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-reward-scale-05
|
| 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-reward-scale-05
|
| 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: 0.5142
|
| 21 |
+
- Rewards/chosen: -1.9575
|
| 22 |
+
- Rewards/rejected: -3.2488
|
| 23 |
+
- Rewards/accuracies: 0.7198
|
| 24 |
+
- Rewards/margins: 1.2913
|
| 25 |
+
- Logps/rejected: -573.1553
|
| 26 |
+
- Logps/chosen: -455.7142
|
| 27 |
+
- Logits/rejected: 4.0394
|
| 28 |
+
- Logits/chosen: 3.0591
|
| 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 |
+
| 0.6718 | 0.1143 | 50 | 0.6623 | -0.0203 | -0.1341 | 0.6595 | 0.1139 | -261.6897 | -261.9885 | -2.5701 | -2.6137 |
|
| 66 |
+
| 0.5837 | 0.2286 | 100 | 0.5855 | -0.7924 | -1.5143 | 0.6595 | 0.7219 | -399.7070 | -339.2018 | -0.2076 | -0.5772 |
|
| 67 |
+
| 0.5481 | 0.3429 | 150 | 0.5498 | -1.3440 | -2.2915 | 0.6940 | 0.9475 | -477.4253 | -394.3634 | 2.4336 | 1.6485 |
|
| 68 |
+
| 0.5289 | 0.4571 | 200 | 0.5409 | -1.6426 | -2.7545 | 0.6983 | 1.1119 | -523.7253 | -424.2230 | 3.5667 | 2.6372 |
|
| 69 |
+
| 0.54 | 0.5714 | 250 | 0.5280 | -1.5391 | -2.7058 | 0.7026 | 1.1667 | -518.8563 | -413.8667 | 2.6437 | 1.4166 |
|
| 70 |
+
| 0.5147 | 0.6857 | 300 | 0.5204 | -1.8487 | -3.0990 | 0.7112 | 1.2504 | -558.1808 | -444.8300 | 3.7771 | 2.6969 |
|
| 71 |
+
| 0.5033 | 0.8 | 350 | 0.5168 | -1.8756 | -3.1461 | 0.7414 | 1.2705 | -562.8871 | -447.5259 | 3.8637 | 2.7948 |
|
| 72 |
+
| 0.52 | 0.9143 | 400 | 0.5142 | -1.9575 | -3.2488 | 0.7198 | 1.2913 | -573.1553 | -455.7142 | 4.0394 | 3.0591 |
|
| 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": 0.5629221896557841,
|
| 5 |
+
"train_runtime": 11730.276,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.774,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 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": 0.5629221896557841,
|
| 5 |
+
"train_runtime": 11730.276,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.774,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 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": 4.698039299265843,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.654461622238159,
|
| 16 |
+
"logits/rejected": -2.6349782943725586,
|
| 17 |
+
"logps/chosen": -287.92938232421875,
|
| 18 |
+
"logps/rejected": -266.479248046875,
|
| 19 |
+
"loss": 0.6932,
|
| 20 |
+
"rewards/accuracies": 0.4000000059604645,
|
| 21 |
+
"rewards/chosen": -0.00014368369011208415,
|
| 22 |
+
"rewards/margins": -0.00023763992066960782,
|
| 23 |
+
"rewards/rejected": 9.395615052198991e-05,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 4.697985911511002,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.655208110809326,
|
| 31 |
+
"logits/rejected": -2.623675584793091,
|
| 32 |
+
"logps/chosen": -271.43304443359375,
|
| 33 |
+
"logps/rejected": -260.2268981933594,
|
| 34 |
+
"loss": 0.6927,
|
| 35 |
+
"rewards/accuracies": 0.581250011920929,
|
| 36 |
+
"rewards/chosen": 0.002769916784018278,
|
| 37 |
+
"rewards/margins": 0.0018681169021874666,
|
| 38 |
+
"rewards/rejected": 0.0009017998236231506,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 4.400775078217756,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6501145362854004,
|
| 46 |
+
"logits/rejected": -2.5595479011535645,
|
| 47 |
+
"logps/chosen": -271.99346923828125,
|
| 48 |
+
"logps/rejected": -254.75973510742188,
|
| 49 |
+
"loss": 0.6903,
|
| 50 |
+
"rewards/accuracies": 0.737500011920929,
|
| 51 |
+
"rewards/chosen": 0.018953289836645126,
|
| 52 |
+
"rewards/margins": 0.01143287681043148,
|
| 53 |
+
"rewards/rejected": 0.007520412094891071,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 4.224410491085399,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.610691547393799,
|
| 61 |
+
"logits/rejected": -2.596449375152588,
|
| 62 |
+
"logps/chosen": -262.10076904296875,
|
| 63 |
+
"logps/rejected": -255.63131713867188,
|
| 64 |
+
"loss": 0.6838,
|
| 65 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 66 |
+
"rewards/chosen": 0.03673754259943962,
|
| 67 |
+
"rewards/margins": 0.03850514069199562,
|
| 68 |
+
"rewards/rejected": -0.0017675980925559998,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 5.26619927871512,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.59323787689209,
|
| 76 |
+
"logits/rejected": -2.5254805088043213,
|
| 77 |
+
"logps/chosen": -288.6198425292969,
|
| 78 |
+
"logps/rejected": -263.8785095214844,
|
| 79 |
+
"loss": 0.6718,
|
| 80 |
+
"rewards/accuracies": 0.75,
|
| 81 |
+
"rewards/chosen": 0.008304540999233723,
|
| 82 |
+
"rewards/margins": 0.10466225445270538,
|
| 83 |
+
"rewards/rejected": -0.09635771811008453,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.613680839538574,
|
| 89 |
+
"eval_logits/rejected": -2.570056200027466,
|
| 90 |
+
"eval_logps/chosen": -261.988525390625,
|
| 91 |
+
"eval_logps/rejected": -261.6897277832031,
|
| 92 |
+
"eval_loss": 0.6623334884643555,
|
| 93 |
+
"eval_rewards/accuracies": 0.6594827771186829,
|
| 94 |
+
"eval_rewards/chosen": -0.020272402092814445,
|
| 95 |
+
"eval_rewards/margins": 0.11385296285152435,
|
| 96 |
+
"eval_rewards/rejected": -0.13412536680698395,
|
| 97 |
+
"eval_runtime": 94.9862,
|
| 98 |
+
"eval_samples_per_second": 19.276,
|
| 99 |
+
"eval_steps_per_second": 0.305,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 6.932576674963505,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.6151022911071777,
|
| 107 |
+
"logits/rejected": -2.563739776611328,
|
| 108 |
+
"logps/chosen": -303.95672607421875,
|
| 109 |
+
"logps/rejected": -269.7010498046875,
|
| 110 |
+
"loss": 0.6526,
|
| 111 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 112 |
+
"rewards/chosen": -0.04470657557249069,
|
| 113 |
+
"rewards/margins": 0.18772244453430176,
|
| 114 |
+
"rewards/rejected": -0.23242902755737305,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 10.775877567674739,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.5894806385040283,
|
| 122 |
+
"logits/rejected": -2.5019643306732178,
|
| 123 |
+
"logps/chosen": -320.10382080078125,
|
| 124 |
+
"logps/rejected": -296.3632507324219,
|
| 125 |
+
"loss": 0.6419,
|
| 126 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 127 |
+
"rewards/chosen": -0.1674882471561432,
|
| 128 |
+
"rewards/margins": 0.30879640579223633,
|
| 129 |
+
"rewards/rejected": -0.47628456354141235,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 8.903665160814306,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.577723979949951,
|
| 137 |
+
"logits/rejected": -2.5037548542022705,
|
| 138 |
+
"logps/chosen": -308.7467346191406,
|
| 139 |
+
"logps/rejected": -314.86712646484375,
|
| 140 |
+
"loss": 0.6373,
|
| 141 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 142 |
+
"rewards/chosen": -0.41550832986831665,
|
| 143 |
+
"rewards/margins": 0.2658213675022125,
|
| 144 |
+
"rewards/rejected": -0.6813297271728516,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 13.187818767437374,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -1.7901182174682617,
|
| 152 |
+
"logits/rejected": -1.6417770385742188,
|
| 153 |
+
"logps/chosen": -292.3073425292969,
|
| 154 |
+
"logps/rejected": -347.75286865234375,
|
| 155 |
+
"loss": 0.6171,
|
| 156 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 157 |
+
"rewards/chosen": -0.507135272026062,
|
| 158 |
+
"rewards/margins": 0.4524330496788025,
|
| 159 |
+
"rewards/rejected": -0.9595683813095093,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 18.455967644618873,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -1.0523039102554321,
|
| 167 |
+
"logits/rejected": -0.7287885546684265,
|
| 168 |
+
"logps/chosen": -313.5967712402344,
|
| 169 |
+
"logps/rejected": -351.8410949707031,
|
| 170 |
+
"loss": 0.5837,
|
| 171 |
+
"rewards/accuracies": 0.71875,
|
| 172 |
+
"rewards/chosen": -0.5234693884849548,
|
| 173 |
+
"rewards/margins": 0.6886458396911621,
|
| 174 |
+
"rewards/rejected": -1.2121151685714722,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -0.5771592259407043,
|
| 180 |
+
"eval_logits/rejected": -0.20762209594249725,
|
| 181 |
+
"eval_logps/chosen": -339.2017517089844,
|
| 182 |
+
"eval_logps/rejected": -399.70703125,
|
| 183 |
+
"eval_loss": 0.5854533314704895,
|
| 184 |
+
"eval_rewards/accuracies": 0.6594827771186829,
|
| 185 |
+
"eval_rewards/chosen": -0.7924044132232666,
|
| 186 |
+
"eval_rewards/margins": 0.7218940258026123,
|
| 187 |
+
"eval_rewards/rejected": -1.5142985582351685,
|
| 188 |
+
"eval_runtime": 93.9256,
|
| 189 |
+
"eval_samples_per_second": 19.494,
|
| 190 |
+
"eval_steps_per_second": 0.309,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 20.061982050119,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": -0.07258351147174835,
|
| 198 |
+
"logits/rejected": 0.5934920310974121,
|
| 199 |
+
"logps/chosen": -378.9754333496094,
|
| 200 |
+
"logps/rejected": -401.3129577636719,
|
| 201 |
+
"loss": 0.5614,
|
| 202 |
+
"rewards/accuracies": 0.768750011920929,
|
| 203 |
+
"rewards/chosen": -0.7271420955657959,
|
| 204 |
+
"rewards/margins": 0.9416759610176086,
|
| 205 |
+
"rewards/rejected": -1.6688178777694702,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 19.936668067723307,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": 0.4706380367279053,
|
| 213 |
+
"logits/rejected": 0.9462782740592957,
|
| 214 |
+
"logps/chosen": -361.02105712890625,
|
| 215 |
+
"logps/rejected": -419.98455810546875,
|
| 216 |
+
"loss": 0.5824,
|
| 217 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 218 |
+
"rewards/chosen": -0.8818947076797485,
|
| 219 |
+
"rewards/margins": 0.8208815455436707,
|
| 220 |
+
"rewards/rejected": -1.702776312828064,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 19.241668087031446,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": 0.959698498249054,
|
| 228 |
+
"logits/rejected": 1.6635558605194092,
|
| 229 |
+
"logps/chosen": -376.22796630859375,
|
| 230 |
+
"logps/rejected": -403.8036193847656,
|
| 231 |
+
"loss": 0.5653,
|
| 232 |
+
"rewards/accuracies": 0.643750011920929,
|
| 233 |
+
"rewards/chosen": -1.0274622440338135,
|
| 234 |
+
"rewards/margins": 0.6216489672660828,
|
| 235 |
+
"rewards/rejected": -1.6491111516952515,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 14.721763869214124,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": 1.4800783395767212,
|
| 243 |
+
"logits/rejected": 2.052093029022217,
|
| 244 |
+
"logps/chosen": -443.85260009765625,
|
| 245 |
+
"logps/rejected": -478.54522705078125,
|
| 246 |
+
"loss": 0.5663,
|
| 247 |
+
"rewards/accuracies": 0.71875,
|
| 248 |
+
"rewards/chosen": -1.2601962089538574,
|
| 249 |
+
"rewards/margins": 0.9224704504013062,
|
| 250 |
+
"rewards/rejected": -2.182666540145874,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 18.13176648345513,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": 1.882054328918457,
|
| 258 |
+
"logits/rejected": 2.4975225925445557,
|
| 259 |
+
"logps/chosen": -417.7086486816406,
|
| 260 |
+
"logps/rejected": -473.6971740722656,
|
| 261 |
+
"loss": 0.5481,
|
| 262 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 263 |
+
"rewards/chosen": -1.2165459394454956,
|
| 264 |
+
"rewards/margins": 0.8646339178085327,
|
| 265 |
+
"rewards/rejected": -2.0811798572540283,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": 1.6485474109649658,
|
| 271 |
+
"eval_logits/rejected": 2.4336249828338623,
|
| 272 |
+
"eval_logps/chosen": -394.3633728027344,
|
| 273 |
+
"eval_logps/rejected": -477.4252624511719,
|
| 274 |
+
"eval_loss": 0.5498274564743042,
|
| 275 |
+
"eval_rewards/accuracies": 0.693965494632721,
|
| 276 |
+
"eval_rewards/chosen": -1.3440207242965698,
|
| 277 |
+
"eval_rewards/margins": 0.9474601745605469,
|
| 278 |
+
"eval_rewards/rejected": -2.291480779647827,
|
| 279 |
+
"eval_runtime": 94.1545,
|
| 280 |
+
"eval_samples_per_second": 19.447,
|
| 281 |
+
"eval_steps_per_second": 0.308,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 25.73162601058156,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": 0.9372028112411499,
|
| 289 |
+
"logits/rejected": 2.038717746734619,
|
| 290 |
+
"logps/chosen": -432.73846435546875,
|
| 291 |
+
"logps/rejected": -488.79315185546875,
|
| 292 |
+
"loss": 0.5381,
|
| 293 |
+
"rewards/accuracies": 0.71875,
|
| 294 |
+
"rewards/chosen": -1.2358500957489014,
|
| 295 |
+
"rewards/margins": 0.9547734260559082,
|
| 296 |
+
"rewards/rejected": -2.1906232833862305,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 22.65560343905013,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": 1.915405511856079,
|
| 304 |
+
"logits/rejected": 2.607452869415283,
|
| 305 |
+
"logps/chosen": -435.5879821777344,
|
| 306 |
+
"logps/rejected": -534.2739868164062,
|
| 307 |
+
"loss": 0.5443,
|
| 308 |
+
"rewards/accuracies": 0.71875,
|
| 309 |
+
"rewards/chosen": -1.4189122915267944,
|
| 310 |
+
"rewards/margins": 1.1850039958953857,
|
| 311 |
+
"rewards/rejected": -2.6039161682128906,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 23.399452190018653,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": 2.181429624557495,
|
| 319 |
+
"logits/rejected": 3.391533374786377,
|
| 320 |
+
"logps/chosen": -434.38287353515625,
|
| 321 |
+
"logps/rejected": -499.68768310546875,
|
| 322 |
+
"loss": 0.5411,
|
| 323 |
+
"rewards/accuracies": 0.800000011920929,
|
| 324 |
+
"rewards/chosen": -1.3194260597229004,
|
| 325 |
+
"rewards/margins": 1.2128746509552002,
|
| 326 |
+
"rewards/rejected": -2.5323009490966797,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 20.359816434441168,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": 1.7659626007080078,
|
| 334 |
+
"logits/rejected": 3.134269952774048,
|
| 335 |
+
"logps/chosen": -422.34130859375,
|
| 336 |
+
"logps/rejected": -507.91058349609375,
|
| 337 |
+
"loss": 0.5419,
|
| 338 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 339 |
+
"rewards/chosen": -1.6150566339492798,
|
| 340 |
+
"rewards/margins": 1.1147284507751465,
|
| 341 |
+
"rewards/rejected": -2.729785203933716,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 19.72704748785004,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": 2.624401569366455,
|
| 349 |
+
"logits/rejected": 3.5372815132141113,
|
| 350 |
+
"logps/chosen": -405.11773681640625,
|
| 351 |
+
"logps/rejected": -503.3150939941406,
|
| 352 |
+
"loss": 0.5289,
|
| 353 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 354 |
+
"rewards/chosen": -1.5478131771087646,
|
| 355 |
+
"rewards/margins": 1.097571611404419,
|
| 356 |
+
"rewards/rejected": -2.6453845500946045,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": 2.6372268199920654,
|
| 362 |
+
"eval_logits/rejected": 3.5667386054992676,
|
| 363 |
+
"eval_logps/chosen": -424.2229919433594,
|
| 364 |
+
"eval_logps/rejected": -523.725341796875,
|
| 365 |
+
"eval_loss": 0.5408812165260315,
|
| 366 |
+
"eval_rewards/accuracies": 0.6982758641242981,
|
| 367 |
+
"eval_rewards/chosen": -1.642616629600525,
|
| 368 |
+
"eval_rewards/margins": 1.111864686012268,
|
| 369 |
+
"eval_rewards/rejected": -2.754481554031372,
|
| 370 |
+
"eval_runtime": 94.7878,
|
| 371 |
+
"eval_samples_per_second": 19.317,
|
| 372 |
+
"eval_steps_per_second": 0.306,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 22.170086283776122,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": 2.3618617057800293,
|
| 380 |
+
"logits/rejected": 3.551234722137451,
|
| 381 |
+
"logps/chosen": -444.750244140625,
|
| 382 |
+
"logps/rejected": -544.8475341796875,
|
| 383 |
+
"loss": 0.5315,
|
| 384 |
+
"rewards/accuracies": 0.71875,
|
| 385 |
+
"rewards/chosen": -1.5890655517578125,
|
| 386 |
+
"rewards/margins": 1.23880934715271,
|
| 387 |
+
"rewards/rejected": -2.8278753757476807,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 20.88239892734878,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": 2.4734537601470947,
|
| 395 |
+
"logits/rejected": 3.3690295219421387,
|
| 396 |
+
"logps/chosen": -479.52294921875,
|
| 397 |
+
"logps/rejected": -594.2535400390625,
|
| 398 |
+
"loss": 0.542,
|
| 399 |
+
"rewards/accuracies": 0.71875,
|
| 400 |
+
"rewards/chosen": -1.751405119895935,
|
| 401 |
+
"rewards/margins": 1.2897474765777588,
|
| 402 |
+
"rewards/rejected": -3.0411527156829834,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 17.747062332553714,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": 1.912064790725708,
|
| 410 |
+
"logits/rejected": 2.7421936988830566,
|
| 411 |
+
"logps/chosen": -438.20208740234375,
|
| 412 |
+
"logps/rejected": -520.5653076171875,
|
| 413 |
+
"loss": 0.5653,
|
| 414 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 415 |
+
"rewards/chosen": -1.534075140953064,
|
| 416 |
+
"rewards/margins": 0.8896983861923218,
|
| 417 |
+
"rewards/rejected": -2.4237735271453857,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 18.030085534322904,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": 2.9936442375183105,
|
| 425 |
+
"logits/rejected": 4.284924507141113,
|
| 426 |
+
"logps/chosen": -503.80523681640625,
|
| 427 |
+
"logps/rejected": -576.6605834960938,
|
| 428 |
+
"loss": 0.5468,
|
| 429 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 430 |
+
"rewards/chosen": -1.9611694812774658,
|
| 431 |
+
"rewards/margins": 1.1891307830810547,
|
| 432 |
+
"rewards/rejected": -3.1503000259399414,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 24.141080092797733,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": 2.715038299560547,
|
| 440 |
+
"logits/rejected": 3.2168846130371094,
|
| 441 |
+
"logps/chosen": -476.09295654296875,
|
| 442 |
+
"logps/rejected": -553.4679565429688,
|
| 443 |
+
"loss": 0.54,
|
| 444 |
+
"rewards/accuracies": 0.65625,
|
| 445 |
+
"rewards/chosen": -1.9882991313934326,
|
| 446 |
+
"rewards/margins": 0.7364269495010376,
|
| 447 |
+
"rewards/rejected": -2.7247262001037598,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": 1.416563630104065,
|
| 453 |
+
"eval_logits/rejected": 2.6436657905578613,
|
| 454 |
+
"eval_logps/chosen": -413.8667297363281,
|
| 455 |
+
"eval_logps/rejected": -518.8563232421875,
|
| 456 |
+
"eval_loss": 0.5280377268791199,
|
| 457 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 458 |
+
"eval_rewards/chosen": -1.5390541553497314,
|
| 459 |
+
"eval_rewards/margins": 1.1667375564575195,
|
| 460 |
+
"eval_rewards/rejected": -2.70579195022583,
|
| 461 |
+
"eval_runtime": 93.2943,
|
| 462 |
+
"eval_samples_per_second": 19.626,
|
| 463 |
+
"eval_steps_per_second": 0.311,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 19.312331708156304,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": 1.422081708908081,
|
| 471 |
+
"logits/rejected": 2.8245503902435303,
|
| 472 |
+
"logps/chosen": -426.8099670410156,
|
| 473 |
+
"logps/rejected": -459.8812561035156,
|
| 474 |
+
"loss": 0.5455,
|
| 475 |
+
"rewards/accuracies": 0.71875,
|
| 476 |
+
"rewards/chosen": -1.5098531246185303,
|
| 477 |
+
"rewards/margins": 0.8642712831497192,
|
| 478 |
+
"rewards/rejected": -2.374124526977539,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 17.556239848475172,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": 1.7777824401855469,
|
| 486 |
+
"logits/rejected": 2.890540599822998,
|
| 487 |
+
"logps/chosen": -466.5252990722656,
|
| 488 |
+
"logps/rejected": -499.5956115722656,
|
| 489 |
+
"loss": 0.5345,
|
| 490 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 491 |
+
"rewards/chosen": -1.783983588218689,
|
| 492 |
+
"rewards/margins": 0.7515356540679932,
|
| 493 |
+
"rewards/rejected": -2.5355191230773926,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 16.478054364358847,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": 2.3310110569000244,
|
| 501 |
+
"logits/rejected": 3.422163486480713,
|
| 502 |
+
"logps/chosen": -419.81512451171875,
|
| 503 |
+
"logps/rejected": -533.2318115234375,
|
| 504 |
+
"loss": 0.5402,
|
| 505 |
+
"rewards/accuracies": 0.737500011920929,
|
| 506 |
+
"rewards/chosen": -1.7008750438690186,
|
| 507 |
+
"rewards/margins": 1.2960573434829712,
|
| 508 |
+
"rewards/rejected": -2.9969325065612793,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 21.127488932057656,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": 2.3282363414764404,
|
| 516 |
+
"logits/rejected": 3.5149283409118652,
|
| 517 |
+
"logps/chosen": -463.74200439453125,
|
| 518 |
+
"logps/rejected": -535.0039672851562,
|
| 519 |
+
"loss": 0.5179,
|
| 520 |
+
"rewards/accuracies": 0.71875,
|
| 521 |
+
"rewards/chosen": -1.7781498432159424,
|
| 522 |
+
"rewards/margins": 1.0403541326522827,
|
| 523 |
+
"rewards/rejected": -2.8185038566589355,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 21.3761678454725,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": 2.3439698219299316,
|
| 531 |
+
"logits/rejected": 3.820255994796753,
|
| 532 |
+
"logps/chosen": -497.107666015625,
|
| 533 |
+
"logps/rejected": -575.4114990234375,
|
| 534 |
+
"loss": 0.5147,
|
| 535 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 536 |
+
"rewards/chosen": -1.8375751972198486,
|
| 537 |
+
"rewards/margins": 1.2724038362503052,
|
| 538 |
+
"rewards/rejected": -3.1099791526794434,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": 2.696895122528076,
|
| 544 |
+
"eval_logits/rejected": 3.777069330215454,
|
| 545 |
+
"eval_logps/chosen": -444.8299560546875,
|
| 546 |
+
"eval_logps/rejected": -558.1808471679688,
|
| 547 |
+
"eval_loss": 0.5203906297683716,
|
| 548 |
+
"eval_rewards/accuracies": 0.7112069129943848,
|
| 549 |
+
"eval_rewards/chosen": -1.8486860990524292,
|
| 550 |
+
"eval_rewards/margins": 1.2503504753112793,
|
| 551 |
+
"eval_rewards/rejected": -3.099036455154419,
|
| 552 |
+
"eval_runtime": 93.3957,
|
| 553 |
+
"eval_samples_per_second": 19.605,
|
| 554 |
+
"eval_steps_per_second": 0.311,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 21.989201062769567,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": 2.4984259605407715,
|
| 562 |
+
"logits/rejected": 3.986057996749878,
|
| 563 |
+
"logps/chosen": -490.80255126953125,
|
| 564 |
+
"logps/rejected": -586.6756591796875,
|
| 565 |
+
"loss": 0.5285,
|
| 566 |
+
"rewards/accuracies": 0.768750011920929,
|
| 567 |
+
"rewards/chosen": -1.775442361831665,
|
| 568 |
+
"rewards/margins": 1.4152400493621826,
|
| 569 |
+
"rewards/rejected": -3.1906826496124268,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 22.181868921213347,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": 2.8167648315429688,
|
| 577 |
+
"logits/rejected": 4.0772247314453125,
|
| 578 |
+
"logps/chosen": -484.88201904296875,
|
| 579 |
+
"logps/rejected": -577.0535888671875,
|
| 580 |
+
"loss": 0.533,
|
| 581 |
+
"rewards/accuracies": 0.71875,
|
| 582 |
+
"rewards/chosen": -1.8776414394378662,
|
| 583 |
+
"rewards/margins": 1.1431219577789307,
|
| 584 |
+
"rewards/rejected": -3.0207631587982178,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 21.423699121141595,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": 2.9268698692321777,
|
| 592 |
+
"logits/rejected": 3.6747875213623047,
|
| 593 |
+
"logps/chosen": -416.2300720214844,
|
| 594 |
+
"logps/rejected": -506.71966552734375,
|
| 595 |
+
"loss": 0.5385,
|
| 596 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 597 |
+
"rewards/chosen": -1.7054163217544556,
|
| 598 |
+
"rewards/margins": 0.9577444195747375,
|
| 599 |
+
"rewards/rejected": -2.663160800933838,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 22.736161602909004,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": 2.4303653240203857,
|
| 607 |
+
"logits/rejected": 3.7860004901885986,
|
| 608 |
+
"logps/chosen": -426.480224609375,
|
| 609 |
+
"logps/rejected": -512.9637451171875,
|
| 610 |
+
"loss": 0.5184,
|
| 611 |
+
"rewards/accuracies": 0.762499988079071,
|
| 612 |
+
"rewards/chosen": -1.8047637939453125,
|
| 613 |
+
"rewards/margins": 1.1890126466751099,
|
| 614 |
+
"rewards/rejected": -2.9937758445739746,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 23.477309940264664,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": 2.426809310913086,
|
| 622 |
+
"logits/rejected": 4.1134934425354,
|
| 623 |
+
"logps/chosen": -493.06414794921875,
|
| 624 |
+
"logps/rejected": -577.4049072265625,
|
| 625 |
+
"loss": 0.5033,
|
| 626 |
+
"rewards/accuracies": 0.768750011920929,
|
| 627 |
+
"rewards/chosen": -1.8289105892181396,
|
| 628 |
+
"rewards/margins": 1.2613272666931152,
|
| 629 |
+
"rewards/rejected": -3.090238094329834,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": 2.7948429584503174,
|
| 635 |
+
"eval_logits/rejected": 3.863651990890503,
|
| 636 |
+
"eval_logps/chosen": -447.5259094238281,
|
| 637 |
+
"eval_logps/rejected": -562.8871459960938,
|
| 638 |
+
"eval_loss": 0.5168038606643677,
|
| 639 |
+
"eval_rewards/accuracies": 0.7413793206214905,
|
| 640 |
+
"eval_rewards/chosen": -1.8756457567214966,
|
| 641 |
+
"eval_rewards/margins": 1.2704533338546753,
|
| 642 |
+
"eval_rewards/rejected": -3.146099328994751,
|
| 643 |
+
"eval_runtime": 92.678,
|
| 644 |
+
"eval_samples_per_second": 19.757,
|
| 645 |
+
"eval_steps_per_second": 0.313,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 22.089570552576745,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": 3.1440396308898926,
|
| 653 |
+
"logits/rejected": 4.210689067840576,
|
| 654 |
+
"logps/chosen": -474.7491760253906,
|
| 655 |
+
"logps/rejected": -611.8050537109375,
|
| 656 |
+
"loss": 0.5176,
|
| 657 |
+
"rewards/accuracies": 0.800000011920929,
|
| 658 |
+
"rewards/chosen": -1.8223841190338135,
|
| 659 |
+
"rewards/margins": 1.6220409870147705,
|
| 660 |
+
"rewards/rejected": -3.444424867630005,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 26.5380528583338,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": 3.0336647033691406,
|
| 668 |
+
"logits/rejected": 3.8006350994110107,
|
| 669 |
+
"logps/chosen": -432.41278076171875,
|
| 670 |
+
"logps/rejected": -567.8787231445312,
|
| 671 |
+
"loss": 0.5204,
|
| 672 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 673 |
+
"rewards/chosen": -1.7640663385391235,
|
| 674 |
+
"rewards/margins": 1.2935981750488281,
|
| 675 |
+
"rewards/rejected": -3.057664394378662,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 24.05648133941441,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": 3.677473783493042,
|
| 683 |
+
"logits/rejected": 4.557853698730469,
|
| 684 |
+
"logps/chosen": -449.2403259277344,
|
| 685 |
+
"logps/rejected": -553.65966796875,
|
| 686 |
+
"loss": 0.5084,
|
| 687 |
+
"rewards/accuracies": 0.699999988079071,
|
| 688 |
+
"rewards/chosen": -1.9337879419326782,
|
| 689 |
+
"rewards/margins": 1.0252636671066284,
|
| 690 |
+
"rewards/rejected": -2.9590516090393066,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 18.50726505627672,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": 3.2151970863342285,
|
| 698 |
+
"logits/rejected": 4.2298479080200195,
|
| 699 |
+
"logps/chosen": -456.96112060546875,
|
| 700 |
+
"logps/rejected": -567.398681640625,
|
| 701 |
+
"loss": 0.519,
|
| 702 |
+
"rewards/accuracies": 0.675000011920929,
|
| 703 |
+
"rewards/chosen": -1.903903603553772,
|
| 704 |
+
"rewards/margins": 1.2563444375991821,
|
| 705 |
+
"rewards/rejected": -3.160248041152954,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 23.71785447914187,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": 2.804896831512451,
|
| 713 |
+
"logits/rejected": 3.584916353225708,
|
| 714 |
+
"logps/chosen": -462.6809997558594,
|
| 715 |
+
"logps/rejected": -580.8098754882812,
|
| 716 |
+
"loss": 0.52,
|
| 717 |
+
"rewards/accuracies": 0.75,
|
| 718 |
+
"rewards/chosen": -1.885479211807251,
|
| 719 |
+
"rewards/margins": 1.2495968341827393,
|
| 720 |
+
"rewards/rejected": -3.1350760459899902,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": 3.059108257293701,
|
| 726 |
+
"eval_logits/rejected": 4.039414882659912,
|
| 727 |
+
"eval_logps/chosen": -455.7142333984375,
|
| 728 |
+
"eval_logps/rejected": -573.1553344726562,
|
| 729 |
+
"eval_loss": 0.5141626000404358,
|
| 730 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 731 |
+
"eval_rewards/chosen": -1.9575295448303223,
|
| 732 |
+
"eval_rewards/margins": 1.2912517786026,
|
| 733 |
+
"eval_rewards/rejected": -3.248781442642212,
|
| 734 |
+
"eval_runtime": 94.0826,
|
| 735 |
+
"eval_samples_per_second": 19.462,
|
| 736 |
+
"eval_steps_per_second": 0.308,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 24.860041582436438,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": 3.145111560821533,
|
| 744 |
+
"logits/rejected": 4.010159492492676,
|
| 745 |
+
"logps/chosen": -449.4214782714844,
|
| 746 |
+
"logps/rejected": -553.9349365234375,
|
| 747 |
+
"loss": 0.5035,
|
| 748 |
+
"rewards/accuracies": 0.78125,
|
| 749 |
+
"rewards/chosen": -1.891174077987671,
|
| 750 |
+
"rewards/margins": 1.0715255737304688,
|
| 751 |
+
"rewards/rejected": -2.9626996517181396,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 20.06298675421982,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": 2.699260950088501,
|
| 759 |
+
"logits/rejected": 3.91593861579895,
|
| 760 |
+
"logps/chosen": -503.0846252441406,
|
| 761 |
+
"logps/rejected": -636.1543579101562,
|
| 762 |
+
"loss": 0.524,
|
| 763 |
+
"rewards/accuracies": 0.75,
|
| 764 |
+
"rewards/chosen": -1.9241759777069092,
|
| 765 |
+
"rewards/margins": 1.4948225021362305,
|
| 766 |
+
"rewards/rejected": -3.4189987182617188,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 18.461900399401518,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": 3.151226043701172,
|
| 774 |
+
"logits/rejected": 4.009431838989258,
|
| 775 |
+
"logps/chosen": -474.90179443359375,
|
| 776 |
+
"logps/rejected": -583.8973388671875,
|
| 777 |
+
"loss": 0.4974,
|
| 778 |
+
"rewards/accuracies": 0.762499988079071,
|
| 779 |
+
"rewards/chosen": -1.8930785655975342,
|
| 780 |
+
"rewards/margins": 1.3667445182800293,
|
| 781 |
+
"rewards/rejected": -3.2598233222961426,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5629221896557841,
|
| 789 |
+
"train_runtime": 11730.276,
|
| 790 |
+
"train_samples_per_second": 4.774,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 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 |
+
}
|