Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- train_results.json +9 -0
- trainer_state.json +977 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
base_model: microsoft/Phi-3-mini-4k-instruct
|
| 9 |
+
model-index:
|
| 10 |
+
- name: phi3-offline-dpo-lora-noise-0.0-5e-6-42
|
| 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 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/causal/huggingface/runs/ds27l9yx)
|
| 18 |
+
# phi3-offline-dpo-lora-noise-0.0-5e-6-42
|
| 19 |
+
|
| 20 |
+
This model is a fine-tuned version of [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) on the None dataset.
|
| 21 |
+
It achieves the following results on the evaluation set:
|
| 22 |
+
- Loss: 0.6633
|
| 23 |
+
- Rewards/chosen: -0.1262
|
| 24 |
+
- Rewards/rejected: -0.1959
|
| 25 |
+
- Rewards/accuracies: 0.7540
|
| 26 |
+
- Rewards/margins: 0.0697
|
| 27 |
+
- Logps/rejected: -403.3280
|
| 28 |
+
- Logps/chosen: -421.1901
|
| 29 |
+
- Logits/rejected: 12.0952
|
| 30 |
+
- Logits/chosen: 13.8997
|
| 31 |
+
|
| 32 |
+
## Model description
|
| 33 |
+
|
| 34 |
+
More information needed
|
| 35 |
+
|
| 36 |
+
## Intended uses & limitations
|
| 37 |
+
|
| 38 |
+
More information needed
|
| 39 |
+
|
| 40 |
+
## Training and evaluation data
|
| 41 |
+
|
| 42 |
+
More information needed
|
| 43 |
+
|
| 44 |
+
## Training procedure
|
| 45 |
+
|
| 46 |
+
### Training hyperparameters
|
| 47 |
+
|
| 48 |
+
The following hyperparameters were used during training:
|
| 49 |
+
- learning_rate: 5e-06
|
| 50 |
+
- train_batch_size: 4
|
| 51 |
+
- eval_batch_size: 4
|
| 52 |
+
- seed: 42
|
| 53 |
+
- distributed_type: multi-GPU
|
| 54 |
+
- num_devices: 4
|
| 55 |
+
- gradient_accumulation_steps: 4
|
| 56 |
+
- total_train_batch_size: 64
|
| 57 |
+
- total_eval_batch_size: 16
|
| 58 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 59 |
+
- lr_scheduler_type: cosine
|
| 60 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 61 |
+
- num_epochs: 1
|
| 62 |
+
|
| 63 |
+
### Training results
|
| 64 |
+
|
| 65 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 66 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 67 |
+
| 0.6931 | 0.1778 | 100 | 0.6835 | -0.0511 | -0.0728 | 0.6905 | 0.0218 | -391.0186 | -413.6780 | 12.3764 | 14.1803 |
|
| 68 |
+
| 0.689 | 0.3556 | 200 | 0.6682 | -0.1441 | -0.2014 | 0.7460 | 0.0573 | -403.8743 | -422.9761 | 12.1803 | 13.9841 |
|
| 69 |
+
| 0.6923 | 0.5333 | 300 | 0.6673 | -0.1140 | -0.1749 | 0.7897 | 0.0609 | -401.2295 | -419.9747 | 12.1769 | 13.9748 |
|
| 70 |
+
| 0.6914 | 0.7111 | 400 | 0.6655 | -0.1195 | -0.1839 | 0.7698 | 0.0644 | -402.1236 | -420.5240 | 12.1267 | 13.9317 |
|
| 71 |
+
| 0.696 | 0.8889 | 500 | 0.6633 | -0.1262 | -0.1959 | 0.7540 | 0.0697 | -403.3280 | -421.1901 | 12.0952 | 13.8997 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- PEFT 0.7.1
|
| 77 |
+
- Transformers 4.42.3
|
| 78 |
+
- Pytorch 2.3.0+cu121
|
| 79 |
+
- Datasets 2.14.6
|
| 80 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9991111111111111,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6916975294143703,
|
| 5 |
+
"train_runtime": 7520.1237,
|
| 6 |
+
"train_samples": 36000,
|
| 7 |
+
"train_samples_per_second": 4.787,
|
| 8 |
+
"train_steps_per_second": 0.075
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9991111111111111,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6916975294143703,
|
| 5 |
+
"train_runtime": 7520.1237,
|
| 6 |
+
"train_samples": 36000,
|
| 7 |
+
"train_samples_per_second": 4.787,
|
| 8 |
+
"train_steps_per_second": 0.075
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,977 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9991111111111111,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 562,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0017777777777777779,
|
| 13 |
+
"grad_norm": 0.22311078073034826,
|
| 14 |
+
"learning_rate": 8.771929824561404e-08,
|
| 15 |
+
"logits/chosen": 14.539060592651367,
|
| 16 |
+
"logits/rejected": 15.870795249938965,
|
| 17 |
+
"logps/chosen": -470.04345703125,
|
| 18 |
+
"logps/rejected": -509.49163818359375,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.017777777777777778,
|
| 28 |
+
"grad_norm": 0.1976540584216542,
|
| 29 |
+
"learning_rate": 8.771929824561404e-07,
|
| 30 |
+
"logits/chosen": 12.96641731262207,
|
| 31 |
+
"logits/rejected": 13.155448913574219,
|
| 32 |
+
"logps/chosen": -400.9219665527344,
|
| 33 |
+
"logps/rejected": -399.60699462890625,
|
| 34 |
+
"loss": 0.693,
|
| 35 |
+
"rewards/accuracies": 0.4513888955116272,
|
| 36 |
+
"rewards/chosen": -0.00211474671959877,
|
| 37 |
+
"rewards/margins": -0.001011626678518951,
|
| 38 |
+
"rewards/rejected": -0.0011031196918338537,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.035555555555555556,
|
| 43 |
+
"grad_norm": 0.21429385122960512,
|
| 44 |
+
"learning_rate": 1.7543859649122807e-06,
|
| 45 |
+
"logits/chosen": 13.817936897277832,
|
| 46 |
+
"logits/rejected": 13.85405158996582,
|
| 47 |
+
"logps/chosen": -420.9461975097656,
|
| 48 |
+
"logps/rejected": -404.7037658691406,
|
| 49 |
+
"loss": 0.6938,
|
| 50 |
+
"rewards/accuracies": 0.4625000059604645,
|
| 51 |
+
"rewards/chosen": -6.537516310345381e-05,
|
| 52 |
+
"rewards/margins": -0.0014555875677615404,
|
| 53 |
+
"rewards/rejected": 0.0013902120990678668,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.05333333333333334,
|
| 58 |
+
"grad_norm": 0.1898289423532557,
|
| 59 |
+
"learning_rate": 2.631578947368421e-06,
|
| 60 |
+
"logits/chosen": 13.255971908569336,
|
| 61 |
+
"logits/rejected": 13.399693489074707,
|
| 62 |
+
"logps/chosen": -402.4463806152344,
|
| 63 |
+
"logps/rejected": -412.36480712890625,
|
| 64 |
+
"loss": 0.6937,
|
| 65 |
+
"rewards/accuracies": 0.5,
|
| 66 |
+
"rewards/chosen": -0.0020085538271814585,
|
| 67 |
+
"rewards/margins": -0.0013788806973025203,
|
| 68 |
+
"rewards/rejected": -0.00062967324629426,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.07111111111111111,
|
| 73 |
+
"grad_norm": 0.2162185808879369,
|
| 74 |
+
"learning_rate": 3.5087719298245615e-06,
|
| 75 |
+
"logits/chosen": 13.550105094909668,
|
| 76 |
+
"logits/rejected": 13.091300964355469,
|
| 77 |
+
"logps/chosen": -379.90777587890625,
|
| 78 |
+
"logps/rejected": -373.92138671875,
|
| 79 |
+
"loss": 0.6935,
|
| 80 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 81 |
+
"rewards/chosen": 0.0008380211656913161,
|
| 82 |
+
"rewards/margins": -0.00024208976537920535,
|
| 83 |
+
"rewards/rejected": 0.0010801110183820128,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.08888888888888889,
|
| 88 |
+
"grad_norm": 0.22810063623939997,
|
| 89 |
+
"learning_rate": 4.385964912280702e-06,
|
| 90 |
+
"logits/chosen": 13.239973068237305,
|
| 91 |
+
"logits/rejected": 13.241312980651855,
|
| 92 |
+
"logps/chosen": -396.4941101074219,
|
| 93 |
+
"logps/rejected": -430.57049560546875,
|
| 94 |
+
"loss": 0.6932,
|
| 95 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 96 |
+
"rewards/chosen": -0.0004129047447349876,
|
| 97 |
+
"rewards/margins": 4.834262654185295e-05,
|
| 98 |
+
"rewards/rejected": -0.00046124737127684057,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.10666666666666667,
|
| 103 |
+
"grad_norm": 0.2512545993725567,
|
| 104 |
+
"learning_rate": 4.999564631597802e-06,
|
| 105 |
+
"logits/chosen": 13.0819673538208,
|
| 106 |
+
"logits/rejected": 13.337594985961914,
|
| 107 |
+
"logps/chosen": -388.62628173828125,
|
| 108 |
+
"logps/rejected": -414.85382080078125,
|
| 109 |
+
"loss": 0.6928,
|
| 110 |
+
"rewards/accuracies": 0.518750011920929,
|
| 111 |
+
"rewards/chosen": -0.002572793047875166,
|
| 112 |
+
"rewards/margins": 0.0024341598618775606,
|
| 113 |
+
"rewards/rejected": -0.005006952676922083,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.12444444444444444,
|
| 118 |
+
"grad_norm": 0.22761734037588918,
|
| 119 |
+
"learning_rate": 4.991828966534002e-06,
|
| 120 |
+
"logits/chosen": 13.887951850891113,
|
| 121 |
+
"logits/rejected": 13.595651626586914,
|
| 122 |
+
"logps/chosen": -450.8296813964844,
|
| 123 |
+
"logps/rejected": -444.7730407714844,
|
| 124 |
+
"loss": 0.6936,
|
| 125 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 126 |
+
"rewards/chosen": -0.006924263201653957,
|
| 127 |
+
"rewards/margins": -0.001168354763649404,
|
| 128 |
+
"rewards/rejected": -0.005755907855927944,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.14222222222222222,
|
| 133 |
+
"grad_norm": 0.21722031386603705,
|
| 134 |
+
"learning_rate": 4.974452899279292e-06,
|
| 135 |
+
"logits/chosen": 13.351213455200195,
|
| 136 |
+
"logits/rejected": 12.537083625793457,
|
| 137 |
+
"logps/chosen": -425.934814453125,
|
| 138 |
+
"logps/rejected": -383.2563781738281,
|
| 139 |
+
"loss": 0.6932,
|
| 140 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 141 |
+
"rewards/chosen": -0.015285441651940346,
|
| 142 |
+
"rewards/margins": -0.0011091658379882574,
|
| 143 |
+
"rewards/rejected": -0.014176277443766594,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.16,
|
| 148 |
+
"grad_norm": 0.27601757784168,
|
| 149 |
+
"learning_rate": 4.947503654462277e-06,
|
| 150 |
+
"logits/chosen": 13.753236770629883,
|
| 151 |
+
"logits/rejected": 13.5679292678833,
|
| 152 |
+
"logps/chosen": -435.3922424316406,
|
| 153 |
+
"logps/rejected": -426.5663146972656,
|
| 154 |
+
"loss": 0.693,
|
| 155 |
+
"rewards/accuracies": 0.4437499940395355,
|
| 156 |
+
"rewards/chosen": -0.03359255567193031,
|
| 157 |
+
"rewards/margins": -0.0016514979070052505,
|
| 158 |
+
"rewards/rejected": -0.03194105625152588,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.17777777777777778,
|
| 163 |
+
"grad_norm": 0.23435867570612137,
|
| 164 |
+
"learning_rate": 4.911085493475802e-06,
|
| 165 |
+
"logits/chosen": 12.843754768371582,
|
| 166 |
+
"logits/rejected": 12.69409465789795,
|
| 167 |
+
"logps/chosen": -412.6481018066406,
|
| 168 |
+
"logps/rejected": -412.96405029296875,
|
| 169 |
+
"loss": 0.6931,
|
| 170 |
+
"rewards/accuracies": 0.518750011920929,
|
| 171 |
+
"rewards/chosen": -0.0513535737991333,
|
| 172 |
+
"rewards/margins": 0.002654131967574358,
|
| 173 |
+
"rewards/rejected": -0.05400770157575607,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.17777777777777778,
|
| 178 |
+
"eval_logits/chosen": 14.180258750915527,
|
| 179 |
+
"eval_logits/rejected": 12.376383781433105,
|
| 180 |
+
"eval_logps/chosen": -413.67803955078125,
|
| 181 |
+
"eval_logps/rejected": -391.0186462402344,
|
| 182 |
+
"eval_loss": 0.6835460066795349,
|
| 183 |
+
"eval_rewards/accuracies": 0.6904761791229248,
|
| 184 |
+
"eval_rewards/chosen": -0.05108103156089783,
|
| 185 |
+
"eval_rewards/margins": 0.021760080009698868,
|
| 186 |
+
"eval_rewards/rejected": -0.072841115295887,
|
| 187 |
+
"eval_runtime": 90.1605,
|
| 188 |
+
"eval_samples_per_second": 11.091,
|
| 189 |
+
"eval_steps_per_second": 0.699,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.19555555555555557,
|
| 194 |
+
"grad_norm": 0.2803874830493494,
|
| 195 |
+
"learning_rate": 4.86533931110987e-06,
|
| 196 |
+
"logits/chosen": 13.455083847045898,
|
| 197 |
+
"logits/rejected": 13.341041564941406,
|
| 198 |
+
"logps/chosen": -415.30401611328125,
|
| 199 |
+
"logps/rejected": -425.7816467285156,
|
| 200 |
+
"loss": 0.6933,
|
| 201 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 202 |
+
"rewards/chosen": -0.06350487470626831,
|
| 203 |
+
"rewards/margins": 0.00028126564575359225,
|
| 204 |
+
"rewards/rejected": -0.06378613412380219,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.21333333333333335,
|
| 209 |
+
"grad_norm": 0.35378680708306215,
|
| 210 |
+
"learning_rate": 4.810442090457072e-06,
|
| 211 |
+
"logits/chosen": 12.907583236694336,
|
| 212 |
+
"logits/rejected": 13.067194938659668,
|
| 213 |
+
"logps/chosen": -389.30303955078125,
|
| 214 |
+
"logps/rejected": -398.48333740234375,
|
| 215 |
+
"loss": 0.6925,
|
| 216 |
+
"rewards/accuracies": 0.5,
|
| 217 |
+
"rewards/chosen": -0.06600853055715561,
|
| 218 |
+
"rewards/margins": -0.003265662584453821,
|
| 219 |
+
"rewards/rejected": -0.06274287402629852,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2311111111111111,
|
| 224 |
+
"grad_norm": 0.2869743496766919,
|
| 225 |
+
"learning_rate": 4.7466062181993855e-06,
|
| 226 |
+
"logits/chosen": 13.347814559936523,
|
| 227 |
+
"logits/rejected": 13.254618644714355,
|
| 228 |
+
"logps/chosen": -396.9159851074219,
|
| 229 |
+
"logps/rejected": -428.38861083984375,
|
| 230 |
+
"loss": 0.6921,
|
| 231 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 232 |
+
"rewards/chosen": -0.08628938347101212,
|
| 233 |
+
"rewards/margins": 0.005706036929041147,
|
| 234 |
+
"rewards/rejected": -0.09199541807174683,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.24888888888888888,
|
| 239 |
+
"grad_norm": 0.3379629862483891,
|
| 240 |
+
"learning_rate": 4.6740786629253595e-06,
|
| 241 |
+
"logits/chosen": 13.342312812805176,
|
| 242 |
+
"logits/rejected": 13.019983291625977,
|
| 243 |
+
"logps/chosen": -384.1279602050781,
|
| 244 |
+
"logps/rejected": -393.2745666503906,
|
| 245 |
+
"loss": 0.6912,
|
| 246 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 247 |
+
"rewards/chosen": -0.0947018712759018,
|
| 248 |
+
"rewards/margins": 0.0077432007528841496,
|
| 249 |
+
"rewards/rejected": -0.10244506597518921,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.26666666666666666,
|
| 254 |
+
"grad_norm": 0.350219971213616,
|
| 255 |
+
"learning_rate": 4.5931400196566256e-06,
|
| 256 |
+
"logits/chosen": 13.074136734008789,
|
| 257 |
+
"logits/rejected": 13.148015022277832,
|
| 258 |
+
"logps/chosen": -425.1592712402344,
|
| 259 |
+
"logps/rejected": -443.909423828125,
|
| 260 |
+
"loss": 0.6903,
|
| 261 |
+
"rewards/accuracies": 0.512499988079071,
|
| 262 |
+
"rewards/chosen": -0.11653944104909897,
|
| 263 |
+
"rewards/margins": 0.003865143982693553,
|
| 264 |
+
"rewards/rejected": -0.12040458619594574,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.28444444444444444,
|
| 269 |
+
"grad_norm": 0.37627729798456006,
|
| 270 |
+
"learning_rate": 4.504103424280267e-06,
|
| 271 |
+
"logits/chosen": 13.34211540222168,
|
| 272 |
+
"logits/rejected": 13.010579109191895,
|
| 273 |
+
"logps/chosen": -439.13916015625,
|
| 274 |
+
"logps/rejected": -433.8727111816406,
|
| 275 |
+
"loss": 0.6908,
|
| 276 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 277 |
+
"rewards/chosen": -0.16217225790023804,
|
| 278 |
+
"rewards/margins": 0.007592835463583469,
|
| 279 |
+
"rewards/rejected": -0.16976511478424072,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.3022222222222222,
|
| 284 |
+
"grad_norm": 0.3180727458224856,
|
| 285 |
+
"learning_rate": 4.407313342086906e-06,
|
| 286 |
+
"logits/chosen": 13.196157455444336,
|
| 287 |
+
"logits/rejected": 13.456674575805664,
|
| 288 |
+
"logps/chosen": -426.7943420410156,
|
| 289 |
+
"logps/rejected": -437.723388671875,
|
| 290 |
+
"loss": 0.6937,
|
| 291 |
+
"rewards/accuracies": 0.5625,
|
| 292 |
+
"rewards/chosen": -0.16415511071681976,
|
| 293 |
+
"rewards/margins": 0.009541703388094902,
|
| 294 |
+
"rewards/rejected": -0.17369681596755981,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.32,
|
| 299 |
+
"grad_norm": 0.31672781627835284,
|
| 300 |
+
"learning_rate": 4.303144235101412e-06,
|
| 301 |
+
"logits/chosen": 13.912513732910156,
|
| 302 |
+
"logits/rejected": 13.92699146270752,
|
| 303 |
+
"logps/chosen": -400.5416564941406,
|
| 304 |
+
"logps/rejected": -418.99554443359375,
|
| 305 |
+
"loss": 0.6907,
|
| 306 |
+
"rewards/accuracies": 0.5,
|
| 307 |
+
"rewards/chosen": -0.18776951730251312,
|
| 308 |
+
"rewards/margins": 0.00795517023652792,
|
| 309 |
+
"rewards/rejected": -0.1957246959209442,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.3377777777777778,
|
| 314 |
+
"grad_norm": 0.3187670525496018,
|
| 315 |
+
"learning_rate": 4.1919991133620705e-06,
|
| 316 |
+
"logits/chosen": 12.957951545715332,
|
| 317 |
+
"logits/rejected": 12.835040092468262,
|
| 318 |
+
"logps/chosen": -428.75921630859375,
|
| 319 |
+
"logps/rejected": -429.77801513671875,
|
| 320 |
+
"loss": 0.6943,
|
| 321 |
+
"rewards/accuracies": 0.4437499940395355,
|
| 322 |
+
"rewards/chosen": -0.18785884976387024,
|
| 323 |
+
"rewards/margins": -0.00687809195369482,
|
| 324 |
+
"rewards/rejected": -0.18098074197769165,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.35555555555555557,
|
| 329 |
+
"grad_norm": 0.31966290081913296,
|
| 330 |
+
"learning_rate": 4.074307975753044e-06,
|
| 331 |
+
"logits/chosen": 13.527238845825195,
|
| 332 |
+
"logits/rejected": 13.133882522583008,
|
| 333 |
+
"logps/chosen": -411.04052734375,
|
| 334 |
+
"logps/rejected": -405.96820068359375,
|
| 335 |
+
"loss": 0.689,
|
| 336 |
+
"rewards/accuracies": 0.550000011920929,
|
| 337 |
+
"rewards/chosen": -0.1447705328464508,
|
| 338 |
+
"rewards/margins": 0.013265645131468773,
|
| 339 |
+
"rewards/rejected": -0.15803618729114532,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.35555555555555557,
|
| 344 |
+
"eval_logits/chosen": 13.984086036682129,
|
| 345 |
+
"eval_logits/rejected": 12.180322647094727,
|
| 346 |
+
"eval_logps/chosen": -422.97613525390625,
|
| 347 |
+
"eval_logps/rejected": -403.87432861328125,
|
| 348 |
+
"eval_loss": 0.6681899428367615,
|
| 349 |
+
"eval_rewards/accuracies": 0.7460317611694336,
|
| 350 |
+
"eval_rewards/chosen": -0.1440620720386505,
|
| 351 |
+
"eval_rewards/margins": 0.05733573064208031,
|
| 352 |
+
"eval_rewards/rejected": -0.20139780640602112,
|
| 353 |
+
"eval_runtime": 90.195,
|
| 354 |
+
"eval_samples_per_second": 11.087,
|
| 355 |
+
"eval_steps_per_second": 0.698,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.37333333333333335,
|
| 360 |
+
"grad_norm": 0.2899334647869614,
|
| 361 |
+
"learning_rate": 3.950526146422213e-06,
|
| 362 |
+
"logits/chosen": 13.182365417480469,
|
| 363 |
+
"logits/rejected": 13.274526596069336,
|
| 364 |
+
"logps/chosen": -412.0303649902344,
|
| 365 |
+
"logps/rejected": -435.66754150390625,
|
| 366 |
+
"loss": 0.6921,
|
| 367 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 368 |
+
"rewards/chosen": -0.1916799396276474,
|
| 369 |
+
"rewards/margins": -0.0004939109203405678,
|
| 370 |
+
"rewards/rejected": -0.19118604063987732,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.39111111111111113,
|
| 375 |
+
"grad_norm": 0.3113343433252108,
|
| 376 |
+
"learning_rate": 3.821132513220511e-06,
|
| 377 |
+
"logits/chosen": 13.472271919250488,
|
| 378 |
+
"logits/rejected": 12.98419189453125,
|
| 379 |
+
"logps/chosen": -407.59857177734375,
|
| 380 |
+
"logps/rejected": -413.1893615722656,
|
| 381 |
+
"loss": 0.6852,
|
| 382 |
+
"rewards/accuracies": 0.543749988079071,
|
| 383 |
+
"rewards/chosen": -0.18176871538162231,
|
| 384 |
+
"rewards/margins": 0.02464126981794834,
|
| 385 |
+
"rewards/rejected": -0.2064099758863449,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4088888888888889,
|
| 390 |
+
"grad_norm": 0.37212373445120756,
|
| 391 |
+
"learning_rate": 3.686627674977858e-06,
|
| 392 |
+
"logits/chosen": 12.922555923461914,
|
| 393 |
+
"logits/rejected": 13.052223205566406,
|
| 394 |
+
"logps/chosen": -420.0254821777344,
|
| 395 |
+
"logps/rejected": -438.8111267089844,
|
| 396 |
+
"loss": 0.6927,
|
| 397 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 398 |
+
"rewards/chosen": -0.2267444133758545,
|
| 399 |
+
"rewards/margins": 0.00632152333855629,
|
| 400 |
+
"rewards/rejected": -0.23306593298912048,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.4266666666666667,
|
| 405 |
+
"grad_norm": 0.4194286554865179,
|
| 406 |
+
"learning_rate": 3.547532004783539e-06,
|
| 407 |
+
"logits/chosen": 13.442713737487793,
|
| 408 |
+
"logits/rejected": 12.382542610168457,
|
| 409 |
+
"logps/chosen": -448.96044921875,
|
| 410 |
+
"logps/rejected": -424.40631103515625,
|
| 411 |
+
"loss": 0.6883,
|
| 412 |
+
"rewards/accuracies": 0.643750011920929,
|
| 413 |
+
"rewards/chosen": -0.2144048511981964,
|
| 414 |
+
"rewards/margins": 0.04138387367129326,
|
| 415 |
+
"rewards/rejected": -0.2557887136936188,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.4444444444444444,
|
| 420 |
+
"grad_norm": 0.2831313864511856,
|
| 421 |
+
"learning_rate": 3.404383636763809e-06,
|
| 422 |
+
"logits/chosen": 13.222406387329102,
|
| 423 |
+
"logits/rejected": 12.874628067016602,
|
| 424 |
+
"logps/chosen": -460.45782470703125,
|
| 425 |
+
"logps/rejected": -457.3854064941406,
|
| 426 |
+
"loss": 0.6915,
|
| 427 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 428 |
+
"rewards/chosen": -0.22632627189159393,
|
| 429 |
+
"rewards/margins": 0.018462661653757095,
|
| 430 |
+
"rewards/rejected": -0.24478892982006073,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.4622222222222222,
|
| 435 |
+
"grad_norm": 0.3260671824484102,
|
| 436 |
+
"learning_rate": 3.2577363841455063e-06,
|
| 437 |
+
"logits/chosen": 13.391107559204102,
|
| 438 |
+
"logits/rejected": 13.3804931640625,
|
| 439 |
+
"logps/chosen": -380.6063537597656,
|
| 440 |
+
"logps/rejected": -394.7472839355469,
|
| 441 |
+
"loss": 0.6954,
|
| 442 |
+
"rewards/accuracies": 0.53125,
|
| 443 |
+
"rewards/chosen": -0.20835450291633606,
|
| 444 |
+
"rewards/margins": 0.013799709267914295,
|
| 445 |
+
"rewards/rejected": -0.22215421497821808,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.48,
|
| 450 |
+
"grad_norm": 0.3777555315785377,
|
| 451 |
+
"learning_rate": 3.1081575966602627e-06,
|
| 452 |
+
"logits/chosen": 12.585054397583008,
|
| 453 |
+
"logits/rejected": 12.490537643432617,
|
| 454 |
+
"logps/chosen": -439.46844482421875,
|
| 455 |
+
"logps/rejected": -421.29962158203125,
|
| 456 |
+
"loss": 0.6929,
|
| 457 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 458 |
+
"rewards/chosen": -0.19497910141944885,
|
| 459 |
+
"rewards/margins": -0.00843017641454935,
|
| 460 |
+
"rewards/rejected": -0.18654890358448029,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.49777777777777776,
|
| 465 |
+
"grad_norm": 0.2554633592042652,
|
| 466 |
+
"learning_rate": 2.9562259655786067e-06,
|
| 467 |
+
"logits/chosen": 13.010714530944824,
|
| 468 |
+
"logits/rejected": 12.371983528137207,
|
| 469 |
+
"logps/chosen": -389.00665283203125,
|
| 470 |
+
"logps/rejected": -389.6372985839844,
|
| 471 |
+
"loss": 0.6933,
|
| 472 |
+
"rewards/accuracies": 0.518750011920929,
|
| 473 |
+
"rewards/chosen": -0.14177796244621277,
|
| 474 |
+
"rewards/margins": 0.008302886970341206,
|
| 475 |
+
"rewards/rejected": -0.1500808447599411,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.5155555555555555,
|
| 480 |
+
"grad_norm": 0.27965221608690755,
|
| 481 |
+
"learning_rate": 2.802529284865863e-06,
|
| 482 |
+
"logits/chosen": 12.02336597442627,
|
| 483 |
+
"logits/rejected": 12.071900367736816,
|
| 484 |
+
"logps/chosen": -402.3577575683594,
|
| 485 |
+
"logps/rejected": -393.9205017089844,
|
| 486 |
+
"loss": 0.6895,
|
| 487 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 488 |
+
"rewards/chosen": -0.13531939685344696,
|
| 489 |
+
"rewards/margins": 0.004990140907466412,
|
| 490 |
+
"rewards/rejected": -0.14030954241752625,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.5333333333333333,
|
| 495 |
+
"grad_norm": 0.2913462405762709,
|
| 496 |
+
"learning_rate": 2.6476621771214865e-06,
|
| 497 |
+
"logits/chosen": 13.3645601272583,
|
| 498 |
+
"logits/rejected": 13.140283584594727,
|
| 499 |
+
"logps/chosen": -407.8977966308594,
|
| 500 |
+
"logps/rejected": -416.1576232910156,
|
| 501 |
+
"loss": 0.6923,
|
| 502 |
+
"rewards/accuracies": 0.543749988079071,
|
| 503 |
+
"rewards/chosen": -0.1288231909275055,
|
| 504 |
+
"rewards/margins": 0.010681845247745514,
|
| 505 |
+
"rewards/rejected": -0.1395050287246704,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.5333333333333333,
|
| 510 |
+
"eval_logits/chosen": 13.974791526794434,
|
| 511 |
+
"eval_logits/rejected": 12.176884651184082,
|
| 512 |
+
"eval_logps/chosen": -419.9747009277344,
|
| 513 |
+
"eval_logps/rejected": -401.2295227050781,
|
| 514 |
+
"eval_loss": 0.6673460006713867,
|
| 515 |
+
"eval_rewards/accuracies": 0.7896825671195984,
|
| 516 |
+
"eval_rewards/chosen": -0.11404754966497421,
|
| 517 |
+
"eval_rewards/margins": 0.06090213730931282,
|
| 518 |
+
"eval_rewards/rejected": -0.17494967579841614,
|
| 519 |
+
"eval_runtime": 90.2773,
|
| 520 |
+
"eval_samples_per_second": 11.077,
|
| 521 |
+
"eval_steps_per_second": 0.698,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.5511111111111111,
|
| 526 |
+
"grad_norm": 0.340661457280574,
|
| 527 |
+
"learning_rate": 2.4922237930997435e-06,
|
| 528 |
+
"logits/chosen": 12.930778503417969,
|
| 529 |
+
"logits/rejected": 12.75629711151123,
|
| 530 |
+
"logps/chosen": -408.3033142089844,
|
| 531 |
+
"logps/rejected": -429.09954833984375,
|
| 532 |
+
"loss": 0.6891,
|
| 533 |
+
"rewards/accuracies": 0.5,
|
| 534 |
+
"rewards/chosen": -0.1441982239484787,
|
| 535 |
+
"rewards/margins": 0.008970534428954124,
|
| 536 |
+
"rewards/rejected": -0.15316873788833618,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.5688888888888889,
|
| 541 |
+
"grad_norm": 0.3045715209107194,
|
| 542 |
+
"learning_rate": 2.3368154937118355e-06,
|
| 543 |
+
"logits/chosen": 12.338762283325195,
|
| 544 |
+
"logits/rejected": 12.408100128173828,
|
| 545 |
+
"logps/chosen": -417.6708984375,
|
| 546 |
+
"logps/rejected": -436.1878356933594,
|
| 547 |
+
"loss": 0.6906,
|
| 548 |
+
"rewards/accuracies": 0.53125,
|
| 549 |
+
"rewards/chosen": -0.1610310822725296,
|
| 550 |
+
"rewards/margins": 0.003636928740888834,
|
| 551 |
+
"rewards/rejected": -0.164668008685112,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.5866666666666667,
|
| 556 |
+
"grad_norm": 0.3261096004039151,
|
| 557 |
+
"learning_rate": 2.1820385234773604e-06,
|
| 558 |
+
"logits/chosen": 12.432132720947266,
|
| 559 |
+
"logits/rejected": 12.333971977233887,
|
| 560 |
+
"logps/chosen": -371.9541015625,
|
| 561 |
+
"logps/rejected": -377.29327392578125,
|
| 562 |
+
"loss": 0.6938,
|
| 563 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 564 |
+
"rewards/chosen": -0.14137257635593414,
|
| 565 |
+
"rewards/margins": 0.012616041116416454,
|
| 566 |
+
"rewards/rejected": -0.15398862957954407,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.6044444444444445,
|
| 571 |
+
"grad_norm": 0.3167841738260838,
|
| 572 |
+
"learning_rate": 2.02849168442607e-06,
|
| 573 |
+
"logits/chosen": 13.544441223144531,
|
| 574 |
+
"logits/rejected": 13.638467788696289,
|
| 575 |
+
"logps/chosen": -407.0322265625,
|
| 576 |
+
"logps/rejected": -412.5428771972656,
|
| 577 |
+
"loss": 0.6934,
|
| 578 |
+
"rewards/accuracies": 0.543749988079071,
|
| 579 |
+
"rewards/chosen": -0.1435524821281433,
|
| 580 |
+
"rewards/margins": 0.008516514673829079,
|
| 581 |
+
"rewards/rejected": -0.15206900238990784,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.6222222222222222,
|
| 586 |
+
"grad_norm": 0.30772925863397405,
|
| 587 |
+
"learning_rate": 1.876769019449141e-06,
|
| 588 |
+
"logits/chosen": 12.576923370361328,
|
| 589 |
+
"logits/rejected": 12.56715202331543,
|
| 590 |
+
"logps/chosen": -379.2666015625,
|
| 591 |
+
"logps/rejected": -408.4591979980469,
|
| 592 |
+
"loss": 0.6904,
|
| 593 |
+
"rewards/accuracies": 0.550000011920929,
|
| 594 |
+
"rewards/chosen": -0.12999220192432404,
|
| 595 |
+
"rewards/margins": 0.009667792357504368,
|
| 596 |
+
"rewards/rejected": -0.13966000080108643,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.64,
|
| 601 |
+
"grad_norm": 0.3510251282815541,
|
| 602 |
+
"learning_rate": 1.7274575140626318e-06,
|
| 603 |
+
"logits/chosen": 12.569269180297852,
|
| 604 |
+
"logits/rejected": 12.474992752075195,
|
| 605 |
+
"logps/chosen": -426.98651123046875,
|
| 606 |
+
"logps/rejected": -420.2184143066406,
|
| 607 |
+
"loss": 0.693,
|
| 608 |
+
"rewards/accuracies": 0.46875,
|
| 609 |
+
"rewards/chosen": -0.14840544760227203,
|
| 610 |
+
"rewards/margins": -0.0038787845987826586,
|
| 611 |
+
"rewards/rejected": -0.1445266604423523,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.6577777777777778,
|
| 616 |
+
"grad_norm": 0.29359728465406504,
|
| 617 |
+
"learning_rate": 1.5811348254745574e-06,
|
| 618 |
+
"logits/chosen": 13.278701782226562,
|
| 619 |
+
"logits/rejected": 13.138870239257812,
|
| 620 |
+
"logps/chosen": -419.22552490234375,
|
| 621 |
+
"logps/rejected": -429.7049865722656,
|
| 622 |
+
"loss": 0.6907,
|
| 623 |
+
"rewards/accuracies": 0.550000011920929,
|
| 624 |
+
"rewards/chosen": -0.1437017023563385,
|
| 625 |
+
"rewards/margins": 0.006204875651746988,
|
| 626 |
+
"rewards/rejected": -0.14990659058094025,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.6755555555555556,
|
| 631 |
+
"grad_norm": 0.3041311221496272,
|
| 632 |
+
"learning_rate": 1.4383670477413676e-06,
|
| 633 |
+
"logits/chosen": 12.825994491577148,
|
| 634 |
+
"logits/rejected": 12.180424690246582,
|
| 635 |
+
"logps/chosen": -393.3042297363281,
|
| 636 |
+
"logps/rejected": -379.05279541015625,
|
| 637 |
+
"loss": 0.6881,
|
| 638 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 639 |
+
"rewards/chosen": -0.1392853856086731,
|
| 640 |
+
"rewards/margins": 0.007530958391726017,
|
| 641 |
+
"rewards/rejected": -0.14681634306907654,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.6933333333333334,
|
| 646 |
+
"grad_norm": 0.3072757460893545,
|
| 647 |
+
"learning_rate": 1.2997065216600179e-06,
|
| 648 |
+
"logits/chosen": 12.89039134979248,
|
| 649 |
+
"logits/rejected": 13.162782669067383,
|
| 650 |
+
"logps/chosen": -427.39910888671875,
|
| 651 |
+
"logps/rejected": -431.7848205566406,
|
| 652 |
+
"loss": 0.6928,
|
| 653 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 654 |
+
"rewards/chosen": -0.14786246418952942,
|
| 655 |
+
"rewards/margins": 0.0003430729848332703,
|
| 656 |
+
"rewards/rejected": -0.14820551872253418,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.7111111111111111,
|
| 661 |
+
"grad_norm": 0.27986512697276633,
|
| 662 |
+
"learning_rate": 1.165689697868726e-06,
|
| 663 |
+
"logits/chosen": 12.779977798461914,
|
| 664 |
+
"logits/rejected": 12.16929817199707,
|
| 665 |
+
"logps/chosen": -420.75653076171875,
|
| 666 |
+
"logps/rejected": -415.25244140625,
|
| 667 |
+
"loss": 0.6914,
|
| 668 |
+
"rewards/accuracies": 0.53125,
|
| 669 |
+
"rewards/chosen": -0.14449290931224823,
|
| 670 |
+
"rewards/margins": 0.02024017833173275,
|
| 671 |
+
"rewards/rejected": -0.16473311185836792,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.7111111111111111,
|
| 676 |
+
"eval_logits/chosen": 13.931650161743164,
|
| 677 |
+
"eval_logits/rejected": 12.126729011535645,
|
| 678 |
+
"eval_logps/chosen": -420.5239562988281,
|
| 679 |
+
"eval_logps/rejected": -402.1236267089844,
|
| 680 |
+
"eval_loss": 0.665495753288269,
|
| 681 |
+
"eval_rewards/accuracies": 0.7698412537574768,
|
| 682 |
+
"eval_rewards/chosen": -0.11954014003276825,
|
| 683 |
+
"eval_rewards/margins": 0.06435071676969528,
|
| 684 |
+
"eval_rewards/rejected": -0.18389087915420532,
|
| 685 |
+
"eval_runtime": 90.069,
|
| 686 |
+
"eval_samples_per_second": 11.103,
|
| 687 |
+
"eval_steps_per_second": 0.699,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.7288888888888889,
|
| 692 |
+
"grad_norm": 0.3450004634992278,
|
| 693 |
+
"learning_rate": 1.0368350614236685e-06,
|
| 694 |
+
"logits/chosen": 12.588386535644531,
|
| 695 |
+
"logits/rejected": 12.996353149414062,
|
| 696 |
+
"logps/chosen": -403.56207275390625,
|
| 697 |
+
"logps/rejected": -424.3435974121094,
|
| 698 |
+
"loss": 0.6911,
|
| 699 |
+
"rewards/accuracies": 0.5,
|
| 700 |
+
"rewards/chosen": -0.15337641537189484,
|
| 701 |
+
"rewards/margins": 0.0022048167884349823,
|
| 702 |
+
"rewards/rejected": -0.15558123588562012,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.7466666666666667,
|
| 707 |
+
"grad_norm": 0.2947231406284332,
|
| 708 |
+
"learning_rate": 9.136411258810229e-07,
|
| 709 |
+
"logits/chosen": 13.508552551269531,
|
| 710 |
+
"logits/rejected": 13.856636047363281,
|
| 711 |
+
"logps/chosen": -397.31103515625,
|
| 712 |
+
"logps/rejected": -415.14794921875,
|
| 713 |
+
"loss": 0.6916,
|
| 714 |
+
"rewards/accuracies": 0.53125,
|
| 715 |
+
"rewards/chosen": -0.14091812074184418,
|
| 716 |
+
"rewards/margins": 0.009603964164853096,
|
| 717 |
+
"rewards/rejected": -0.15052208304405212,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.7644444444444445,
|
| 722 |
+
"grad_norm": 0.2949706409885573,
|
| 723 |
+
"learning_rate": 7.965845046448659e-07,
|
| 724 |
+
"logits/chosen": 12.716641426086426,
|
| 725 |
+
"logits/rejected": 12.739425659179688,
|
| 726 |
+
"logps/chosen": -413.5958557128906,
|
| 727 |
+
"logps/rejected": -418.07379150390625,
|
| 728 |
+
"loss": 0.6924,
|
| 729 |
+
"rewards/accuracies": 0.543749988079071,
|
| 730 |
+
"rewards/chosen": -0.13801425695419312,
|
| 731 |
+
"rewards/margins": 0.0007665277225896716,
|
| 732 |
+
"rewards/rejected": -0.13878078758716583,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.7822222222222223,
|
| 737 |
+
"grad_norm": 0.29714148256405104,
|
| 738 |
+
"learning_rate": 6.861180670424983e-07,
|
| 739 |
+
"logits/chosen": 13.228073120117188,
|
| 740 |
+
"logits/rejected": 12.513433456420898,
|
| 741 |
+
"logps/chosen": -436.03326416015625,
|
| 742 |
+
"logps/rejected": -432.7162170410156,
|
| 743 |
+
"loss": 0.6882,
|
| 744 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 745 |
+
"rewards/chosen": -0.14261861145496368,
|
| 746 |
+
"rewards/margins": 0.018722299486398697,
|
| 747 |
+
"rewards/rejected": -0.16134092211723328,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.8,
|
| 752 |
+
"grad_norm": 0.35583565190086214,
|
| 753 |
+
"learning_rate": 5.826691862609987e-07,
|
| 754 |
+
"logits/chosen": 13.008562088012695,
|
| 755 |
+
"logits/rejected": 12.5421724319458,
|
| 756 |
+
"logps/chosen": -393.10614013671875,
|
| 757 |
+
"logps/rejected": -396.7547607421875,
|
| 758 |
+
"loss": 0.6858,
|
| 759 |
+
"rewards/accuracies": 0.543749988079071,
|
| 760 |
+
"rewards/chosen": -0.14102980494499207,
|
| 761 |
+
"rewards/margins": 0.006347469985485077,
|
| 762 |
+
"rewards/rejected": -0.14737728238105774,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.8177777777777778,
|
| 767 |
+
"grad_norm": 1.2849378777119373,
|
| 768 |
+
"learning_rate": 4.866380859233891e-07,
|
| 769 |
+
"logits/chosen": 12.925387382507324,
|
| 770 |
+
"logits/rejected": 13.122329711914062,
|
| 771 |
+
"logps/chosen": -406.62255859375,
|
| 772 |
+
"logps/rejected": -432.58575439453125,
|
| 773 |
+
"loss": 0.6911,
|
| 774 |
+
"rewards/accuracies": 0.543749988079071,
|
| 775 |
+
"rewards/chosen": -0.13780517876148224,
|
| 776 |
+
"rewards/margins": 0.009602868929505348,
|
| 777 |
+
"rewards/rejected": -0.14740803837776184,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.8355555555555556,
|
| 782 |
+
"grad_norm": 0.3135496868277975,
|
| 783 |
+
"learning_rate": 3.98396291701183e-07,
|
| 784 |
+
"logits/chosen": 13.169825553894043,
|
| 785 |
+
"logits/rejected": 12.919093132019043,
|
| 786 |
+
"logps/chosen": -420.40594482421875,
|
| 787 |
+
"logps/rejected": -424.626220703125,
|
| 788 |
+
"loss": 0.6887,
|
| 789 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 790 |
+
"rewards/chosen": -0.14023754000663757,
|
| 791 |
+
"rewards/margins": 0.012833138927817345,
|
| 792 |
+
"rewards/rejected": -0.15307065844535828,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.8533333333333334,
|
| 797 |
+
"grad_norm": 0.3340559430991029,
|
| 798 |
+
"learning_rate": 3.1828519395374095e-07,
|
| 799 |
+
"logits/chosen": 13.121709823608398,
|
| 800 |
+
"logits/rejected": 13.285209655761719,
|
| 801 |
+
"logps/chosen": -428.4640197753906,
|
| 802 |
+
"logps/rejected": -445.868408203125,
|
| 803 |
+
"loss": 0.6934,
|
| 804 |
+
"rewards/accuracies": 0.53125,
|
| 805 |
+
"rewards/chosen": -0.1603536158800125,
|
| 806 |
+
"rewards/margins": 0.00828765518963337,
|
| 807 |
+
"rewards/rejected": -0.16864125430583954,
|
| 808 |
+
"step": 480
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 0.8711111111111111,
|
| 812 |
+
"grad_norm": 0.32372602686992025,
|
| 813 |
+
"learning_rate": 2.466147269552893e-07,
|
| 814 |
+
"logits/chosen": 13.549275398254395,
|
| 815 |
+
"logits/rejected": 13.303945541381836,
|
| 816 |
+
"logps/chosen": -398.70245361328125,
|
| 817 |
+
"logps/rejected": -403.9294128417969,
|
| 818 |
+
"loss": 0.6932,
|
| 819 |
+
"rewards/accuracies": 0.59375,
|
| 820 |
+
"rewards/chosen": -0.1653539538383484,
|
| 821 |
+
"rewards/margins": 0.015345364809036255,
|
| 822 |
+
"rewards/rejected": -0.18069931864738464,
|
| 823 |
+
"step": 490
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 0.8888888888888888,
|
| 827 |
+
"grad_norm": 0.28909373758975987,
|
| 828 |
+
"learning_rate": 1.8366216981942632e-07,
|
| 829 |
+
"logits/chosen": 12.993377685546875,
|
| 830 |
+
"logits/rejected": 13.172945976257324,
|
| 831 |
+
"logps/chosen": -441.27288818359375,
|
| 832 |
+
"logps/rejected": -444.38983154296875,
|
| 833 |
+
"loss": 0.696,
|
| 834 |
+
"rewards/accuracies": 0.543749988079071,
|
| 835 |
+
"rewards/chosen": -0.17415288090705872,
|
| 836 |
+
"rewards/margins": -0.0011280607432126999,
|
| 837 |
+
"rewards/rejected": -0.17302480340003967,
|
| 838 |
+
"step": 500
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 0.8888888888888888,
|
| 842 |
+
"eval_logits/chosen": 13.899726867675781,
|
| 843 |
+
"eval_logits/rejected": 12.095231056213379,
|
| 844 |
+
"eval_logps/chosen": -421.1900939941406,
|
| 845 |
+
"eval_logps/rejected": -403.3279724121094,
|
| 846 |
+
"eval_loss": 0.6633419990539551,
|
| 847 |
+
"eval_rewards/accuracies": 0.7539682388305664,
|
| 848 |
+
"eval_rewards/chosen": -0.1262015700340271,
|
| 849 |
+
"eval_rewards/margins": 0.06973244994878769,
|
| 850 |
+
"eval_rewards/rejected": -0.19593402743339539,
|
| 851 |
+
"eval_runtime": 90.1435,
|
| 852 |
+
"eval_samples_per_second": 11.093,
|
| 853 |
+
"eval_steps_per_second": 0.699,
|
| 854 |
+
"step": 500
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 0.9066666666666666,
|
| 858 |
+
"grad_norm": 0.44905140288995754,
|
| 859 |
+
"learning_rate": 1.296710737600934e-07,
|
| 860 |
+
"logits/chosen": 12.890707015991211,
|
| 861 |
+
"logits/rejected": 12.480443000793457,
|
| 862 |
+
"logps/chosen": -397.1539611816406,
|
| 863 |
+
"logps/rejected": -404.51470947265625,
|
| 864 |
+
"loss": 0.6926,
|
| 865 |
+
"rewards/accuracies": 0.53125,
|
| 866 |
+
"rewards/chosen": -0.14948078989982605,
|
| 867 |
+
"rewards/margins": 0.0065328641794621944,
|
| 868 |
+
"rewards/rejected": -0.1560136377811432,
|
| 869 |
+
"step": 510
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 0.9244444444444444,
|
| 873 |
+
"grad_norm": 0.3012451953425647,
|
| 874 |
+
"learning_rate": 8.485031983924558e-08,
|
| 875 |
+
"logits/chosen": 13.72374153137207,
|
| 876 |
+
"logits/rejected": 14.290632247924805,
|
| 877 |
+
"logps/chosen": -409.4125061035156,
|
| 878 |
+
"logps/rejected": -427.1560974121094,
|
| 879 |
+
"loss": 0.6908,
|
| 880 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 881 |
+
"rewards/chosen": -0.15699708461761475,
|
| 882 |
+
"rewards/margins": -0.00222613662481308,
|
| 883 |
+
"rewards/rejected": -0.15477094054222107,
|
| 884 |
+
"step": 520
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 0.9422222222222222,
|
| 888 |
+
"grad_norm": 0.3269177206871498,
|
| 889 |
+
"learning_rate": 4.93733108466013e-08,
|
| 890 |
+
"logits/chosen": 12.78927993774414,
|
| 891 |
+
"logits/rejected": 13.388028144836426,
|
| 892 |
+
"logps/chosen": -428.2759704589844,
|
| 893 |
+
"logps/rejected": -456.5814514160156,
|
| 894 |
+
"loss": 0.6895,
|
| 895 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 896 |
+
"rewards/chosen": -0.15250369906425476,
|
| 897 |
+
"rewards/margins": 0.005445868708193302,
|
| 898 |
+
"rewards/rejected": -0.15794958174228668,
|
| 899 |
+
"step": 530
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.96,
|
| 903 |
+
"grad_norm": 0.33934624748600184,
|
| 904 |
+
"learning_rate": 2.3377300437934236e-08,
|
| 905 |
+
"logits/chosen": 13.640138626098633,
|
| 906 |
+
"logits/rejected": 13.247881889343262,
|
| 907 |
+
"logps/chosen": -384.22723388671875,
|
| 908 |
+
"logps/rejected": -377.16546630859375,
|
| 909 |
+
"loss": 0.6933,
|
| 910 |
+
"rewards/accuracies": 0.5625,
|
| 911 |
+
"rewards/chosen": -0.14129753410816193,
|
| 912 |
+
"rewards/margins": 0.019017567858099937,
|
| 913 |
+
"rewards/rejected": -0.16031508147716522,
|
| 914 |
+
"step": 540
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 0.9777777777777777,
|
| 918 |
+
"grad_norm": 0.31794002037631625,
|
| 919 |
+
"learning_rate": 6.962862127343206e-09,
|
| 920 |
+
"logits/chosen": 13.490945816040039,
|
| 921 |
+
"logits/rejected": 13.803362846374512,
|
| 922 |
+
"logps/chosen": -422.62939453125,
|
| 923 |
+
"logps/rejected": -427.26959228515625,
|
| 924 |
+
"loss": 0.6894,
|
| 925 |
+
"rewards/accuracies": 0.512499988079071,
|
| 926 |
+
"rewards/chosen": -0.167506605386734,
|
| 927 |
+
"rewards/margins": 0.010436911135911942,
|
| 928 |
+
"rewards/rejected": -0.17794351279735565,
|
| 929 |
+
"step": 550
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.9955555555555555,
|
| 933 |
+
"grad_norm": 0.29176173477304573,
|
| 934 |
+
"learning_rate": 1.9350018786556956e-10,
|
| 935 |
+
"logits/chosen": 13.073209762573242,
|
| 936 |
+
"logits/rejected": 13.310659408569336,
|
| 937 |
+
"logps/chosen": -431.8447265625,
|
| 938 |
+
"logps/rejected": -418.0799865722656,
|
| 939 |
+
"loss": 0.693,
|
| 940 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 941 |
+
"rewards/chosen": -0.1620911806821823,
|
| 942 |
+
"rewards/margins": -0.006803811527788639,
|
| 943 |
+
"rewards/rejected": -0.15528738498687744,
|
| 944 |
+
"step": 560
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 0.9991111111111111,
|
| 948 |
+
"step": 562,
|
| 949 |
+
"total_flos": 0.0,
|
| 950 |
+
"train_loss": 0.6916975294143703,
|
| 951 |
+
"train_runtime": 7520.1237,
|
| 952 |
+
"train_samples_per_second": 4.787,
|
| 953 |
+
"train_steps_per_second": 0.075
|
| 954 |
+
}
|
| 955 |
+
],
|
| 956 |
+
"logging_steps": 10,
|
| 957 |
+
"max_steps": 562,
|
| 958 |
+
"num_input_tokens_seen": 0,
|
| 959 |
+
"num_train_epochs": 1,
|
| 960 |
+
"save_steps": 100,
|
| 961 |
+
"stateful_callbacks": {
|
| 962 |
+
"TrainerControl": {
|
| 963 |
+
"args": {
|
| 964 |
+
"should_epoch_stop": false,
|
| 965 |
+
"should_evaluate": false,
|
| 966 |
+
"should_log": false,
|
| 967 |
+
"should_save": true,
|
| 968 |
+
"should_training_stop": true
|
| 969 |
+
},
|
| 970 |
+
"attributes": {}
|
| 971 |
+
}
|
| 972 |
+
},
|
| 973 |
+
"total_flos": 0.0,
|
| 974 |
+
"train_batch_size": 4,
|
| 975 |
+
"trial_name": null,
|
| 976 |
+
"trial_params": null
|
| 977 |
+
}
|