Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +815 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full-prometheus_consistent-reward-scale-1-rpo
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full-prometheus_consistent-reward-scale-1-rpo
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.0340
|
| 21 |
+
- Rewards/chosen: -0.0965
|
| 22 |
+
- Rewards/rejected: -0.3938
|
| 23 |
+
- Rewards/accuracies: 0.7414
|
| 24 |
+
- Rewards/margins: 0.2973
|
| 25 |
+
- Logps/rejected: -258.4546
|
| 26 |
+
- Logps/chosen: -285.2520
|
| 27 |
+
- Logits/rejected: -2.1563
|
| 28 |
+
- Logits/chosen: -2.3140
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 8
|
| 50 |
+
- seed: 55
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
+
- total_eval_batch_size: 64
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.048 | 0.1143 | 50 | 0.0428 | 0.0615 | -0.0563 | 0.7026 | 0.1178 | -224.7078 | -269.4535 | -2.4545 | -2.5520 |
|
| 66 |
+
| 0.0399 | 0.2286 | 100 | 0.0385 | -0.0747 | -0.3118 | 0.75 | 0.2371 | -250.2514 | -283.0706 | -1.9893 | -2.1601 |
|
| 67 |
+
| 0.0367 | 0.3429 | 150 | 0.0371 | -0.1934 | -0.4431 | 0.7672 | 0.2497 | -263.3893 | -294.9446 | -2.3057 | -2.4218 |
|
| 68 |
+
| 0.0375 | 0.4571 | 200 | 0.0353 | -0.0541 | -0.3320 | 0.7672 | 0.2779 | -252.2786 | -281.0130 | -2.1436 | -2.2907 |
|
| 69 |
+
| 0.0371 | 0.5714 | 250 | 0.0344 | -0.0812 | -0.3496 | 0.7629 | 0.2684 | -254.0325 | -283.7219 | -2.2615 | -2.3785 |
|
| 70 |
+
| 0.0345 | 0.6857 | 300 | 0.0341 | -0.0682 | -0.3495 | 0.7457 | 0.2813 | -254.0265 | -282.4234 | -2.2130 | -2.3475 |
|
| 71 |
+
| 0.0373 | 0.8 | 350 | 0.0341 | -0.0908 | -0.3849 | 0.7414 | 0.2941 | -257.5619 | -284.6819 | -2.1788 | -2.3321 |
|
| 72 |
+
| 0.0367 | 0.9143 | 400 | 0.0340 | -0.0965 | -0.3938 | 0.7414 | 0.2973 | -258.4546 | -285.2520 | -2.1563 | -2.3140 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- Transformers 4.44.0.dev0
|
| 78 |
+
- Pytorch 2.1.2
|
| 79 |
+
- Datasets 2.20.0
|
| 80 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.038760803656938146,
|
| 5 |
+
"train_runtime": 11420.4923,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.903,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.0.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.038760803656938146,
|
| 5 |
+
"train_runtime": 11420.4923,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.903,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9988571428571429,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 437,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022857142857142857,
|
| 13 |
+
"grad_norm": 3.2870738344316774,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.700951099395752,
|
| 16 |
+
"logits/rejected": -2.6252384185791016,
|
| 17 |
+
"logps/chosen": -301.28485107421875,
|
| 18 |
+
"logps/rejected": -281.7684020996094,
|
| 19 |
+
"loss": 0.0556,
|
| 20 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 21 |
+
"rewards/chosen": 1.3337824384507257e-05,
|
| 22 |
+
"rewards/margins": 0.00020079524256289005,
|
| 23 |
+
"rewards/rejected": -0.0001874573645181954,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 2.5214503350239097,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.641275405883789,
|
| 31 |
+
"logits/rejected": -2.606044292449951,
|
| 32 |
+
"logps/chosen": -278.9186706542969,
|
| 33 |
+
"logps/rejected": -254.6912384033203,
|
| 34 |
+
"loss": 0.0552,
|
| 35 |
+
"rewards/accuracies": 0.550000011920929,
|
| 36 |
+
"rewards/chosen": 0.002525148680433631,
|
| 37 |
+
"rewards/margins": 0.0017468780279159546,
|
| 38 |
+
"rewards/rejected": 0.0007782711763866246,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 2.7331110185448306,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6379804611206055,
|
| 46 |
+
"logits/rejected": -2.617119550704956,
|
| 47 |
+
"logps/chosen": -263.29449462890625,
|
| 48 |
+
"logps/rejected": -263.40509033203125,
|
| 49 |
+
"loss": 0.0535,
|
| 50 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 51 |
+
"rewards/chosen": 0.01388559676706791,
|
| 52 |
+
"rewards/margins": 0.008006090298295021,
|
| 53 |
+
"rewards/rejected": 0.00587950786575675,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 2.830903095609413,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6505517959594727,
|
| 61 |
+
"logits/rejected": -2.5873801708221436,
|
| 62 |
+
"logps/chosen": -290.2440185546875,
|
| 63 |
+
"logps/rejected": -268.03863525390625,
|
| 64 |
+
"loss": 0.0502,
|
| 65 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 66 |
+
"rewards/chosen": 0.037532415241003036,
|
| 67 |
+
"rewards/margins": 0.04160536080598831,
|
| 68 |
+
"rewards/rejected": -0.004072942305356264,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 6.7770519006865815,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.6873106956481934,
|
| 76 |
+
"logits/rejected": -2.6081948280334473,
|
| 77 |
+
"logps/chosen": -290.84735107421875,
|
| 78 |
+
"logps/rejected": -249.636474609375,
|
| 79 |
+
"loss": 0.048,
|
| 80 |
+
"rewards/accuracies": 0.675000011920929,
|
| 81 |
+
"rewards/chosen": 0.05462931841611862,
|
| 82 |
+
"rewards/margins": 0.08235044032335281,
|
| 83 |
+
"rewards/rejected": -0.027721121907234192,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.552030086517334,
|
| 89 |
+
"eval_logits/rejected": -2.4545059204101562,
|
| 90 |
+
"eval_logps/chosen": -269.4535217285156,
|
| 91 |
+
"eval_logps/rejected": -224.7078094482422,
|
| 92 |
+
"eval_loss": 0.04283829778432846,
|
| 93 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 94 |
+
"eval_rewards/chosen": 0.06150422990322113,
|
| 95 |
+
"eval_rewards/margins": 0.11782074719667435,
|
| 96 |
+
"eval_rewards/rejected": -0.05631651729345322,
|
| 97 |
+
"eval_runtime": 91.0985,
|
| 98 |
+
"eval_samples_per_second": 20.099,
|
| 99 |
+
"eval_steps_per_second": 0.318,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 3.063883877478527,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.5849452018737793,
|
| 107 |
+
"logits/rejected": -2.476766347885132,
|
| 108 |
+
"logps/chosen": -280.53155517578125,
|
| 109 |
+
"logps/rejected": -259.91339111328125,
|
| 110 |
+
"loss": 0.0438,
|
| 111 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 112 |
+
"rewards/chosen": -0.00025545060634613037,
|
| 113 |
+
"rewards/margins": 0.11335714906454086,
|
| 114 |
+
"rewards/rejected": -0.1136125922203064,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 3.212195825091894,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.527914524078369,
|
| 122 |
+
"logits/rejected": -2.468449354171753,
|
| 123 |
+
"logps/chosen": -273.3289489746094,
|
| 124 |
+
"logps/rejected": -270.74847412109375,
|
| 125 |
+
"loss": 0.0414,
|
| 126 |
+
"rewards/accuracies": 0.668749988079071,
|
| 127 |
+
"rewards/chosen": 0.024492865428328514,
|
| 128 |
+
"rewards/margins": 0.1357099711894989,
|
| 129 |
+
"rewards/rejected": -0.11121710389852524,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 2.9871578323865613,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.435612678527832,
|
| 137 |
+
"logits/rejected": -2.3714098930358887,
|
| 138 |
+
"logps/chosen": -269.2018127441406,
|
| 139 |
+
"logps/rejected": -276.71612548828125,
|
| 140 |
+
"loss": 0.0424,
|
| 141 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 142 |
+
"rewards/chosen": -0.02172762155532837,
|
| 143 |
+
"rewards/margins": 0.19234482944011688,
|
| 144 |
+
"rewards/rejected": -0.21407246589660645,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 7.2824339333866,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -2.3872673511505127,
|
| 152 |
+
"logits/rejected": -2.300445318222046,
|
| 153 |
+
"logps/chosen": -275.88641357421875,
|
| 154 |
+
"logps/rejected": -275.3648681640625,
|
| 155 |
+
"loss": 0.0423,
|
| 156 |
+
"rewards/accuracies": 0.768750011920929,
|
| 157 |
+
"rewards/chosen": 0.003988614305853844,
|
| 158 |
+
"rewards/margins": 0.20215804874897003,
|
| 159 |
+
"rewards/rejected": -0.19816945493221283,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 3.336291542294826,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -2.3031492233276367,
|
| 167 |
+
"logits/rejected": -2.170569658279419,
|
| 168 |
+
"logps/chosen": -289.19390869140625,
|
| 169 |
+
"logps/rejected": -258.89019775390625,
|
| 170 |
+
"loss": 0.0399,
|
| 171 |
+
"rewards/accuracies": 0.800000011920929,
|
| 172 |
+
"rewards/chosen": -0.028075385838747025,
|
| 173 |
+
"rewards/margins": 0.2263123095035553,
|
| 174 |
+
"rewards/rejected": -0.25438767671585083,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -2.1601011753082275,
|
| 180 |
+
"eval_logits/rejected": -1.989274263381958,
|
| 181 |
+
"eval_logps/chosen": -283.0706481933594,
|
| 182 |
+
"eval_logps/rejected": -250.25137329101562,
|
| 183 |
+
"eval_loss": 0.03850938379764557,
|
| 184 |
+
"eval_rewards/accuracies": 0.75,
|
| 185 |
+
"eval_rewards/chosen": -0.07466702908277512,
|
| 186 |
+
"eval_rewards/margins": 0.23708537220954895,
|
| 187 |
+
"eval_rewards/rejected": -0.3117523789405823,
|
| 188 |
+
"eval_runtime": 91.3269,
|
| 189 |
+
"eval_samples_per_second": 20.049,
|
| 190 |
+
"eval_steps_per_second": 0.318,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 3.177801599242846,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": -2.190119504928589,
|
| 198 |
+
"logits/rejected": -2.104891061782837,
|
| 199 |
+
"logps/chosen": -322.0883483886719,
|
| 200 |
+
"logps/rejected": -316.6219482421875,
|
| 201 |
+
"loss": 0.0407,
|
| 202 |
+
"rewards/accuracies": 0.65625,
|
| 203 |
+
"rewards/chosen": -0.05784715339541435,
|
| 204 |
+
"rewards/margins": 0.12927614152431488,
|
| 205 |
+
"rewards/rejected": -0.18712328374385834,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 3.162850908555599,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": -2.1028637886047363,
|
| 213 |
+
"logits/rejected": -1.9295556545257568,
|
| 214 |
+
"logps/chosen": -316.31634521484375,
|
| 215 |
+
"logps/rejected": -253.72195434570312,
|
| 216 |
+
"loss": 0.0376,
|
| 217 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 218 |
+
"rewards/chosen": -0.05411834642291069,
|
| 219 |
+
"rewards/margins": 0.16595926880836487,
|
| 220 |
+
"rewards/rejected": -0.22007760405540466,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 3.4487456432507195,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": -2.1982297897338867,
|
| 228 |
+
"logits/rejected": -2.1048200130462646,
|
| 229 |
+
"logps/chosen": -314.8988037109375,
|
| 230 |
+
"logps/rejected": -276.97674560546875,
|
| 231 |
+
"loss": 0.0373,
|
| 232 |
+
"rewards/accuracies": 0.643750011920929,
|
| 233 |
+
"rewards/chosen": -0.09742867946624756,
|
| 234 |
+
"rewards/margins": 0.16763603687286377,
|
| 235 |
+
"rewards/rejected": -0.26506468653678894,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 3.9041569099317925,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": -2.3467020988464355,
|
| 243 |
+
"logits/rejected": -2.3200185298919678,
|
| 244 |
+
"logps/chosen": -315.0827941894531,
|
| 245 |
+
"logps/rejected": -290.0022277832031,
|
| 246 |
+
"loss": 0.0375,
|
| 247 |
+
"rewards/accuracies": 0.706250011920929,
|
| 248 |
+
"rewards/chosen": 0.0036486200988292694,
|
| 249 |
+
"rewards/margins": 0.18455657362937927,
|
| 250 |
+
"rewards/rejected": -0.1809079647064209,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 3.1721506985515817,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": -2.452073097229004,
|
| 258 |
+
"logits/rejected": -2.371002674102783,
|
| 259 |
+
"logps/chosen": -296.94329833984375,
|
| 260 |
+
"logps/rejected": -307.01837158203125,
|
| 261 |
+
"loss": 0.0367,
|
| 262 |
+
"rewards/accuracies": 0.71875,
|
| 263 |
+
"rewards/chosen": -0.09957350790500641,
|
| 264 |
+
"rewards/margins": 0.2076205015182495,
|
| 265 |
+
"rewards/rejected": -0.3071940243244171,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": -2.4217660427093506,
|
| 271 |
+
"eval_logits/rejected": -2.3057165145874023,
|
| 272 |
+
"eval_logps/chosen": -294.94464111328125,
|
| 273 |
+
"eval_logps/rejected": -263.3892822265625,
|
| 274 |
+
"eval_loss": 0.037056975066661835,
|
| 275 |
+
"eval_rewards/accuracies": 0.767241358757019,
|
| 276 |
+
"eval_rewards/chosen": -0.19340714812278748,
|
| 277 |
+
"eval_rewards/margins": 0.24972420930862427,
|
| 278 |
+
"eval_rewards/rejected": -0.44313138723373413,
|
| 279 |
+
"eval_runtime": 91.4385,
|
| 280 |
+
"eval_samples_per_second": 20.024,
|
| 281 |
+
"eval_steps_per_second": 0.317,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 3.2106332008237772,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": -2.493586540222168,
|
| 289 |
+
"logits/rejected": -2.4255142211914062,
|
| 290 |
+
"logps/chosen": -316.421630859375,
|
| 291 |
+
"logps/rejected": -294.204345703125,
|
| 292 |
+
"loss": 0.0372,
|
| 293 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 294 |
+
"rewards/chosen": -0.17081338167190552,
|
| 295 |
+
"rewards/margins": 0.21480269730091095,
|
| 296 |
+
"rewards/rejected": -0.38561612367630005,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 3.305163626494215,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": -2.473874568939209,
|
| 304 |
+
"logits/rejected": -2.4223687648773193,
|
| 305 |
+
"logps/chosen": -309.713623046875,
|
| 306 |
+
"logps/rejected": -291.6881103515625,
|
| 307 |
+
"loss": 0.0373,
|
| 308 |
+
"rewards/accuracies": 0.737500011920929,
|
| 309 |
+
"rewards/chosen": -0.056086935102939606,
|
| 310 |
+
"rewards/margins": 0.2034837305545807,
|
| 311 |
+
"rewards/rejected": -0.2595706582069397,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 3.211306956770945,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": -2.486910104751587,
|
| 319 |
+
"logits/rejected": -2.3967292308807373,
|
| 320 |
+
"logps/chosen": -334.092041015625,
|
| 321 |
+
"logps/rejected": -292.355712890625,
|
| 322 |
+
"loss": 0.0373,
|
| 323 |
+
"rewards/accuracies": 0.793749988079071,
|
| 324 |
+
"rewards/chosen": -0.03912293165922165,
|
| 325 |
+
"rewards/margins": 0.2849285304546356,
|
| 326 |
+
"rewards/rejected": -0.3240514397621155,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 3.4199998150652564,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": -2.507291078567505,
|
| 334 |
+
"logits/rejected": -2.4220941066741943,
|
| 335 |
+
"logps/chosen": -285.0407409667969,
|
| 336 |
+
"logps/rejected": -261.77392578125,
|
| 337 |
+
"loss": 0.0365,
|
| 338 |
+
"rewards/accuracies": 0.737500011920929,
|
| 339 |
+
"rewards/chosen": -0.08783163130283356,
|
| 340 |
+
"rewards/margins": 0.1849258691072464,
|
| 341 |
+
"rewards/rejected": -0.27275753021240234,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 2.966792213983748,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": -2.508540630340576,
|
| 349 |
+
"logits/rejected": -2.376711130142212,
|
| 350 |
+
"logps/chosen": -271.0993347167969,
|
| 351 |
+
"logps/rejected": -254.6019744873047,
|
| 352 |
+
"loss": 0.0375,
|
| 353 |
+
"rewards/accuracies": 0.71875,
|
| 354 |
+
"rewards/chosen": -0.02756611444056034,
|
| 355 |
+
"rewards/margins": 0.2296477109193802,
|
| 356 |
+
"rewards/rejected": -0.257213830947876,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": -2.290728807449341,
|
| 362 |
+
"eval_logits/rejected": -2.1435999870300293,
|
| 363 |
+
"eval_logps/chosen": -281.01300048828125,
|
| 364 |
+
"eval_logps/rejected": -252.2786407470703,
|
| 365 |
+
"eval_loss": 0.03527820482850075,
|
| 366 |
+
"eval_rewards/accuracies": 0.767241358757019,
|
| 367 |
+
"eval_rewards/chosen": -0.054090581834316254,
|
| 368 |
+
"eval_rewards/margins": 0.2779344916343689,
|
| 369 |
+
"eval_rewards/rejected": -0.33202511072158813,
|
| 370 |
+
"eval_runtime": 91.8763,
|
| 371 |
+
"eval_samples_per_second": 19.929,
|
| 372 |
+
"eval_steps_per_second": 0.316,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 3.030911456207702,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": -2.2013955116271973,
|
| 380 |
+
"logits/rejected": -2.0953996181488037,
|
| 381 |
+
"logps/chosen": -276.37432861328125,
|
| 382 |
+
"logps/rejected": -280.0777893066406,
|
| 383 |
+
"loss": 0.0362,
|
| 384 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 385 |
+
"rewards/chosen": -0.12920817732810974,
|
| 386 |
+
"rewards/margins": 0.2187701165676117,
|
| 387 |
+
"rewards/rejected": -0.3479783535003662,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 3.521691949347102,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": -2.1590688228607178,
|
| 395 |
+
"logits/rejected": -2.0675222873687744,
|
| 396 |
+
"logps/chosen": -301.6694030761719,
|
| 397 |
+
"logps/rejected": -281.8822326660156,
|
| 398 |
+
"loss": 0.0387,
|
| 399 |
+
"rewards/accuracies": 0.737500011920929,
|
| 400 |
+
"rewards/chosen": -0.16603410243988037,
|
| 401 |
+
"rewards/margins": 0.1909538209438324,
|
| 402 |
+
"rewards/rejected": -0.35698795318603516,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 3.271109178578225,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": -2.195312023162842,
|
| 410 |
+
"logits/rejected": -2.095759391784668,
|
| 411 |
+
"logps/chosen": -293.96038818359375,
|
| 412 |
+
"logps/rejected": -271.3498229980469,
|
| 413 |
+
"loss": 0.0366,
|
| 414 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 415 |
+
"rewards/chosen": -0.1290125697851181,
|
| 416 |
+
"rewards/margins": 0.19161902368068695,
|
| 417 |
+
"rewards/rejected": -0.32063156366348267,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 3.788405528579881,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": -2.3450160026550293,
|
| 425 |
+
"logits/rejected": -2.233682870864868,
|
| 426 |
+
"logps/chosen": -311.3532409667969,
|
| 427 |
+
"logps/rejected": -312.9554138183594,
|
| 428 |
+
"loss": 0.0359,
|
| 429 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 430 |
+
"rewards/chosen": -0.12643404304981232,
|
| 431 |
+
"rewards/margins": 0.2889019250869751,
|
| 432 |
+
"rewards/rejected": -0.4153359532356262,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 2.975794895949038,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": -2.361924648284912,
|
| 440 |
+
"logits/rejected": -2.305227041244507,
|
| 441 |
+
"logps/chosen": -292.22430419921875,
|
| 442 |
+
"logps/rejected": -304.44610595703125,
|
| 443 |
+
"loss": 0.0371,
|
| 444 |
+
"rewards/accuracies": 0.737500011920929,
|
| 445 |
+
"rewards/chosen": -0.11598072946071625,
|
| 446 |
+
"rewards/margins": 0.22647562623023987,
|
| 447 |
+
"rewards/rejected": -0.3424563705921173,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": -2.378539800643921,
|
| 453 |
+
"eval_logits/rejected": -2.2614917755126953,
|
| 454 |
+
"eval_logps/chosen": -283.72186279296875,
|
| 455 |
+
"eval_logps/rejected": -254.03253173828125,
|
| 456 |
+
"eval_loss": 0.0344076044857502,
|
| 457 |
+
"eval_rewards/accuracies": 0.7629310488700867,
|
| 458 |
+
"eval_rewards/chosen": -0.08117903023958206,
|
| 459 |
+
"eval_rewards/margins": 0.2683848738670349,
|
| 460 |
+
"eval_rewards/rejected": -0.3495638966560364,
|
| 461 |
+
"eval_runtime": 90.3763,
|
| 462 |
+
"eval_samples_per_second": 20.26,
|
| 463 |
+
"eval_steps_per_second": 0.321,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 3.4496793515012127,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": -2.378929615020752,
|
| 471 |
+
"logits/rejected": -2.367784023284912,
|
| 472 |
+
"logps/chosen": -292.9862365722656,
|
| 473 |
+
"logps/rejected": -303.8517150878906,
|
| 474 |
+
"loss": 0.0378,
|
| 475 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 476 |
+
"rewards/chosen": -0.1319950819015503,
|
| 477 |
+
"rewards/margins": 0.17309217154979706,
|
| 478 |
+
"rewards/rejected": -0.30508726835250854,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 3.309695226000979,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": -2.4140243530273438,
|
| 486 |
+
"logits/rejected": -2.2664575576782227,
|
| 487 |
+
"logps/chosen": -284.2572326660156,
|
| 488 |
+
"logps/rejected": -267.2565612792969,
|
| 489 |
+
"loss": 0.0344,
|
| 490 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 491 |
+
"rewards/chosen": -0.11731357872486115,
|
| 492 |
+
"rewards/margins": 0.2501775622367859,
|
| 493 |
+
"rewards/rejected": -0.3674911856651306,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 3.6035910497695895,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": -2.3951687812805176,
|
| 501 |
+
"logits/rejected": -2.2973971366882324,
|
| 502 |
+
"logps/chosen": -286.1737365722656,
|
| 503 |
+
"logps/rejected": -301.73150634765625,
|
| 504 |
+
"loss": 0.0359,
|
| 505 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 506 |
+
"rewards/chosen": -0.13135617971420288,
|
| 507 |
+
"rewards/margins": 0.25489118695259094,
|
| 508 |
+
"rewards/rejected": -0.38624733686447144,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 4.1650025414210665,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": -2.3873696327209473,
|
| 516 |
+
"logits/rejected": -2.2862212657928467,
|
| 517 |
+
"logps/chosen": -316.511962890625,
|
| 518 |
+
"logps/rejected": -304.41741943359375,
|
| 519 |
+
"loss": 0.0349,
|
| 520 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 521 |
+
"rewards/chosen": -0.15606102347373962,
|
| 522 |
+
"rewards/margins": 0.2326282560825348,
|
| 523 |
+
"rewards/rejected": -0.3886892795562744,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 2.987221364071687,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": -2.387683391571045,
|
| 531 |
+
"logits/rejected": -2.2113590240478516,
|
| 532 |
+
"logps/chosen": -317.22491455078125,
|
| 533 |
+
"logps/rejected": -293.2046813964844,
|
| 534 |
+
"loss": 0.0345,
|
| 535 |
+
"rewards/accuracies": 0.762499988079071,
|
| 536 |
+
"rewards/chosen": -0.1531573385000229,
|
| 537 |
+
"rewards/margins": 0.25246819853782654,
|
| 538 |
+
"rewards/rejected": -0.40562552213668823,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": -2.3475093841552734,
|
| 544 |
+
"eval_logits/rejected": -2.2130320072174072,
|
| 545 |
+
"eval_logps/chosen": -282.4234313964844,
|
| 546 |
+
"eval_logps/rejected": -254.02647399902344,
|
| 547 |
+
"eval_loss": 0.03413059934973717,
|
| 548 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 549 |
+
"eval_rewards/chosen": -0.06819500029087067,
|
| 550 |
+
"eval_rewards/margins": 0.28130805492401123,
|
| 551 |
+
"eval_rewards/rejected": -0.3495030105113983,
|
| 552 |
+
"eval_runtime": 90.5949,
|
| 553 |
+
"eval_samples_per_second": 20.211,
|
| 554 |
+
"eval_steps_per_second": 0.32,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 2.83762479667315,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": -2.3416621685028076,
|
| 562 |
+
"logits/rejected": -2.2815122604370117,
|
| 563 |
+
"logps/chosen": -267.0755310058594,
|
| 564 |
+
"logps/rejected": -267.9870300292969,
|
| 565 |
+
"loss": 0.0367,
|
| 566 |
+
"rewards/accuracies": 0.706250011920929,
|
| 567 |
+
"rewards/chosen": -0.10771362483501434,
|
| 568 |
+
"rewards/margins": 0.20020349323749542,
|
| 569 |
+
"rewards/rejected": -0.30791711807250977,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 3.3425045367761492,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": -2.3024659156799316,
|
| 577 |
+
"logits/rejected": -2.2111096382141113,
|
| 578 |
+
"logps/chosen": -257.63531494140625,
|
| 579 |
+
"logps/rejected": -272.8826904296875,
|
| 580 |
+
"loss": 0.0364,
|
| 581 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 582 |
+
"rewards/chosen": -0.13352924585342407,
|
| 583 |
+
"rewards/margins": 0.18541577458381653,
|
| 584 |
+
"rewards/rejected": -0.3189450204372406,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 2.769882287509293,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": -2.4251229763031006,
|
| 592 |
+
"logits/rejected": -2.29219388961792,
|
| 593 |
+
"logps/chosen": -307.23333740234375,
|
| 594 |
+
"logps/rejected": -280.0276794433594,
|
| 595 |
+
"loss": 0.0348,
|
| 596 |
+
"rewards/accuracies": 0.78125,
|
| 597 |
+
"rewards/chosen": -0.08015141636133194,
|
| 598 |
+
"rewards/margins": 0.24224629998207092,
|
| 599 |
+
"rewards/rejected": -0.3223976790904999,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 3.0959592695193763,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": -2.3518755435943604,
|
| 607 |
+
"logits/rejected": -2.3008193969726562,
|
| 608 |
+
"logps/chosen": -277.1438293457031,
|
| 609 |
+
"logps/rejected": -279.57427978515625,
|
| 610 |
+
"loss": 0.0353,
|
| 611 |
+
"rewards/accuracies": 0.768750011920929,
|
| 612 |
+
"rewards/chosen": -0.11192858219146729,
|
| 613 |
+
"rewards/margins": 0.2635765075683594,
|
| 614 |
+
"rewards/rejected": -0.3755050301551819,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 3.411772564497867,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": -2.4039082527160645,
|
| 622 |
+
"logits/rejected": -2.2675061225891113,
|
| 623 |
+
"logps/chosen": -300.5389099121094,
|
| 624 |
+
"logps/rejected": -285.20794677734375,
|
| 625 |
+
"loss": 0.0373,
|
| 626 |
+
"rewards/accuracies": 0.6875,
|
| 627 |
+
"rewards/chosen": -0.10831709951162338,
|
| 628 |
+
"rewards/margins": 0.20078572630882263,
|
| 629 |
+
"rewards/rejected": -0.3091028332710266,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": -2.3320775032043457,
|
| 635 |
+
"eval_logits/rejected": -2.178847312927246,
|
| 636 |
+
"eval_logps/chosen": -284.68194580078125,
|
| 637 |
+
"eval_logps/rejected": -257.5619201660156,
|
| 638 |
+
"eval_loss": 0.03409451246261597,
|
| 639 |
+
"eval_rewards/accuracies": 0.7413793206214905,
|
| 640 |
+
"eval_rewards/chosen": -0.09078007191419601,
|
| 641 |
+
"eval_rewards/margins": 0.2940778136253357,
|
| 642 |
+
"eval_rewards/rejected": -0.38485783338546753,
|
| 643 |
+
"eval_runtime": 90.2939,
|
| 644 |
+
"eval_samples_per_second": 20.278,
|
| 645 |
+
"eval_steps_per_second": 0.321,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 2.905033636374422,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": -2.4123928546905518,
|
| 653 |
+
"logits/rejected": -2.2830798625946045,
|
| 654 |
+
"logps/chosen": -295.31842041015625,
|
| 655 |
+
"logps/rejected": -286.4782409667969,
|
| 656 |
+
"loss": 0.0342,
|
| 657 |
+
"rewards/accuracies": 0.8125,
|
| 658 |
+
"rewards/chosen": -0.09913380444049835,
|
| 659 |
+
"rewards/margins": 0.25478315353393555,
|
| 660 |
+
"rewards/rejected": -0.3539169430732727,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 3.0252642177231617,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": -2.3916738033294678,
|
| 668 |
+
"logits/rejected": -2.244600296020508,
|
| 669 |
+
"logps/chosen": -278.41217041015625,
|
| 670 |
+
"logps/rejected": -273.110595703125,
|
| 671 |
+
"loss": 0.034,
|
| 672 |
+
"rewards/accuracies": 0.78125,
|
| 673 |
+
"rewards/chosen": -0.07350718975067139,
|
| 674 |
+
"rewards/margins": 0.273443341255188,
|
| 675 |
+
"rewards/rejected": -0.346950501203537,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 3.1477774029837864,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": -2.4462573528289795,
|
| 683 |
+
"logits/rejected": -2.309199094772339,
|
| 684 |
+
"logps/chosen": -329.63348388671875,
|
| 685 |
+
"logps/rejected": -289.4853515625,
|
| 686 |
+
"loss": 0.0352,
|
| 687 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 688 |
+
"rewards/chosen": -0.11577162891626358,
|
| 689 |
+
"rewards/margins": 0.20508745312690735,
|
| 690 |
+
"rewards/rejected": -0.32085904479026794,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 3.5873354511870312,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": -2.403637170791626,
|
| 698 |
+
"logits/rejected": -2.321964740753174,
|
| 699 |
+
"logps/chosen": -318.7025146484375,
|
| 700 |
+
"logps/rejected": -324.0418395996094,
|
| 701 |
+
"loss": 0.0358,
|
| 702 |
+
"rewards/accuracies": 0.706250011920929,
|
| 703 |
+
"rewards/chosen": -0.07329531759023666,
|
| 704 |
+
"rewards/margins": 0.21939659118652344,
|
| 705 |
+
"rewards/rejected": -0.2926918864250183,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 5.030788992854401,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": -2.3374335765838623,
|
| 713 |
+
"logits/rejected": -2.3087658882141113,
|
| 714 |
+
"logps/chosen": -292.0116271972656,
|
| 715 |
+
"logps/rejected": -298.6130065917969,
|
| 716 |
+
"loss": 0.0367,
|
| 717 |
+
"rewards/accuracies": 0.762499988079071,
|
| 718 |
+
"rewards/chosen": -0.13278374075889587,
|
| 719 |
+
"rewards/margins": 0.18760989606380463,
|
| 720 |
+
"rewards/rejected": -0.3203936815261841,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": -2.3139522075653076,
|
| 726 |
+
"eval_logits/rejected": -2.1562979221343994,
|
| 727 |
+
"eval_logps/chosen": -285.25201416015625,
|
| 728 |
+
"eval_logps/rejected": -258.45458984375,
|
| 729 |
+
"eval_loss": 0.034024760127067566,
|
| 730 |
+
"eval_rewards/accuracies": 0.7413793206214905,
|
| 731 |
+
"eval_rewards/chosen": -0.09648066759109497,
|
| 732 |
+
"eval_rewards/margins": 0.2973036468029022,
|
| 733 |
+
"eval_rewards/rejected": -0.3937843143939972,
|
| 734 |
+
"eval_runtime": 90.5725,
|
| 735 |
+
"eval_samples_per_second": 20.216,
|
| 736 |
+
"eval_steps_per_second": 0.32,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 3.2590392659418335,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": -2.3900809288024902,
|
| 744 |
+
"logits/rejected": -2.28753924369812,
|
| 745 |
+
"logps/chosen": -301.38275146484375,
|
| 746 |
+
"logps/rejected": -294.66497802734375,
|
| 747 |
+
"loss": 0.0352,
|
| 748 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 749 |
+
"rewards/chosen": -0.13202829658985138,
|
| 750 |
+
"rewards/margins": 0.21732142567634583,
|
| 751 |
+
"rewards/rejected": -0.349349707365036,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 3.0742600085581127,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": -2.326852798461914,
|
| 759 |
+
"logits/rejected": -2.221895694732666,
|
| 760 |
+
"logps/chosen": -288.41644287109375,
|
| 761 |
+
"logps/rejected": -271.6980895996094,
|
| 762 |
+
"loss": 0.0345,
|
| 763 |
+
"rewards/accuracies": 0.800000011920929,
|
| 764 |
+
"rewards/chosen": -0.056683529168367386,
|
| 765 |
+
"rewards/margins": 0.32042860984802246,
|
| 766 |
+
"rewards/rejected": -0.37711212038993835,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 3.069144533981935,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": -2.312039613723755,
|
| 774 |
+
"logits/rejected": -2.213510751724243,
|
| 775 |
+
"logps/chosen": -293.47271728515625,
|
| 776 |
+
"logps/rejected": -299.9048156738281,
|
| 777 |
+
"loss": 0.0344,
|
| 778 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 779 |
+
"rewards/chosen": -0.11240877956151962,
|
| 780 |
+
"rewards/margins": 0.24994345009326935,
|
| 781 |
+
"rewards/rejected": -0.3623522222042084,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.038760803656938146,
|
| 789 |
+
"train_runtime": 11420.4923,
|
| 790 |
+
"train_samples_per_second": 4.903,
|
| 791 |
+
"train_steps_per_second": 0.038
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 437,
|
| 796 |
+
"num_input_tokens_seen": 0,
|
| 797 |
+
"num_train_epochs": 1,
|
| 798 |
+
"save_steps": 100,
|
| 799 |
+
"stateful_callbacks": {
|
| 800 |
+
"TrainerControl": {
|
| 801 |
+
"args": {
|
| 802 |
+
"should_epoch_stop": false,
|
| 803 |
+
"should_evaluate": false,
|
| 804 |
+
"should_log": false,
|
| 805 |
+
"should_save": true,
|
| 806 |
+
"should_training_stop": true
|
| 807 |
+
},
|
| 808 |
+
"attributes": {}
|
| 809 |
+
}
|
| 810 |
+
},
|
| 811 |
+
"total_flos": 0.0,
|
| 812 |
+
"train_batch_size": 8,
|
| 813 |
+
"trial_name": null,
|
| 814 |
+
"trial_params": null
|
| 815 |
+
}
|