Model save
Browse files- README.md +78 -0
- adapter_model.safetensors +1 -1
- all_results.json +21 -0
- eval_results.json +16 -0
- train_results.json +8 -0
- trainer_state.json +1050 -0
README.md
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
base_model: mistralai/Mistral-7B-v0.1
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-gpo-iter2
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-7b-gpo-iter2
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.0114
|
| 22 |
+
- Rewards/chosen: -0.0874
|
| 23 |
+
- Rewards/rejected: -0.0645
|
| 24 |
+
- Rewards/accuracies: 0.3940
|
| 25 |
+
- Rewards/margins: -0.0229
|
| 26 |
+
- Logps/rejected: -264.6114
|
| 27 |
+
- Logps/chosen: -288.2511
|
| 28 |
+
- Logits/rejected: -2.1907
|
| 29 |
+
- Logits/chosen: -2.3882
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-06
|
| 49 |
+
- train_batch_size: 1
|
| 50 |
+
- eval_batch_size: 2
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 2
|
| 55 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
+
- lr_scheduler_type: cosine
|
| 57 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 2
|
| 59 |
+
|
| 60 |
+
### Training results
|
| 61 |
+
|
| 62 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 63 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 64 |
+
| 0.0012 | 0.3 | 100 | 0.0016 | -0.0164 | -0.0160 | 0.5035 | -0.0005 | -259.7555 | -281.1500 | -2.1644 | -2.3583 |
|
| 65 |
+
| 0.0011 | 0.61 | 200 | 0.0018 | -0.0088 | -0.0077 | 0.4815 | -0.0011 | -258.9317 | -280.3858 | -2.1837 | -2.3781 |
|
| 66 |
+
| 0.0015 | 0.91 | 300 | 0.0019 | -0.0167 | -0.0149 | 0.4805 | -0.0017 | -259.6521 | -281.1740 | -2.1796 | -2.3740 |
|
| 67 |
+
| 0.0397 | 1.22 | 400 | 0.0074 | -0.0779 | -0.0627 | 0.4160 | -0.0151 | -264.4323 | -287.2935 | -2.1632 | -2.3568 |
|
| 68 |
+
| 0.0305 | 1.52 | 500 | 0.0117 | -0.0898 | -0.0668 | 0.3945 | -0.0230 | -264.8388 | -288.4842 | -2.1902 | -2.3875 |
|
| 69 |
+
| 0.0366 | 1.82 | 600 | 0.0115 | -0.0876 | -0.0647 | 0.4000 | -0.0230 | -264.6301 | -288.2723 | -2.1900 | -2.3873 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- PEFT 0.7.1
|
| 75 |
+
- Transformers 4.36.2
|
| 76 |
+
- Pytorch 2.1.2+cu118
|
| 77 |
+
- Datasets 2.14.6
|
| 78 |
+
- Tokenizers 0.15.2
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eeca74f490c374bfa634438505679bb07a5398dfd257f26fcee76c900b989388
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"eval_logits/chosen": -2.388157844543457,
|
| 4 |
+
"eval_logits/rejected": -2.1907498836517334,
|
| 5 |
+
"eval_logps/chosen": -288.2511291503906,
|
| 6 |
+
"eval_logps/rejected": -264.61138916015625,
|
| 7 |
+
"eval_loss": 0.011413076892495155,
|
| 8 |
+
"eval_rewards/accuracies": 0.39399999380111694,
|
| 9 |
+
"eval_rewards/chosen": -0.08743663877248764,
|
| 10 |
+
"eval_rewards/margins": -0.022926397621631622,
|
| 11 |
+
"eval_rewards/rejected": -0.06451024115085602,
|
| 12 |
+
"eval_runtime": 1422.7975,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 1.406,
|
| 15 |
+
"eval_steps_per_second": 0.703,
|
| 16 |
+
"train_loss": 0.014390302914878392,
|
| 17 |
+
"train_runtime": 11750.1681,
|
| 18 |
+
"train_samples": 61135,
|
| 19 |
+
"train_samples_per_second": 0.112,
|
| 20 |
+
"train_steps_per_second": 0.056
|
| 21 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"eval_logits/chosen": -2.388157844543457,
|
| 4 |
+
"eval_logits/rejected": -2.1907498836517334,
|
| 5 |
+
"eval_logps/chosen": -288.2511291503906,
|
| 6 |
+
"eval_logps/rejected": -264.61138916015625,
|
| 7 |
+
"eval_loss": 0.011413076892495155,
|
| 8 |
+
"eval_rewards/accuracies": 0.39399999380111694,
|
| 9 |
+
"eval_rewards/chosen": -0.08743663877248764,
|
| 10 |
+
"eval_rewards/margins": -0.022926397621631622,
|
| 11 |
+
"eval_rewards/rejected": -0.06451024115085602,
|
| 12 |
+
"eval_runtime": 1422.7975,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 1.406,
|
| 15 |
+
"eval_steps_per_second": 0.703
|
| 16 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"train_loss": 0.014390302914878392,
|
| 4 |
+
"train_runtime": 11750.1681,
|
| 5 |
+
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 0.112,
|
| 7 |
+
"train_steps_per_second": 0.056
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1050 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 658,
|
| 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": 7.575757575757576e-08,
|
| 14 |
+
"logits/chosen": -3.0401854515075684,
|
| 15 |
+
"logits/rejected": -2.9935803413391113,
|
| 16 |
+
"logps/chosen": -33.6235466003418,
|
| 17 |
+
"logps/rejected": -31.46235466003418,
|
| 18 |
+
"loss": 0.0011,
|
| 19 |
+
"rewards/accuracies": 0.0,
|
| 20 |
+
"rewards/chosen": 0.0,
|
| 21 |
+
"rewards/margins": 0.0,
|
| 22 |
+
"rewards/rejected": 0.0,
|
| 23 |
+
"step": 1
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.03,
|
| 27 |
+
"learning_rate": 7.575757575757576e-07,
|
| 28 |
+
"logits/chosen": -3.0845224857330322,
|
| 29 |
+
"logits/rejected": -2.939974784851074,
|
| 30 |
+
"logps/chosen": -187.66502380371094,
|
| 31 |
+
"logps/rejected": -172.21742248535156,
|
| 32 |
+
"loss": 0.0011,
|
| 33 |
+
"rewards/accuracies": 0.4444444477558136,
|
| 34 |
+
"rewards/chosen": 4.157695730100386e-05,
|
| 35 |
+
"rewards/margins": -0.0002048778405878693,
|
| 36 |
+
"rewards/rejected": 0.0002464548160787672,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.06,
|
| 41 |
+
"learning_rate": 1.5151515151515152e-06,
|
| 42 |
+
"logits/chosen": -2.975489616394043,
|
| 43 |
+
"logits/rejected": -2.8098347187042236,
|
| 44 |
+
"logps/chosen": -101.97224426269531,
|
| 45 |
+
"logps/rejected": -79.41976928710938,
|
| 46 |
+
"loss": 0.001,
|
| 47 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 48 |
+
"rewards/chosen": 0.0006187028484418988,
|
| 49 |
+
"rewards/margins": 0.0002881823165807873,
|
| 50 |
+
"rewards/rejected": 0.0003305206191726029,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.09,
|
| 55 |
+
"learning_rate": 2.2727272727272728e-06,
|
| 56 |
+
"logits/chosen": -2.9616599082946777,
|
| 57 |
+
"logits/rejected": -2.853991746902466,
|
| 58 |
+
"logps/chosen": -93.24244689941406,
|
| 59 |
+
"logps/rejected": -93.37409973144531,
|
| 60 |
+
"loss": 0.0011,
|
| 61 |
+
"rewards/accuracies": 0.5,
|
| 62 |
+
"rewards/chosen": 0.0018816631054505706,
|
| 63 |
+
"rewards/margins": 0.000262443587416783,
|
| 64 |
+
"rewards/rejected": 0.001619219547137618,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.12,
|
| 69 |
+
"learning_rate": 3.0303030303030305e-06,
|
| 70 |
+
"logits/chosen": -2.957383871078491,
|
| 71 |
+
"logits/rejected": -2.8873894214630127,
|
| 72 |
+
"logps/chosen": -98.40808868408203,
|
| 73 |
+
"logps/rejected": -90.6903305053711,
|
| 74 |
+
"loss": 0.0011,
|
| 75 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 76 |
+
"rewards/chosen": 0.0038822232745587826,
|
| 77 |
+
"rewards/margins": 0.0006505203782580793,
|
| 78 |
+
"rewards/rejected": 0.0032317028380930424,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.15,
|
| 83 |
+
"learning_rate": 3.7878787878787882e-06,
|
| 84 |
+
"logits/chosen": -2.793536424636841,
|
| 85 |
+
"logits/rejected": -2.701014995574951,
|
| 86 |
+
"logps/chosen": -76.89764404296875,
|
| 87 |
+
"logps/rejected": -76.88601684570312,
|
| 88 |
+
"loss": 0.0009,
|
| 89 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 90 |
+
"rewards/chosen": 0.0003440352447796613,
|
| 91 |
+
"rewards/margins": 0.003660330083221197,
|
| 92 |
+
"rewards/rejected": -0.003316295798867941,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.18,
|
| 97 |
+
"learning_rate": 4.5454545454545455e-06,
|
| 98 |
+
"logits/chosen": -2.8517611026763916,
|
| 99 |
+
"logits/rejected": -2.701003313064575,
|
| 100 |
+
"logps/chosen": -111.06632232666016,
|
| 101 |
+
"logps/rejected": -97.74620056152344,
|
| 102 |
+
"loss": 0.001,
|
| 103 |
+
"rewards/accuracies": 0.4000000059604645,
|
| 104 |
+
"rewards/chosen": 0.0011594895040616393,
|
| 105 |
+
"rewards/margins": 0.00348883168771863,
|
| 106 |
+
"rewards/rejected": -0.0023293420672416687,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.21,
|
| 111 |
+
"learning_rate": 4.999436790436924e-06,
|
| 112 |
+
"logits/chosen": -3.074284791946411,
|
| 113 |
+
"logits/rejected": -2.859107494354248,
|
| 114 |
+
"logps/chosen": -175.5392608642578,
|
| 115 |
+
"logps/rejected": -165.30654907226562,
|
| 116 |
+
"loss": 0.0011,
|
| 117 |
+
"rewards/accuracies": 0.5,
|
| 118 |
+
"rewards/chosen": 0.0042682006023824215,
|
| 119 |
+
"rewards/margins": 0.004156609531491995,
|
| 120 |
+
"rewards/rejected": 0.00011159134737681597,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.24,
|
| 125 |
+
"learning_rate": 4.993103596812269e-06,
|
| 126 |
+
"logits/chosen": -2.7597286701202393,
|
| 127 |
+
"logits/rejected": -2.6338610649108887,
|
| 128 |
+
"logps/chosen": -117.37264251708984,
|
| 129 |
+
"logps/rejected": -96.65496826171875,
|
| 130 |
+
"loss": 0.0014,
|
| 131 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 132 |
+
"rewards/chosen": -0.004695770796388388,
|
| 133 |
+
"rewards/margins": -0.0023340280167758465,
|
| 134 |
+
"rewards/rejected": -0.002361743012443185,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.27,
|
| 139 |
+
"learning_rate": 4.979751088147192e-06,
|
| 140 |
+
"logits/chosen": -2.7671074867248535,
|
| 141 |
+
"logits/rejected": -2.5734329223632812,
|
| 142 |
+
"logps/chosen": -124.0124740600586,
|
| 143 |
+
"logps/rejected": -117.49992370605469,
|
| 144 |
+
"loss": 0.0013,
|
| 145 |
+
"rewards/accuracies": 0.550000011920929,
|
| 146 |
+
"rewards/chosen": -0.0014447696739807725,
|
| 147 |
+
"rewards/margins": 0.006112386472523212,
|
| 148 |
+
"rewards/rejected": -0.0075571550987660885,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.3,
|
| 153 |
+
"learning_rate": 4.95941685833271e-06,
|
| 154 |
+
"logits/chosen": -2.841676712036133,
|
| 155 |
+
"logits/rejected": -2.6840147972106934,
|
| 156 |
+
"logps/chosen": -136.9166259765625,
|
| 157 |
+
"logps/rejected": -142.25997924804688,
|
| 158 |
+
"loss": 0.0012,
|
| 159 |
+
"rewards/accuracies": 0.550000011920929,
|
| 160 |
+
"rewards/chosen": -0.005834830459207296,
|
| 161 |
+
"rewards/margins": 0.0021175735164433718,
|
| 162 |
+
"rewards/rejected": -0.007952402345836163,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.3,
|
| 167 |
+
"eval_logits/chosen": -2.3582723140716553,
|
| 168 |
+
"eval_logits/rejected": -2.1644442081451416,
|
| 169 |
+
"eval_logps/chosen": -281.1500244140625,
|
| 170 |
+
"eval_logps/rejected": -259.7554626464844,
|
| 171 |
+
"eval_loss": 0.001627120072953403,
|
| 172 |
+
"eval_rewards/accuracies": 0.5034999847412109,
|
| 173 |
+
"eval_rewards/chosen": -0.016425278037786484,
|
| 174 |
+
"eval_rewards/margins": -0.00047451091813854873,
|
| 175 |
+
"eval_rewards/rejected": -0.0159507654607296,
|
| 176 |
+
"eval_runtime": 1420.6531,
|
| 177 |
+
"eval_samples_per_second": 1.408,
|
| 178 |
+
"eval_steps_per_second": 0.704,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.33,
|
| 183 |
+
"learning_rate": 4.9321581582449365e-06,
|
| 184 |
+
"logits/chosen": -2.8993427753448486,
|
| 185 |
+
"logits/rejected": -2.786620616912842,
|
| 186 |
+
"logps/chosen": -118.65157318115234,
|
| 187 |
+
"logps/rejected": -127.34912109375,
|
| 188 |
+
"loss": 0.0011,
|
| 189 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 190 |
+
"rewards/chosen": -0.0042579686269164085,
|
| 191 |
+
"rewards/margins": 0.004396714735776186,
|
| 192 |
+
"rewards/rejected": -0.008654682897031307,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.36,
|
| 197 |
+
"learning_rate": 4.898051734555676e-06,
|
| 198 |
+
"logits/chosen": -3.0458617210388184,
|
| 199 |
+
"logits/rejected": -2.99739670753479,
|
| 200 |
+
"logps/chosen": -95.90953063964844,
|
| 201 |
+
"logps/rejected": -102.84895324707031,
|
| 202 |
+
"loss": 0.0013,
|
| 203 |
+
"rewards/accuracies": 0.550000011920929,
|
| 204 |
+
"rewards/chosen": -0.004254769999533892,
|
| 205 |
+
"rewards/margins": -0.0005313962465152144,
|
| 206 |
+
"rewards/rejected": -0.0037233736366033554,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.4,
|
| 211 |
+
"learning_rate": 4.857193613652711e-06,
|
| 212 |
+
"logits/chosen": -3.0125811100006104,
|
| 213 |
+
"logits/rejected": -2.8767762184143066,
|
| 214 |
+
"logps/chosen": -66.88941955566406,
|
| 215 |
+
"logps/rejected": -71.67733001708984,
|
| 216 |
+
"loss": 0.0012,
|
| 217 |
+
"rewards/accuracies": 0.5,
|
| 218 |
+
"rewards/chosen": 0.004949026275426149,
|
| 219 |
+
"rewards/margins": 0.0032058332581073046,
|
| 220 |
+
"rewards/rejected": 0.001743193482980132,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.43,
|
| 225 |
+
"learning_rate": 4.809698831278217e-06,
|
| 226 |
+
"logits/chosen": -2.9552953243255615,
|
| 227 |
+
"logits/rejected": -2.7150585651397705,
|
| 228 |
+
"logps/chosen": -75.46265411376953,
|
| 229 |
+
"logps/rejected": -72.64187622070312,
|
| 230 |
+
"loss": 0.001,
|
| 231 |
+
"rewards/accuracies": 0.800000011920929,
|
| 232 |
+
"rewards/chosen": 0.00027392804622650146,
|
| 233 |
+
"rewards/margins": 0.016122380271553993,
|
| 234 |
+
"rewards/rejected": -0.01584845408797264,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.46,
|
| 239 |
+
"learning_rate": 4.7557011086464634e-06,
|
| 240 |
+
"logits/chosen": -2.9132320880889893,
|
| 241 |
+
"logits/rejected": -2.64630126953125,
|
| 242 |
+
"logps/chosen": -76.73291778564453,
|
| 243 |
+
"logps/rejected": -97.9271011352539,
|
| 244 |
+
"loss": 0.0013,
|
| 245 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 246 |
+
"rewards/chosen": -0.00335231376811862,
|
| 247 |
+
"rewards/margins": 0.01173357479274273,
|
| 248 |
+
"rewards/rejected": -0.015085889026522636,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.49,
|
| 253 |
+
"learning_rate": 4.695352475952706e-06,
|
| 254 |
+
"logits/chosen": -3.013190746307373,
|
| 255 |
+
"logits/rejected": -2.926192045211792,
|
| 256 |
+
"logps/chosen": -73.78190612792969,
|
| 257 |
+
"logps/rejected": -84.00215148925781,
|
| 258 |
+
"loss": 0.0016,
|
| 259 |
+
"rewards/accuracies": 0.699999988079071,
|
| 260 |
+
"rewards/chosen": -0.006495082285255194,
|
| 261 |
+
"rewards/margins": 0.01742926612496376,
|
| 262 |
+
"rewards/rejected": -0.023924347013235092,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.52,
|
| 267 |
+
"learning_rate": 4.6288228443332786e-06,
|
| 268 |
+
"logits/chosen": -2.892408847808838,
|
| 269 |
+
"logits/rejected": -2.785109758377075,
|
| 270 |
+
"logps/chosen": -88.50574493408203,
|
| 271 |
+
"logps/rejected": -63.979164123535156,
|
| 272 |
+
"loss": 0.0014,
|
| 273 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 274 |
+
"rewards/chosen": -0.015971587970852852,
|
| 275 |
+
"rewards/margins": 0.0037319730035960674,
|
| 276 |
+
"rewards/rejected": -0.019703561440110207,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.55,
|
| 281 |
+
"learning_rate": 4.556299527482029e-06,
|
| 282 |
+
"logits/chosen": -3.024970293045044,
|
| 283 |
+
"logits/rejected": -2.9614624977111816,
|
| 284 |
+
"logps/chosen": -95.4378433227539,
|
| 285 |
+
"logps/rejected": -86.34746551513672,
|
| 286 |
+
"loss": 0.0009,
|
| 287 |
+
"rewards/accuracies": 0.75,
|
| 288 |
+
"rewards/chosen": 0.0011760194320231676,
|
| 289 |
+
"rewards/margins": 0.010892638936638832,
|
| 290 |
+
"rewards/rejected": -0.009716618806123734,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.58,
|
| 295 |
+
"learning_rate": 4.4779867142699715e-06,
|
| 296 |
+
"logits/chosen": -3.0479438304901123,
|
| 297 |
+
"logits/rejected": -2.903231620788574,
|
| 298 |
+
"logps/chosen": -138.85531616210938,
|
| 299 |
+
"logps/rejected": -146.4200439453125,
|
| 300 |
+
"loss": 0.0017,
|
| 301 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 302 |
+
"rewards/chosen": 0.013675776310265064,
|
| 303 |
+
"rewards/margins": 0.003341741394251585,
|
| 304 |
+
"rewards/rejected": 0.010334033519029617,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.61,
|
| 309 |
+
"learning_rate": 4.394104893853007e-06,
|
| 310 |
+
"logits/chosen": -2.9416115283966064,
|
| 311 |
+
"logits/rejected": -2.8143601417541504,
|
| 312 |
+
"logps/chosen": -85.66619110107422,
|
| 313 |
+
"logps/rejected": -84.1985092163086,
|
| 314 |
+
"loss": 0.0011,
|
| 315 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 316 |
+
"rewards/chosen": 0.003594033420085907,
|
| 317 |
+
"rewards/margins": 0.005169789306819439,
|
| 318 |
+
"rewards/rejected": -0.0015757561195641756,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.61,
|
| 323 |
+
"eval_logits/chosen": -2.378091812133789,
|
| 324 |
+
"eval_logits/rejected": -2.183748960494995,
|
| 325 |
+
"eval_logps/chosen": -280.3858337402344,
|
| 326 |
+
"eval_logps/rejected": -258.9316711425781,
|
| 327 |
+
"eval_loss": 0.0018177788006141782,
|
| 328 |
+
"eval_rewards/accuracies": 0.4814999997615814,
|
| 329 |
+
"eval_rewards/chosen": -0.008783474564552307,
|
| 330 |
+
"eval_rewards/margins": -0.0010705279419198632,
|
| 331 |
+
"eval_rewards/rejected": -0.007712946273386478,
|
| 332 |
+
"eval_runtime": 1420.2588,
|
| 333 |
+
"eval_samples_per_second": 1.408,
|
| 334 |
+
"eval_steps_per_second": 0.704,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.64,
|
| 339 |
+
"learning_rate": 4.3048902348863116e-06,
|
| 340 |
+
"logits/chosen": -2.9989707469940186,
|
| 341 |
+
"logits/rejected": -2.8650078773498535,
|
| 342 |
+
"logps/chosen": -102.37027740478516,
|
| 343 |
+
"logps/rejected": -118.4636001586914,
|
| 344 |
+
"loss": 0.0012,
|
| 345 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 346 |
+
"rewards/chosen": 0.001182993990369141,
|
| 347 |
+
"rewards/margins": 0.004763666074723005,
|
| 348 |
+
"rewards/rejected": -0.003580672200769186,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.67,
|
| 353 |
+
"learning_rate": 4.210593920593201e-06,
|
| 354 |
+
"logits/chosen": -2.9398374557495117,
|
| 355 |
+
"logits/rejected": -2.8116965293884277,
|
| 356 |
+
"logps/chosen": -70.69538879394531,
|
| 357 |
+
"logps/rejected": -75.53007507324219,
|
| 358 |
+
"loss": 0.0013,
|
| 359 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 360 |
+
"rewards/chosen": 0.005323179066181183,
|
| 361 |
+
"rewards/margins": 0.004396263975650072,
|
| 362 |
+
"rewards/rejected": 0.0009269150905311108,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.7,
|
| 367 |
+
"learning_rate": 4.111481441560598e-06,
|
| 368 |
+
"logits/chosen": -2.972592830657959,
|
| 369 |
+
"logits/rejected": -2.852461099624634,
|
| 370 |
+
"logps/chosen": -160.5238037109375,
|
| 371 |
+
"logps/rejected": -129.62954711914062,
|
| 372 |
+
"loss": 0.0013,
|
| 373 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 374 |
+
"rewards/chosen": 0.00035965832648798823,
|
| 375 |
+
"rewards/margins": 0.005134374834597111,
|
| 376 |
+
"rewards/rejected": -0.004774716217070818,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.73,
|
| 381 |
+
"learning_rate": 4.007831848252212e-06,
|
| 382 |
+
"logits/chosen": -3.1746320724487305,
|
| 383 |
+
"logits/rejected": -2.9887349605560303,
|
| 384 |
+
"logps/chosen": -124.45001220703125,
|
| 385 |
+
"logps/rejected": -101.69696807861328,
|
| 386 |
+
"loss": 0.0009,
|
| 387 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 388 |
+
"rewards/chosen": 0.019443057477474213,
|
| 389 |
+
"rewards/margins": 0.012666161172091961,
|
| 390 |
+
"rewards/rejected": 0.0067768944427371025,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.76,
|
| 395 |
+
"learning_rate": 3.899936965343989e-06,
|
| 396 |
+
"logits/chosen": -3.0935516357421875,
|
| 397 |
+
"logits/rejected": -3.028534412384033,
|
| 398 |
+
"logps/chosen": -131.24337768554688,
|
| 399 |
+
"logps/rejected": -116.88172912597656,
|
| 400 |
+
"loss": 0.0016,
|
| 401 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 402 |
+
"rewards/chosen": 0.005977076943963766,
|
| 403 |
+
"rewards/margins": 0.005966751836240292,
|
| 404 |
+
"rewards/rejected": 1.0324455615773331e-05,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.79,
|
| 409 |
+
"learning_rate": 3.7881005700938635e-06,
|
| 410 |
+
"logits/chosen": -2.7336764335632324,
|
| 411 |
+
"logits/rejected": -2.7675654888153076,
|
| 412 |
+
"logps/chosen": -95.82691955566406,
|
| 413 |
+
"logps/rejected": -99.57430267333984,
|
| 414 |
+
"loss": 0.0008,
|
| 415 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 416 |
+
"rewards/chosen": 0.012143732979893684,
|
| 417 |
+
"rewards/margins": 0.008911015465855598,
|
| 418 |
+
"rewards/rejected": 0.003232717514038086,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.82,
|
| 423 |
+
"learning_rate": 3.6726375370590927e-06,
|
| 424 |
+
"logits/chosen": -2.959117889404297,
|
| 425 |
+
"logits/rejected": -2.841716766357422,
|
| 426 |
+
"logps/chosen": -132.79884338378906,
|
| 427 |
+
"logps/rejected": -122.66941833496094,
|
| 428 |
+
"loss": 0.0015,
|
| 429 |
+
"rewards/accuracies": 0.550000011920929,
|
| 430 |
+
"rewards/chosen": 0.003037696471437812,
|
| 431 |
+
"rewards/margins": 0.00202806293964386,
|
| 432 |
+
"rewards/rejected": 0.0010096338810399175,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.85,
|
| 437 |
+
"learning_rate": 3.553872951569236e-06,
|
| 438 |
+
"logits/chosen": -3.0903332233428955,
|
| 439 |
+
"logits/rejected": -3.009970188140869,
|
| 440 |
+
"logps/chosen": -90.32659912109375,
|
| 441 |
+
"logps/rejected": -92.47020721435547,
|
| 442 |
+
"loss": 0.0012,
|
| 443 |
+
"rewards/accuracies": 0.5,
|
| 444 |
+
"rewards/chosen": 0.005868277512490749,
|
| 445 |
+
"rewards/margins": 0.011209800839424133,
|
| 446 |
+
"rewards/rejected": -0.005341522395610809,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.88,
|
| 451 |
+
"learning_rate": 3.432141194450772e-06,
|
| 452 |
+
"logits/chosen": -2.920701503753662,
|
| 453 |
+
"logits/rejected": -2.721304416656494,
|
| 454 |
+
"logps/chosen": -111.44612121582031,
|
| 455 |
+
"logps/rejected": -111.91583251953125,
|
| 456 |
+
"loss": 0.0007,
|
| 457 |
+
"rewards/accuracies": 0.699999988079071,
|
| 458 |
+
"rewards/chosen": -0.004392694681882858,
|
| 459 |
+
"rewards/margins": 0.016653168946504593,
|
| 460 |
+
"rewards/rejected": -0.02104586362838745,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.91,
|
| 465 |
+
"learning_rate": 3.307785000580313e-06,
|
| 466 |
+
"logits/chosen": -2.757887363433838,
|
| 467 |
+
"logits/rejected": -2.5684075355529785,
|
| 468 |
+
"logps/chosen": -128.5594024658203,
|
| 469 |
+
"logps/rejected": -115.97843933105469,
|
| 470 |
+
"loss": 0.0015,
|
| 471 |
+
"rewards/accuracies": 0.4000000059604645,
|
| 472 |
+
"rewards/chosen": -0.006182619370520115,
|
| 473 |
+
"rewards/margins": 0.0031226237770169973,
|
| 474 |
+
"rewards/rejected": -0.009305243380367756,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.91,
|
| 479 |
+
"eval_logits/chosen": -2.374006748199463,
|
| 480 |
+
"eval_logits/rejected": -2.1796298027038574,
|
| 481 |
+
"eval_logps/chosen": -281.17401123046875,
|
| 482 |
+
"eval_logps/rejected": -259.6520690917969,
|
| 483 |
+
"eval_loss": 0.0019075373420491815,
|
| 484 |
+
"eval_rewards/accuracies": 0.4805000126361847,
|
| 485 |
+
"eval_rewards/chosen": -0.016665350645780563,
|
| 486 |
+
"eval_rewards/margins": -0.0017483622068539262,
|
| 487 |
+
"eval_rewards/rejected": -0.014916987158358097,
|
| 488 |
+
"eval_runtime": 1421.116,
|
| 489 |
+
"eval_samples_per_second": 1.407,
|
| 490 |
+
"eval_steps_per_second": 0.704,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.94,
|
| 495 |
+
"learning_rate": 3.1811544939170573e-06,
|
| 496 |
+
"logits/chosen": -2.7975335121154785,
|
| 497 |
+
"logits/rejected": -2.667598009109497,
|
| 498 |
+
"logps/chosen": -161.1660919189453,
|
| 499 |
+
"logps/rejected": -133.93838500976562,
|
| 500 |
+
"loss": 0.0012,
|
| 501 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 502 |
+
"rewards/chosen": -0.010928490199148655,
|
| 503 |
+
"rewards/margins": 0.00381992245092988,
|
| 504 |
+
"rewards/rejected": -0.014748412184417248,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.97,
|
| 509 |
+
"learning_rate": 3.052606201731325e-06,
|
| 510 |
+
"logits/chosen": -2.93320894241333,
|
| 511 |
+
"logits/rejected": -2.7140159606933594,
|
| 512 |
+
"logps/chosen": -79.52749633789062,
|
| 513 |
+
"logps/rejected": -84.45579528808594,
|
| 514 |
+
"loss": 0.0008,
|
| 515 |
+
"rewards/accuracies": 0.699999988079071,
|
| 516 |
+
"rewards/chosen": -0.0026647502090781927,
|
| 517 |
+
"rewards/margins": 0.009594769217073917,
|
| 518 |
+
"rewards/rejected": -0.01225951872766018,
|
| 519 |
+
"step": 320
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 1.0,
|
| 523 |
+
"learning_rate": 2.9225020508046233e-06,
|
| 524 |
+
"logits/chosen": -2.901916742324829,
|
| 525 |
+
"logits/rejected": -2.736692190170288,
|
| 526 |
+
"logps/chosen": -115.4826431274414,
|
| 527 |
+
"logps/rejected": -130.37356567382812,
|
| 528 |
+
"loss": 0.0313,
|
| 529 |
+
"rewards/accuracies": 0.699999988079071,
|
| 530 |
+
"rewards/chosen": 0.017747070640325546,
|
| 531 |
+
"rewards/margins": 0.05266844481229782,
|
| 532 |
+
"rewards/rejected": -0.034921370446681976,
|
| 533 |
+
"step": 330
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 1.03,
|
| 537 |
+
"learning_rate": 2.7912083484274266e-06,
|
| 538 |
+
"logits/chosen": -2.7787222862243652,
|
| 539 |
+
"logits/rejected": -2.63339900970459,
|
| 540 |
+
"logps/chosen": -137.5430450439453,
|
| 541 |
+
"logps/rejected": -127.07392883300781,
|
| 542 |
+
"loss": 0.0436,
|
| 543 |
+
"rewards/accuracies": 0.800000011920929,
|
| 544 |
+
"rewards/chosen": 0.0036686602979898453,
|
| 545 |
+
"rewards/margins": 0.05118006467819214,
|
| 546 |
+
"rewards/rejected": -0.04751140996813774,
|
| 547 |
+
"step": 340
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 1.06,
|
| 551 |
+
"learning_rate": 2.659094751063666e-06,
|
| 552 |
+
"logits/chosen": -2.9705729484558105,
|
| 553 |
+
"logits/rejected": -2.832822322845459,
|
| 554 |
+
"logps/chosen": -143.50698852539062,
|
| 555 |
+
"logps/rejected": -140.02352905273438,
|
| 556 |
+
"loss": 0.0133,
|
| 557 |
+
"rewards/accuracies": 0.75,
|
| 558 |
+
"rewards/chosen": 0.008256057277321815,
|
| 559 |
+
"rewards/margins": 0.05859958380460739,
|
| 560 |
+
"rewards/rejected": -0.05034352466464043,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 1.09,
|
| 565 |
+
"learning_rate": 2.526533223585641e-06,
|
| 566 |
+
"logits/chosen": -2.8163113594055176,
|
| 567 |
+
"logits/rejected": -2.697537660598755,
|
| 568 |
+
"logps/chosen": -85.76821899414062,
|
| 569 |
+
"logps/rejected": -81.71627807617188,
|
| 570 |
+
"loss": 0.0073,
|
| 571 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 572 |
+
"rewards/chosen": 0.010457667522132397,
|
| 573 |
+
"rewards/margins": 0.053769804537296295,
|
| 574 |
+
"rewards/rejected": -0.043312136083841324,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 1.12,
|
| 579 |
+
"learning_rate": 2.39389699200963e-06,
|
| 580 |
+
"logits/chosen": -3.000455141067505,
|
| 581 |
+
"logits/rejected": -2.8073432445526123,
|
| 582 |
+
"logps/chosen": -107.18000793457031,
|
| 583 |
+
"logps/rejected": -118.2176284790039,
|
| 584 |
+
"loss": 0.019,
|
| 585 |
+
"rewards/accuracies": 0.800000011920929,
|
| 586 |
+
"rewards/chosen": 0.008137053810060024,
|
| 587 |
+
"rewards/margins": 0.07659684866666794,
|
| 588 |
+
"rewards/rejected": -0.06845979392528534,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 1.16,
|
| 593 |
+
"learning_rate": 2.2615594926807554e-06,
|
| 594 |
+
"logits/chosen": -2.9433727264404297,
|
| 595 |
+
"logits/rejected": -2.817605495452881,
|
| 596 |
+
"logps/chosen": -84.96995544433594,
|
| 597 |
+
"logps/rejected": -95.87684631347656,
|
| 598 |
+
"loss": 0.0277,
|
| 599 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 600 |
+
"rewards/chosen": 0.01651051640510559,
|
| 601 |
+
"rewards/margins": 0.06000148132443428,
|
| 602 |
+
"rewards/rejected": -0.04349096864461899,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 1.19,
|
| 607 |
+
"learning_rate": 2.129893320865672e-06,
|
| 608 |
+
"logits/chosen": -3.0510168075561523,
|
| 609 |
+
"logits/rejected": -2.938931465148926,
|
| 610 |
+
"logps/chosen": -76.3388442993164,
|
| 611 |
+
"logps/rejected": -62.002540588378906,
|
| 612 |
+
"loss": 0.0302,
|
| 613 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 614 |
+
"rewards/chosen": 0.04483187943696976,
|
| 615 |
+
"rewards/margins": 0.10971853882074356,
|
| 616 |
+
"rewards/rejected": -0.0648866593837738,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.22,
|
| 621 |
+
"learning_rate": 1.9992691817133025e-06,
|
| 622 |
+
"logits/chosen": -3.05735445022583,
|
| 623 |
+
"logits/rejected": -2.815657138824463,
|
| 624 |
+
"logps/chosen": -251.65219116210938,
|
| 625 |
+
"logps/rejected": -255.02658081054688,
|
| 626 |
+
"loss": 0.0397,
|
| 627 |
+
"rewards/accuracies": 0.75,
|
| 628 |
+
"rewards/chosen": -0.03865816444158554,
|
| 629 |
+
"rewards/margins": 0.04983866214752197,
|
| 630 |
+
"rewards/rejected": -0.08849682658910751,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 1.22,
|
| 635 |
+
"eval_logits/chosen": -2.356828451156616,
|
| 636 |
+
"eval_logits/rejected": -2.1631951332092285,
|
| 637 |
+
"eval_logps/chosen": -287.29345703125,
|
| 638 |
+
"eval_logps/rejected": -264.43231201171875,
|
| 639 |
+
"eval_loss": 0.007440177723765373,
|
| 640 |
+
"eval_rewards/accuracies": 0.41600000858306885,
|
| 641 |
+
"eval_rewards/chosen": -0.0778598040342331,
|
| 642 |
+
"eval_rewards/margins": -0.015140472911298275,
|
| 643 |
+
"eval_rewards/rejected": -0.06271932274103165,
|
| 644 |
+
"eval_runtime": 1421.1528,
|
| 645 |
+
"eval_samples_per_second": 1.407,
|
| 646 |
+
"eval_steps_per_second": 0.704,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 1.25,
|
| 651 |
+
"learning_rate": 1.8700548465371877e-06,
|
| 652 |
+
"logits/chosen": -3.085784435272217,
|
| 653 |
+
"logits/rejected": -2.8534493446350098,
|
| 654 |
+
"logps/chosen": -150.0517120361328,
|
| 655 |
+
"logps/rejected": -130.98390197753906,
|
| 656 |
+
"loss": 0.0304,
|
| 657 |
+
"rewards/accuracies": 0.75,
|
| 658 |
+
"rewards/chosen": -0.026314150542020798,
|
| 659 |
+
"rewards/margins": 0.05200232192873955,
|
| 660 |
+
"rewards/rejected": -0.07831647992134094,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 1.28,
|
| 665 |
+
"learning_rate": 1.742614117358029e-06,
|
| 666 |
+
"logits/chosen": -2.8179147243499756,
|
| 667 |
+
"logits/rejected": -2.72475528717041,
|
| 668 |
+
"logps/chosen": -85.7872543334961,
|
| 669 |
+
"logps/rejected": -98.0990219116211,
|
| 670 |
+
"loss": 0.0839,
|
| 671 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 672 |
+
"rewards/chosen": -0.052498847246170044,
|
| 673 |
+
"rewards/margins": 0.009759088046848774,
|
| 674 |
+
"rewards/rejected": -0.06225793436169624,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.31,
|
| 679 |
+
"learning_rate": 1.617305802621748e-06,
|
| 680 |
+
"logits/chosen": -2.978959083557129,
|
| 681 |
+
"logits/rejected": -2.9247641563415527,
|
| 682 |
+
"logps/chosen": -94.34339904785156,
|
| 683 |
+
"logps/rejected": -103.36552429199219,
|
| 684 |
+
"loss": 0.0221,
|
| 685 |
+
"rewards/accuracies": 0.800000011920929,
|
| 686 |
+
"rewards/chosen": 0.004420995246618986,
|
| 687 |
+
"rewards/margins": 0.09026098251342773,
|
| 688 |
+
"rewards/rejected": -0.08583998680114746,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.34,
|
| 693 |
+
"learning_rate": 1.4944827069769125e-06,
|
| 694 |
+
"logits/chosen": -2.94343900680542,
|
| 695 |
+
"logits/rejected": -2.8201680183410645,
|
| 696 |
+
"logps/chosen": -164.69044494628906,
|
| 697 |
+
"logps/rejected": -141.93528747558594,
|
| 698 |
+
"loss": 0.004,
|
| 699 |
+
"rewards/accuracies": 0.550000011920929,
|
| 700 |
+
"rewards/chosen": -0.01349672395735979,
|
| 701 |
+
"rewards/margins": 0.0265080276876688,
|
| 702 |
+
"rewards/rejected": -0.040004752576351166,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.37,
|
| 707 |
+
"learning_rate": 1.3744906379558165e-06,
|
| 708 |
+
"logits/chosen": -2.9248745441436768,
|
| 709 |
+
"logits/rejected": -2.9322562217712402,
|
| 710 |
+
"logps/chosen": -103.63645935058594,
|
| 711 |
+
"logps/rejected": -123.60746765136719,
|
| 712 |
+
"loss": 0.0383,
|
| 713 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 714 |
+
"rewards/chosen": 0.014714512042701244,
|
| 715 |
+
"rewards/margins": 0.11696286499500275,
|
| 716 |
+
"rewards/rejected": -0.10224835574626923,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.4,
|
| 721 |
+
"learning_rate": 1.257667432355893e-06,
|
| 722 |
+
"logits/chosen": -2.974544048309326,
|
| 723 |
+
"logits/rejected": -2.780351161956787,
|
| 724 |
+
"logps/chosen": -76.25318145751953,
|
| 725 |
+
"logps/rejected": -95.62451934814453,
|
| 726 |
+
"loss": 0.0157,
|
| 727 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 728 |
+
"rewards/chosen": 0.0021516154520213604,
|
| 729 |
+
"rewards/margins": 0.10200606286525726,
|
| 730 |
+
"rewards/rejected": -0.09985444694757462,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 1.43,
|
| 735 |
+
"learning_rate": 1.1443420050626624e-06,
|
| 736 |
+
"logits/chosen": -3.061354875564575,
|
| 737 |
+
"logits/rejected": -2.9399447441101074,
|
| 738 |
+
"logps/chosen": -71.61457824707031,
|
| 739 |
+
"logps/rejected": -85.53517150878906,
|
| 740 |
+
"loss": 0.047,
|
| 741 |
+
"rewards/accuracies": 0.800000011920929,
|
| 742 |
+
"rewards/chosen": -0.00521849375218153,
|
| 743 |
+
"rewards/margins": 0.06701686233282089,
|
| 744 |
+
"rewards/rejected": -0.0722353607416153,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.46,
|
| 749 |
+
"learning_rate": 1.0348334229922677e-06,
|
| 750 |
+
"logits/chosen": -3.0284695625305176,
|
| 751 |
+
"logits/rejected": -2.7861952781677246,
|
| 752 |
+
"logps/chosen": -84.50325012207031,
|
| 753 |
+
"logps/rejected": -80.56702423095703,
|
| 754 |
+
"loss": 0.0331,
|
| 755 |
+
"rewards/accuracies": 0.75,
|
| 756 |
+
"rewards/chosen": 0.010606551542878151,
|
| 757 |
+
"rewards/margins": 0.08576520532369614,
|
| 758 |
+
"rewards/rejected": -0.07515865564346313,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.49,
|
| 763 |
+
"learning_rate": 9.294500067608941e-07,
|
| 764 |
+
"logits/chosen": -2.929229974746704,
|
| 765 |
+
"logits/rejected": -2.727160692214966,
|
| 766 |
+
"logps/chosen": -106.28855895996094,
|
| 767 |
+
"logps/rejected": -96.00597381591797,
|
| 768 |
+
"loss": 0.0183,
|
| 769 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 770 |
+
"rewards/chosen": -0.032337434589862823,
|
| 771 |
+
"rewards/margins": 0.024760348722338676,
|
| 772 |
+
"rewards/rejected": -0.05709778517484665,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 1.52,
|
| 777 |
+
"learning_rate": 8.284884626103165e-07,
|
| 778 |
+
"logits/chosen": -2.9482462406158447,
|
| 779 |
+
"logits/rejected": -2.7884809970855713,
|
| 780 |
+
"logps/chosen": -77.4437255859375,
|
| 781 |
+
"logps/rejected": -83.18536376953125,
|
| 782 |
+
"loss": 0.0305,
|
| 783 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 784 |
+
"rewards/chosen": 0.025916963815689087,
|
| 785 |
+
"rewards/margins": 0.11217900365591049,
|
| 786 |
+
"rewards/rejected": -0.086262047290802,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 1.52,
|
| 791 |
+
"eval_logits/chosen": -2.3874802589416504,
|
| 792 |
+
"eval_logits/rejected": -2.1901655197143555,
|
| 793 |
+
"eval_logps/chosen": -288.4842224121094,
|
| 794 |
+
"eval_logps/rejected": -264.83880615234375,
|
| 795 |
+
"eval_loss": 0.011713836342096329,
|
| 796 |
+
"eval_rewards/accuracies": 0.3944999873638153,
|
| 797 |
+
"eval_rewards/chosen": -0.08976726979017258,
|
| 798 |
+
"eval_rewards/margins": -0.022982925176620483,
|
| 799 |
+
"eval_rewards/rejected": -0.06678435206413269,
|
| 800 |
+
"eval_runtime": 1421.6286,
|
| 801 |
+
"eval_samples_per_second": 1.407,
|
| 802 |
+
"eval_steps_per_second": 0.703,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 1.55,
|
| 807 |
+
"learning_rate": 7.322330470336314e-07,
|
| 808 |
+
"logits/chosen": -3.0825541019439697,
|
| 809 |
+
"logits/rejected": -2.9925718307495117,
|
| 810 |
+
"logps/chosen": -95.14485931396484,
|
| 811 |
+
"logps/rejected": -104.95149993896484,
|
| 812 |
+
"loss": 0.0318,
|
| 813 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 814 |
+
"rewards/chosen": 0.004621284082531929,
|
| 815 |
+
"rewards/margins": 0.12566979229450226,
|
| 816 |
+
"rewards/rejected": -0.12104851007461548,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 1.58,
|
| 821 |
+
"learning_rate": 6.409547664531734e-07,
|
| 822 |
+
"logits/chosen": -2.9336235523223877,
|
| 823 |
+
"logits/rejected": -2.8529655933380127,
|
| 824 |
+
"logps/chosen": -130.2211151123047,
|
| 825 |
+
"logps/rejected": -122.72911071777344,
|
| 826 |
+
"loss": 0.0063,
|
| 827 |
+
"rewards/accuracies": 0.800000011920929,
|
| 828 |
+
"rewards/chosen": 0.002978086471557617,
|
| 829 |
+
"rewards/margins": 0.05790657550096512,
|
| 830 |
+
"rewards/rejected": -0.0549284890294075,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.61,
|
| 835 |
+
"learning_rate": 5.549106142039018e-07,
|
| 836 |
+
"logits/chosen": -3.001368522644043,
|
| 837 |
+
"logits/rejected": -2.9931020736694336,
|
| 838 |
+
"logps/chosen": -122.59478759765625,
|
| 839 |
+
"logps/rejected": -139.2528839111328,
|
| 840 |
+
"loss": 0.0343,
|
| 841 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 842 |
+
"rewards/chosen": -0.03542733192443848,
|
| 843 |
+
"rewards/margins": 0.028016597032546997,
|
| 844 |
+
"rewards/rejected": -0.06344393640756607,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 1.64,
|
| 849 |
+
"learning_rate": 4.743428469705336e-07,
|
| 850 |
+
"logits/chosen": -2.895446300506592,
|
| 851 |
+
"logits/rejected": -2.805732011795044,
|
| 852 |
+
"logps/chosen": -77.88665771484375,
|
| 853 |
+
"logps/rejected": -86.38732147216797,
|
| 854 |
+
"loss": 0.0152,
|
| 855 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 856 |
+
"rewards/chosen": 0.012998471967875957,
|
| 857 |
+
"rewards/margins": 0.09006098657846451,
|
| 858 |
+
"rewards/rejected": -0.07706250250339508,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 1.67,
|
| 863 |
+
"learning_rate": 3.994783027156143e-07,
|
| 864 |
+
"logits/chosen": -2.8439111709594727,
|
| 865 |
+
"logits/rejected": -2.765376567840576,
|
| 866 |
+
"logps/chosen": -124.60064697265625,
|
| 867 |
+
"logps/rejected": -115.8946533203125,
|
| 868 |
+
"loss": 0.0116,
|
| 869 |
+
"rewards/accuracies": 0.800000011920929,
|
| 870 |
+
"rewards/chosen": -0.0016067728865891695,
|
| 871 |
+
"rewards/margins": 0.08273078501224518,
|
| 872 |
+
"rewards/rejected": -0.08433755487203598,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.7,
|
| 877 |
+
"learning_rate": 3.3052776201888266e-07,
|
| 878 |
+
"logits/chosen": -2.8251261711120605,
|
| 879 |
+
"logits/rejected": -2.69478178024292,
|
| 880 |
+
"logps/chosen": -85.72721862792969,
|
| 881 |
+
"logps/rejected": -84.4536361694336,
|
| 882 |
+
"loss": 0.0146,
|
| 883 |
+
"rewards/accuracies": 0.75,
|
| 884 |
+
"rewards/chosen": 0.007551318965852261,
|
| 885 |
+
"rewards/margins": 0.08319269120693207,
|
| 886 |
+
"rewards/rejected": -0.07564137130975723,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 1.73,
|
| 891 |
+
"learning_rate": 2.676853546260791e-07,
|
| 892 |
+
"logits/chosen": -2.8717586994171143,
|
| 893 |
+
"logits/rejected": -2.805624485015869,
|
| 894 |
+
"logps/chosen": -102.53792572021484,
|
| 895 |
+
"logps/rejected": -89.75617980957031,
|
| 896 |
+
"loss": 0.0086,
|
| 897 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 898 |
+
"rewards/chosen": -0.008890565484762192,
|
| 899 |
+
"rewards/margins": 0.04497462511062622,
|
| 900 |
+
"rewards/rejected": -0.05386519432067871,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 1.76,
|
| 905 |
+
"learning_rate": 2.111280128780638e-07,
|
| 906 |
+
"logits/chosen": -3.0598039627075195,
|
| 907 |
+
"logits/rejected": -2.9559364318847656,
|
| 908 |
+
"logps/chosen": -84.91885375976562,
|
| 909 |
+
"logps/rejected": -75.99688720703125,
|
| 910 |
+
"loss": 0.0162,
|
| 911 |
+
"rewards/accuracies": 0.75,
|
| 912 |
+
"rewards/chosen": -0.025450268760323524,
|
| 913 |
+
"rewards/margins": 0.03124385140836239,
|
| 914 |
+
"rewards/rejected": -0.056694112718105316,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 1.79,
|
| 919 |
+
"learning_rate": 1.610149735590949e-07,
|
| 920 |
+
"logits/chosen": -2.929579496383667,
|
| 921 |
+
"logits/rejected": -2.8060431480407715,
|
| 922 |
+
"logps/chosen": -60.950294494628906,
|
| 923 |
+
"logps/rejected": -59.30121994018555,
|
| 924 |
+
"loss": 0.0299,
|
| 925 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 926 |
+
"rewards/chosen": -0.03121623769402504,
|
| 927 |
+
"rewards/margins": 0.015263142995536327,
|
| 928 |
+
"rewards/rejected": -0.04647938907146454,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 1.82,
|
| 933 |
+
"learning_rate": 1.1748732956682023e-07,
|
| 934 |
+
"logits/chosen": -3.107717275619507,
|
| 935 |
+
"logits/rejected": -2.9001567363739014,
|
| 936 |
+
"logps/chosen": -90.45118713378906,
|
| 937 |
+
"logps/rejected": -84.7816390991211,
|
| 938 |
+
"loss": 0.0366,
|
| 939 |
+
"rewards/accuracies": 0.75,
|
| 940 |
+
"rewards/chosen": -0.021300766617059708,
|
| 941 |
+
"rewards/margins": 0.0555335208773613,
|
| 942 |
+
"rewards/rejected": -0.0768342837691307,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 1.82,
|
| 947 |
+
"eval_logits/chosen": -2.3873283863067627,
|
| 948 |
+
"eval_logits/rejected": -2.1899969577789307,
|
| 949 |
+
"eval_logps/chosen": -288.27227783203125,
|
| 950 |
+
"eval_logps/rejected": -264.630126953125,
|
| 951 |
+
"eval_loss": 0.011489564552903175,
|
| 952 |
+
"eval_rewards/accuracies": 0.4000000059604645,
|
| 953 |
+
"eval_rewards/chosen": -0.08764798194169998,
|
| 954 |
+
"eval_rewards/margins": -0.02295052260160446,
|
| 955 |
+
"eval_rewards/rejected": -0.06469745188951492,
|
| 956 |
+
"eval_runtime": 1422.4891,
|
| 957 |
+
"eval_samples_per_second": 1.406,
|
| 958 |
+
"eval_steps_per_second": 0.703,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 1.85,
|
| 963 |
+
"learning_rate": 8.066763266625283e-08,
|
| 964 |
+
"logits/chosen": -2.9506607055664062,
|
| 965 |
+
"logits/rejected": -2.825108051300049,
|
| 966 |
+
"logps/chosen": -100.47017669677734,
|
| 967 |
+
"logps/rejected": -127.78892517089844,
|
| 968 |
+
"loss": 0.0349,
|
| 969 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 970 |
+
"rewards/chosen": 0.004581383429467678,
|
| 971 |
+
"rewards/margins": 0.1253204494714737,
|
| 972 |
+
"rewards/rejected": -0.12073905766010284,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 1.88,
|
| 977 |
+
"learning_rate": 5.065954844616722e-08,
|
| 978 |
+
"logits/chosen": -2.9829344749450684,
|
| 979 |
+
"logits/rejected": -2.7227559089660645,
|
| 980 |
+
"logps/chosen": -101.5233154296875,
|
| 981 |
+
"logps/rejected": -99.42388916015625,
|
| 982 |
+
"loss": 0.027,
|
| 983 |
+
"rewards/accuracies": 0.699999988079071,
|
| 984 |
+
"rewards/chosen": -0.0389946773648262,
|
| 985 |
+
"rewards/margins": 0.016298165544867516,
|
| 986 |
+
"rewards/rejected": -0.05529283359646797,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 1.91,
|
| 991 |
+
"learning_rate": 2.7547564449386666e-08,
|
| 992 |
+
"logits/chosen": -2.7666523456573486,
|
| 993 |
+
"logits/rejected": -2.719158887863159,
|
| 994 |
+
"logps/chosen": -215.42599487304688,
|
| 995 |
+
"logps/rejected": -217.2696075439453,
|
| 996 |
+
"loss": 0.0308,
|
| 997 |
+
"rewards/accuracies": 0.699999988079071,
|
| 998 |
+
"rewards/chosen": -0.015156986191868782,
|
| 999 |
+
"rewards/margins": 0.07782644778490067,
|
| 1000 |
+
"rewards/rejected": -0.0929834321141243,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 1.95,
|
| 1005 |
+
"learning_rate": 1.1396752298723501e-08,
|
| 1006 |
+
"logits/chosen": -2.9985458850860596,
|
| 1007 |
+
"logits/rejected": -2.927014112472534,
|
| 1008 |
+
"logps/chosen": -109.7343521118164,
|
| 1009 |
+
"logps/rejected": -111.255859375,
|
| 1010 |
+
"loss": 0.0344,
|
| 1011 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1012 |
+
"rewards/chosen": 0.020465310662984848,
|
| 1013 |
+
"rewards/margins": 0.10245206207036972,
|
| 1014 |
+
"rewards/rejected": -0.08198676258325577,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 1.98,
|
| 1019 |
+
"learning_rate": 2.252584488296461e-09,
|
| 1020 |
+
"logits/chosen": -2.9206175804138184,
|
| 1021 |
+
"logits/rejected": -2.8202061653137207,
|
| 1022 |
+
"logps/chosen": -157.34161376953125,
|
| 1023 |
+
"logps/rejected": -171.29029846191406,
|
| 1024 |
+
"loss": 0.0403,
|
| 1025 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1026 |
+
"rewards/chosen": -0.005368991754949093,
|
| 1027 |
+
"rewards/margins": 0.11630947887897491,
|
| 1028 |
+
"rewards/rejected": -0.12167844921350479,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 2.0,
|
| 1033 |
+
"step": 658,
|
| 1034 |
+
"total_flos": 0.0,
|
| 1035 |
+
"train_loss": 0.014390302914878392,
|
| 1036 |
+
"train_runtime": 11750.1681,
|
| 1037 |
+
"train_samples_per_second": 0.112,
|
| 1038 |
+
"train_steps_per_second": 0.056
|
| 1039 |
+
}
|
| 1040 |
+
],
|
| 1041 |
+
"logging_steps": 10,
|
| 1042 |
+
"max_steps": 658,
|
| 1043 |
+
"num_input_tokens_seen": 0,
|
| 1044 |
+
"num_train_epochs": 2,
|
| 1045 |
+
"save_steps": 100,
|
| 1046 |
+
"total_flos": 0.0,
|
| 1047 |
+
"train_batch_size": 1,
|
| 1048 |
+
"trial_name": null,
|
| 1049 |
+
"trial_params": null
|
| 1050 |
+
}
|