Model save
Browse files- README.md +83 -0
- adapter_model.safetensors +1 -1
- all_results.json +21 -0
- eval_results.json +16 -0
- train_results.json +8 -0
- trainer_state.json +2310 -0
README.md
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: stabilityai/StableBeluga-13B
|
| 3 |
+
tags:
|
| 4 |
+
- generated_from_trainer
|
| 5 |
+
model-index:
|
| 6 |
+
- name: PE-13b-lora
|
| 7 |
+
results: []
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 11 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 12 |
+
|
| 13 |
+
# PE-13b-lora
|
| 14 |
+
|
| 15 |
+
This model is a fine-tuned version of [stabilityai/StableBeluga-13B](https://huggingface.co/stabilityai/StableBeluga-13B) on an unknown dataset.
|
| 16 |
+
It achieves the following results on the evaluation set:
|
| 17 |
+
- Loss: 0.5704
|
| 18 |
+
- Rewards/chosen: 0.1581
|
| 19 |
+
- Rewards/rejected: -0.1076
|
| 20 |
+
- Rewards/accuracies: 0.9472
|
| 21 |
+
- Rewards/margins: 0.2658
|
| 22 |
+
- Logps/rejected: -73.1769
|
| 23 |
+
- Logps/chosen: -90.4042
|
| 24 |
+
- Logits/rejected: -1.7758
|
| 25 |
+
- Logits/chosen: -2.0462
|
| 26 |
+
|
| 27 |
+
## Model description
|
| 28 |
+
|
| 29 |
+
More information needed
|
| 30 |
+
|
| 31 |
+
## Intended uses & limitations
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Training and evaluation data
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training procedure
|
| 40 |
+
|
| 41 |
+
### Training hyperparameters
|
| 42 |
+
|
| 43 |
+
The following hyperparameters were used during training:
|
| 44 |
+
- learning_rate: 5e-07
|
| 45 |
+
- train_batch_size: 6
|
| 46 |
+
- eval_batch_size: 4
|
| 47 |
+
- seed: 42
|
| 48 |
+
- distributed_type: multi-GPU
|
| 49 |
+
- num_devices: 8
|
| 50 |
+
- gradient_accumulation_steps: 2
|
| 51 |
+
- total_train_batch_size: 96
|
| 52 |
+
- total_eval_batch_size: 32
|
| 53 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 54 |
+
- lr_scheduler_type: linear
|
| 55 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 56 |
+
- num_epochs: 1
|
| 57 |
+
|
| 58 |
+
### Training results
|
| 59 |
+
|
| 60 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 61 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 62 |
+
| 0.693 | 0.07 | 100 | 0.6933 | -0.0008 | -0.0005 | 0.4889 | -0.0003 | -72.1053 | -91.9932 | -1.7861 | -2.0525 |
|
| 63 |
+
| 0.69 | 0.14 | 200 | 0.6901 | 0.0031 | -0.0015 | 0.5611 | 0.0046 | -72.1153 | -91.9544 | -1.7859 | -2.0524 |
|
| 64 |
+
| 0.6842 | 0.21 | 300 | 0.6832 | 0.0139 | -0.0056 | 0.6917 | 0.0195 | -72.1567 | -91.8467 | -1.7847 | -2.0513 |
|
| 65 |
+
| 0.672 | 0.27 | 400 | 0.6718 | 0.0281 | -0.0131 | 0.8250 | 0.0412 | -72.2312 | -91.7049 | -1.7836 | -2.0504 |
|
| 66 |
+
| 0.6563 | 0.34 | 500 | 0.6575 | 0.0498 | -0.0211 | 0.8861 | 0.0709 | -72.3116 | -91.4876 | -1.7821 | -2.0494 |
|
| 67 |
+
| 0.6437 | 0.41 | 600 | 0.6416 | 0.0705 | -0.0340 | 0.9111 | 0.1044 | -72.4401 | -91.2810 | -1.7807 | -2.0486 |
|
| 68 |
+
| 0.6261 | 0.48 | 700 | 0.6277 | 0.0885 | -0.0435 | 0.9250 | 0.1320 | -72.5355 | -91.1010 | -1.7796 | -2.0478 |
|
| 69 |
+
| 0.6117 | 0.55 | 800 | 0.6127 | 0.1097 | -0.0567 | 0.9222 | 0.1664 | -72.6675 | -90.8891 | -1.7786 | -2.0474 |
|
| 70 |
+
| 0.6002 | 0.62 | 900 | 0.6019 | 0.1226 | -0.0683 | 0.9278 | 0.1909 | -72.7836 | -90.7598 | -1.7777 | -2.0468 |
|
| 71 |
+
| 0.5912 | 0.68 | 1000 | 0.5912 | 0.1344 | -0.0805 | 0.9333 | 0.2148 | -72.9053 | -90.6422 | -1.7770 | -2.0466 |
|
| 72 |
+
| 0.5822 | 0.75 | 1100 | 0.5822 | 0.1441 | -0.0909 | 0.9472 | 0.2350 | -73.0092 | -90.5447 | -1.7763 | -2.0462 |
|
| 73 |
+
| 0.5789 | 0.82 | 1200 | 0.5759 | 0.1517 | -0.0992 | 0.9333 | 0.2509 | -73.0923 | -90.4690 | -1.7763 | -2.0465 |
|
| 74 |
+
| 0.5689 | 0.89 | 1300 | 0.5722 | 0.1555 | -0.1033 | 0.9500 | 0.2588 | -73.1332 | -90.4305 | -1.7762 | -2.0465 |
|
| 75 |
+
| 0.5694 | 0.96 | 1400 | 0.5702 | 0.1579 | -0.1066 | 0.9417 | 0.2644 | -73.1662 | -90.4070 | -1.7761 | -2.0465 |
|
| 76 |
+
|
| 77 |
+
|
| 78 |
+
### Framework versions
|
| 79 |
+
|
| 80 |
+
- Transformers 4.35.0
|
| 81 |
+
- Pytorch 2.1.1+cu121
|
| 82 |
+
- Datasets 2.14.6
|
| 83 |
+
- Tokenizers 0.14.1
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 209758976
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f9a1af2ba9753d0ecbe37035f3e479048c69509166da286ba458b139a0ec623
|
| 3 |
size 209758976
|
all_results.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -2.0462207794189453,
|
| 4 |
+
"eval_logits/rejected": -1.7758138179779053,
|
| 5 |
+
"eval_logps/chosen": -90.40424346923828,
|
| 6 |
+
"eval_logps/rejected": -73.17691802978516,
|
| 7 |
+
"eval_loss": 0.5703843832015991,
|
| 8 |
+
"eval_rewards/accuracies": 0.9472222328186035,
|
| 9 |
+
"eval_rewards/chosen": 0.1581471860408783,
|
| 10 |
+
"eval_rewards/margins": 0.2657936215400696,
|
| 11 |
+
"eval_rewards/rejected": -0.10764642059803009,
|
| 12 |
+
"eval_runtime": 118.208,
|
| 13 |
+
"eval_samples": 2862,
|
| 14 |
+
"eval_samples_per_second": 24.212,
|
| 15 |
+
"eval_steps_per_second": 0.761,
|
| 16 |
+
"train_loss": 0.6280729855576607,
|
| 17 |
+
"train_runtime": 9689.6427,
|
| 18 |
+
"train_samples": 140201,
|
| 19 |
+
"train_samples_per_second": 14.469,
|
| 20 |
+
"train_steps_per_second": 0.151
|
| 21 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -2.0462207794189453,
|
| 4 |
+
"eval_logits/rejected": -1.7758138179779053,
|
| 5 |
+
"eval_logps/chosen": -90.40424346923828,
|
| 6 |
+
"eval_logps/rejected": -73.17691802978516,
|
| 7 |
+
"eval_loss": 0.5703843832015991,
|
| 8 |
+
"eval_rewards/accuracies": 0.9472222328186035,
|
| 9 |
+
"eval_rewards/chosen": 0.1581471860408783,
|
| 10 |
+
"eval_rewards/margins": 0.2657936215400696,
|
| 11 |
+
"eval_rewards/rejected": -0.10764642059803009,
|
| 12 |
+
"eval_runtime": 118.208,
|
| 13 |
+
"eval_samples": 2862,
|
| 14 |
+
"eval_samples_per_second": 24.212,
|
| 15 |
+
"eval_steps_per_second": 0.761
|
| 16 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6280729855576607,
|
| 4 |
+
"train_runtime": 9689.6427,
|
| 5 |
+
"train_samples": 140201,
|
| 6 |
+
"train_samples_per_second": 14.469,
|
| 7 |
+
"train_steps_per_second": 0.151
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2310 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.999657651489216,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 1460,
|
| 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": 3.424657534246575e-09,
|
| 14 |
+
"logits/chosen": -1.796067237854004,
|
| 15 |
+
"logits/rejected": -1.6250377893447876,
|
| 16 |
+
"logps/chosen": -84.08734130859375,
|
| 17 |
+
"logps/rejected": -66.90229797363281,
|
| 18 |
+
"loss": 0.6931,
|
| 19 |
+
"rewards/accuracies": 0.0,
|
| 20 |
+
"rewards/chosen": 0.0,
|
| 21 |
+
"rewards/margins": 0.0,
|
| 22 |
+
"rewards/rejected": 0.0,
|
| 23 |
+
"step": 1
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.01,
|
| 27 |
+
"learning_rate": 3.424657534246575e-08,
|
| 28 |
+
"logits/chosen": -1.7872660160064697,
|
| 29 |
+
"logits/rejected": -1.5217690467834473,
|
| 30 |
+
"logps/chosen": -91.57577514648438,
|
| 31 |
+
"logps/rejected": -78.510498046875,
|
| 32 |
+
"loss": 0.6935,
|
| 33 |
+
"rewards/accuracies": 0.4166666567325592,
|
| 34 |
+
"rewards/chosen": 0.00252585974521935,
|
| 35 |
+
"rewards/margins": 0.003410403151065111,
|
| 36 |
+
"rewards/rejected": -0.0008845434640534222,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.01,
|
| 41 |
+
"learning_rate": 6.84931506849315e-08,
|
| 42 |
+
"logits/chosen": -1.9197826385498047,
|
| 43 |
+
"logits/rejected": -1.6265367269515991,
|
| 44 |
+
"logps/chosen": -96.18563079833984,
|
| 45 |
+
"logps/rejected": -73.0329818725586,
|
| 46 |
+
"loss": 0.6941,
|
| 47 |
+
"rewards/accuracies": 0.4333333373069763,
|
| 48 |
+
"rewards/chosen": -0.0032726190984249115,
|
| 49 |
+
"rewards/margins": -0.004598576575517654,
|
| 50 |
+
"rewards/rejected": 0.0013259568950161338,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.02,
|
| 55 |
+
"learning_rate": 1.0273972602739725e-07,
|
| 56 |
+
"logits/chosen": -1.8969062566757202,
|
| 57 |
+
"logits/rejected": -1.5750768184661865,
|
| 58 |
+
"logps/chosen": -96.1051254272461,
|
| 59 |
+
"logps/rejected": -74.69762420654297,
|
| 60 |
+
"loss": 0.6929,
|
| 61 |
+
"rewards/accuracies": 0.5416666865348816,
|
| 62 |
+
"rewards/chosen": -0.0011650085216388106,
|
| 63 |
+
"rewards/margins": 0.001948213903233409,
|
| 64 |
+
"rewards/rejected": -0.0031132223084568977,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.03,
|
| 69 |
+
"learning_rate": 1.36986301369863e-07,
|
| 70 |
+
"logits/chosen": -1.9271103143692017,
|
| 71 |
+
"logits/rejected": -1.6277456283569336,
|
| 72 |
+
"logps/chosen": -96.0181884765625,
|
| 73 |
+
"logps/rejected": -79.47406005859375,
|
| 74 |
+
"loss": 0.6927,
|
| 75 |
+
"rewards/accuracies": 0.5166666507720947,
|
| 76 |
+
"rewards/chosen": -0.001698245876468718,
|
| 77 |
+
"rewards/margins": -0.0010797118302434683,
|
| 78 |
+
"rewards/rejected": -0.0006185341626405716,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.03,
|
| 83 |
+
"learning_rate": 1.7123287671232875e-07,
|
| 84 |
+
"logits/chosen": -1.9380648136138916,
|
| 85 |
+
"logits/rejected": -1.6861069202423096,
|
| 86 |
+
"logps/chosen": -93.2973861694336,
|
| 87 |
+
"logps/rejected": -76.30517578125,
|
| 88 |
+
"loss": 0.6927,
|
| 89 |
+
"rewards/accuracies": 0.34166663885116577,
|
| 90 |
+
"rewards/chosen": -0.0030570379458367825,
|
| 91 |
+
"rewards/margins": -0.005480821710079908,
|
| 92 |
+
"rewards/rejected": 0.002423783764243126,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.04,
|
| 97 |
+
"learning_rate": 2.054794520547945e-07,
|
| 98 |
+
"logits/chosen": -1.832024335861206,
|
| 99 |
+
"logits/rejected": -1.511380910873413,
|
| 100 |
+
"logps/chosen": -98.84480285644531,
|
| 101 |
+
"logps/rejected": -75.86248779296875,
|
| 102 |
+
"loss": 0.6931,
|
| 103 |
+
"rewards/accuracies": 0.4500000476837158,
|
| 104 |
+
"rewards/chosen": -0.003249814035370946,
|
| 105 |
+
"rewards/margins": -0.00364103470928967,
|
| 106 |
+
"rewards/rejected": 0.00039122122689150274,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.05,
|
| 111 |
+
"learning_rate": 2.3972602739726023e-07,
|
| 112 |
+
"logits/chosen": -1.9178974628448486,
|
| 113 |
+
"logits/rejected": -1.6177875995635986,
|
| 114 |
+
"logps/chosen": -92.13301086425781,
|
| 115 |
+
"logps/rejected": -75.75408172607422,
|
| 116 |
+
"loss": 0.6927,
|
| 117 |
+
"rewards/accuracies": 0.5333333611488342,
|
| 118 |
+
"rewards/chosen": 0.0013952379813417792,
|
| 119 |
+
"rewards/margins": 0.0021971219684928656,
|
| 120 |
+
"rewards/rejected": -0.0008018844528123736,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.05,
|
| 125 |
+
"learning_rate": 2.73972602739726e-07,
|
| 126 |
+
"logits/chosen": -1.938677191734314,
|
| 127 |
+
"logits/rejected": -1.6598854064941406,
|
| 128 |
+
"logps/chosen": -88.68348693847656,
|
| 129 |
+
"logps/rejected": -74.39127349853516,
|
| 130 |
+
"loss": 0.6929,
|
| 131 |
+
"rewards/accuracies": 0.5333333015441895,
|
| 132 |
+
"rewards/chosen": 0.0035478367935866117,
|
| 133 |
+
"rewards/margins": 0.003417719155550003,
|
| 134 |
+
"rewards/rejected": 0.00013011766714043915,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.06,
|
| 139 |
+
"learning_rate": 3.0821917808219176e-07,
|
| 140 |
+
"logits/chosen": -1.832851767539978,
|
| 141 |
+
"logits/rejected": -1.5398364067077637,
|
| 142 |
+
"logps/chosen": -92.82958984375,
|
| 143 |
+
"logps/rejected": -73.15336608886719,
|
| 144 |
+
"loss": 0.6927,
|
| 145 |
+
"rewards/accuracies": 0.5833333730697632,
|
| 146 |
+
"rewards/chosen": 0.00075482705142349,
|
| 147 |
+
"rewards/margins": 0.0024536100681871176,
|
| 148 |
+
"rewards/rejected": -0.001698783366009593,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.07,
|
| 153 |
+
"learning_rate": 3.424657534246575e-07,
|
| 154 |
+
"logits/chosen": -1.8959871530532837,
|
| 155 |
+
"logits/rejected": -1.6301053762435913,
|
| 156 |
+
"logps/chosen": -92.27674865722656,
|
| 157 |
+
"logps/rejected": -74.60256958007812,
|
| 158 |
+
"loss": 0.693,
|
| 159 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 160 |
+
"rewards/chosen": 0.0006887881900183856,
|
| 161 |
+
"rewards/margins": 0.0005900462856516242,
|
| 162 |
+
"rewards/rejected": 9.874170791590586e-05,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.07,
|
| 167 |
+
"eval_logits/chosen": -2.0524938106536865,
|
| 168 |
+
"eval_logits/rejected": -1.7860654592514038,
|
| 169 |
+
"eval_logps/chosen": -91.99323272705078,
|
| 170 |
+
"eval_logps/rejected": -72.10526275634766,
|
| 171 |
+
"eval_loss": 0.693277895450592,
|
| 172 |
+
"eval_rewards/accuracies": 0.4888888895511627,
|
| 173 |
+
"eval_rewards/chosen": -0.0007517762714996934,
|
| 174 |
+
"eval_rewards/margins": -0.00027055441751144826,
|
| 175 |
+
"eval_rewards/rejected": -0.00048122191219590604,
|
| 176 |
+
"eval_runtime": 117.2952,
|
| 177 |
+
"eval_samples_per_second": 24.4,
|
| 178 |
+
"eval_steps_per_second": 0.767,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.08,
|
| 183 |
+
"learning_rate": 3.767123287671233e-07,
|
| 184 |
+
"logits/chosen": -1.947257399559021,
|
| 185 |
+
"logits/rejected": -1.6791489124298096,
|
| 186 |
+
"logps/chosen": -93.37996673583984,
|
| 187 |
+
"logps/rejected": -72.86904907226562,
|
| 188 |
+
"loss": 0.6936,
|
| 189 |
+
"rewards/accuracies": 0.5083333849906921,
|
| 190 |
+
"rewards/chosen": -0.002063233172520995,
|
| 191 |
+
"rewards/margins": 0.0007919662748463452,
|
| 192 |
+
"rewards/rejected": -0.0028551991563290358,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.08,
|
| 197 |
+
"learning_rate": 4.10958904109589e-07,
|
| 198 |
+
"logits/chosen": -1.8895361423492432,
|
| 199 |
+
"logits/rejected": -1.587282419204712,
|
| 200 |
+
"logps/chosen": -93.67366027832031,
|
| 201 |
+
"logps/rejected": -72.37762451171875,
|
| 202 |
+
"loss": 0.6921,
|
| 203 |
+
"rewards/accuracies": 0.491666704416275,
|
| 204 |
+
"rewards/chosen": 0.0028340499848127365,
|
| 205 |
+
"rewards/margins": 0.0019404724007472396,
|
| 206 |
+
"rewards/rejected": 0.000893576827365905,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.09,
|
| 211 |
+
"learning_rate": 4.4520547945205477e-07,
|
| 212 |
+
"logits/chosen": -1.9517349004745483,
|
| 213 |
+
"logits/rejected": -1.663013219833374,
|
| 214 |
+
"logps/chosen": -84.09037780761719,
|
| 215 |
+
"logps/rejected": -72.67787170410156,
|
| 216 |
+
"loss": 0.6928,
|
| 217 |
+
"rewards/accuracies": 0.46666663885116577,
|
| 218 |
+
"rewards/chosen": 0.0004304441681597382,
|
| 219 |
+
"rewards/margins": -0.0008509824983775616,
|
| 220 |
+
"rewards/rejected": 0.001281426870264113,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.1,
|
| 225 |
+
"learning_rate": 4.794520547945205e-07,
|
| 226 |
+
"logits/chosen": -1.947576880455017,
|
| 227 |
+
"logits/rejected": -1.722400426864624,
|
| 228 |
+
"logps/chosen": -89.95055389404297,
|
| 229 |
+
"logps/rejected": -76.27583312988281,
|
| 230 |
+
"loss": 0.6919,
|
| 231 |
+
"rewards/accuracies": 0.625,
|
| 232 |
+
"rewards/chosen": 0.004270606208592653,
|
| 233 |
+
"rewards/margins": 0.009159665554761887,
|
| 234 |
+
"rewards/rejected": -0.004889058880507946,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.1,
|
| 239 |
+
"learning_rate": 4.984779299847793e-07,
|
| 240 |
+
"logits/chosen": -1.9852988719940186,
|
| 241 |
+
"logits/rejected": -1.6872488260269165,
|
| 242 |
+
"logps/chosen": -90.7852783203125,
|
| 243 |
+
"logps/rejected": -73.9708251953125,
|
| 244 |
+
"loss": 0.6915,
|
| 245 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 246 |
+
"rewards/chosen": 0.0020091754850000143,
|
| 247 |
+
"rewards/margins": 0.0034632813185453415,
|
| 248 |
+
"rewards/rejected": -0.0014541053678840399,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.11,
|
| 253 |
+
"learning_rate": 4.946727549467275e-07,
|
| 254 |
+
"logits/chosen": -1.9156850576400757,
|
| 255 |
+
"logits/rejected": -1.5840113162994385,
|
| 256 |
+
"logps/chosen": -97.8337173461914,
|
| 257 |
+
"logps/rejected": -73.20053100585938,
|
| 258 |
+
"loss": 0.6921,
|
| 259 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 260 |
+
"rewards/chosen": 0.0025704619474709034,
|
| 261 |
+
"rewards/margins": 0.002448607701808214,
|
| 262 |
+
"rewards/rejected": 0.00012185415107524022,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.12,
|
| 267 |
+
"learning_rate": 4.908675799086758e-07,
|
| 268 |
+
"logits/chosen": -1.803034782409668,
|
| 269 |
+
"logits/rejected": -1.4945720434188843,
|
| 270 |
+
"logps/chosen": -96.1871337890625,
|
| 271 |
+
"logps/rejected": -72.80181121826172,
|
| 272 |
+
"loss": 0.6921,
|
| 273 |
+
"rewards/accuracies": 0.5333333611488342,
|
| 274 |
+
"rewards/chosen": 0.003732017008587718,
|
| 275 |
+
"rewards/margins": 0.003292496781796217,
|
| 276 |
+
"rewards/rejected": 0.0004395198484417051,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.12,
|
| 281 |
+
"learning_rate": 4.87062404870624e-07,
|
| 282 |
+
"logits/chosen": -1.8866764307022095,
|
| 283 |
+
"logits/rejected": -1.5327541828155518,
|
| 284 |
+
"logps/chosen": -97.74296569824219,
|
| 285 |
+
"logps/rejected": -76.4982681274414,
|
| 286 |
+
"loss": 0.6909,
|
| 287 |
+
"rewards/accuracies": 0.5916666984558105,
|
| 288 |
+
"rewards/chosen": 0.004220059607177973,
|
| 289 |
+
"rewards/margins": 0.006730073597282171,
|
| 290 |
+
"rewards/rejected": -0.002510013757273555,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.13,
|
| 295 |
+
"learning_rate": 4.832572298325722e-07,
|
| 296 |
+
"logits/chosen": -1.945041298866272,
|
| 297 |
+
"logits/rejected": -1.6189870834350586,
|
| 298 |
+
"logps/chosen": -94.37440490722656,
|
| 299 |
+
"logps/rejected": -73.1822738647461,
|
| 300 |
+
"loss": 0.6904,
|
| 301 |
+
"rewards/accuracies": 0.574999988079071,
|
| 302 |
+
"rewards/chosen": 0.004208459984511137,
|
| 303 |
+
"rewards/margins": 0.005117190536111593,
|
| 304 |
+
"rewards/rejected": -0.0009087308426387608,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.14,
|
| 309 |
+
"learning_rate": 4.794520547945205e-07,
|
| 310 |
+
"logits/chosen": -1.9501270055770874,
|
| 311 |
+
"logits/rejected": -1.6498991250991821,
|
| 312 |
+
"logps/chosen": -93.06495666503906,
|
| 313 |
+
"logps/rejected": -72.0920181274414,
|
| 314 |
+
"loss": 0.69,
|
| 315 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 316 |
+
"rewards/chosen": 0.0078277587890625,
|
| 317 |
+
"rewards/margins": 0.010002164170145988,
|
| 318 |
+
"rewards/rejected": -0.0021744065452367067,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.14,
|
| 323 |
+
"eval_logits/chosen": -2.052419900894165,
|
| 324 |
+
"eval_logits/rejected": -1.7859160900115967,
|
| 325 |
+
"eval_logps/chosen": -91.95441436767578,
|
| 326 |
+
"eval_logps/rejected": -72.11531066894531,
|
| 327 |
+
"eval_loss": 0.6901015043258667,
|
| 328 |
+
"eval_rewards/accuracies": 0.5611110925674438,
|
| 329 |
+
"eval_rewards/chosen": 0.0031298992689698935,
|
| 330 |
+
"eval_rewards/margins": 0.004615597892552614,
|
| 331 |
+
"eval_rewards/rejected": -0.0014856986235827208,
|
| 332 |
+
"eval_runtime": 117.9228,
|
| 333 |
+
"eval_samples_per_second": 24.27,
|
| 334 |
+
"eval_steps_per_second": 0.763,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.14,
|
| 339 |
+
"learning_rate": 4.756468797564688e-07,
|
| 340 |
+
"logits/chosen": -1.9345300197601318,
|
| 341 |
+
"logits/rejected": -1.6839030981063843,
|
| 342 |
+
"logps/chosen": -92.93646240234375,
|
| 343 |
+
"logps/rejected": -75.16416931152344,
|
| 344 |
+
"loss": 0.6903,
|
| 345 |
+
"rewards/accuracies": 0.6166666746139526,
|
| 346 |
+
"rewards/chosen": 0.0068414295092225075,
|
| 347 |
+
"rewards/margins": 0.005408720578998327,
|
| 348 |
+
"rewards/rejected": 0.001432707766070962,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.15,
|
| 353 |
+
"learning_rate": 4.71841704718417e-07,
|
| 354 |
+
"logits/chosen": -1.9018728733062744,
|
| 355 |
+
"logits/rejected": -1.599656343460083,
|
| 356 |
+
"logps/chosen": -94.06185913085938,
|
| 357 |
+
"logps/rejected": -72.92243957519531,
|
| 358 |
+
"loss": 0.6888,
|
| 359 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 360 |
+
"rewards/chosen": 0.00332554685883224,
|
| 361 |
+
"rewards/margins": 0.007436770014464855,
|
| 362 |
+
"rewards/rejected": -0.004111223388463259,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.16,
|
| 367 |
+
"learning_rate": 4.680365296803653e-07,
|
| 368 |
+
"logits/chosen": -1.879314661026001,
|
| 369 |
+
"logits/rejected": -1.5864207744598389,
|
| 370 |
+
"logps/chosen": -93.1510238647461,
|
| 371 |
+
"logps/rejected": -74.74366760253906,
|
| 372 |
+
"loss": 0.6898,
|
| 373 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 374 |
+
"rewards/chosen": 0.007647272199392319,
|
| 375 |
+
"rewards/margins": 0.00648108497262001,
|
| 376 |
+
"rewards/rejected": 0.0011661878088489175,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.16,
|
| 381 |
+
"learning_rate": 4.642313546423135e-07,
|
| 382 |
+
"logits/chosen": -1.975441336631775,
|
| 383 |
+
"logits/rejected": -1.7306878566741943,
|
| 384 |
+
"logps/chosen": -86.38069915771484,
|
| 385 |
+
"logps/rejected": -75.68878173828125,
|
| 386 |
+
"loss": 0.6884,
|
| 387 |
+
"rewards/accuracies": 0.5749999284744263,
|
| 388 |
+
"rewards/chosen": 0.007093862630426884,
|
| 389 |
+
"rewards/margins": 0.008020764216780663,
|
| 390 |
+
"rewards/rejected": -0.0009269017027691007,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.17,
|
| 395 |
+
"learning_rate": 4.604261796042618e-07,
|
| 396 |
+
"logits/chosen": -1.843636155128479,
|
| 397 |
+
"logits/rejected": -1.5651946067810059,
|
| 398 |
+
"logps/chosen": -92.92024230957031,
|
| 399 |
+
"logps/rejected": -75.78328704833984,
|
| 400 |
+
"loss": 0.6864,
|
| 401 |
+
"rewards/accuracies": 0.6333333849906921,
|
| 402 |
+
"rewards/chosen": 0.007120449095964432,
|
| 403 |
+
"rewards/margins": 0.011716886423528194,
|
| 404 |
+
"rewards/rejected": -0.004596438258886337,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.18,
|
| 409 |
+
"learning_rate": 4.5662100456621e-07,
|
| 410 |
+
"logits/chosen": -1.8188960552215576,
|
| 411 |
+
"logits/rejected": -1.5440260171890259,
|
| 412 |
+
"logps/chosen": -91.01820373535156,
|
| 413 |
+
"logps/rejected": -72.49656677246094,
|
| 414 |
+
"loss": 0.687,
|
| 415 |
+
"rewards/accuracies": 0.6166666150093079,
|
| 416 |
+
"rewards/chosen": 0.007922597229480743,
|
| 417 |
+
"rewards/margins": 0.010534586384892464,
|
| 418 |
+
"rewards/rejected": -0.002611987991258502,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.18,
|
| 423 |
+
"learning_rate": 4.528158295281583e-07,
|
| 424 |
+
"logits/chosen": -1.931947946548462,
|
| 425 |
+
"logits/rejected": -1.6909589767456055,
|
| 426 |
+
"logps/chosen": -88.5571517944336,
|
| 427 |
+
"logps/rejected": -72.00991821289062,
|
| 428 |
+
"loss": 0.6866,
|
| 429 |
+
"rewards/accuracies": 0.5999999642372131,
|
| 430 |
+
"rewards/chosen": 0.005753463599830866,
|
| 431 |
+
"rewards/margins": 0.010802066884934902,
|
| 432 |
+
"rewards/rejected": -0.005048603750765324,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.19,
|
| 437 |
+
"learning_rate": 4.490106544901065e-07,
|
| 438 |
+
"logits/chosen": -1.9274402856826782,
|
| 439 |
+
"logits/rejected": -1.6220242977142334,
|
| 440 |
+
"logps/chosen": -91.9419174194336,
|
| 441 |
+
"logps/rejected": -77.3694076538086,
|
| 442 |
+
"loss": 0.6851,
|
| 443 |
+
"rewards/accuracies": 0.7083333134651184,
|
| 444 |
+
"rewards/chosen": 0.014447471126914024,
|
| 445 |
+
"rewards/margins": 0.018269026651978493,
|
| 446 |
+
"rewards/rejected": -0.0038215541280806065,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.2,
|
| 451 |
+
"learning_rate": 4.4520547945205477e-07,
|
| 452 |
+
"logits/chosen": -1.8854271173477173,
|
| 453 |
+
"logits/rejected": -1.6241188049316406,
|
| 454 |
+
"logps/chosen": -92.77326965332031,
|
| 455 |
+
"logps/rejected": -77.81119537353516,
|
| 456 |
+
"loss": 0.6838,
|
| 457 |
+
"rewards/accuracies": 0.64166659116745,
|
| 458 |
+
"rewards/chosen": 0.013706192374229431,
|
| 459 |
+
"rewards/margins": 0.015467122197151184,
|
| 460 |
+
"rewards/rejected": -0.0017609309870749712,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.21,
|
| 465 |
+
"learning_rate": 4.41400304414003e-07,
|
| 466 |
+
"logits/chosen": -1.8017442226409912,
|
| 467 |
+
"logits/rejected": -1.5209267139434814,
|
| 468 |
+
"logps/chosen": -90.86246490478516,
|
| 469 |
+
"logps/rejected": -72.70795440673828,
|
| 470 |
+
"loss": 0.6842,
|
| 471 |
+
"rewards/accuracies": 0.699999988079071,
|
| 472 |
+
"rewards/chosen": 0.013066952116787434,
|
| 473 |
+
"rewards/margins": 0.018115142360329628,
|
| 474 |
+
"rewards/rejected": -0.00504818931221962,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.21,
|
| 479 |
+
"eval_logits/chosen": -2.051283121109009,
|
| 480 |
+
"eval_logits/rejected": -1.7846735715866089,
|
| 481 |
+
"eval_logps/chosen": -91.84673309326172,
|
| 482 |
+
"eval_logps/rejected": -72.15672302246094,
|
| 483 |
+
"eval_loss": 0.6832027435302734,
|
| 484 |
+
"eval_rewards/accuracies": 0.6916666626930237,
|
| 485 |
+
"eval_rewards/chosen": 0.013898174278438091,
|
| 486 |
+
"eval_rewards/margins": 0.01952529139816761,
|
| 487 |
+
"eval_rewards/rejected": -0.005627114325761795,
|
| 488 |
+
"eval_runtime": 116.9802,
|
| 489 |
+
"eval_samples_per_second": 24.466,
|
| 490 |
+
"eval_steps_per_second": 0.769,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.21,
|
| 495 |
+
"learning_rate": 4.375951293759513e-07,
|
| 496 |
+
"logits/chosen": -1.8816429376602173,
|
| 497 |
+
"logits/rejected": -1.5877922773361206,
|
| 498 |
+
"logps/chosen": -92.76510620117188,
|
| 499 |
+
"logps/rejected": -77.3439712524414,
|
| 500 |
+
"loss": 0.6827,
|
| 501 |
+
"rewards/accuracies": 0.7166666388511658,
|
| 502 |
+
"rewards/chosen": 0.016739103943109512,
|
| 503 |
+
"rewards/margins": 0.020335419103503227,
|
| 504 |
+
"rewards/rejected": -0.003596315626055002,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.22,
|
| 509 |
+
"learning_rate": 4.337899543378995e-07,
|
| 510 |
+
"logits/chosen": -1.8938591480255127,
|
| 511 |
+
"logits/rejected": -1.6317039728164673,
|
| 512 |
+
"logps/chosen": -93.17677307128906,
|
| 513 |
+
"logps/rejected": -74.86593627929688,
|
| 514 |
+
"loss": 0.6818,
|
| 515 |
+
"rewards/accuracies": 0.699999988079071,
|
| 516 |
+
"rewards/chosen": 0.01637251302599907,
|
| 517 |
+
"rewards/margins": 0.019245153293013573,
|
| 518 |
+
"rewards/rejected": -0.0028726388700306416,
|
| 519 |
+
"step": 320
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 0.23,
|
| 523 |
+
"learning_rate": 4.2998477929984777e-07,
|
| 524 |
+
"logits/chosen": -1.8730627298355103,
|
| 525 |
+
"logits/rejected": -1.6184980869293213,
|
| 526 |
+
"logps/chosen": -91.06995391845703,
|
| 527 |
+
"logps/rejected": -72.5287857055664,
|
| 528 |
+
"loss": 0.6808,
|
| 529 |
+
"rewards/accuracies": 0.7666667103767395,
|
| 530 |
+
"rewards/chosen": 0.01837759278714657,
|
| 531 |
+
"rewards/margins": 0.02310130000114441,
|
| 532 |
+
"rewards/rejected": -0.004723704420030117,
|
| 533 |
+
"step": 330
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.23,
|
| 537 |
+
"learning_rate": 4.26179604261796e-07,
|
| 538 |
+
"logits/chosen": -1.8828544616699219,
|
| 539 |
+
"logits/rejected": -1.576556921005249,
|
| 540 |
+
"logps/chosen": -92.8334732055664,
|
| 541 |
+
"logps/rejected": -74.53410339355469,
|
| 542 |
+
"loss": 0.6788,
|
| 543 |
+
"rewards/accuracies": 0.8166666030883789,
|
| 544 |
+
"rewards/chosen": 0.023710301145911217,
|
| 545 |
+
"rewards/margins": 0.03184016793966293,
|
| 546 |
+
"rewards/rejected": -0.008129866793751717,
|
| 547 |
+
"step": 340
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 0.24,
|
| 551 |
+
"learning_rate": 4.223744292237443e-07,
|
| 552 |
+
"logits/chosen": -1.9196786880493164,
|
| 553 |
+
"logits/rejected": -1.625704050064087,
|
| 554 |
+
"logps/chosen": -90.74723815917969,
|
| 555 |
+
"logps/rejected": -76.68948364257812,
|
| 556 |
+
"loss": 0.6777,
|
| 557 |
+
"rewards/accuracies": 0.8083333969116211,
|
| 558 |
+
"rewards/chosen": 0.026256781071424484,
|
| 559 |
+
"rewards/margins": 0.03269972652196884,
|
| 560 |
+
"rewards/rejected": -0.00644295010715723,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 0.25,
|
| 565 |
+
"learning_rate": 4.185692541856925e-07,
|
| 566 |
+
"logits/chosen": -1.981180191040039,
|
| 567 |
+
"logits/rejected": -1.6961545944213867,
|
| 568 |
+
"logps/chosen": -93.02009582519531,
|
| 569 |
+
"logps/rejected": -76.1031723022461,
|
| 570 |
+
"loss": 0.6762,
|
| 571 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 572 |
+
"rewards/chosen": 0.03220153599977493,
|
| 573 |
+
"rewards/margins": 0.03597740828990936,
|
| 574 |
+
"rewards/rejected": -0.0037758699618279934,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 0.25,
|
| 579 |
+
"learning_rate": 4.1476407914764077e-07,
|
| 580 |
+
"logits/chosen": -1.8603204488754272,
|
| 581 |
+
"logits/rejected": -1.5979934930801392,
|
| 582 |
+
"logps/chosen": -94.58716583251953,
|
| 583 |
+
"logps/rejected": -75.44288635253906,
|
| 584 |
+
"loss": 0.6759,
|
| 585 |
+
"rewards/accuracies": 0.841666579246521,
|
| 586 |
+
"rewards/chosen": 0.021503183990716934,
|
| 587 |
+
"rewards/margins": 0.03375691920518875,
|
| 588 |
+
"rewards/rejected": -0.012253734283149242,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 0.26,
|
| 593 |
+
"learning_rate": 4.10958904109589e-07,
|
| 594 |
+
"logits/chosen": -1.9121630191802979,
|
| 595 |
+
"logits/rejected": -1.613445520401001,
|
| 596 |
+
"logps/chosen": -93.34215545654297,
|
| 597 |
+
"logps/rejected": -74.89527893066406,
|
| 598 |
+
"loss": 0.6739,
|
| 599 |
+
"rewards/accuracies": 0.824999988079071,
|
| 600 |
+
"rewards/chosen": 0.02675667405128479,
|
| 601 |
+
"rewards/margins": 0.03688011318445206,
|
| 602 |
+
"rewards/rejected": -0.010123440995812416,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 0.27,
|
| 607 |
+
"learning_rate": 4.071537290715373e-07,
|
| 608 |
+
"logits/chosen": -1.9191780090332031,
|
| 609 |
+
"logits/rejected": -1.532434105873108,
|
| 610 |
+
"logps/chosen": -94.9774169921875,
|
| 611 |
+
"logps/rejected": -73.49574279785156,
|
| 612 |
+
"loss": 0.6744,
|
| 613 |
+
"rewards/accuracies": 0.8666666150093079,
|
| 614 |
+
"rewards/chosen": 0.03616553544998169,
|
| 615 |
+
"rewards/margins": 0.048189498484134674,
|
| 616 |
+
"rewards/rejected": -0.012023964896798134,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.27,
|
| 621 |
+
"learning_rate": 4.033485540334855e-07,
|
| 622 |
+
"logits/chosen": -1.8837106227874756,
|
| 623 |
+
"logits/rejected": -1.6179090738296509,
|
| 624 |
+
"logps/chosen": -93.18563079833984,
|
| 625 |
+
"logps/rejected": -75.48190307617188,
|
| 626 |
+
"loss": 0.672,
|
| 627 |
+
"rewards/accuracies": 0.8333333730697632,
|
| 628 |
+
"rewards/chosen": 0.03295673802495003,
|
| 629 |
+
"rewards/margins": 0.044657547026872635,
|
| 630 |
+
"rewards/rejected": -0.011700802482664585,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.27,
|
| 635 |
+
"eval_logits/chosen": -2.0504438877105713,
|
| 636 |
+
"eval_logits/rejected": -1.78355073928833,
|
| 637 |
+
"eval_logps/chosen": -91.70490264892578,
|
| 638 |
+
"eval_logps/rejected": -72.23121643066406,
|
| 639 |
+
"eval_loss": 0.671801745891571,
|
| 640 |
+
"eval_rewards/accuracies": 0.824999988079071,
|
| 641 |
+
"eval_rewards/chosen": 0.028079798445105553,
|
| 642 |
+
"eval_rewards/margins": 0.041155941784381866,
|
| 643 |
+
"eval_rewards/rejected": -0.013076143339276314,
|
| 644 |
+
"eval_runtime": 122.6508,
|
| 645 |
+
"eval_samples_per_second": 23.335,
|
| 646 |
+
"eval_steps_per_second": 0.734,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.28,
|
| 651 |
+
"learning_rate": 3.9954337899543377e-07,
|
| 652 |
+
"logits/chosen": -1.9980239868164062,
|
| 653 |
+
"logits/rejected": -1.710172414779663,
|
| 654 |
+
"logps/chosen": -90.13986206054688,
|
| 655 |
+
"logps/rejected": -74.13956451416016,
|
| 656 |
+
"loss": 0.6726,
|
| 657 |
+
"rewards/accuracies": 0.7833333015441895,
|
| 658 |
+
"rewards/chosen": 0.026057172566652298,
|
| 659 |
+
"rewards/margins": 0.03604119271039963,
|
| 660 |
+
"rewards/rejected": -0.00998402014374733,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.29,
|
| 665 |
+
"learning_rate": 3.95738203957382e-07,
|
| 666 |
+
"logits/chosen": -1.9292805194854736,
|
| 667 |
+
"logits/rejected": -1.6599111557006836,
|
| 668 |
+
"logps/chosen": -89.1054458618164,
|
| 669 |
+
"logps/rejected": -76.2321548461914,
|
| 670 |
+
"loss": 0.67,
|
| 671 |
+
"rewards/accuracies": 0.875,
|
| 672 |
+
"rewards/chosen": 0.03600526601076126,
|
| 673 |
+
"rewards/margins": 0.04854360967874527,
|
| 674 |
+
"rewards/rejected": -0.012538343667984009,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.29,
|
| 679 |
+
"learning_rate": 3.919330289193303e-07,
|
| 680 |
+
"logits/chosen": -1.8889102935791016,
|
| 681 |
+
"logits/rejected": -1.5878620147705078,
|
| 682 |
+
"logps/chosen": -94.20903015136719,
|
| 683 |
+
"logps/rejected": -75.64437103271484,
|
| 684 |
+
"loss": 0.6677,
|
| 685 |
+
"rewards/accuracies": 0.8833333253860474,
|
| 686 |
+
"rewards/chosen": 0.04072072356939316,
|
| 687 |
+
"rewards/margins": 0.05353847146034241,
|
| 688 |
+
"rewards/rejected": -0.012817745096981525,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.3,
|
| 693 |
+
"learning_rate": 3.881278538812785e-07,
|
| 694 |
+
"logits/chosen": -1.889439344406128,
|
| 695 |
+
"logits/rejected": -1.5668418407440186,
|
| 696 |
+
"logps/chosen": -92.68305206298828,
|
| 697 |
+
"logps/rejected": -71.53545379638672,
|
| 698 |
+
"loss": 0.6653,
|
| 699 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 700 |
+
"rewards/chosen": 0.042834434658288956,
|
| 701 |
+
"rewards/margins": 0.05957134813070297,
|
| 702 |
+
"rewards/rejected": -0.016736917197704315,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.31,
|
| 707 |
+
"learning_rate": 3.8432267884322677e-07,
|
| 708 |
+
"logits/chosen": -1.8060448169708252,
|
| 709 |
+
"logits/rejected": -1.5509716272354126,
|
| 710 |
+
"logps/chosen": -88.83539581298828,
|
| 711 |
+
"logps/rejected": -71.70087432861328,
|
| 712 |
+
"loss": 0.6645,
|
| 713 |
+
"rewards/accuracies": 0.8416666984558105,
|
| 714 |
+
"rewards/chosen": 0.04349132627248764,
|
| 715 |
+
"rewards/margins": 0.05466403439640999,
|
| 716 |
+
"rewards/rejected": -0.011172705329954624,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.31,
|
| 721 |
+
"learning_rate": 3.80517503805175e-07,
|
| 722 |
+
"logits/chosen": -1.8319886922836304,
|
| 723 |
+
"logits/rejected": -1.5673277378082275,
|
| 724 |
+
"logps/chosen": -88.6297378540039,
|
| 725 |
+
"logps/rejected": -74.78864288330078,
|
| 726 |
+
"loss": 0.6631,
|
| 727 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 728 |
+
"rewards/chosen": 0.044696420431137085,
|
| 729 |
+
"rewards/margins": 0.0604400709271431,
|
| 730 |
+
"rewards/rejected": -0.015743646770715714,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.32,
|
| 735 |
+
"learning_rate": 3.767123287671233e-07,
|
| 736 |
+
"logits/chosen": -1.9542433023452759,
|
| 737 |
+
"logits/rejected": -1.641196608543396,
|
| 738 |
+
"logps/chosen": -99.97395324707031,
|
| 739 |
+
"logps/rejected": -78.014404296875,
|
| 740 |
+
"loss": 0.6626,
|
| 741 |
+
"rewards/accuracies": 0.9166666865348816,
|
| 742 |
+
"rewards/chosen": 0.047287195920944214,
|
| 743 |
+
"rewards/margins": 0.06728993356227875,
|
| 744 |
+
"rewards/rejected": -0.020002741366624832,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 0.33,
|
| 749 |
+
"learning_rate": 3.729071537290715e-07,
|
| 750 |
+
"logits/chosen": -1.916868805885315,
|
| 751 |
+
"logits/rejected": -1.6233441829681396,
|
| 752 |
+
"logps/chosen": -97.3001937866211,
|
| 753 |
+
"logps/rejected": -76.58578491210938,
|
| 754 |
+
"loss": 0.6619,
|
| 755 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 756 |
+
"rewards/chosen": 0.05124374479055405,
|
| 757 |
+
"rewards/margins": 0.07088983058929443,
|
| 758 |
+
"rewards/rejected": -0.019646091386675835,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 0.34,
|
| 763 |
+
"learning_rate": 3.6910197869101977e-07,
|
| 764 |
+
"logits/chosen": -1.9672422409057617,
|
| 765 |
+
"logits/rejected": -1.700531244277954,
|
| 766 |
+
"logps/chosen": -89.29413604736328,
|
| 767 |
+
"logps/rejected": -72.65511322021484,
|
| 768 |
+
"loss": 0.66,
|
| 769 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 770 |
+
"rewards/chosen": 0.050498634576797485,
|
| 771 |
+
"rewards/margins": 0.06749961525201797,
|
| 772 |
+
"rewards/rejected": -0.01700098067522049,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 0.34,
|
| 777 |
+
"learning_rate": 3.65296803652968e-07,
|
| 778 |
+
"logits/chosen": -1.833929419517517,
|
| 779 |
+
"logits/rejected": -1.5389481782913208,
|
| 780 |
+
"logps/chosen": -85.8098373413086,
|
| 781 |
+
"logps/rejected": -72.5254135131836,
|
| 782 |
+
"loss": 0.6563,
|
| 783 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 784 |
+
"rewards/chosen": 0.05660303682088852,
|
| 785 |
+
"rewards/margins": 0.0805855318903923,
|
| 786 |
+
"rewards/rejected": -0.023982489481568336,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 0.34,
|
| 791 |
+
"eval_logits/chosen": -2.0494003295898438,
|
| 792 |
+
"eval_logits/rejected": -1.7820732593536377,
|
| 793 |
+
"eval_logps/chosen": -91.48755645751953,
|
| 794 |
+
"eval_logps/rejected": -72.31159973144531,
|
| 795 |
+
"eval_loss": 0.6574758887290955,
|
| 796 |
+
"eval_rewards/accuracies": 0.8861111402511597,
|
| 797 |
+
"eval_rewards/chosen": 0.049815867096185684,
|
| 798 |
+
"eval_rewards/margins": 0.0709303691983223,
|
| 799 |
+
"eval_rewards/rejected": -0.021114489063620567,
|
| 800 |
+
"eval_runtime": 126.0719,
|
| 801 |
+
"eval_samples_per_second": 22.701,
|
| 802 |
+
"eval_steps_per_second": 0.714,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 0.35,
|
| 807 |
+
"learning_rate": 3.614916286149163e-07,
|
| 808 |
+
"logits/chosen": -1.913190484046936,
|
| 809 |
+
"logits/rejected": -1.6543632745742798,
|
| 810 |
+
"logps/chosen": -89.34779357910156,
|
| 811 |
+
"logps/rejected": -71.7219009399414,
|
| 812 |
+
"loss": 0.6567,
|
| 813 |
+
"rewards/accuracies": 0.8916667103767395,
|
| 814 |
+
"rewards/chosen": 0.05136920139193535,
|
| 815 |
+
"rewards/margins": 0.07187855988740921,
|
| 816 |
+
"rewards/rejected": -0.02050935849547386,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 0.36,
|
| 821 |
+
"learning_rate": 3.576864535768645e-07,
|
| 822 |
+
"logits/chosen": -1.8918750286102295,
|
| 823 |
+
"logits/rejected": -1.621498465538025,
|
| 824 |
+
"logps/chosen": -92.99454498291016,
|
| 825 |
+
"logps/rejected": -72.34910583496094,
|
| 826 |
+
"loss": 0.6569,
|
| 827 |
+
"rewards/accuracies": 0.841666579246521,
|
| 828 |
+
"rewards/chosen": 0.05106909200549126,
|
| 829 |
+
"rewards/margins": 0.0683208703994751,
|
| 830 |
+
"rewards/rejected": -0.01725177839398384,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 0.36,
|
| 835 |
+
"learning_rate": 3.5388127853881277e-07,
|
| 836 |
+
"logits/chosen": -1.8910505771636963,
|
| 837 |
+
"logits/rejected": -1.6094516515731812,
|
| 838 |
+
"logps/chosen": -92.9715347290039,
|
| 839 |
+
"logps/rejected": -75.84952545166016,
|
| 840 |
+
"loss": 0.6559,
|
| 841 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 842 |
+
"rewards/chosen": 0.05335830897092819,
|
| 843 |
+
"rewards/margins": 0.07476408034563065,
|
| 844 |
+
"rewards/rejected": -0.021405773237347603,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 0.37,
|
| 849 |
+
"learning_rate": 3.50076103500761e-07,
|
| 850 |
+
"logits/chosen": -1.8273894786834717,
|
| 851 |
+
"logits/rejected": -1.5480557680130005,
|
| 852 |
+
"logps/chosen": -93.09770202636719,
|
| 853 |
+
"logps/rejected": -77.72835540771484,
|
| 854 |
+
"loss": 0.6531,
|
| 855 |
+
"rewards/accuracies": 0.8833333253860474,
|
| 856 |
+
"rewards/chosen": 0.06002092361450195,
|
| 857 |
+
"rewards/margins": 0.08498416841030121,
|
| 858 |
+
"rewards/rejected": -0.02496323548257351,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 0.38,
|
| 863 |
+
"learning_rate": 3.462709284627093e-07,
|
| 864 |
+
"logits/chosen": -1.7952553033828735,
|
| 865 |
+
"logits/rejected": -1.5454210042953491,
|
| 866 |
+
"logps/chosen": -92.24395751953125,
|
| 867 |
+
"logps/rejected": -74.6566390991211,
|
| 868 |
+
"loss": 0.6514,
|
| 869 |
+
"rewards/accuracies": 0.8666666746139526,
|
| 870 |
+
"rewards/chosen": 0.06232692673802376,
|
| 871 |
+
"rewards/margins": 0.08038316667079926,
|
| 872 |
+
"rewards/rejected": -0.018056249246001244,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 0.38,
|
| 877 |
+
"learning_rate": 3.424657534246575e-07,
|
| 878 |
+
"logits/chosen": -1.8805965185165405,
|
| 879 |
+
"logits/rejected": -1.586458444595337,
|
| 880 |
+
"logps/chosen": -96.27812194824219,
|
| 881 |
+
"logps/rejected": -75.61478424072266,
|
| 882 |
+
"loss": 0.6487,
|
| 883 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 884 |
+
"rewards/chosen": 0.06480798870325089,
|
| 885 |
+
"rewards/margins": 0.09341150522232056,
|
| 886 |
+
"rewards/rejected": -0.02860351838171482,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 0.39,
|
| 891 |
+
"learning_rate": 3.3866057838660576e-07,
|
| 892 |
+
"logits/chosen": -1.8029924631118774,
|
| 893 |
+
"logits/rejected": -1.5792688131332397,
|
| 894 |
+
"logps/chosen": -87.8210678100586,
|
| 895 |
+
"logps/rejected": -74.4839096069336,
|
| 896 |
+
"loss": 0.651,
|
| 897 |
+
"rewards/accuracies": 0.8749998807907104,
|
| 898 |
+
"rewards/chosen": 0.05512385442852974,
|
| 899 |
+
"rewards/margins": 0.07892224937677383,
|
| 900 |
+
"rewards/rejected": -0.023798387497663498,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 0.4,
|
| 905 |
+
"learning_rate": 3.34855403348554e-07,
|
| 906 |
+
"logits/chosen": -1.913487434387207,
|
| 907 |
+
"logits/rejected": -1.612151861190796,
|
| 908 |
+
"logps/chosen": -94.51315307617188,
|
| 909 |
+
"logps/rejected": -74.9591293334961,
|
| 910 |
+
"loss": 0.6465,
|
| 911 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 912 |
+
"rewards/chosen": 0.0840396136045456,
|
| 913 |
+
"rewards/margins": 0.107123002409935,
|
| 914 |
+
"rewards/rejected": -0.023083383217453957,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 0.4,
|
| 919 |
+
"learning_rate": 3.310502283105023e-07,
|
| 920 |
+
"logits/chosen": -1.9018150568008423,
|
| 921 |
+
"logits/rejected": -1.6232668161392212,
|
| 922 |
+
"logps/chosen": -93.73323822021484,
|
| 923 |
+
"logps/rejected": -75.38792419433594,
|
| 924 |
+
"loss": 0.6435,
|
| 925 |
+
"rewards/accuracies": 0.8666666150093079,
|
| 926 |
+
"rewards/chosen": 0.07307516038417816,
|
| 927 |
+
"rewards/margins": 0.10251389443874359,
|
| 928 |
+
"rewards/rejected": -0.029438745230436325,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.41,
|
| 933 |
+
"learning_rate": 3.272450532724505e-07,
|
| 934 |
+
"logits/chosen": -1.8873398303985596,
|
| 935 |
+
"logits/rejected": -1.6442911624908447,
|
| 936 |
+
"logps/chosen": -90.5989761352539,
|
| 937 |
+
"logps/rejected": -77.07279968261719,
|
| 938 |
+
"loss": 0.6437,
|
| 939 |
+
"rewards/accuracies": 0.966666579246521,
|
| 940 |
+
"rewards/chosen": 0.08184785395860672,
|
| 941 |
+
"rewards/margins": 0.11672016233205795,
|
| 942 |
+
"rewards/rejected": -0.034872300922870636,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 0.41,
|
| 947 |
+
"eval_logits/chosen": -2.048579692840576,
|
| 948 |
+
"eval_logits/rejected": -1.780737280845642,
|
| 949 |
+
"eval_logps/chosen": -91.28104400634766,
|
| 950 |
+
"eval_logps/rejected": -72.44007110595703,
|
| 951 |
+
"eval_loss": 0.6415905356407166,
|
| 952 |
+
"eval_rewards/accuracies": 0.9111111164093018,
|
| 953 |
+
"eval_rewards/chosen": 0.07046664506196976,
|
| 954 |
+
"eval_rewards/margins": 0.10442798584699631,
|
| 955 |
+
"eval_rewards/rejected": -0.03396133333444595,
|
| 956 |
+
"eval_runtime": 117.9826,
|
| 957 |
+
"eval_samples_per_second": 24.258,
|
| 958 |
+
"eval_steps_per_second": 0.763,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.42,
|
| 963 |
+
"learning_rate": 3.2343987823439876e-07,
|
| 964 |
+
"logits/chosen": -1.9880199432373047,
|
| 965 |
+
"logits/rejected": -1.676412582397461,
|
| 966 |
+
"logps/chosen": -91.00142669677734,
|
| 967 |
+
"logps/rejected": -73.52288818359375,
|
| 968 |
+
"loss": 0.6401,
|
| 969 |
+
"rewards/accuracies": 0.9333332777023315,
|
| 970 |
+
"rewards/chosen": 0.07217723876237869,
|
| 971 |
+
"rewards/margins": 0.1089334487915039,
|
| 972 |
+
"rewards/rejected": -0.036756210029125214,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 0.42,
|
| 977 |
+
"learning_rate": 3.19634703196347e-07,
|
| 978 |
+
"logits/chosen": -1.8773486614227295,
|
| 979 |
+
"logits/rejected": -1.573994755744934,
|
| 980 |
+
"logps/chosen": -89.79063415527344,
|
| 981 |
+
"logps/rejected": -73.29415130615234,
|
| 982 |
+
"loss": 0.6391,
|
| 983 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 984 |
+
"rewards/chosen": 0.06451607495546341,
|
| 985 |
+
"rewards/margins": 0.09732060134410858,
|
| 986 |
+
"rewards/rejected": -0.032804541289806366,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 0.43,
|
| 991 |
+
"learning_rate": 3.158295281582953e-07,
|
| 992 |
+
"logits/chosen": -1.8495439291000366,
|
| 993 |
+
"logits/rejected": -1.5227216482162476,
|
| 994 |
+
"logps/chosen": -93.12200164794922,
|
| 995 |
+
"logps/rejected": -74.12937927246094,
|
| 996 |
+
"loss": 0.6371,
|
| 997 |
+
"rewards/accuracies": 0.9333333969116211,
|
| 998 |
+
"rewards/chosen": 0.08449102938175201,
|
| 999 |
+
"rewards/margins": 0.11922381818294525,
|
| 1000 |
+
"rewards/rejected": -0.034732796251773834,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 0.44,
|
| 1005 |
+
"learning_rate": 3.120243531202435e-07,
|
| 1006 |
+
"logits/chosen": -1.93155837059021,
|
| 1007 |
+
"logits/rejected": -1.6498839855194092,
|
| 1008 |
+
"logps/chosen": -94.86695861816406,
|
| 1009 |
+
"logps/rejected": -76.8067626953125,
|
| 1010 |
+
"loss": 0.6354,
|
| 1011 |
+
"rewards/accuracies": 0.8833333253860474,
|
| 1012 |
+
"rewards/chosen": 0.08870759606361389,
|
| 1013 |
+
"rewards/margins": 0.12238001823425293,
|
| 1014 |
+
"rewards/rejected": -0.03367242217063904,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 0.45,
|
| 1019 |
+
"learning_rate": 3.0821917808219176e-07,
|
| 1020 |
+
"logits/chosen": -1.8748143911361694,
|
| 1021 |
+
"logits/rejected": -1.6098239421844482,
|
| 1022 |
+
"logps/chosen": -91.36283874511719,
|
| 1023 |
+
"logps/rejected": -73.6336669921875,
|
| 1024 |
+
"loss": 0.6328,
|
| 1025 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 1026 |
+
"rewards/chosen": 0.0779472142457962,
|
| 1027 |
+
"rewards/margins": 0.1164277195930481,
|
| 1028 |
+
"rewards/rejected": -0.038480497896671295,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 0.45,
|
| 1033 |
+
"learning_rate": 3.0441400304414e-07,
|
| 1034 |
+
"logits/chosen": -1.8293393850326538,
|
| 1035 |
+
"logits/rejected": -1.4808156490325928,
|
| 1036 |
+
"logps/chosen": -94.16930389404297,
|
| 1037 |
+
"logps/rejected": -72.66043853759766,
|
| 1038 |
+
"loss": 0.6341,
|
| 1039 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 1040 |
+
"rewards/chosen": 0.09546177834272385,
|
| 1041 |
+
"rewards/margins": 0.1290528029203415,
|
| 1042 |
+
"rewards/rejected": -0.033591024577617645,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 0.46,
|
| 1047 |
+
"learning_rate": 3.006088280060883e-07,
|
| 1048 |
+
"logits/chosen": -1.8831846714019775,
|
| 1049 |
+
"logits/rejected": -1.577532172203064,
|
| 1050 |
+
"logps/chosen": -95.71855926513672,
|
| 1051 |
+
"logps/rejected": -74.64360046386719,
|
| 1052 |
+
"loss": 0.6305,
|
| 1053 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1054 |
+
"rewards/chosen": 0.09222020953893661,
|
| 1055 |
+
"rewards/margins": 0.1298314929008484,
|
| 1056 |
+
"rewards/rejected": -0.037611283361911774,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 0.47,
|
| 1061 |
+
"learning_rate": 2.968036529680365e-07,
|
| 1062 |
+
"logits/chosen": -1.884874701499939,
|
| 1063 |
+
"logits/rejected": -1.56434965133667,
|
| 1064 |
+
"logps/chosen": -91.71931457519531,
|
| 1065 |
+
"logps/rejected": -74.07929992675781,
|
| 1066 |
+
"loss": 0.6298,
|
| 1067 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1068 |
+
"rewards/chosen": 0.09030507504940033,
|
| 1069 |
+
"rewards/margins": 0.13949953019618988,
|
| 1070 |
+
"rewards/rejected": -0.04919447377324104,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 0.47,
|
| 1075 |
+
"learning_rate": 2.9299847792998476e-07,
|
| 1076 |
+
"logits/chosen": -1.939512848854065,
|
| 1077 |
+
"logits/rejected": -1.6418523788452148,
|
| 1078 |
+
"logps/chosen": -88.81291198730469,
|
| 1079 |
+
"logps/rejected": -71.72727966308594,
|
| 1080 |
+
"loss": 0.6255,
|
| 1081 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1082 |
+
"rewards/chosen": 0.1102442592382431,
|
| 1083 |
+
"rewards/margins": 0.14661459624767303,
|
| 1084 |
+
"rewards/rejected": -0.036370351910591125,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 0.48,
|
| 1089 |
+
"learning_rate": 2.89193302891933e-07,
|
| 1090 |
+
"logits/chosen": -1.8088592290878296,
|
| 1091 |
+
"logits/rejected": -1.554652452468872,
|
| 1092 |
+
"logps/chosen": -88.36043548583984,
|
| 1093 |
+
"logps/rejected": -74.16236877441406,
|
| 1094 |
+
"loss": 0.6261,
|
| 1095 |
+
"rewards/accuracies": 0.9416666030883789,
|
| 1096 |
+
"rewards/chosen": 0.09904900938272476,
|
| 1097 |
+
"rewards/margins": 0.13188037276268005,
|
| 1098 |
+
"rewards/rejected": -0.032831382006406784,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 0.48,
|
| 1103 |
+
"eval_logits/chosen": -2.0478439331054688,
|
| 1104 |
+
"eval_logits/rejected": -1.7795759439468384,
|
| 1105 |
+
"eval_logps/chosen": -91.1009750366211,
|
| 1106 |
+
"eval_logps/rejected": -72.53546905517578,
|
| 1107 |
+
"eval_loss": 0.6277271509170532,
|
| 1108 |
+
"eval_rewards/accuracies": 0.925000011920929,
|
| 1109 |
+
"eval_rewards/chosen": 0.08847405016422272,
|
| 1110 |
+
"eval_rewards/margins": 0.13197554647922516,
|
| 1111 |
+
"eval_rewards/rejected": -0.04350150376558304,
|
| 1112 |
+
"eval_runtime": 117.9463,
|
| 1113 |
+
"eval_samples_per_second": 24.265,
|
| 1114 |
+
"eval_steps_per_second": 0.763,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 0.49,
|
| 1119 |
+
"learning_rate": 2.853881278538813e-07,
|
| 1120 |
+
"logits/chosen": -1.8632869720458984,
|
| 1121 |
+
"logits/rejected": -1.598181962966919,
|
| 1122 |
+
"logps/chosen": -89.87626647949219,
|
| 1123 |
+
"logps/rejected": -74.79205322265625,
|
| 1124 |
+
"loss": 0.6285,
|
| 1125 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1126 |
+
"rewards/chosen": 0.09145402163267136,
|
| 1127 |
+
"rewards/margins": 0.13780589401721954,
|
| 1128 |
+
"rewards/rejected": -0.04635186120867729,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 0.49,
|
| 1133 |
+
"learning_rate": 2.815829528158295e-07,
|
| 1134 |
+
"logits/chosen": -1.8058583736419678,
|
| 1135 |
+
"logits/rejected": -1.5232179164886475,
|
| 1136 |
+
"logps/chosen": -87.31756591796875,
|
| 1137 |
+
"logps/rejected": -72.79600524902344,
|
| 1138 |
+
"loss": 0.6247,
|
| 1139 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1140 |
+
"rewards/chosen": 0.09649817645549774,
|
| 1141 |
+
"rewards/margins": 0.1376573145389557,
|
| 1142 |
+
"rewards/rejected": -0.041159145534038544,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 0.5,
|
| 1147 |
+
"learning_rate": 2.7777777777777776e-07,
|
| 1148 |
+
"logits/chosen": -1.8448421955108643,
|
| 1149 |
+
"logits/rejected": -1.6317142248153687,
|
| 1150 |
+
"logps/chosen": -87.55412292480469,
|
| 1151 |
+
"logps/rejected": -75.91432189941406,
|
| 1152 |
+
"loss": 0.6245,
|
| 1153 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1154 |
+
"rewards/chosen": 0.10163182020187378,
|
| 1155 |
+
"rewards/margins": 0.14317932724952698,
|
| 1156 |
+
"rewards/rejected": -0.0415474958717823,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 0.51,
|
| 1161 |
+
"learning_rate": 2.73972602739726e-07,
|
| 1162 |
+
"logits/chosen": -1.800663709640503,
|
| 1163 |
+
"logits/rejected": -1.5165865421295166,
|
| 1164 |
+
"logps/chosen": -92.58036041259766,
|
| 1165 |
+
"logps/rejected": -73.22102355957031,
|
| 1166 |
+
"loss": 0.6216,
|
| 1167 |
+
"rewards/accuracies": 0.908333420753479,
|
| 1168 |
+
"rewards/chosen": 0.09606580436229706,
|
| 1169 |
+
"rewards/margins": 0.14485225081443787,
|
| 1170 |
+
"rewards/rejected": -0.04878643900156021,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 0.51,
|
| 1175 |
+
"learning_rate": 2.701674277016743e-07,
|
| 1176 |
+
"logits/chosen": -1.884385347366333,
|
| 1177 |
+
"logits/rejected": -1.6005618572235107,
|
| 1178 |
+
"logps/chosen": -94.69025421142578,
|
| 1179 |
+
"logps/rejected": -74.93373107910156,
|
| 1180 |
+
"loss": 0.6225,
|
| 1181 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1182 |
+
"rewards/chosen": 0.11928985267877579,
|
| 1183 |
+
"rewards/margins": 0.16091035306453705,
|
| 1184 |
+
"rewards/rejected": -0.04162050783634186,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 0.52,
|
| 1189 |
+
"learning_rate": 2.663622526636225e-07,
|
| 1190 |
+
"logits/chosen": -1.8744373321533203,
|
| 1191 |
+
"logits/rejected": -1.5992999076843262,
|
| 1192 |
+
"logps/chosen": -93.0952377319336,
|
| 1193 |
+
"logps/rejected": -76.86521911621094,
|
| 1194 |
+
"loss": 0.6187,
|
| 1195 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1196 |
+
"rewards/chosen": 0.12268207967281342,
|
| 1197 |
+
"rewards/margins": 0.15957853198051453,
|
| 1198 |
+
"rewards/rejected": -0.036896444857120514,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 0.53,
|
| 1203 |
+
"learning_rate": 2.6255707762557076e-07,
|
| 1204 |
+
"logits/chosen": -1.9210466146469116,
|
| 1205 |
+
"logits/rejected": -1.6140285730361938,
|
| 1206 |
+
"logps/chosen": -94.59263610839844,
|
| 1207 |
+
"logps/rejected": -75.74002838134766,
|
| 1208 |
+
"loss": 0.6174,
|
| 1209 |
+
"rewards/accuracies": 0.908333420753479,
|
| 1210 |
+
"rewards/chosen": 0.09473783522844315,
|
| 1211 |
+
"rewards/margins": 0.14541365206241608,
|
| 1212 |
+
"rewards/rejected": -0.050675809383392334,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 0.53,
|
| 1217 |
+
"learning_rate": 2.58751902587519e-07,
|
| 1218 |
+
"logits/chosen": -1.9041986465454102,
|
| 1219 |
+
"logits/rejected": -1.643701195716858,
|
| 1220 |
+
"logps/chosen": -92.38731384277344,
|
| 1221 |
+
"logps/rejected": -76.4394302368164,
|
| 1222 |
+
"loss": 0.6147,
|
| 1223 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1224 |
+
"rewards/chosen": 0.09374178946018219,
|
| 1225 |
+
"rewards/margins": 0.14521454274654388,
|
| 1226 |
+
"rewards/rejected": -0.05147276073694229,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 0.54,
|
| 1231 |
+
"learning_rate": 2.549467275494673e-07,
|
| 1232 |
+
"logits/chosen": -1.8866844177246094,
|
| 1233 |
+
"logits/rejected": -1.5806655883789062,
|
| 1234 |
+
"logps/chosen": -89.06953430175781,
|
| 1235 |
+
"logps/rejected": -75.25362396240234,
|
| 1236 |
+
"loss": 0.6127,
|
| 1237 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1238 |
+
"rewards/chosen": 0.1129077672958374,
|
| 1239 |
+
"rewards/margins": 0.1759023219347,
|
| 1240 |
+
"rewards/rejected": -0.06299454718828201,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 0.55,
|
| 1245 |
+
"learning_rate": 2.511415525114155e-07,
|
| 1246 |
+
"logits/chosen": -1.9603450298309326,
|
| 1247 |
+
"logits/rejected": -1.658216118812561,
|
| 1248 |
+
"logps/chosen": -91.58163452148438,
|
| 1249 |
+
"logps/rejected": -74.63312530517578,
|
| 1250 |
+
"loss": 0.6117,
|
| 1251 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1252 |
+
"rewards/chosen": 0.1321893036365509,
|
| 1253 |
+
"rewards/margins": 0.18407562375068665,
|
| 1254 |
+
"rewards/rejected": -0.051886312663555145,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 0.55,
|
| 1259 |
+
"eval_logits/chosen": -2.047370672225952,
|
| 1260 |
+
"eval_logits/rejected": -1.7785520553588867,
|
| 1261 |
+
"eval_logps/chosen": -90.88909149169922,
|
| 1262 |
+
"eval_logps/rejected": -72.66747283935547,
|
| 1263 |
+
"eval_loss": 0.6126503348350525,
|
| 1264 |
+
"eval_rewards/accuracies": 0.9222221970558167,
|
| 1265 |
+
"eval_rewards/chosen": 0.10966197401285172,
|
| 1266 |
+
"eval_rewards/margins": 0.16636402904987335,
|
| 1267 |
+
"eval_rewards/rejected": -0.05670207738876343,
|
| 1268 |
+
"eval_runtime": 117.9996,
|
| 1269 |
+
"eval_samples_per_second": 24.254,
|
| 1270 |
+
"eval_steps_per_second": 0.763,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 0.55,
|
| 1275 |
+
"learning_rate": 2.4733637747336376e-07,
|
| 1276 |
+
"logits/chosen": -1.9528766870498657,
|
| 1277 |
+
"logits/rejected": -1.594366192817688,
|
| 1278 |
+
"logps/chosen": -97.9262466430664,
|
| 1279 |
+
"logps/rejected": -73.47969055175781,
|
| 1280 |
+
"loss": 0.6145,
|
| 1281 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 1282 |
+
"rewards/chosen": 0.12691722810268402,
|
| 1283 |
+
"rewards/margins": 0.1831960827112198,
|
| 1284 |
+
"rewards/rejected": -0.056278862059116364,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 0.56,
|
| 1289 |
+
"learning_rate": 2.43531202435312e-07,
|
| 1290 |
+
"logits/chosen": -1.9507348537445068,
|
| 1291 |
+
"logits/rejected": -1.6255791187286377,
|
| 1292 |
+
"logps/chosen": -93.60541534423828,
|
| 1293 |
+
"logps/rejected": -74.82536315917969,
|
| 1294 |
+
"loss": 0.6094,
|
| 1295 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 1296 |
+
"rewards/chosen": 0.11346453428268433,
|
| 1297 |
+
"rewards/margins": 0.17843127250671387,
|
| 1298 |
+
"rewards/rejected": -0.06496672332286835,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 0.57,
|
| 1303 |
+
"learning_rate": 2.3972602739726023e-07,
|
| 1304 |
+
"logits/chosen": -1.7865006923675537,
|
| 1305 |
+
"logits/rejected": -1.5033773183822632,
|
| 1306 |
+
"logps/chosen": -89.35474395751953,
|
| 1307 |
+
"logps/rejected": -70.51929473876953,
|
| 1308 |
+
"loss": 0.6081,
|
| 1309 |
+
"rewards/accuracies": 0.9333332777023315,
|
| 1310 |
+
"rewards/chosen": 0.1159995049238205,
|
| 1311 |
+
"rewards/margins": 0.1914626657962799,
|
| 1312 |
+
"rewards/rejected": -0.07546313852071762,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 0.58,
|
| 1317 |
+
"learning_rate": 2.359208523592085e-07,
|
| 1318 |
+
"logits/chosen": -1.9572938680648804,
|
| 1319 |
+
"logits/rejected": -1.6931850910186768,
|
| 1320 |
+
"logps/chosen": -90.90480041503906,
|
| 1321 |
+
"logps/rejected": -77.85359191894531,
|
| 1322 |
+
"loss": 0.6088,
|
| 1323 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 1324 |
+
"rewards/chosen": 0.10320776700973511,
|
| 1325 |
+
"rewards/margins": 0.16814057528972626,
|
| 1326 |
+
"rewards/rejected": -0.06493280827999115,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 0.58,
|
| 1331 |
+
"learning_rate": 2.3211567732115676e-07,
|
| 1332 |
+
"logits/chosen": -2.0167603492736816,
|
| 1333 |
+
"logits/rejected": -1.715150237083435,
|
| 1334 |
+
"logps/chosen": -90.37574768066406,
|
| 1335 |
+
"logps/rejected": -77.0097427368164,
|
| 1336 |
+
"loss": 0.6039,
|
| 1337 |
+
"rewards/accuracies": 0.9333332777023315,
|
| 1338 |
+
"rewards/chosen": 0.1519448608160019,
|
| 1339 |
+
"rewards/margins": 0.21030330657958984,
|
| 1340 |
+
"rewards/rejected": -0.058358438313007355,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 0.59,
|
| 1345 |
+
"learning_rate": 2.28310502283105e-07,
|
| 1346 |
+
"logits/chosen": -1.8651390075683594,
|
| 1347 |
+
"logits/rejected": -1.5632587671279907,
|
| 1348 |
+
"logps/chosen": -94.64379119873047,
|
| 1349 |
+
"logps/rejected": -75.53591918945312,
|
| 1350 |
+
"loss": 0.6078,
|
| 1351 |
+
"rewards/accuracies": 0.9416667222976685,
|
| 1352 |
+
"rewards/chosen": 0.13561630249023438,
|
| 1353 |
+
"rewards/margins": 0.19979830086231232,
|
| 1354 |
+
"rewards/rejected": -0.06418199837207794,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 0.6,
|
| 1359 |
+
"learning_rate": 2.2450532724505325e-07,
|
| 1360 |
+
"logits/chosen": -1.9391330480575562,
|
| 1361 |
+
"logits/rejected": -1.663351058959961,
|
| 1362 |
+
"logps/chosen": -86.51952362060547,
|
| 1363 |
+
"logps/rejected": -73.44007873535156,
|
| 1364 |
+
"loss": 0.6057,
|
| 1365 |
+
"rewards/accuracies": 0.9333332777023315,
|
| 1366 |
+
"rewards/chosen": 0.1343904286623001,
|
| 1367 |
+
"rewards/margins": 0.1892845332622528,
|
| 1368 |
+
"rewards/rejected": -0.054894138127565384,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 0.6,
|
| 1373 |
+
"learning_rate": 2.207001522070015e-07,
|
| 1374 |
+
"logits/chosen": -1.842283010482788,
|
| 1375 |
+
"logits/rejected": -1.5807665586471558,
|
| 1376 |
+
"logps/chosen": -90.40550231933594,
|
| 1377 |
+
"logps/rejected": -77.0767593383789,
|
| 1378 |
+
"loss": 0.6045,
|
| 1379 |
+
"rewards/accuracies": 0.9249998927116394,
|
| 1380 |
+
"rewards/chosen": 0.13356170058250427,
|
| 1381 |
+
"rewards/margins": 0.1978417932987213,
|
| 1382 |
+
"rewards/rejected": -0.06428009271621704,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 0.61,
|
| 1387 |
+
"learning_rate": 2.1689497716894975e-07,
|
| 1388 |
+
"logits/chosen": -1.913578748703003,
|
| 1389 |
+
"logits/rejected": -1.648938536643982,
|
| 1390 |
+
"logps/chosen": -87.49850463867188,
|
| 1391 |
+
"logps/rejected": -74.8587646484375,
|
| 1392 |
+
"loss": 0.6026,
|
| 1393 |
+
"rewards/accuracies": 0.9333332777023315,
|
| 1394 |
+
"rewards/chosen": 0.13597458600997925,
|
| 1395 |
+
"rewards/margins": 0.20577768981456757,
|
| 1396 |
+
"rewards/rejected": -0.06980310380458832,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 0.62,
|
| 1401 |
+
"learning_rate": 2.13089802130898e-07,
|
| 1402 |
+
"logits/chosen": -1.8712265491485596,
|
| 1403 |
+
"logits/rejected": -1.5505828857421875,
|
| 1404 |
+
"logps/chosen": -92.89669036865234,
|
| 1405 |
+
"logps/rejected": -77.97479248046875,
|
| 1406 |
+
"loss": 0.6002,
|
| 1407 |
+
"rewards/accuracies": 0.9416667222976685,
|
| 1408 |
+
"rewards/chosen": 0.1514441967010498,
|
| 1409 |
+
"rewards/margins": 0.2083124816417694,
|
| 1410 |
+
"rewards/rejected": -0.056868284940719604,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 0.62,
|
| 1415 |
+
"eval_logits/chosen": -2.046788454055786,
|
| 1416 |
+
"eval_logits/rejected": -1.7776765823364258,
|
| 1417 |
+
"eval_logps/chosen": -90.75981140136719,
|
| 1418 |
+
"eval_logps/rejected": -72.78363037109375,
|
| 1419 |
+
"eval_loss": 0.6019285321235657,
|
| 1420 |
+
"eval_rewards/accuracies": 0.9277777671813965,
|
| 1421 |
+
"eval_rewards/chosen": 0.12258908152580261,
|
| 1422 |
+
"eval_rewards/margins": 0.1909066140651703,
|
| 1423 |
+
"eval_rewards/rejected": -0.06831753998994827,
|
| 1424 |
+
"eval_runtime": 117.8634,
|
| 1425 |
+
"eval_samples_per_second": 24.282,
|
| 1426 |
+
"eval_steps_per_second": 0.764,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 0.62,
|
| 1431 |
+
"learning_rate": 2.0928462709284625e-07,
|
| 1432 |
+
"logits/chosen": -1.7578074932098389,
|
| 1433 |
+
"logits/rejected": -1.445682406425476,
|
| 1434 |
+
"logps/chosen": -92.38932800292969,
|
| 1435 |
+
"logps/rejected": -74.73234558105469,
|
| 1436 |
+
"loss": 0.5977,
|
| 1437 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 1438 |
+
"rewards/chosen": 0.14447703957557678,
|
| 1439 |
+
"rewards/margins": 0.2118067443370819,
|
| 1440 |
+
"rewards/rejected": -0.06732969731092453,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 0.63,
|
| 1445 |
+
"learning_rate": 2.054794520547945e-07,
|
| 1446 |
+
"logits/chosen": -1.9356153011322021,
|
| 1447 |
+
"logits/rejected": -1.6309950351715088,
|
| 1448 |
+
"logps/chosen": -90.90232849121094,
|
| 1449 |
+
"logps/rejected": -75.2989730834961,
|
| 1450 |
+
"loss": 0.5962,
|
| 1451 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1452 |
+
"rewards/chosen": 0.14640390872955322,
|
| 1453 |
+
"rewards/margins": 0.21756890416145325,
|
| 1454 |
+
"rewards/rejected": -0.07116499543190002,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 0.64,
|
| 1459 |
+
"learning_rate": 2.0167427701674275e-07,
|
| 1460 |
+
"logits/chosen": -1.903969407081604,
|
| 1461 |
+
"logits/rejected": -1.6307262182235718,
|
| 1462 |
+
"logps/chosen": -91.58943939208984,
|
| 1463 |
+
"logps/rejected": -75.18601989746094,
|
| 1464 |
+
"loss": 0.6003,
|
| 1465 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1466 |
+
"rewards/chosen": 0.13933506608009338,
|
| 1467 |
+
"rewards/margins": 0.20189881324768066,
|
| 1468 |
+
"rewards/rejected": -0.06256375461816788,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 0.64,
|
| 1473 |
+
"learning_rate": 1.97869101978691e-07,
|
| 1474 |
+
"logits/chosen": -1.866371512413025,
|
| 1475 |
+
"logits/rejected": -1.6297776699066162,
|
| 1476 |
+
"logps/chosen": -91.3348617553711,
|
| 1477 |
+
"logps/rejected": -75.30091857910156,
|
| 1478 |
+
"loss": 0.5968,
|
| 1479 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1480 |
+
"rewards/chosen": 0.16422715783119202,
|
| 1481 |
+
"rewards/margins": 0.22424063086509705,
|
| 1482 |
+
"rewards/rejected": -0.06001347303390503,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 0.65,
|
| 1487 |
+
"learning_rate": 1.9406392694063925e-07,
|
| 1488 |
+
"logits/chosen": -1.860414743423462,
|
| 1489 |
+
"logits/rejected": -1.5813719034194946,
|
| 1490 |
+
"logps/chosen": -94.010498046875,
|
| 1491 |
+
"logps/rejected": -73.02679443359375,
|
| 1492 |
+
"loss": 0.595,
|
| 1493 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 1494 |
+
"rewards/chosen": 0.14462696015834808,
|
| 1495 |
+
"rewards/margins": 0.22170260548591614,
|
| 1496 |
+
"rewards/rejected": -0.07707564532756805,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 0.66,
|
| 1501 |
+
"learning_rate": 1.902587519025875e-07,
|
| 1502 |
+
"logits/chosen": -1.755253791809082,
|
| 1503 |
+
"logits/rejected": -1.5063341856002808,
|
| 1504 |
+
"logps/chosen": -89.60089111328125,
|
| 1505 |
+
"logps/rejected": -75.72676086425781,
|
| 1506 |
+
"loss": 0.5942,
|
| 1507 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1508 |
+
"rewards/chosen": 0.12570711970329285,
|
| 1509 |
+
"rewards/margins": 0.19667670130729675,
|
| 1510 |
+
"rewards/rejected": -0.0709695890545845,
|
| 1511 |
+
"step": 960
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 0.66,
|
| 1515 |
+
"learning_rate": 1.8645357686453575e-07,
|
| 1516 |
+
"logits/chosen": -1.9189532995224,
|
| 1517 |
+
"logits/rejected": -1.60482919216156,
|
| 1518 |
+
"logps/chosen": -91.0718765258789,
|
| 1519 |
+
"logps/rejected": -76.86122131347656,
|
| 1520 |
+
"loss": 0.5919,
|
| 1521 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1522 |
+
"rewards/chosen": 0.14830803871154785,
|
| 1523 |
+
"rewards/margins": 0.22014153003692627,
|
| 1524 |
+
"rewards/rejected": -0.07183349877595901,
|
| 1525 |
+
"step": 970
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 0.67,
|
| 1529 |
+
"learning_rate": 1.82648401826484e-07,
|
| 1530 |
+
"logits/chosen": -1.9582935571670532,
|
| 1531 |
+
"logits/rejected": -1.6904990673065186,
|
| 1532 |
+
"logps/chosen": -92.6358871459961,
|
| 1533 |
+
"logps/rejected": -75.13966369628906,
|
| 1534 |
+
"loss": 0.593,
|
| 1535 |
+
"rewards/accuracies": 0.9333333969116211,
|
| 1536 |
+
"rewards/chosen": 0.12001453340053558,
|
| 1537 |
+
"rewards/margins": 0.20278160274028778,
|
| 1538 |
+
"rewards/rejected": -0.0827670693397522,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 0.68,
|
| 1543 |
+
"learning_rate": 1.7884322678843225e-07,
|
| 1544 |
+
"logits/chosen": -1.7855488061904907,
|
| 1545 |
+
"logits/rejected": -1.498254418373108,
|
| 1546 |
+
"logps/chosen": -89.93516540527344,
|
| 1547 |
+
"logps/rejected": -71.06876373291016,
|
| 1548 |
+
"loss": 0.5923,
|
| 1549 |
+
"rewards/accuracies": 0.9666666984558105,
|
| 1550 |
+
"rewards/chosen": 0.13533183932304382,
|
| 1551 |
+
"rewards/margins": 0.2196374386548996,
|
| 1552 |
+
"rewards/rejected": -0.08430557698011398,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 0.68,
|
| 1557 |
+
"learning_rate": 1.750380517503805e-07,
|
| 1558 |
+
"logits/chosen": -1.9469058513641357,
|
| 1559 |
+
"logits/rejected": -1.655491828918457,
|
| 1560 |
+
"logps/chosen": -92.86201477050781,
|
| 1561 |
+
"logps/rejected": -77.54034423828125,
|
| 1562 |
+
"loss": 0.5912,
|
| 1563 |
+
"rewards/accuracies": 0.9499999284744263,
|
| 1564 |
+
"rewards/chosen": 0.14615695178508759,
|
| 1565 |
+
"rewards/margins": 0.22164049744606018,
|
| 1566 |
+
"rewards/rejected": -0.0754835233092308,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 0.68,
|
| 1571 |
+
"eval_logits/chosen": -2.046569585800171,
|
| 1572 |
+
"eval_logits/rejected": -1.7769988775253296,
|
| 1573 |
+
"eval_logps/chosen": -90.64215850830078,
|
| 1574 |
+
"eval_logps/rejected": -72.90531921386719,
|
| 1575 |
+
"eval_loss": 0.5911818742752075,
|
| 1576 |
+
"eval_rewards/accuracies": 0.9333333373069763,
|
| 1577 |
+
"eval_rewards/chosen": 0.13435469567775726,
|
| 1578 |
+
"eval_rewards/margins": 0.21484099328517914,
|
| 1579 |
+
"eval_rewards/rejected": -0.08048629015684128,
|
| 1580 |
+
"eval_runtime": 117.9339,
|
| 1581 |
+
"eval_samples_per_second": 24.268,
|
| 1582 |
+
"eval_steps_per_second": 0.763,
|
| 1583 |
+
"step": 1000
|
| 1584 |
+
},
|
| 1585 |
+
{
|
| 1586 |
+
"epoch": 0.69,
|
| 1587 |
+
"learning_rate": 1.7123287671232875e-07,
|
| 1588 |
+
"logits/chosen": -1.940800666809082,
|
| 1589 |
+
"logits/rejected": -1.6664402484893799,
|
| 1590 |
+
"logps/chosen": -86.94990539550781,
|
| 1591 |
+
"logps/rejected": -73.69987487792969,
|
| 1592 |
+
"loss": 0.5931,
|
| 1593 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1594 |
+
"rewards/chosen": 0.09665495157241821,
|
| 1595 |
+
"rewards/margins": 0.18213674426078796,
|
| 1596 |
+
"rewards/rejected": -0.08548180013895035,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 0.7,
|
| 1601 |
+
"learning_rate": 1.67427701674277e-07,
|
| 1602 |
+
"logits/chosen": -1.9208215475082397,
|
| 1603 |
+
"logits/rejected": -1.6697795391082764,
|
| 1604 |
+
"logps/chosen": -90.12159729003906,
|
| 1605 |
+
"logps/rejected": -75.43165588378906,
|
| 1606 |
+
"loss": 0.5936,
|
| 1607 |
+
"rewards/accuracies": 0.9083331823348999,
|
| 1608 |
+
"rewards/chosen": 0.1299724578857422,
|
| 1609 |
+
"rewards/margins": 0.21237368881702423,
|
| 1610 |
+
"rewards/rejected": -0.08240120857954025,
|
| 1611 |
+
"step": 1020
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 0.71,
|
| 1615 |
+
"learning_rate": 1.6362252663622525e-07,
|
| 1616 |
+
"logits/chosen": -1.8031762838363647,
|
| 1617 |
+
"logits/rejected": -1.552947759628296,
|
| 1618 |
+
"logps/chosen": -93.24359130859375,
|
| 1619 |
+
"logps/rejected": -73.16694641113281,
|
| 1620 |
+
"loss": 0.595,
|
| 1621 |
+
"rewards/accuracies": 0.9333333969116211,
|
| 1622 |
+
"rewards/chosen": 0.14134086668491364,
|
| 1623 |
+
"rewards/margins": 0.22129371762275696,
|
| 1624 |
+
"rewards/rejected": -0.07995286583900452,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 0.71,
|
| 1629 |
+
"learning_rate": 1.598173515981735e-07,
|
| 1630 |
+
"logits/chosen": -1.9094417095184326,
|
| 1631 |
+
"logits/rejected": -1.6032779216766357,
|
| 1632 |
+
"logps/chosen": -95.291015625,
|
| 1633 |
+
"logps/rejected": -76.30043029785156,
|
| 1634 |
+
"loss": 0.588,
|
| 1635 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1636 |
+
"rewards/chosen": 0.17578403651714325,
|
| 1637 |
+
"rewards/margins": 0.24385884404182434,
|
| 1638 |
+
"rewards/rejected": -0.0680748000741005,
|
| 1639 |
+
"step": 1040
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 0.72,
|
| 1643 |
+
"learning_rate": 1.5601217656012175e-07,
|
| 1644 |
+
"logits/chosen": -1.8715407848358154,
|
| 1645 |
+
"logits/rejected": -1.5754361152648926,
|
| 1646 |
+
"logps/chosen": -93.08331298828125,
|
| 1647 |
+
"logps/rejected": -71.55984497070312,
|
| 1648 |
+
"loss": 0.5885,
|
| 1649 |
+
"rewards/accuracies": 0.908333420753479,
|
| 1650 |
+
"rewards/chosen": 0.16208195686340332,
|
| 1651 |
+
"rewards/margins": 0.24299950897693634,
|
| 1652 |
+
"rewards/rejected": -0.08091756701469421,
|
| 1653 |
+
"step": 1050
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 0.73,
|
| 1657 |
+
"learning_rate": 1.5220700152207e-07,
|
| 1658 |
+
"logits/chosen": -1.9729465246200562,
|
| 1659 |
+
"logits/rejected": -1.6472351551055908,
|
| 1660 |
+
"logps/chosen": -96.95377349853516,
|
| 1661 |
+
"logps/rejected": -74.73013305664062,
|
| 1662 |
+
"loss": 0.5829,
|
| 1663 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 1664 |
+
"rewards/chosen": 0.17102012038230896,
|
| 1665 |
+
"rewards/margins": 0.26194968819618225,
|
| 1666 |
+
"rewards/rejected": -0.09092956781387329,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 0.73,
|
| 1671 |
+
"learning_rate": 1.4840182648401825e-07,
|
| 1672 |
+
"logits/chosen": -1.87423574924469,
|
| 1673 |
+
"logits/rejected": -1.5873512029647827,
|
| 1674 |
+
"logps/chosen": -92.32403564453125,
|
| 1675 |
+
"logps/rejected": -77.21125793457031,
|
| 1676 |
+
"loss": 0.5842,
|
| 1677 |
+
"rewards/accuracies": 0.966666579246521,
|
| 1678 |
+
"rewards/chosen": 0.1644188016653061,
|
| 1679 |
+
"rewards/margins": 0.27040743827819824,
|
| 1680 |
+
"rewards/rejected": -0.10598863661289215,
|
| 1681 |
+
"step": 1070
|
| 1682 |
+
},
|
| 1683 |
+
{
|
| 1684 |
+
"epoch": 0.74,
|
| 1685 |
+
"learning_rate": 1.445966514459665e-07,
|
| 1686 |
+
"logits/chosen": -1.8510887622833252,
|
| 1687 |
+
"logits/rejected": -1.5857640504837036,
|
| 1688 |
+
"logps/chosen": -92.05542755126953,
|
| 1689 |
+
"logps/rejected": -77.68433380126953,
|
| 1690 |
+
"loss": 0.5877,
|
| 1691 |
+
"rewards/accuracies": 0.9333333969116211,
|
| 1692 |
+
"rewards/chosen": 0.146043062210083,
|
| 1693 |
+
"rewards/margins": 0.22200524806976318,
|
| 1694 |
+
"rewards/rejected": -0.07596220076084137,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 0.75,
|
| 1699 |
+
"learning_rate": 1.4079147640791475e-07,
|
| 1700 |
+
"logits/chosen": -1.8213762044906616,
|
| 1701 |
+
"logits/rejected": -1.5474189519882202,
|
| 1702 |
+
"logps/chosen": -91.6738052368164,
|
| 1703 |
+
"logps/rejected": -74.36346435546875,
|
| 1704 |
+
"loss": 0.5796,
|
| 1705 |
+
"rewards/accuracies": 0.9833332896232605,
|
| 1706 |
+
"rewards/chosen": 0.1721878945827484,
|
| 1707 |
+
"rewards/margins": 0.26567989587783813,
|
| 1708 |
+
"rewards/rejected": -0.09349202364683151,
|
| 1709 |
+
"step": 1090
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 0.75,
|
| 1713 |
+
"learning_rate": 1.36986301369863e-07,
|
| 1714 |
+
"logits/chosen": -1.8960554599761963,
|
| 1715 |
+
"logits/rejected": -1.5947376489639282,
|
| 1716 |
+
"logps/chosen": -95.58143615722656,
|
| 1717 |
+
"logps/rejected": -75.37010192871094,
|
| 1718 |
+
"loss": 0.5822,
|
| 1719 |
+
"rewards/accuracies": 0.9166666865348816,
|
| 1720 |
+
"rewards/chosen": 0.18196699023246765,
|
| 1721 |
+
"rewards/margins": 0.25899866223335266,
|
| 1722 |
+
"rewards/rejected": -0.0770316869020462,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 0.75,
|
| 1727 |
+
"eval_logits/chosen": -2.0461771488189697,
|
| 1728 |
+
"eval_logits/rejected": -1.7763375043869019,
|
| 1729 |
+
"eval_logps/chosen": -90.54474639892578,
|
| 1730 |
+
"eval_logps/rejected": -73.00917053222656,
|
| 1731 |
+
"eval_loss": 0.5822051763534546,
|
| 1732 |
+
"eval_rewards/accuracies": 0.9472222328186035,
|
| 1733 |
+
"eval_rewards/chosen": 0.14409679174423218,
|
| 1734 |
+
"eval_rewards/margins": 0.23496907949447632,
|
| 1735 |
+
"eval_rewards/rejected": -0.09087225794792175,
|
| 1736 |
+
"eval_runtime": 117.8174,
|
| 1737 |
+
"eval_samples_per_second": 24.292,
|
| 1738 |
+
"eval_steps_per_second": 0.764,
|
| 1739 |
+
"step": 1100
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 0.76,
|
| 1743 |
+
"learning_rate": 1.3318112633181125e-07,
|
| 1744 |
+
"logits/chosen": -1.893864393234253,
|
| 1745 |
+
"logits/rejected": -1.5917404890060425,
|
| 1746 |
+
"logps/chosen": -90.5267105102539,
|
| 1747 |
+
"logps/rejected": -73.85464477539062,
|
| 1748 |
+
"loss": 0.5809,
|
| 1749 |
+
"rewards/accuracies": 0.9666666984558105,
|
| 1750 |
+
"rewards/chosen": 0.18791857361793518,
|
| 1751 |
+
"rewards/margins": 0.26672154664993286,
|
| 1752 |
+
"rewards/rejected": -0.07880295813083649,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 0.77,
|
| 1757 |
+
"learning_rate": 1.293759512937595e-07,
|
| 1758 |
+
"logits/chosen": -1.851485013961792,
|
| 1759 |
+
"logits/rejected": -1.5561844110488892,
|
| 1760 |
+
"logps/chosen": -91.46730041503906,
|
| 1761 |
+
"logps/rejected": -75.5124740600586,
|
| 1762 |
+
"loss": 0.5803,
|
| 1763 |
+
"rewards/accuracies": 0.9083333015441895,
|
| 1764 |
+
"rewards/chosen": 0.15885117650032043,
|
| 1765 |
+
"rewards/margins": 0.24810326099395752,
|
| 1766 |
+
"rewards/rejected": -0.08925210684537888,
|
| 1767 |
+
"step": 1120
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 0.77,
|
| 1771 |
+
"learning_rate": 1.2557077625570775e-07,
|
| 1772 |
+
"logits/chosen": -1.9113174676895142,
|
| 1773 |
+
"logits/rejected": -1.6106466054916382,
|
| 1774 |
+
"logps/chosen": -88.4524917602539,
|
| 1775 |
+
"logps/rejected": -71.1933364868164,
|
| 1776 |
+
"loss": 0.5805,
|
| 1777 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1778 |
+
"rewards/chosen": 0.1571049988269806,
|
| 1779 |
+
"rewards/margins": 0.23198041319847107,
|
| 1780 |
+
"rewards/rejected": -0.07487543672323227,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 0.78,
|
| 1785 |
+
"learning_rate": 1.21765601217656e-07,
|
| 1786 |
+
"logits/chosen": -1.9750347137451172,
|
| 1787 |
+
"logits/rejected": -1.6996724605560303,
|
| 1788 |
+
"logps/chosen": -91.44354248046875,
|
| 1789 |
+
"logps/rejected": -77.03483581542969,
|
| 1790 |
+
"loss": 0.5768,
|
| 1791 |
+
"rewards/accuracies": 0.9416666030883789,
|
| 1792 |
+
"rewards/chosen": 0.14498132467269897,
|
| 1793 |
+
"rewards/margins": 0.23422233760356903,
|
| 1794 |
+
"rewards/rejected": -0.08924100548028946,
|
| 1795 |
+
"step": 1140
|
| 1796 |
+
},
|
| 1797 |
+
{
|
| 1798 |
+
"epoch": 0.79,
|
| 1799 |
+
"learning_rate": 1.1796042617960425e-07,
|
| 1800 |
+
"logits/chosen": -1.9331929683685303,
|
| 1801 |
+
"logits/rejected": -1.6460412740707397,
|
| 1802 |
+
"logps/chosen": -84.95231628417969,
|
| 1803 |
+
"logps/rejected": -73.00121307373047,
|
| 1804 |
+
"loss": 0.5759,
|
| 1805 |
+
"rewards/accuracies": 0.9416667222976685,
|
| 1806 |
+
"rewards/chosen": 0.17358574271202087,
|
| 1807 |
+
"rewards/margins": 0.2643177807331085,
|
| 1808 |
+
"rewards/rejected": -0.09073203802108765,
|
| 1809 |
+
"step": 1150
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 0.79,
|
| 1813 |
+
"learning_rate": 1.141552511415525e-07,
|
| 1814 |
+
"logits/chosen": -1.9368797540664673,
|
| 1815 |
+
"logits/rejected": -1.6109968423843384,
|
| 1816 |
+
"logps/chosen": -95.30493927001953,
|
| 1817 |
+
"logps/rejected": -79.38502502441406,
|
| 1818 |
+
"loss": 0.5725,
|
| 1819 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 1820 |
+
"rewards/chosen": 0.16715973615646362,
|
| 1821 |
+
"rewards/margins": 0.26029014587402344,
|
| 1822 |
+
"rewards/rejected": -0.09313040971755981,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 0.8,
|
| 1827 |
+
"learning_rate": 1.1035007610350075e-07,
|
| 1828 |
+
"logits/chosen": -1.9934200048446655,
|
| 1829 |
+
"logits/rejected": -1.7067596912384033,
|
| 1830 |
+
"logps/chosen": -91.30809020996094,
|
| 1831 |
+
"logps/rejected": -75.1029052734375,
|
| 1832 |
+
"loss": 0.577,
|
| 1833 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1834 |
+
"rewards/chosen": 0.1644877940416336,
|
| 1835 |
+
"rewards/margins": 0.2513524889945984,
|
| 1836 |
+
"rewards/rejected": -0.08686470985412598,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 0.81,
|
| 1841 |
+
"learning_rate": 1.06544901065449e-07,
|
| 1842 |
+
"logits/chosen": -1.8585717678070068,
|
| 1843 |
+
"logits/rejected": -1.593147873878479,
|
| 1844 |
+
"logps/chosen": -94.06452941894531,
|
| 1845 |
+
"logps/rejected": -78.18238830566406,
|
| 1846 |
+
"loss": 0.5798,
|
| 1847 |
+
"rewards/accuracies": 0.9166666865348816,
|
| 1848 |
+
"rewards/chosen": 0.1385236382484436,
|
| 1849 |
+
"rewards/margins": 0.2384444922208786,
|
| 1850 |
+
"rewards/rejected": -0.09992088377475739,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 0.81,
|
| 1855 |
+
"learning_rate": 1.0273972602739725e-07,
|
| 1856 |
+
"logits/chosen": -1.921449065208435,
|
| 1857 |
+
"logits/rejected": -1.6535272598266602,
|
| 1858 |
+
"logps/chosen": -87.8677749633789,
|
| 1859 |
+
"logps/rejected": -78.9840087890625,
|
| 1860 |
+
"loss": 0.5767,
|
| 1861 |
+
"rewards/accuracies": 0.9666666984558105,
|
| 1862 |
+
"rewards/chosen": 0.14824248850345612,
|
| 1863 |
+
"rewards/margins": 0.25046244263648987,
|
| 1864 |
+
"rewards/rejected": -0.10221991688013077,
|
| 1865 |
+
"step": 1190
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 0.82,
|
| 1869 |
+
"learning_rate": 9.89345509893455e-08,
|
| 1870 |
+
"logits/chosen": -1.8294260501861572,
|
| 1871 |
+
"logits/rejected": -1.5607296228408813,
|
| 1872 |
+
"logps/chosen": -94.44920349121094,
|
| 1873 |
+
"logps/rejected": -77.64144134521484,
|
| 1874 |
+
"loss": 0.5789,
|
| 1875 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 1876 |
+
"rewards/chosen": 0.16983038187026978,
|
| 1877 |
+
"rewards/margins": 0.2730123996734619,
|
| 1878 |
+
"rewards/rejected": -0.10318204015493393,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 0.82,
|
| 1883 |
+
"eval_logits/chosen": -2.0464720726013184,
|
| 1884 |
+
"eval_logits/rejected": -1.7763383388519287,
|
| 1885 |
+
"eval_logps/chosen": -90.46904754638672,
|
| 1886 |
+
"eval_logps/rejected": -73.09234619140625,
|
| 1887 |
+
"eval_loss": 0.5758996605873108,
|
| 1888 |
+
"eval_rewards/accuracies": 0.9333333373069763,
|
| 1889 |
+
"eval_rewards/chosen": 0.1516665369272232,
|
| 1890 |
+
"eval_rewards/margins": 0.2508557140827179,
|
| 1891 |
+
"eval_rewards/rejected": -0.0991891473531723,
|
| 1892 |
+
"eval_runtime": 117.7184,
|
| 1893 |
+
"eval_samples_per_second": 24.312,
|
| 1894 |
+
"eval_steps_per_second": 0.765,
|
| 1895 |
+
"step": 1200
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 0.83,
|
| 1899 |
+
"learning_rate": 9.512937595129374e-08,
|
| 1900 |
+
"logits/chosen": -1.8505995273590088,
|
| 1901 |
+
"logits/rejected": -1.570237398147583,
|
| 1902 |
+
"logps/chosen": -95.08009338378906,
|
| 1903 |
+
"logps/rejected": -75.39155578613281,
|
| 1904 |
+
"loss": 0.5739,
|
| 1905 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 1906 |
+
"rewards/chosen": 0.14622533321380615,
|
| 1907 |
+
"rewards/margins": 0.23580579459667206,
|
| 1908 |
+
"rewards/rejected": -0.08958044648170471,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 0.84,
|
| 1913 |
+
"learning_rate": 9.1324200913242e-08,
|
| 1914 |
+
"logits/chosen": -1.8380448818206787,
|
| 1915 |
+
"logits/rejected": -1.5534820556640625,
|
| 1916 |
+
"logps/chosen": -89.47401428222656,
|
| 1917 |
+
"logps/rejected": -73.60565948486328,
|
| 1918 |
+
"loss": 0.5772,
|
| 1919 |
+
"rewards/accuracies": 0.9416667222976685,
|
| 1920 |
+
"rewards/chosen": 0.1780690848827362,
|
| 1921 |
+
"rewards/margins": 0.2615908980369568,
|
| 1922 |
+
"rewards/rejected": -0.08352181315422058,
|
| 1923 |
+
"step": 1220
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 0.84,
|
| 1927 |
+
"learning_rate": 8.751902587519024e-08,
|
| 1928 |
+
"logits/chosen": -1.975015640258789,
|
| 1929 |
+
"logits/rejected": -1.6878995895385742,
|
| 1930 |
+
"logps/chosen": -93.83460998535156,
|
| 1931 |
+
"logps/rejected": -76.86270904541016,
|
| 1932 |
+
"loss": 0.576,
|
| 1933 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1934 |
+
"rewards/chosen": 0.17377497255802155,
|
| 1935 |
+
"rewards/margins": 0.25830045342445374,
|
| 1936 |
+
"rewards/rejected": -0.08452550321817398,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 0.85,
|
| 1941 |
+
"learning_rate": 8.37138508371385e-08,
|
| 1942 |
+
"logits/chosen": -1.8430579900741577,
|
| 1943 |
+
"logits/rejected": -1.5395987033843994,
|
| 1944 |
+
"logps/chosen": -92.52012634277344,
|
| 1945 |
+
"logps/rejected": -79.96121215820312,
|
| 1946 |
+
"loss": 0.5745,
|
| 1947 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 1948 |
+
"rewards/chosen": 0.15905624628067017,
|
| 1949 |
+
"rewards/margins": 0.25326135754585266,
|
| 1950 |
+
"rewards/rejected": -0.0942051112651825,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 0.86,
|
| 1955 |
+
"learning_rate": 7.990867579908676e-08,
|
| 1956 |
+
"logits/chosen": -1.8872146606445312,
|
| 1957 |
+
"logits/rejected": -1.5909473896026611,
|
| 1958 |
+
"logps/chosen": -89.97090148925781,
|
| 1959 |
+
"logps/rejected": -77.28529357910156,
|
| 1960 |
+
"loss": 0.5727,
|
| 1961 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1962 |
+
"rewards/chosen": 0.16229455173015594,
|
| 1963 |
+
"rewards/margins": 0.2598329186439514,
|
| 1964 |
+
"rewards/rejected": -0.09753839671611786,
|
| 1965 |
+
"step": 1250
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 0.86,
|
| 1969 |
+
"learning_rate": 7.6103500761035e-08,
|
| 1970 |
+
"logits/chosen": -1.8499984741210938,
|
| 1971 |
+
"logits/rejected": -1.5525275468826294,
|
| 1972 |
+
"logps/chosen": -93.28520202636719,
|
| 1973 |
+
"logps/rejected": -74.24462890625,
|
| 1974 |
+
"loss": 0.5731,
|
| 1975 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1976 |
+
"rewards/chosen": 0.14929968118667603,
|
| 1977 |
+
"rewards/margins": 0.2529754042625427,
|
| 1978 |
+
"rewards/rejected": -0.10367570072412491,
|
| 1979 |
+
"step": 1260
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 0.87,
|
| 1983 |
+
"learning_rate": 7.229832572298326e-08,
|
| 1984 |
+
"logits/chosen": -2.009666681289673,
|
| 1985 |
+
"logits/rejected": -1.7177025079727173,
|
| 1986 |
+
"logps/chosen": -89.4050521850586,
|
| 1987 |
+
"logps/rejected": -79.06332397460938,
|
| 1988 |
+
"loss": 0.5712,
|
| 1989 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1990 |
+
"rewards/chosen": 0.1704416573047638,
|
| 1991 |
+
"rewards/margins": 0.27492305636405945,
|
| 1992 |
+
"rewards/rejected": -0.10448137670755386,
|
| 1993 |
+
"step": 1270
|
| 1994 |
+
},
|
| 1995 |
+
{
|
| 1996 |
+
"epoch": 0.88,
|
| 1997 |
+
"learning_rate": 6.84931506849315e-08,
|
| 1998 |
+
"logits/chosen": -1.81943678855896,
|
| 1999 |
+
"logits/rejected": -1.5563119649887085,
|
| 2000 |
+
"logps/chosen": -88.48675537109375,
|
| 2001 |
+
"logps/rejected": -72.3918685913086,
|
| 2002 |
+
"loss": 0.5729,
|
| 2003 |
+
"rewards/accuracies": 0.8916667103767395,
|
| 2004 |
+
"rewards/chosen": 0.13842932879924774,
|
| 2005 |
+
"rewards/margins": 0.24020667374134064,
|
| 2006 |
+
"rewards/rejected": -0.10177735984325409,
|
| 2007 |
+
"step": 1280
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 0.88,
|
| 2011 |
+
"learning_rate": 6.468797564687976e-08,
|
| 2012 |
+
"logits/chosen": -1.8918097019195557,
|
| 2013 |
+
"logits/rejected": -1.6081184148788452,
|
| 2014 |
+
"logps/chosen": -87.69499969482422,
|
| 2015 |
+
"logps/rejected": -71.82918548583984,
|
| 2016 |
+
"loss": 0.5745,
|
| 2017 |
+
"rewards/accuracies": 0.9333332777023315,
|
| 2018 |
+
"rewards/chosen": 0.17439430952072144,
|
| 2019 |
+
"rewards/margins": 0.2683844268321991,
|
| 2020 |
+
"rewards/rejected": -0.09399012476205826,
|
| 2021 |
+
"step": 1290
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 0.89,
|
| 2025 |
+
"learning_rate": 6.0882800608828e-08,
|
| 2026 |
+
"logits/chosen": -1.7783477306365967,
|
| 2027 |
+
"logits/rejected": -1.4948413372039795,
|
| 2028 |
+
"logps/chosen": -89.221435546875,
|
| 2029 |
+
"logps/rejected": -72.36725616455078,
|
| 2030 |
+
"loss": 0.5689,
|
| 2031 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 2032 |
+
"rewards/chosen": 0.16254201531410217,
|
| 2033 |
+
"rewards/margins": 0.27369171380996704,
|
| 2034 |
+
"rewards/rejected": -0.11114968359470367,
|
| 2035 |
+
"step": 1300
|
| 2036 |
+
},
|
| 2037 |
+
{
|
| 2038 |
+
"epoch": 0.89,
|
| 2039 |
+
"eval_logits/chosen": -2.0464510917663574,
|
| 2040 |
+
"eval_logits/rejected": -1.7761657238006592,
|
| 2041 |
+
"eval_logps/chosen": -90.43045043945312,
|
| 2042 |
+
"eval_logps/rejected": -73.13316345214844,
|
| 2043 |
+
"eval_loss": 0.5722280740737915,
|
| 2044 |
+
"eval_rewards/accuracies": 0.949999988079071,
|
| 2045 |
+
"eval_rewards/chosen": 0.15552671253681183,
|
| 2046 |
+
"eval_rewards/margins": 0.2587975263595581,
|
| 2047 |
+
"eval_rewards/rejected": -0.10327085852622986,
|
| 2048 |
+
"eval_runtime": 118.3212,
|
| 2049 |
+
"eval_samples_per_second": 24.188,
|
| 2050 |
+
"eval_steps_per_second": 0.761,
|
| 2051 |
+
"step": 1300
|
| 2052 |
+
},
|
| 2053 |
+
{
|
| 2054 |
+
"epoch": 0.9,
|
| 2055 |
+
"learning_rate": 5.707762557077625e-08,
|
| 2056 |
+
"logits/chosen": -1.7794984579086304,
|
| 2057 |
+
"logits/rejected": -1.5128498077392578,
|
| 2058 |
+
"logps/chosen": -94.00843048095703,
|
| 2059 |
+
"logps/rejected": -75.231201171875,
|
| 2060 |
+
"loss": 0.5719,
|
| 2061 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2062 |
+
"rewards/chosen": 0.18278029561042786,
|
| 2063 |
+
"rewards/margins": 0.2786504328250885,
|
| 2064 |
+
"rewards/rejected": -0.09587012976408005,
|
| 2065 |
+
"step": 1310
|
| 2066 |
+
},
|
| 2067 |
+
{
|
| 2068 |
+
"epoch": 0.9,
|
| 2069 |
+
"learning_rate": 5.32724505327245e-08,
|
| 2070 |
+
"logits/chosen": -1.9731611013412476,
|
| 2071 |
+
"logits/rejected": -1.6866945028305054,
|
| 2072 |
+
"logps/chosen": -93.30955505371094,
|
| 2073 |
+
"logps/rejected": -78.57191467285156,
|
| 2074 |
+
"loss": 0.571,
|
| 2075 |
+
"rewards/accuracies": 0.9583333134651184,
|
| 2076 |
+
"rewards/chosen": 0.15462180972099304,
|
| 2077 |
+
"rewards/margins": 0.26251837611198425,
|
| 2078 |
+
"rewards/rejected": -0.10789655148983002,
|
| 2079 |
+
"step": 1320
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 0.91,
|
| 2083 |
+
"learning_rate": 4.946727549467275e-08,
|
| 2084 |
+
"logits/chosen": -1.9133844375610352,
|
| 2085 |
+
"logits/rejected": -1.636694312095642,
|
| 2086 |
+
"logps/chosen": -88.56840515136719,
|
| 2087 |
+
"logps/rejected": -76.69215393066406,
|
| 2088 |
+
"loss": 0.5716,
|
| 2089 |
+
"rewards/accuracies": 0.966666579246521,
|
| 2090 |
+
"rewards/chosen": 0.15411558747291565,
|
| 2091 |
+
"rewards/margins": 0.26415151357650757,
|
| 2092 |
+
"rewards/rejected": -0.11003589630126953,
|
| 2093 |
+
"step": 1330
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 0.92,
|
| 2097 |
+
"learning_rate": 4.5662100456621e-08,
|
| 2098 |
+
"logits/chosen": -1.8889293670654297,
|
| 2099 |
+
"logits/rejected": -1.5668977499008179,
|
| 2100 |
+
"logps/chosen": -93.04912567138672,
|
| 2101 |
+
"logps/rejected": -74.54277038574219,
|
| 2102 |
+
"loss": 0.5671,
|
| 2103 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 2104 |
+
"rewards/chosen": 0.1801871955394745,
|
| 2105 |
+
"rewards/margins": 0.2698620557785034,
|
| 2106 |
+
"rewards/rejected": -0.08967487514019012,
|
| 2107 |
+
"step": 1340
|
| 2108 |
+
},
|
| 2109 |
+
{
|
| 2110 |
+
"epoch": 0.92,
|
| 2111 |
+
"learning_rate": 4.185692541856925e-08,
|
| 2112 |
+
"logits/chosen": -1.853014588356018,
|
| 2113 |
+
"logits/rejected": -1.6049798727035522,
|
| 2114 |
+
"logps/chosen": -92.84537506103516,
|
| 2115 |
+
"logps/rejected": -76.91346740722656,
|
| 2116 |
+
"loss": 0.5734,
|
| 2117 |
+
"rewards/accuracies": 0.9333332777023315,
|
| 2118 |
+
"rewards/chosen": 0.1830371469259262,
|
| 2119 |
+
"rewards/margins": 0.27097687125205994,
|
| 2120 |
+
"rewards/rejected": -0.08793972432613373,
|
| 2121 |
+
"step": 1350
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 0.93,
|
| 2125 |
+
"learning_rate": 3.80517503805175e-08,
|
| 2126 |
+
"logits/chosen": -1.8321574926376343,
|
| 2127 |
+
"logits/rejected": -1.559768795967102,
|
| 2128 |
+
"logps/chosen": -90.3908462524414,
|
| 2129 |
+
"logps/rejected": -77.5791015625,
|
| 2130 |
+
"loss": 0.5728,
|
| 2131 |
+
"rewards/accuracies": 0.9333333969116211,
|
| 2132 |
+
"rewards/chosen": 0.15050294995307922,
|
| 2133 |
+
"rewards/margins": 0.2644408941268921,
|
| 2134 |
+
"rewards/rejected": -0.11393795162439346,
|
| 2135 |
+
"step": 1360
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 0.94,
|
| 2139 |
+
"learning_rate": 3.424657534246575e-08,
|
| 2140 |
+
"logits/chosen": -1.8819458484649658,
|
| 2141 |
+
"logits/rejected": -1.5830302238464355,
|
| 2142 |
+
"logps/chosen": -96.20283508300781,
|
| 2143 |
+
"logps/rejected": -79.02750396728516,
|
| 2144 |
+
"loss": 0.57,
|
| 2145 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 2146 |
+
"rewards/chosen": 0.1495673507452011,
|
| 2147 |
+
"rewards/margins": 0.26126623153686523,
|
| 2148 |
+
"rewards/rejected": -0.11169885098934174,
|
| 2149 |
+
"step": 1370
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 0.94,
|
| 2153 |
+
"learning_rate": 3.0441400304414e-08,
|
| 2154 |
+
"logits/chosen": -1.8712704181671143,
|
| 2155 |
+
"logits/rejected": -1.5640711784362793,
|
| 2156 |
+
"logps/chosen": -87.56044006347656,
|
| 2157 |
+
"logps/rejected": -73.2048110961914,
|
| 2158 |
+
"loss": 0.5701,
|
| 2159 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2160 |
+
"rewards/chosen": 0.15671199560165405,
|
| 2161 |
+
"rewards/margins": 0.2683314383029938,
|
| 2162 |
+
"rewards/rejected": -0.11161943525075912,
|
| 2163 |
+
"step": 1380
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 0.95,
|
| 2167 |
+
"learning_rate": 2.663622526636225e-08,
|
| 2168 |
+
"logits/chosen": -1.9705963134765625,
|
| 2169 |
+
"logits/rejected": -1.6459972858428955,
|
| 2170 |
+
"logps/chosen": -93.333251953125,
|
| 2171 |
+
"logps/rejected": -75.71430969238281,
|
| 2172 |
+
"loss": 0.5669,
|
| 2173 |
+
"rewards/accuracies": 0.9833332896232605,
|
| 2174 |
+
"rewards/chosen": 0.22007617354393005,
|
| 2175 |
+
"rewards/margins": 0.3152759373188019,
|
| 2176 |
+
"rewards/rejected": -0.09519973397254944,
|
| 2177 |
+
"step": 1390
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 0.96,
|
| 2181 |
+
"learning_rate": 2.28310502283105e-08,
|
| 2182 |
+
"logits/chosen": -1.9400713443756104,
|
| 2183 |
+
"logits/rejected": -1.6670045852661133,
|
| 2184 |
+
"logps/chosen": -90.7359619140625,
|
| 2185 |
+
"logps/rejected": -72.49601745605469,
|
| 2186 |
+
"loss": 0.5694,
|
| 2187 |
+
"rewards/accuracies": 0.9416667222976685,
|
| 2188 |
+
"rewards/chosen": 0.18735943734645844,
|
| 2189 |
+
"rewards/margins": 0.2668268084526062,
|
| 2190 |
+
"rewards/rejected": -0.07946738600730896,
|
| 2191 |
+
"step": 1400
|
| 2192 |
+
},
|
| 2193 |
+
{
|
| 2194 |
+
"epoch": 0.96,
|
| 2195 |
+
"eval_logits/chosen": -2.0464956760406494,
|
| 2196 |
+
"eval_logits/rejected": -1.776126742362976,
|
| 2197 |
+
"eval_logps/chosen": -90.40695190429688,
|
| 2198 |
+
"eval_logps/rejected": -73.16618347167969,
|
| 2199 |
+
"eval_loss": 0.5701765418052673,
|
| 2200 |
+
"eval_rewards/accuracies": 0.9416666626930237,
|
| 2201 |
+
"eval_rewards/chosen": 0.15787601470947266,
|
| 2202 |
+
"eval_rewards/margins": 0.26444879174232483,
|
| 2203 |
+
"eval_rewards/rejected": -0.10657278448343277,
|
| 2204 |
+
"eval_runtime": 118.3051,
|
| 2205 |
+
"eval_samples_per_second": 24.192,
|
| 2206 |
+
"eval_steps_per_second": 0.761,
|
| 2207 |
+
"step": 1400
|
| 2208 |
+
},
|
| 2209 |
+
{
|
| 2210 |
+
"epoch": 0.97,
|
| 2211 |
+
"learning_rate": 1.902587519025875e-08,
|
| 2212 |
+
"logits/chosen": -1.9084300994873047,
|
| 2213 |
+
"logits/rejected": -1.5941137075424194,
|
| 2214 |
+
"logps/chosen": -91.3935546875,
|
| 2215 |
+
"logps/rejected": -76.21849060058594,
|
| 2216 |
+
"loss": 0.5703,
|
| 2217 |
+
"rewards/accuracies": 0.9416666030883789,
|
| 2218 |
+
"rewards/chosen": 0.18224991858005524,
|
| 2219 |
+
"rewards/margins": 0.2809165418148041,
|
| 2220 |
+
"rewards/rejected": -0.09866663068532944,
|
| 2221 |
+
"step": 1410
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 0.97,
|
| 2225 |
+
"learning_rate": 1.5220700152207e-08,
|
| 2226 |
+
"logits/chosen": -1.8584785461425781,
|
| 2227 |
+
"logits/rejected": -1.5252989530563354,
|
| 2228 |
+
"logps/chosen": -94.4938735961914,
|
| 2229 |
+
"logps/rejected": -76.2500228881836,
|
| 2230 |
+
"loss": 0.5705,
|
| 2231 |
+
"rewards/accuracies": 0.9166666269302368,
|
| 2232 |
+
"rewards/chosen": 0.17414768040180206,
|
| 2233 |
+
"rewards/margins": 0.2775163948535919,
|
| 2234 |
+
"rewards/rejected": -0.10336872190237045,
|
| 2235 |
+
"step": 1420
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 0.98,
|
| 2239 |
+
"learning_rate": 1.141552511415525e-08,
|
| 2240 |
+
"logits/chosen": -1.8703104257583618,
|
| 2241 |
+
"logits/rejected": -1.5645638704299927,
|
| 2242 |
+
"logps/chosen": -92.91007232666016,
|
| 2243 |
+
"logps/rejected": -72.98766326904297,
|
| 2244 |
+
"loss": 0.5703,
|
| 2245 |
+
"rewards/accuracies": 0.9416666030883789,
|
| 2246 |
+
"rewards/chosen": 0.1589893251657486,
|
| 2247 |
+
"rewards/margins": 0.26451554894447327,
|
| 2248 |
+
"rewards/rejected": -0.10552623122930527,
|
| 2249 |
+
"step": 1430
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 0.99,
|
| 2253 |
+
"learning_rate": 7.6103500761035e-09,
|
| 2254 |
+
"logits/chosen": -1.9101779460906982,
|
| 2255 |
+
"logits/rejected": -1.6234633922576904,
|
| 2256 |
+
"logps/chosen": -88.99760437011719,
|
| 2257 |
+
"logps/rejected": -73.46092224121094,
|
| 2258 |
+
"loss": 0.5694,
|
| 2259 |
+
"rewards/accuracies": 0.9333333969116211,
|
| 2260 |
+
"rewards/chosen": 0.17230169475078583,
|
| 2261 |
+
"rewards/margins": 0.25403517484664917,
|
| 2262 |
+
"rewards/rejected": -0.08173345029354095,
|
| 2263 |
+
"step": 1440
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 0.99,
|
| 2267 |
+
"learning_rate": 3.80517503805175e-09,
|
| 2268 |
+
"logits/chosen": -1.837728500366211,
|
| 2269 |
+
"logits/rejected": -1.6049699783325195,
|
| 2270 |
+
"logps/chosen": -92.38600158691406,
|
| 2271 |
+
"logps/rejected": -78.56689453125,
|
| 2272 |
+
"loss": 0.571,
|
| 2273 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2274 |
+
"rewards/chosen": 0.18492794036865234,
|
| 2275 |
+
"rewards/margins": 0.28263044357299805,
|
| 2276 |
+
"rewards/rejected": -0.0977025032043457,
|
| 2277 |
+
"step": 1450
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 1.0,
|
| 2281 |
+
"learning_rate": 0.0,
|
| 2282 |
+
"logits/chosen": -1.8631808757781982,
|
| 2283 |
+
"logits/rejected": -1.5711183547973633,
|
| 2284 |
+
"logps/chosen": -90.29685974121094,
|
| 2285 |
+
"logps/rejected": -76.57972717285156,
|
| 2286 |
+
"loss": 0.5684,
|
| 2287 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 2288 |
+
"rewards/chosen": 0.18871954083442688,
|
| 2289 |
+
"rewards/margins": 0.3021387457847595,
|
| 2290 |
+
"rewards/rejected": -0.11341919749975204,
|
| 2291 |
+
"step": 1460
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 1.0,
|
| 2295 |
+
"step": 1460,
|
| 2296 |
+
"total_flos": 0.0,
|
| 2297 |
+
"train_loss": 0.6280729855576607,
|
| 2298 |
+
"train_runtime": 9689.6427,
|
| 2299 |
+
"train_samples_per_second": 14.469,
|
| 2300 |
+
"train_steps_per_second": 0.151
|
| 2301 |
+
}
|
| 2302 |
+
],
|
| 2303 |
+
"logging_steps": 10,
|
| 2304 |
+
"max_steps": 1460,
|
| 2305 |
+
"num_train_epochs": 1,
|
| 2306 |
+
"save_steps": 100,
|
| 2307 |
+
"total_flos": 0.0,
|
| 2308 |
+
"trial_name": null,
|
| 2309 |
+
"trial_params": null
|
| 2310 |
+
}
|