejected": -221.15725708007812,
|
| 968 |
+
"loss": 1905.6564,
|
| 969 |
+
"rewards/accuracies": 0.668749988079071,
|
| 970 |
+
"rewards/chosen": 0.10678652673959732,
|
| 971 |
+
"rewards/margins": 0.10849614441394806,
|
| 972 |
+
"rewards/rejected": -0.001709613250568509,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.98,
|
| 977 |
+
"learning_rate": 2.9624063858701006e-06,
|
| 978 |
+
"logits/chosen": 0.2729710638523102,
|
| 979 |
+
"logits/rejected": 0.4474371075630188,
|
| 980 |
+
"logps/chosen": -252.75082397460938,
|
| 981 |
+
"logps/rejected": -210.4240264892578,
|
| 982 |
+
"loss": 1794.8119,
|
| 983 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 984 |
+
"rewards/chosen": 0.11081516742706299,
|
| 985 |
+
"rewards/margins": 0.1245477944612503,
|
| 986 |
+
"rewards/rejected": -0.013732627034187317,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 2.02,
|
| 991 |
+
"learning_rate": 2.8935036507343185e-06,
|
| 992 |
+
"logits/chosen": 0.20590758323669434,
|
| 993 |
+
"logits/rejected": 0.4346593916416168,
|
| 994 |
+
"logps/chosen": -262.78900146484375,
|
| 995 |
+
"logps/rejected": -220.1951141357422,
|
| 996 |
+
"loss": 1844.0162,
|
| 997 |
+
"rewards/accuracies": 0.643750011920929,
|
| 998 |
+
"rewards/chosen": 0.11512579768896103,
|
| 999 |
+
"rewards/margins": 0.12753179669380188,
|
| 1000 |
+
"rewards/rejected": -0.012406004592776299,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 2.05,
|
| 1005 |
+
"learning_rate": 2.8242929794487926e-06,
|
| 1006 |
+
"logits/chosen": 0.18648579716682434,
|
| 1007 |
+
"logits/rejected": 0.4544282555580139,
|
| 1008 |
+
"logps/chosen": -268.1901550292969,
|
| 1009 |
+
"logps/rejected": -227.70193481445312,
|
| 1010 |
+
"loss": 1771.6072,
|
| 1011 |
+
"rewards/accuracies": 0.6875,
|
| 1012 |
+
"rewards/chosen": 0.1294766366481781,
|
| 1013 |
+
"rewards/margins": 0.14083310961723328,
|
| 1014 |
+
"rewards/rejected": -0.011356466449797153,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 2.08,
|
| 1019 |
+
"learning_rate": 2.7548285328010984e-06,
|
| 1020 |
+
"logits/chosen": 0.26485538482666016,
|
| 1021 |
+
"logits/rejected": 0.36643001437187195,
|
| 1022 |
+
"logps/chosen": -250.9102783203125,
|
| 1023 |
+
"logps/rejected": -210.93991088867188,
|
| 1024 |
+
"loss": 1854.1963,
|
| 1025 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1026 |
+
"rewards/chosen": 0.12583324313163757,
|
| 1027 |
+
"rewards/margins": 0.11704309284687042,
|
| 1028 |
+
"rewards/rejected": 0.008790150284767151,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 2.11,
|
| 1033 |
+
"learning_rate": 2.6851646701706306e-06,
|
| 1034 |
+
"logits/chosen": 0.1529732346534729,
|
| 1035 |
+
"logits/rejected": 0.33808109164237976,
|
| 1036 |
+
"logps/chosen": -280.79803466796875,
|
| 1037 |
+
"logps/rejected": -242.85238647460938,
|
| 1038 |
+
"loss": 1942.4141,
|
| 1039 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 1040 |
+
"rewards/chosen": 0.11561117321252823,
|
| 1041 |
+
"rewards/margins": 0.10545969009399414,
|
| 1042 |
+
"rewards/rejected": 0.010151493363082409,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 2.14,
|
| 1047 |
+
"learning_rate": 2.6153559069897007e-06,
|
| 1048 |
+
"logits/chosen": 0.24622070789337158,
|
| 1049 |
+
"logits/rejected": 0.3663380444049835,
|
| 1050 |
+
"logps/chosen": -240.98788452148438,
|
| 1051 |
+
"logps/rejected": -203.41390991210938,
|
| 1052 |
+
"loss": 1846.866,
|
| 1053 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1054 |
+
"rewards/chosen": 0.10691628605127335,
|
| 1055 |
+
"rewards/margins": 0.11943434178829193,
|
| 1056 |
+
"rewards/rejected": -0.012518051080405712,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 2.18,
|
| 1061 |
+
"learning_rate": 2.5454568720824937e-06,
|
| 1062 |
+
"logits/chosen": 0.14400359988212585,
|
| 1063 |
+
"logits/rejected": 0.4130190908908844,
|
| 1064 |
+
"logps/chosen": -238.1641082763672,
|
| 1065 |
+
"logps/rejected": -194.319580078125,
|
| 1066 |
+
"loss": 1685.1135,
|
| 1067 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1068 |
+
"rewards/chosen": 0.1472087949514389,
|
| 1069 |
+
"rewards/margins": 0.141302227973938,
|
| 1070 |
+
"rewards/rejected": 0.005906577687710524,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 2.21,
|
| 1075 |
+
"learning_rate": 2.4755222649153014e-06,
|
| 1076 |
+
"logits/chosen": 0.19388817250728607,
|
| 1077 |
+
"logits/rejected": 0.38251611590385437,
|
| 1078 |
+
"logps/chosen": -235.1546173095703,
|
| 1079 |
+
"logps/rejected": -187.75552368164062,
|
| 1080 |
+
"loss": 1779.3379,
|
| 1081 |
+
"rewards/accuracies": 0.6875,
|
| 1082 |
+
"rewards/chosen": 0.1266268938779831,
|
| 1083 |
+
"rewards/margins": 0.13590212166309357,
|
| 1084 |
+
"rewards/rejected": -0.009275219403207302,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 2.24,
|
| 1089 |
+
"learning_rate": 2.4056068127914803e-06,
|
| 1090 |
+
"logits/chosen": 0.21440652012825012,
|
| 1091 |
+
"logits/rejected": 0.46250271797180176,
|
| 1092 |
+
"logps/chosen": -270.1151123046875,
|
| 1093 |
+
"logps/rejected": -226.77603149414062,
|
| 1094 |
+
"loss": 1751.5762,
|
| 1095 |
+
"rewards/accuracies": 0.668749988079071,
|
| 1096 |
+
"rewards/chosen": 0.12278898805379868,
|
| 1097 |
+
"rewards/margins": 0.15427103638648987,
|
| 1098 |
+
"rewards/rejected": -0.03148204833269119,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 2.24,
|
| 1103 |
+
"eval_logits/chosen": 0.39058271050453186,
|
| 1104 |
+
"eval_logits/rejected": 0.45453789830207825,
|
| 1105 |
+
"eval_logps/chosen": -271.09881591796875,
|
| 1106 |
+
"eval_logps/rejected": -248.2779541015625,
|
| 1107 |
+
"eval_loss": 2531.35498046875,
|
| 1108 |
+
"eval_rewards/accuracies": 0.5180000066757202,
|
| 1109 |
+
"eval_rewards/chosen": -0.14484626054763794,
|
| 1110 |
+
"eval_rewards/margins": 0.0025505598168820143,
|
| 1111 |
+
"eval_rewards/rejected": -0.14739681780338287,
|
| 1112 |
+
"eval_runtime": 411.7503,
|
| 1113 |
+
"eval_samples_per_second": 4.857,
|
| 1114 |
+
"eval_steps_per_second": 1.214,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 2.27,
|
| 1119 |
+
"learning_rate": 2.3357652280246125e-06,
|
| 1120 |
+
"logits/chosen": 0.22305497527122498,
|
| 1121 |
+
"logits/rejected": 0.4246363043785095,
|
| 1122 |
+
"logps/chosen": -244.64987182617188,
|
| 1123 |
+
"logps/rejected": -204.69613647460938,
|
| 1124 |
+
"loss": 1997.1527,
|
| 1125 |
+
"rewards/accuracies": 0.637499988079071,
|
| 1126 |
+
"rewards/chosen": 0.10293789952993393,
|
| 1127 |
+
"rewards/margins": 0.09320969879627228,
|
| 1128 |
+
"rewards/rejected": 0.009728200733661652,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 2.3,
|
| 1133 |
+
"learning_rate": 2.2660521651234036e-06,
|
| 1134 |
+
"logits/chosen": 0.25884026288986206,
|
| 1135 |
+
"logits/rejected": 0.44062429666519165,
|
| 1136 |
+
"logps/chosen": -264.1467590332031,
|
| 1137 |
+
"logps/rejected": -212.3171844482422,
|
| 1138 |
+
"loss": 1703.6529,
|
| 1139 |
+
"rewards/accuracies": 0.6875,
|
| 1140 |
+
"rewards/chosen": 0.14418098330497742,
|
| 1141 |
+
"rewards/margins": 0.1545061618089676,
|
| 1142 |
+
"rewards/rejected": -0.0103251738473773,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 2.34,
|
| 1147 |
+
"learning_rate": 2.1965221780218173e-06,
|
| 1148 |
+
"logits/chosen": 0.16370469331741333,
|
| 1149 |
+
"logits/rejected": 0.34310778975486755,
|
| 1150 |
+
"logps/chosen": -260.98602294921875,
|
| 1151 |
+
"logps/rejected": -223.19790649414062,
|
| 1152 |
+
"loss": 1666.0893,
|
| 1153 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 1154 |
+
"rewards/chosen": 0.14241552352905273,
|
| 1155 |
+
"rewards/margins": 0.14847032725811005,
|
| 1156 |
+
"rewards/rejected": -0.006054792553186417,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 2.37,
|
| 1161 |
+
"learning_rate": 2.1272296773879107e-06,
|
| 1162 |
+
"logits/chosen": 0.14960137009620667,
|
| 1163 |
+
"logits/rejected": 0.40142565965652466,
|
| 1164 |
+
"logps/chosen": -237.9215087890625,
|
| 1165 |
+
"logps/rejected": -199.8916015625,
|
| 1166 |
+
"loss": 1738.3658,
|
| 1167 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1168 |
+
"rewards/chosen": 0.13602004945278168,
|
| 1169 |
+
"rewards/margins": 0.14607930183410645,
|
| 1170 |
+
"rewards/rejected": -0.010059243068099022,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 2.4,
|
| 1175 |
+
"learning_rate": 2.058228888044788e-06,
|
| 1176 |
+
"logits/chosen": 0.2741914689540863,
|
| 1177 |
+
"logits/rejected": 0.3872790038585663,
|
| 1178 |
+
"logps/chosen": -223.29269409179688,
|
| 1179 |
+
"logps/rejected": -191.80650329589844,
|
| 1180 |
+
"loss": 1910.0223,
|
| 1181 |
+
"rewards/accuracies": 0.606249988079071,
|
| 1182 |
+
"rewards/chosen": 0.10628987848758698,
|
| 1183 |
+
"rewards/margins": 0.11196577548980713,
|
| 1184 |
+
"rewards/rejected": -0.00567590119317174,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 2.43,
|
| 1189 |
+
"learning_rate": 1.989573806536978e-06,
|
| 1190 |
+
"logits/chosen": 0.18802039325237274,
|
| 1191 |
+
"logits/rejected": 0.4464651048183441,
|
| 1192 |
+
"logps/chosen": -256.7940979003906,
|
| 1193 |
+
"logps/rejected": -208.06430053710938,
|
| 1194 |
+
"loss": 1710.393,
|
| 1195 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1196 |
+
"rewards/chosen": 0.15197846293449402,
|
| 1197 |
+
"rewards/margins": 0.1547323763370514,
|
| 1198 |
+
"rewards/rejected": -0.002753905486315489,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 2.46,
|
| 1203 |
+
"learning_rate": 1.921318158875459e-06,
|
| 1204 |
+
"logits/chosen": 0.2419842779636383,
|
| 1205 |
+
"logits/rejected": 0.4348203241825104,
|
| 1206 |
+
"logps/chosen": -249.1728973388672,
|
| 1207 |
+
"logps/rejected": -207.82229614257812,
|
| 1208 |
+
"loss": 1884.0729,
|
| 1209 |
+
"rewards/accuracies": 0.65625,
|
| 1210 |
+
"rewards/chosen": 0.12094120681285858,
|
| 1211 |
+
"rewards/margins": 0.12061796337366104,
|
| 1212 |
+
"rewards/rejected": 0.0003232499584555626,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 2.5,
|
| 1217 |
+
"learning_rate": 1.8535153584943915e-06,
|
| 1218 |
+
"logits/chosen": 0.2110445499420166,
|
| 1219 |
+
"logits/rejected": 0.4180677533149719,
|
| 1220 |
+
"logps/chosen": -257.97967529296875,
|
| 1221 |
+
"logps/rejected": -223.7866668701172,
|
| 1222 |
+
"loss": 1903.2082,
|
| 1223 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 1224 |
+
"rewards/chosen": 0.12379207462072372,
|
| 1225 |
+
"rewards/margins": 0.1165727972984314,
|
| 1226 |
+
"rewards/rejected": 0.007219274528324604,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.53,
|
| 1231 |
+
"learning_rate": 1.7862184644524422e-06,
|
| 1232 |
+
"logits/chosen": 0.2049965113401413,
|
| 1233 |
+
"logits/rejected": 0.4525831639766693,
|
| 1234 |
+
"logps/chosen": -256.8290710449219,
|
| 1235 |
+
"logps/rejected": -224.7571563720703,
|
| 1236 |
+
"loss": 1728.5514,
|
| 1237 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1238 |
+
"rewards/chosen": 0.14060300588607788,
|
| 1239 |
+
"rewards/margins": 0.1399126946926117,
|
| 1240 |
+
"rewards/rejected": 0.0006903231260366738,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.56,
|
| 1245 |
+
"learning_rate": 1.7194801399114471e-06,
|
| 1246 |
+
"logits/chosen": 0.1295299082994461,
|
| 1247 |
+
"logits/rejected": 0.38342705368995667,
|
| 1248 |
+
"logps/chosen": -265.95697021484375,
|
| 1249 |
+
"logps/rejected": -224.35922241210938,
|
| 1250 |
+
"loss": 1711.1711,
|
| 1251 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1252 |
+
"rewards/chosen": 0.1655883491039276,
|
| 1253 |
+
"rewards/margins": 0.14990000426769257,
|
| 1254 |
+
"rewards/rejected": 0.015688356012105942,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 2.56,
|
| 1259 |
+
"eval_logits/chosen": 0.37592813372612,
|
| 1260 |
+
"eval_logits/rejected": 0.44024261832237244,
|
| 1261 |
+
"eval_logps/chosen": -271.4833984375,
|
| 1262 |
+
"eval_logps/rejected": -248.64402770996094,
|
| 1263 |
+
"eval_loss": 2536.2451171875,
|
| 1264 |
+
"eval_rewards/accuracies": 0.5145000219345093,
|
| 1265 |
+
"eval_rewards/chosen": -0.14869214594364166,
|
| 1266 |
+
"eval_rewards/margins": 0.0023655793629586697,
|
| 1267 |
+
"eval_rewards/rejected": -0.15105770528316498,
|
| 1268 |
+
"eval_runtime": 412.1594,
|
| 1269 |
+
"eval_samples_per_second": 4.852,
|
| 1270 |
+
"eval_steps_per_second": 1.213,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 2.59,
|
| 1275 |
+
"learning_rate": 1.6533526109248632e-06,
|
| 1276 |
+
"logits/chosen": 0.22814953327178955,
|
| 1277 |
+
"logits/rejected": 0.44856709241867065,
|
| 1278 |
+
"logps/chosen": -264.6068115234375,
|
| 1279 |
+
"logps/rejected": -218.6339874267578,
|
| 1280 |
+
"loss": 1694.8418,
|
| 1281 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1282 |
+
"rewards/chosen": 0.1343715339899063,
|
| 1283 |
+
"rewards/margins": 0.15012042224407196,
|
| 1284 |
+
"rewards/rejected": -0.015748897567391396,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 2.62,
|
| 1289 |
+
"learning_rate": 1.5878876255682951e-06,
|
| 1290 |
+
"logits/chosen": 0.25933316349983215,
|
| 1291 |
+
"logits/rejected": 0.428168922662735,
|
| 1292 |
+
"logps/chosen": -271.8929748535156,
|
| 1293 |
+
"logps/rejected": -222.19677734375,
|
| 1294 |
+
"loss": 1799.3105,
|
| 1295 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1296 |
+
"rewards/chosen": 0.13050588965415955,
|
| 1297 |
+
"rewards/margins": 0.12607057392597198,
|
| 1298 |
+
"rewards/rejected": 0.00443531759083271,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.66,
|
| 1303 |
+
"learning_rate": 1.5231364134440485e-06,
|
| 1304 |
+
"logits/chosen": 0.16305533051490784,
|
| 1305 |
+
"logits/rejected": 0.37009209394454956,
|
| 1306 |
+
"logps/chosen": -239.7517547607422,
|
| 1307 |
+
"logps/rejected": -201.59359741210938,
|
| 1308 |
+
"loss": 1877.6279,
|
| 1309 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 1310 |
+
"rewards/chosen": 0.13841071724891663,
|
| 1311 |
+
"rewards/margins": 0.12821929156780243,
|
| 1312 |
+
"rewards/rejected": 0.0101914182305336,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 2.69,
|
| 1317 |
+
"learning_rate": 1.4591496455914292e-06,
|
| 1318 |
+
"logits/chosen": 0.23846232891082764,
|
| 1319 |
+
"logits/rejected": 0.44133177399635315,
|
| 1320 |
+
"logps/chosen": -269.0438232421875,
|
| 1321 |
+
"logps/rejected": -235.73202514648438,
|
| 1322 |
+
"loss": 1826.1201,
|
| 1323 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1324 |
+
"rewards/chosen": 0.12339715659618378,
|
| 1325 |
+
"rewards/margins": 0.12225867807865143,
|
| 1326 |
+
"rewards/rejected": 0.0011384893441572785,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 2.72,
|
| 1331 |
+
"learning_rate": 1.395977394834132e-06,
|
| 1332 |
+
"logits/chosen": 0.2057628184556961,
|
| 1333 |
+
"logits/rejected": 0.457078218460083,
|
| 1334 |
+
"logps/chosen": -222.45474243164062,
|
| 1335 |
+
"logps/rejected": -178.09011840820312,
|
| 1336 |
+
"loss": 1674.3545,
|
| 1337 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1338 |
+
"rewards/chosen": 0.14612336456775665,
|
| 1339 |
+
"rewards/margins": 0.1515127718448639,
|
| 1340 |
+
"rewards/rejected": -0.005389401223510504,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.75,
|
| 1345 |
+
"learning_rate": 1.3336690965957733e-06,
|
| 1346 |
+
"logits/chosen": 0.23303177952766418,
|
| 1347 |
+
"logits/rejected": 0.38948389887809753,
|
| 1348 |
+
"logps/chosen": -276.028564453125,
|
| 1349 |
+
"logps/rejected": -242.33120727539062,
|
| 1350 |
+
"loss": 1734.6137,
|
| 1351 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1352 |
+
"rewards/chosen": 0.13224345445632935,
|
| 1353 |
+
"rewards/margins": 0.1367109715938568,
|
| 1354 |
+
"rewards/rejected": -0.004467502236366272,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 2.78,
|
| 1359 |
+
"learning_rate": 1.2722735102142192e-06,
|
| 1360 |
+
"logits/chosen": 0.18385744094848633,
|
| 1361 |
+
"logits/rejected": 0.483973890542984,
|
| 1362 |
+
"logps/chosen": -234.6072540283203,
|
| 1363 |
+
"logps/rejected": -188.79147338867188,
|
| 1364 |
+
"loss": 1524.1409,
|
| 1365 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1366 |
+
"rewards/chosen": 0.1483391523361206,
|
| 1367 |
+
"rewards/margins": 0.16734978556632996,
|
| 1368 |
+
"rewards/rejected": -0.019010629504919052,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 2.82,
|
| 1373 |
+
"learning_rate": 1.2118386807849733e-06,
|
| 1374 |
+
"logits/chosen": 0.23927922546863556,
|
| 1375 |
+
"logits/rejected": 0.39777225255966187,
|
| 1376 |
+
"logps/chosen": -247.7241668701172,
|
| 1377 |
+
"logps/rejected": -209.2562255859375,
|
| 1378 |
+
"loss": 1804.3422,
|
| 1379 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 1380 |
+
"rewards/chosen": 0.11453523486852646,
|
| 1381 |
+
"rewards/margins": 0.13239163160324097,
|
| 1382 |
+
"rewards/rejected": -0.0178564190864563,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.85,
|
| 1387 |
+
"learning_rate": 1.1524119015635116e-06,
|
| 1388 |
+
"logits/chosen": 0.14219129085540771,
|
| 1389 |
+
"logits/rejected": 0.3456658124923706,
|
| 1390 |
+
"logps/chosen": -244.8931121826172,
|
| 1391 |
+
"logps/rejected": -194.36337280273438,
|
| 1392 |
+
"loss": 1655.0236,
|
| 1393 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1394 |
+
"rewards/chosen": 0.17296788096427917,
|
| 1395 |
+
"rewards/margins": 0.17738190293312073,
|
| 1396 |
+
"rewards/rejected": -0.004414013121277094,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.88,
|
| 1401 |
+
"learning_rate": 1.0940396769559584e-06,
|
| 1402 |
+
"logits/chosen": 0.23592603206634521,
|
| 1403 |
+
"logits/rejected": 0.3958457410335541,
|
| 1404 |
+
"logps/chosen": -242.33761596679688,
|
| 1405 |
+
"logps/rejected": -191.34255981445312,
|
| 1406 |
+
"loss": 1894.4447,
|
| 1407 |
+
"rewards/accuracies": 0.625,
|
| 1408 |
+
"rewards/chosen": 0.11816737800836563,
|
| 1409 |
+
"rewards/margins": 0.1154218316078186,
|
| 1410 |
+
"rewards/rejected": 0.002745547564700246,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 2.88,
|
| 1415 |
+
"eval_logits/chosen": 0.3617529273033142,
|
| 1416 |
+
"eval_logits/rejected": 0.42615193128585815,
|
| 1417 |
+
"eval_logps/chosen": -272.0999755859375,
|
| 1418 |
+
"eval_logps/rejected": -249.2417449951172,
|
| 1419 |
+
"eval_loss": 2542.6298828125,
|
| 1420 |
+
"eval_rewards/accuracies": 0.5235000252723694,
|
| 1421 |
+
"eval_rewards/chosen": -0.15485772490501404,
|
| 1422 |
+
"eval_rewards/margins": 0.002177263842895627,
|
| 1423 |
+
"eval_rewards/rejected": -0.1570349782705307,
|
| 1424 |
+
"eval_runtime": 411.8184,
|
| 1425 |
+
"eval_samples_per_second": 4.857,
|
| 1426 |
+
"eval_steps_per_second": 1.214,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 2.91,
|
| 1431 |
+
"learning_rate": 1.036767686127079e-06,
|
| 1432 |
+
"logits/chosen": 0.26092058420181274,
|
| 1433 |
+
"logits/rejected": 0.4202706217765808,
|
| 1434 |
+
"logps/chosen": -271.7019958496094,
|
| 1435 |
+
"logps/rejected": -229.44772338867188,
|
| 1436 |
+
"loss": 1612.1519,
|
| 1437 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1438 |
+
"rewards/chosen": 0.1469217836856842,
|
| 1439 |
+
"rewards/margins": 0.1562378704547882,
|
| 1440 |
+
"rewards/rejected": -0.009316088631749153,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.94,
|
| 1445 |
+
"learning_rate": 9.806407472540644e-07,
|
| 1446 |
+
"logits/chosen": 0.16799505054950714,
|
| 1447 |
+
"logits/rejected": 0.41259631514549255,
|
| 1448 |
+
"logps/chosen": -245.1134033203125,
|
| 1449 |
+
"logps/rejected": -193.49716186523438,
|
| 1450 |
+
"loss": 1752.974,
|
| 1451 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1452 |
+
"rewards/chosen": 0.1420082449913025,
|
| 1453 |
+
"rewards/margins": 0.16994443535804749,
|
| 1454 |
+
"rewards/rejected": -0.027936194092035294,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.98,
|
| 1459 |
+
"learning_rate": 9.257027824540823e-07,
|
| 1460 |
+
"logits/chosen": 0.15648850798606873,
|
| 1461 |
+
"logits/rejected": 0.3334013819694519,
|
| 1462 |
+
"logps/chosen": -237.5020294189453,
|
| 1463 |
+
"logps/rejected": -190.00733947753906,
|
| 1464 |
+
"loss": 1678.6684,
|
| 1465 |
+
"rewards/accuracies": 0.75,
|
| 1466 |
+
"rewards/chosen": 0.1168346032500267,
|
| 1467 |
+
"rewards/margins": 0.14195187389850616,
|
| 1468 |
+
"rewards/rejected": -0.02511727437376976,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 3.01,
|
| 1473 |
+
"learning_rate": 8.719967834130385e-07,
|
| 1474 |
+
"logits/chosen": 0.17559798061847687,
|
| 1475 |
+
"logits/rejected": 0.38583293557167053,
|
| 1476 |
+
"logps/chosen": -226.3419647216797,
|
| 1477 |
+
"logps/rejected": -192.8260498046875,
|
| 1478 |
+
"loss": 1782.9469,
|
| 1479 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1480 |
+
"rewards/chosen": 0.13087774813175201,
|
| 1481 |
+
"rewards/margins": 0.14290814101696014,
|
| 1482 |
+
"rewards/rejected": -0.012030376121401787,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 3.04,
|
| 1487 |
+
"learning_rate": 8.195647777424479e-07,
|
| 1488 |
+
"logits/chosen": 0.19569836556911469,
|
| 1489 |
+
"logits/rejected": 0.420011043548584,
|
| 1490 |
+
"logps/chosen": -243.2490692138672,
|
| 1491 |
+
"logps/rejected": -200.2803497314453,
|
| 1492 |
+
"loss": 1728.6527,
|
| 1493 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1494 |
+
"rewards/chosen": 0.1271076500415802,
|
| 1495 |
+
"rewards/margins": 0.14936277270317078,
|
| 1496 |
+
"rewards/rejected": -0.02225511521100998,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 3.07,
|
| 1501 |
+
"learning_rate": 7.684477960907422e-07,
|
| 1502 |
+
"logits/chosen": 0.2223655879497528,
|
| 1503 |
+
"logits/rejected": 0.36084768176078796,
|
| 1504 |
+
"logps/chosen": -259.83001708984375,
|
| 1505 |
+
"logps/rejected": -217.5656280517578,
|
| 1506 |
+
"loss": 1787.5324,
|
| 1507 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1508 |
+
"rewards/chosen": 0.13919194042682648,
|
| 1509 |
+
"rewards/margins": 0.12921075522899628,
|
| 1510 |
+
"rewards/rejected": 0.009981190785765648,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 3.1,
|
| 1515 |
+
"learning_rate": 7.186858400347455e-07,
|
| 1516 |
+
"logits/chosen": 0.15693505108356476,
|
| 1517 |
+
"logits/rejected": 0.35678160190582275,
|
| 1518 |
+
"logps/chosen": -266.93768310546875,
|
| 1519 |
+
"logps/rejected": -233.291259765625,
|
| 1520 |
+
"loss": 1748.1148,
|
| 1521 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1522 |
+
"rewards/chosen": 0.12252594530582428,
|
| 1523 |
+
"rewards/margins": 0.13253001868724823,
|
| 1524 |
+
"rewards/rejected": -0.010004105977714062,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 3.14,
|
| 1529 |
+
"learning_rate": 6.703178507764618e-07,
|
| 1530 |
+
"logits/chosen": 0.21335263550281525,
|
| 1531 |
+
"logits/rejected": 0.3721942901611328,
|
| 1532 |
+
"logps/chosen": -250.0565948486328,
|
| 1533 |
+
"logps/rejected": -213.24624633789062,
|
| 1534 |
+
"loss": 1792.84,
|
| 1535 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 1536 |
+
"rewards/chosen": 0.1454544961452484,
|
| 1537 |
+
"rewards/margins": 0.12651817500591278,
|
| 1538 |
+
"rewards/rejected": 0.018936317414045334,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 3.17,
|
| 1543 |
+
"learning_rate": 6.233816786696414e-07,
|
| 1544 |
+
"logits/chosen": 0.17327558994293213,
|
| 1545 |
+
"logits/rejected": 0.41259893774986267,
|
| 1546 |
+
"logps/chosen": -242.4098358154297,
|
| 1547 |
+
"logps/rejected": -191.53460693359375,
|
| 1548 |
+
"loss": 1636.6745,
|
| 1549 |
+
"rewards/accuracies": 0.6875,
|
| 1550 |
+
"rewards/chosen": 0.16031518578529358,
|
| 1551 |
+
"rewards/margins": 0.1668807417154312,
|
| 1552 |
+
"rewards/rejected": -0.006565576884895563,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 3.2,
|
| 1557 |
+
"learning_rate": 5.77914053600005e-07,
|
| 1558 |
+
"logits/chosen": 0.15771836042404175,
|
| 1559 |
+
"logits/rejected": 0.384365975856781,
|
| 1560 |
+
"logps/chosen": -247.66213989257812,
|
| 1561 |
+
"logps/rejected": -207.638427734375,
|
| 1562 |
+
"loss": 1798.5389,
|
| 1563 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1564 |
+
"rewards/chosen": 0.13780926167964935,
|
| 1565 |
+
"rewards/margins": 0.1279788464307785,
|
| 1566 |
+
"rewards/rejected": 0.00983042549341917,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 3.2,
|
| 1571 |
+
"eval_logits/chosen": 0.3559388518333435,
|
| 1572 |
+
"eval_logits/rejected": 0.42022594809532166,
|
| 1573 |
+
"eval_logps/chosen": -272.4200439453125,
|
| 1574 |
+
"eval_logps/rejected": -249.57797241210938,
|
| 1575 |
+
"eval_loss": 2542.728759765625,
|
| 1576 |
+
"eval_rewards/accuracies": 0.5205000042915344,
|
| 1577 |
+
"eval_rewards/chosen": -0.15805859863758087,
|
| 1578 |
+
"eval_rewards/margins": 0.0023383116349577904,
|
| 1579 |
+
"eval_rewards/rejected": -0.16039690375328064,
|
| 1580 |
+
"eval_runtime": 411.7713,
|
| 1581 |
+
"eval_samples_per_second": 4.857,
|
| 1582 |
+
"eval_steps_per_second": 1.214,
|
| 1583 |
+
"step": 1000
|
| 1584 |
+
},
|
| 1585 |
+
{
|
| 1586 |
+
"epoch": 3.23,
|
| 1587 |
+
"learning_rate": 5.339505562422851e-07,
|
| 1588 |
+
"logits/chosen": 0.18551324307918549,
|
| 1589 |
+
"logits/rejected": 0.3874700367450714,
|
| 1590 |
+
"logps/chosen": -258.8070983886719,
|
| 1591 |
+
"logps/rejected": -208.31655883789062,
|
| 1592 |
+
"loss": 1684.7502,
|
| 1593 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1594 |
+
"rewards/chosen": 0.13547667860984802,
|
| 1595 |
+
"rewards/margins": 0.14734682440757751,
|
| 1596 |
+
"rewards/rejected": -0.011870155110955238,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 3.26,
|
| 1601 |
+
"learning_rate": 4.915255902165734e-07,
|
| 1602 |
+
"logits/chosen": 0.16962969303131104,
|
| 1603 |
+
"logits/rejected": 0.3440442383289337,
|
| 1604 |
+
"logps/chosen": -257.40386962890625,
|
| 1605 |
+
"logps/rejected": -213.15682983398438,
|
| 1606 |
+
"loss": 1727.4756,
|
| 1607 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1608 |
+
"rewards/chosen": 0.1502576470375061,
|
| 1609 |
+
"rewards/margins": 0.15001319348812103,
|
| 1610 |
+
"rewards/rejected": 0.000244450056925416,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 3.3,
|
| 1615 |
+
"learning_rate": 4.506723551657879e-07,
|
| 1616 |
+
"logits/chosen": 0.08985213190317154,
|
| 1617 |
+
"logits/rejected": 0.3018781542778015,
|
| 1618 |
+
"logps/chosen": -238.0452117919922,
|
| 1619 |
+
"logps/rejected": -200.73129272460938,
|
| 1620 |
+
"loss": 1795.9609,
|
| 1621 |
+
"rewards/accuracies": 0.6875,
|
| 1622 |
+
"rewards/chosen": 0.1125258356332779,
|
| 1623 |
+
"rewards/margins": 0.11527595669031143,
|
| 1624 |
+
"rewards/rejected": -0.0027501187287271023,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 3.33,
|
| 1629 |
+
"learning_rate": 4.11422820775299e-07,
|
| 1630 |
+
"logits/chosen": 0.18122617900371552,
|
| 1631 |
+
"logits/rejected": 0.3776589035987854,
|
| 1632 |
+
"logps/chosen": -273.4202575683594,
|
| 1633 |
+
"logps/rejected": -214.76803588867188,
|
| 1634 |
+
"loss": 1575.2917,
|
| 1635 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1636 |
+
"rewards/chosen": 0.14816126227378845,
|
| 1637 |
+
"rewards/margins": 0.1801377683877945,
|
| 1638 |
+
"rewards/rejected": -0.03197649493813515,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 3.36,
|
| 1643 |
+
"learning_rate": 3.7380770175506397e-07,
|
| 1644 |
+
"logits/chosen": 0.18917641043663025,
|
| 1645 |
+
"logits/rejected": 0.3701549172401428,
|
| 1646 |
+
"logps/chosen": -236.4595184326172,
|
| 1647 |
+
"logps/rejected": -196.47378540039062,
|
| 1648 |
+
"loss": 1757.2316,
|
| 1649 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1650 |
+
"rewards/chosen": 0.13911323249340057,
|
| 1651 |
+
"rewards/margins": 0.14288835227489471,
|
| 1652 |
+
"rewards/rejected": -0.0037751286290585995,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 3.39,
|
| 1657 |
+
"learning_rate": 3.3785643380384063e-07,
|
| 1658 |
+
"logits/chosen": 0.22518393397331238,
|
| 1659 |
+
"logits/rejected": 0.4300554394721985,
|
| 1660 |
+
"logps/chosen": -264.11102294921875,
|
| 1661 |
+
"logps/rejected": -221.0089569091797,
|
| 1662 |
+
"loss": 1780.259,
|
| 1663 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1664 |
+
"rewards/chosen": 0.12094493955373764,
|
| 1665 |
+
"rewards/margins": 0.13927613198757172,
|
| 1666 |
+
"rewards/rejected": -0.018331199884414673,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 3.42,
|
| 1671 |
+
"learning_rate": 3.0359715057429186e-07,
|
| 1672 |
+
"logits/chosen": 0.15539857745170593,
|
| 1673 |
+
"logits/rejected": 0.330243319272995,
|
| 1674 |
+
"logps/chosen": -249.2795867919922,
|
| 1675 |
+
"logps/rejected": -213.9594268798828,
|
| 1676 |
+
"loss": 1857.3367,
|
| 1677 |
+
"rewards/accuracies": 0.668749988079071,
|
| 1678 |
+
"rewards/chosen": 0.11874745041131973,
|
| 1679 |
+
"rewards/margins": 0.1254003345966339,
|
| 1680 |
+
"rewards/rejected": -0.0066528706811368465,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 3.46,
|
| 1685 |
+
"learning_rate": 2.710566616570048e-07,
|
| 1686 |
+
"logits/chosen": 0.22669236361980438,
|
| 1687 |
+
"logits/rejected": 0.3851965069770813,
|
| 1688 |
+
"logps/chosen": -243.06094360351562,
|
| 1689 |
+
"logps/rejected": -207.26522827148438,
|
| 1690 |
+
"loss": 1802.6373,
|
| 1691 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1692 |
+
"rewards/chosen": 0.11891194432973862,
|
| 1693 |
+
"rewards/margins": 0.1296943724155426,
|
| 1694 |
+
"rewards/rejected": -0.010782415047287941,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 3.49,
|
| 1699 |
+
"learning_rate": 2.40260431600654e-07,
|
| 1700 |
+
"logits/chosen": 0.11191525310277939,
|
| 1701 |
+
"logits/rejected": 0.3566216826438904,
|
| 1702 |
+
"logps/chosen": -258.7408142089844,
|
| 1703 |
+
"logps/rejected": -233.2893524169922,
|
| 1704 |
+
"loss": 1846.4627,
|
| 1705 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1706 |
+
"rewards/chosen": 0.13531699776649475,
|
| 1707 |
+
"rewards/margins": 0.1306256651878357,
|
| 1708 |
+
"rewards/rejected": 0.004691324662417173,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 3.52,
|
| 1713 |
+
"learning_rate": 2.1123255998472952e-07,
|
| 1714 |
+
"logits/chosen": 0.13580968976020813,
|
| 1715 |
+
"logits/rejected": 0.3724610209465027,
|
| 1716 |
+
"logps/chosen": -238.85391235351562,
|
| 1717 |
+
"logps/rejected": -202.35397338867188,
|
| 1718 |
+
"loss": 1834.9711,
|
| 1719 |
+
"rewards/accuracies": 0.65625,
|
| 1720 |
+
"rewards/chosen": 0.1293712556362152,
|
| 1721 |
+
"rewards/margins": 0.12213204801082611,
|
| 1722 |
+
"rewards/rejected": 0.007239216007292271,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 3.52,
|
| 1727 |
+
"eval_logits/chosen": 0.3531816899776459,
|
| 1728 |
+
"eval_logits/rejected": 0.417733371257782,
|
| 1729 |
+
"eval_logps/chosen": -272.4703369140625,
|
| 1730 |
+
"eval_logps/rejected": -249.63449096679688,
|
| 1731 |
+
"eval_loss": 2542.2373046875,
|
| 1732 |
+
"eval_rewards/accuracies": 0.5205000042915344,
|
| 1733 |
+
"eval_rewards/chosen": -0.15856170654296875,
|
| 1734 |
+
"eval_rewards/margins": 0.0024005102459341288,
|
| 1735 |
+
"eval_rewards/rejected": -0.16096222400665283,
|
| 1736 |
+
"eval_runtime": 411.8208,
|
| 1737 |
+
"eval_samples_per_second": 4.856,
|
| 1738 |
+
"eval_steps_per_second": 1.214,
|
| 1739 |
+
"step": 1100
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 3.55,
|
| 1743 |
+
"learning_rate": 1.8399576256041525e-07,
|
| 1744 |
+
"logits/chosen": 0.27907487750053406,
|
| 1745 |
+
"logits/rejected": 0.39952999353408813,
|
| 1746 |
+
"logps/chosen": -293.59210205078125,
|
| 1747 |
+
"logps/rejected": -251.298583984375,
|
| 1748 |
+
"loss": 1979.5523,
|
| 1749 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 1750 |
+
"rewards/chosen": 0.12343426048755646,
|
| 1751 |
+
"rewards/margins": 0.1068718284368515,
|
| 1752 |
+
"rewards/rejected": 0.016562417149543762,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 3.58,
|
| 1757 |
+
"learning_rate": 1.58571353474391e-07,
|
| 1758 |
+
"logits/chosen": 0.22743897140026093,
|
| 1759 |
+
"logits/rejected": 0.40824824571609497,
|
| 1760 |
+
"logps/chosen": -248.27365112304688,
|
| 1761 |
+
"logps/rejected": -217.0345916748047,
|
| 1762 |
+
"loss": 1971.9447,
|
| 1763 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 1764 |
+
"rewards/chosen": 0.10497580468654633,
|
| 1765 |
+
"rewards/margins": 0.10696852207183838,
|
| 1766 |
+
"rewards/rejected": -0.0019927166868001223,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 3.62,
|
| 1771 |
+
"learning_rate": 1.3497922858944857e-07,
|
| 1772 |
+
"logits/chosen": 0.21788544952869415,
|
| 1773 |
+
"logits/rejected": 0.3408287763595581,
|
| 1774 |
+
"logps/chosen": -261.25634765625,
|
| 1775 |
+
"logps/rejected": -226.1531982421875,
|
| 1776 |
+
"loss": 1813.0029,
|
| 1777 |
+
"rewards/accuracies": 0.668749988079071,
|
| 1778 |
+
"rewards/chosen": 0.12613138556480408,
|
| 1779 |
+
"rewards/margins": 0.13940462470054626,
|
| 1780 |
+
"rewards/rejected": -0.013273234479129314,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 3.65,
|
| 1785 |
+
"learning_rate": 1.1323784991499471e-07,
|
| 1786 |
+
"logits/chosen": 0.17835107445716858,
|
| 1787 |
+
"logits/rejected": 0.41546598076820374,
|
| 1788 |
+
"logps/chosen": -254.427978515625,
|
| 1789 |
+
"logps/rejected": -208.0858917236328,
|
| 1790 |
+
"loss": 1688.9838,
|
| 1791 |
+
"rewards/accuracies": 0.75,
|
| 1792 |
+
"rewards/chosen": 0.13813026249408722,
|
| 1793 |
+
"rewards/margins": 0.1528724581003189,
|
| 1794 |
+
"rewards/rejected": -0.014742200262844563,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 3.68,
|
| 1799 |
+
"learning_rate": 9.336423115961002e-08,
|
| 1800 |
+
"logits/chosen": 0.13116911053657532,
|
| 1801 |
+
"logits/rejected": 0.34874704480171204,
|
| 1802 |
+
"logps/chosen": -251.4066162109375,
|
| 1803 |
+
"logps/rejected": -205.0511016845703,
|
| 1804 |
+
"loss": 1693.0176,
|
| 1805 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1806 |
+
"rewards/chosen": 0.14480547606945038,
|
| 1807 |
+
"rewards/margins": 0.1530844271183014,
|
| 1808 |
+
"rewards/rejected": -0.008278947323560715,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.71,
|
| 1813 |
+
"learning_rate": 7.537392441697793e-08,
|
| 1814 |
+
"logits/chosen": 0.1982724368572235,
|
| 1815 |
+
"logits/rejected": 0.3957718014717102,
|
| 1816 |
+
"logps/chosen": -265.66033935546875,
|
| 1817 |
+
"logps/rejected": -223.394775390625,
|
| 1818 |
+
"loss": 1694.5328,
|
| 1819 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1820 |
+
"rewards/chosen": 0.12068979442119598,
|
| 1821 |
+
"rewards/margins": 0.15094828605651855,
|
| 1822 |
+
"rewards/rejected": -0.030258500948548317,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 3.74,
|
| 1827 |
+
"learning_rate": 5.928100799559938e-08,
|
| 1828 |
+
"logits/chosen": 0.2445194274187088,
|
| 1829 |
+
"logits/rejected": 0.34973251819610596,
|
| 1830 |
+
"logps/chosen": -266.8522033691406,
|
| 1831 |
+
"logps/rejected": -214.7125701904297,
|
| 1832 |
+
"loss": 1810.5473,
|
| 1833 |
+
"rewards/accuracies": 0.65625,
|
| 1834 |
+
"rewards/chosen": 0.12299786508083344,
|
| 1835 |
+
"rewards/margins": 0.13123488426208496,
|
| 1836 |
+
"rewards/rejected": -0.008237012661993504,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 3.78,
|
| 1841 |
+
"learning_rate": 4.5098075401815435e-08,
|
| 1842 |
+
"logits/chosen": 0.17370513081550598,
|
| 1843 |
+
"logits/rejected": 0.44302305579185486,
|
| 1844 |
+
"logps/chosen": -251.4030303955078,
|
| 1845 |
+
"logps/rejected": -217.8542938232422,
|
| 1846 |
+
"loss": 1799.0785,
|
| 1847 |
+
"rewards/accuracies": 0.65625,
|
| 1848 |
+
"rewards/chosen": 0.14102791249752045,
|
| 1849 |
+
"rewards/margins": 0.14266334474086761,
|
| 1850 |
+
"rewards/rejected": -0.0016354346880689263,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 3.81,
|
| 1855 |
+
"learning_rate": 3.283622548476445e-08,
|
| 1856 |
+
"logits/chosen": 0.2375222146511078,
|
| 1857 |
+
"logits/rejected": 0.45388326048851013,
|
| 1858 |
+
"logps/chosen": -243.04116821289062,
|
| 1859 |
+
"logps/rejected": -203.55050659179688,
|
| 1860 |
+
"loss": 1673.1176,
|
| 1861 |
+
"rewards/accuracies": 0.643750011920929,
|
| 1862 |
+
"rewards/chosen": 0.13139726221561432,
|
| 1863 |
+
"rewards/margins": 0.14136573672294617,
|
| 1864 |
+
"rewards/rejected": -0.009968474507331848,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 3.84,
|
| 1869 |
+
"learning_rate": 2.250505375098161e-08,
|
| 1870 |
+
"logits/chosen": 0.28876617550849915,
|
| 1871 |
+
"logits/rejected": 0.42698416113853455,
|
| 1872 |
+
"logps/chosen": -248.3870086669922,
|
| 1873 |
+
"logps/rejected": -211.6997833251953,
|
| 1874 |
+
"loss": 1765.5148,
|
| 1875 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1876 |
+
"rewards/chosen": 0.1342637687921524,
|
| 1877 |
+
"rewards/margins": 0.1372515708208084,
|
| 1878 |
+
"rewards/rejected": -0.002987801330164075,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 3.84,
|
| 1883 |
+
"eval_logits/chosen": 0.35145023465156555,
|
| 1884 |
+
"eval_logits/rejected": 0.41600194573402405,
|
| 1885 |
+
"eval_logps/chosen": -272.50103759765625,
|
| 1886 |
+
"eval_logps/rejected": -249.63572692871094,
|
| 1887 |
+
"eval_loss": 2546.17138671875,
|
| 1888 |
+
"eval_rewards/accuracies": 0.5220000147819519,
|
| 1889 |
+
"eval_rewards/chosen": -0.15886859595775604,
|
| 1890 |
+
"eval_rewards/margins": 0.00210593082010746,
|
| 1891 |
+
"eval_rewards/rejected": -0.16097451746463776,
|
| 1892 |
+
"eval_runtime": 411.6701,
|
| 1893 |
+
"eval_samples_per_second": 4.858,
|
| 1894 |
+
"eval_steps_per_second": 1.215,
|
| 1895 |
+
"step": 1200
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 3.87,
|
| 1899 |
+
"learning_rate": 1.4112644855438228e-08,
|
| 1900 |
+
"logits/chosen": 0.18522273004055023,
|
| 1901 |
+
"logits/rejected": 0.3528333306312561,
|
| 1902 |
+
"logps/chosen": -252.374755859375,
|
| 1903 |
+
"logps/rejected": -213.50112915039062,
|
| 1904 |
+
"loss": 1695.0207,
|
| 1905 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1906 |
+
"rewards/chosen": 0.1518661379814148,
|
| 1907 |
+
"rewards/margins": 0.1503283679485321,
|
| 1908 |
+
"rewards/rejected": 0.001537766307592392,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 3.9,
|
| 1913 |
+
"learning_rate": 7.665566274897007e-09,
|
| 1914 |
+
"logits/chosen": 0.14011794328689575,
|
| 1915 |
+
"logits/rejected": 0.45459121465682983,
|
| 1916 |
+
"logps/chosen": -269.1119079589844,
|
| 1917 |
+
"logps/rejected": -213.98684692382812,
|
| 1918 |
+
"loss": 1599.6027,
|
| 1919 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1920 |
+
"rewards/chosen": 0.1741984784603119,
|
| 1921 |
+
"rewards/margins": 0.17417994141578674,
|
| 1922 |
+
"rewards/rejected": 1.8526614439906552e-05,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 3.94,
|
| 1927 |
+
"learning_rate": 3.1688631685364292e-09,
|
| 1928 |
+
"logits/chosen": 0.11093461513519287,
|
| 1929 |
+
"logits/rejected": 0.3354471027851105,
|
| 1930 |
+
"logps/chosen": -240.45419311523438,
|
| 1931 |
+
"logps/rejected": -200.0811004638672,
|
| 1932 |
+
"loss": 1762.0594,
|
| 1933 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1934 |
+
"rewards/chosen": 0.14106154441833496,
|
| 1935 |
+
"rewards/margins": 0.14005112648010254,
|
| 1936 |
+
"rewards/rejected": 0.0010104203829541802,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 3.97,
|
| 1941 |
+
"learning_rate": 6.260544298619664e-10,
|
| 1942 |
+
"logits/chosen": 0.1383497416973114,
|
| 1943 |
+
"logits/rejected": 0.3753938674926758,
|
| 1944 |
+
"logps/chosen": -235.23385620117188,
|
| 1945 |
+
"logps/rejected": -191.31582641601562,
|
| 1946 |
+
"loss": 1685.3523,
|
| 1947 |
+
"rewards/accuracies": 0.71875,
|
| 1948 |
+
"rewards/chosen": 0.15916982293128967,
|
| 1949 |
+
"rewards/margins": 0.1575343906879425,
|
| 1950 |
+
"rewards/rejected": 0.0016354434192180634,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 3.99,
|
| 1955 |
+
"step": 1248,
|
| 1956 |
+
"total_flos": 0.0,
|
| 1957 |
+
"train_loss": 1918.4646684695513,
|
| 1958 |
+
"train_runtime": 14252.7073,
|
| 1959 |
+
"train_samples_per_second": 1.403,
|
| 1960 |
+
"train_steps_per_second": 0.088
|
| 1961 |
+
}
|
| 1962 |
+
],
|
| 1963 |
+
"logging_steps": 10,
|
| 1964 |
+
"max_steps": 1248,
|
| 1965 |
+
"num_input_tokens_seen": 0,
|
| 1966 |
+
"num_train_epochs": 4,
|
| 1967 |
+
"save_steps": 100,
|
| 1968 |
+
"total_flos": 0.0,
|
| 1969 |
+
"train_batch_size": 4,
|
| 1970 |
+
"trial_name": null,
|
| 1971 |
+
"trial_params": null
|
| 1972 |
+
}
|