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-01
|
| 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-01
|
| 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: 0.5422
|
| 21 |
+
- Rewards/chosen: -1.5854
|
| 22 |
+
- Rewards/rejected: -3.0024
|
| 23 |
+
- Rewards/accuracies: 0.7672
|
| 24 |
+
- Rewards/margins: 1.4170
|
| 25 |
+
- Logps/rejected: -519.3195
|
| 26 |
+
- Logps/chosen: -434.1443
|
| 27 |
+
- Logits/rejected: 2.4835
|
| 28 |
+
- Logits/chosen: 0.7325
|
| 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 |
+
| 0.6745 | 0.1143 | 50 | 0.6652 | -0.0219 | -0.1700 | 0.6940 | 0.1481 | -236.0727 | -277.7910 | -2.4379 | -2.5388 |
|
| 66 |
+
| 0.6193 | 0.2286 | 100 | 0.6070 | -0.5960 | -1.2325 | 0.7026 | 0.6364 | -342.3217 | -335.2089 | -2.3205 | -2.4474 |
|
| 67 |
+
| 0.5552 | 0.3429 | 150 | 0.5762 | -1.3758 | -2.5659 | 0.7457 | 1.1901 | -475.6642 | -413.1814 | 1.7135 | 0.6211 |
|
| 68 |
+
| 0.5607 | 0.4571 | 200 | 0.5560 | -1.2277 | -2.4493 | 0.7672 | 1.2216 | -464.0081 | -398.3775 | 0.6492 | -0.7141 |
|
| 69 |
+
| 0.5589 | 0.5714 | 250 | 0.5523 | -1.3250 | -2.6441 | 0.7716 | 1.3191 | -483.4845 | -408.1035 | 1.3227 | -0.2479 |
|
| 70 |
+
| 0.5351 | 0.6857 | 300 | 0.5430 | -1.4547 | -2.8407 | 0.7543 | 1.3859 | -503.1421 | -421.0749 | 1.9525 | 0.2785 |
|
| 71 |
+
| 0.548 | 0.8 | 350 | 0.5426 | -1.5802 | -2.9933 | 0.7716 | 1.4131 | -518.4055 | -433.6277 | 2.3928 | 0.6373 |
|
| 72 |
+
| 0.5602 | 0.9143 | 400 | 0.5422 | -1.5854 | -3.0024 | 0.7672 | 1.4170 | -519.3195 | -434.1443 | 2.4835 | 0.7325 |
|
| 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": 0.5784502373134517,
|
| 5 |
+
"train_runtime": 11646.632,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.808,
|
| 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": 0.5784502373134517,
|
| 5 |
+
"train_runtime": 11646.632,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.808,
|
| 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": 5.116205997832364,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.700634717941284,
|
| 16 |
+
"logits/rejected": -2.624845027923584,
|
| 17 |
+
"logps/chosen": -301.29974365234375,
|
| 18 |
+
"logps/rejected": -281.7809753417969,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.4437499940395355,
|
| 21 |
+
"rewards/chosen": -0.0001358071604045108,
|
| 22 |
+
"rewards/margins": 0.00017720976029522717,
|
| 23 |
+
"rewards/rejected": -0.00031301690614782274,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 4.048845319624005,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6409199237823486,
|
| 31 |
+
"logits/rejected": -2.605693817138672,
|
| 32 |
+
"logps/chosen": -278.9129638671875,
|
| 33 |
+
"logps/rejected": -254.7362060546875,
|
| 34 |
+
"loss": 0.6926,
|
| 35 |
+
"rewards/accuracies": 0.606249988079071,
|
| 36 |
+
"rewards/chosen": 0.002582464599981904,
|
| 37 |
+
"rewards/margins": 0.0022538029588758945,
|
| 38 |
+
"rewards/rejected": 0.0003286615537945181,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 4.418664852115103,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6376564502716064,
|
| 46 |
+
"logits/rejected": -2.617002248764038,
|
| 47 |
+
"logps/chosen": -263.3965148925781,
|
| 48 |
+
"logps/rejected": -263.52264404296875,
|
| 49 |
+
"loss": 0.6898,
|
| 50 |
+
"rewards/accuracies": 0.625,
|
| 51 |
+
"rewards/chosen": 0.01286567747592926,
|
| 52 |
+
"rewards/margins": 0.008161441422998905,
|
| 53 |
+
"rewards/rejected": 0.004704234655946493,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 5.048215894637636,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.647663116455078,
|
| 61 |
+
"logits/rejected": -2.584965229034424,
|
| 62 |
+
"logps/chosen": -290.48455810546875,
|
| 63 |
+
"logps/rejected": -268.39874267578125,
|
| 64 |
+
"loss": 0.6832,
|
| 65 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 66 |
+
"rewards/chosen": 0.03512703627347946,
|
| 67 |
+
"rewards/margins": 0.04280107468366623,
|
| 68 |
+
"rewards/rejected": -0.0076740398071706295,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 7.712133622084281,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.6761856079101562,
|
| 76 |
+
"logits/rejected": -2.5965471267700195,
|
| 77 |
+
"logps/chosen": -294.49273681640625,
|
| 78 |
+
"logps/rejected": -254.78549194335938,
|
| 79 |
+
"loss": 0.6745,
|
| 80 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 81 |
+
"rewards/chosen": 0.018175510689616203,
|
| 82 |
+
"rewards/margins": 0.09738700091838837,
|
| 83 |
+
"rewards/rejected": -0.07921148836612701,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.5388200283050537,
|
| 89 |
+
"eval_logits/rejected": -2.4379258155822754,
|
| 90 |
+
"eval_logps/chosen": -277.7910461425781,
|
| 91 |
+
"eval_logps/rejected": -236.07266235351562,
|
| 92 |
+
"eval_loss": 0.6652289628982544,
|
| 93 |
+
"eval_rewards/accuracies": 0.693965494632721,
|
| 94 |
+
"eval_rewards/chosen": -0.021870704367756844,
|
| 95 |
+
"eval_rewards/margins": 0.14809474349021912,
|
| 96 |
+
"eval_rewards/rejected": -0.16996543109416962,
|
| 97 |
+
"eval_runtime": 91.08,
|
| 98 |
+
"eval_samples_per_second": 20.103,
|
| 99 |
+
"eval_steps_per_second": 0.318,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 6.234314729404,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.577277421951294,
|
| 107 |
+
"logits/rejected": -2.4742989540100098,
|
| 108 |
+
"logps/chosen": -293.23321533203125,
|
| 109 |
+
"logps/rejected": -273.9915466308594,
|
| 110 |
+
"loss": 0.6581,
|
| 111 |
+
"rewards/accuracies": 0.65625,
|
| 112 |
+
"rewards/chosen": -0.12727248668670654,
|
| 113 |
+
"rewards/margins": 0.12712165713310242,
|
| 114 |
+
"rewards/rejected": -0.2543941140174866,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 7.736476543325128,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.597801446914673,
|
| 122 |
+
"logits/rejected": -2.543898105621338,
|
| 123 |
+
"logps/chosen": -289.6143493652344,
|
| 124 |
+
"logps/rejected": -295.7275695800781,
|
| 125 |
+
"loss": 0.6398,
|
| 126 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 127 |
+
"rewards/chosen": -0.13836105167865753,
|
| 128 |
+
"rewards/margins": 0.22264671325683594,
|
| 129 |
+
"rewards/rejected": -0.36100777983665466,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 8.2018046742501,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.6352217197418213,
|
| 137 |
+
"logits/rejected": -2.5961461067199707,
|
| 138 |
+
"logps/chosen": -297.4925842285156,
|
| 139 |
+
"logps/rejected": -316.795166015625,
|
| 140 |
+
"loss": 0.6315,
|
| 141 |
+
"rewards/accuracies": 0.75,
|
| 142 |
+
"rewards/chosen": -0.30463558435440063,
|
| 143 |
+
"rewards/margins": 0.3102269768714905,
|
| 144 |
+
"rewards/rejected": -0.6148625612258911,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 10.097131499300193,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -2.6919240951538086,
|
| 152 |
+
"logits/rejected": -2.6384730339050293,
|
| 153 |
+
"logps/chosen": -297.72381591796875,
|
| 154 |
+
"logps/rejected": -324.560791015625,
|
| 155 |
+
"loss": 0.629,
|
| 156 |
+
"rewards/accuracies": 0.706250011920929,
|
| 157 |
+
"rewards/chosen": -0.21438519656658173,
|
| 158 |
+
"rewards/margins": 0.47574347257614136,
|
| 159 |
+
"rewards/rejected": -0.6901286840438843,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 11.497879496244352,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -2.6669628620147705,
|
| 167 |
+
"logits/rejected": -2.5793137550354004,
|
| 168 |
+
"logps/chosen": -332.02459716796875,
|
| 169 |
+
"logps/rejected": -326.01910400390625,
|
| 170 |
+
"loss": 0.6193,
|
| 171 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 172 |
+
"rewards/chosen": -0.45638221502304077,
|
| 173 |
+
"rewards/margins": 0.46929416060447693,
|
| 174 |
+
"rewards/rejected": -0.9256764650344849,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -2.4473772048950195,
|
| 180 |
+
"eval_logits/rejected": -2.3204526901245117,
|
| 181 |
+
"eval_logps/chosen": -335.20892333984375,
|
| 182 |
+
"eval_logps/rejected": -342.3216552734375,
|
| 183 |
+
"eval_loss": 0.6070090532302856,
|
| 184 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 185 |
+
"eval_rewards/chosen": -0.5960496068000793,
|
| 186 |
+
"eval_rewards/margins": 0.6364051699638367,
|
| 187 |
+
"eval_rewards/rejected": -1.2324548959732056,
|
| 188 |
+
"eval_runtime": 90.7362,
|
| 189 |
+
"eval_samples_per_second": 20.179,
|
| 190 |
+
"eval_steps_per_second": 0.32,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 13.479714305351811,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": -2.1286423206329346,
|
| 198 |
+
"logits/rejected": -1.9770896434783936,
|
| 199 |
+
"logps/chosen": -387.36529541015625,
|
| 200 |
+
"logps/rejected": -417.18206787109375,
|
| 201 |
+
"loss": 0.5999,
|
| 202 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 203 |
+
"rewards/chosen": -0.710616946220398,
|
| 204 |
+
"rewards/margins": 0.4821074903011322,
|
| 205 |
+
"rewards/rejected": -1.192724347114563,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 21.834827090505335,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": -1.3608739376068115,
|
| 213 |
+
"logits/rejected": -0.962149441242218,
|
| 214 |
+
"logps/chosen": -384.6209716796875,
|
| 215 |
+
"logps/rejected": -364.0862731933594,
|
| 216 |
+
"loss": 0.5861,
|
| 217 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 218 |
+
"rewards/chosen": -0.7371653914451599,
|
| 219 |
+
"rewards/margins": 0.5865553617477417,
|
| 220 |
+
"rewards/rejected": -1.3237206935882568,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 17.97880068775228,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": -0.3695625364780426,
|
| 228 |
+
"logits/rejected": 0.15285547077655792,
|
| 229 |
+
"logps/chosen": -405.9339294433594,
|
| 230 |
+
"logps/rejected": -413.42706298828125,
|
| 231 |
+
"loss": 0.5653,
|
| 232 |
+
"rewards/accuracies": 0.643750011920929,
|
| 233 |
+
"rewards/chosen": -1.0077803134918213,
|
| 234 |
+
"rewards/margins": 0.6217871904373169,
|
| 235 |
+
"rewards/rejected": -1.6295677423477173,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 14.856966772188793,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": -0.6858156323432922,
|
| 243 |
+
"logits/rejected": -0.15091530978679657,
|
| 244 |
+
"logps/chosen": -395.23504638671875,
|
| 245 |
+
"logps/rejected": -425.66912841796875,
|
| 246 |
+
"loss": 0.5876,
|
| 247 |
+
"rewards/accuracies": 0.699999988079071,
|
| 248 |
+
"rewards/chosen": -0.7978739738464355,
|
| 249 |
+
"rewards/margins": 0.7397030591964722,
|
| 250 |
+
"rewards/rejected": -1.5375769138336182,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 15.48919645782762,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": 0.14924204349517822,
|
| 258 |
+
"logits/rejected": 0.8237205743789673,
|
| 259 |
+
"logps/chosen": -402.83050537109375,
|
| 260 |
+
"logps/rejected": -480.46337890625,
|
| 261 |
+
"loss": 0.5552,
|
| 262 |
+
"rewards/accuracies": 0.6875,
|
| 263 |
+
"rewards/chosen": -1.1584460735321045,
|
| 264 |
+
"rewards/margins": 0.8831979036331177,
|
| 265 |
+
"rewards/rejected": -2.0416440963745117,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": 0.6210644245147705,
|
| 271 |
+
"eval_logits/rejected": 1.7134919166564941,
|
| 272 |
+
"eval_logps/chosen": -413.1814270019531,
|
| 273 |
+
"eval_logps/rejected": -475.66424560546875,
|
| 274 |
+
"eval_loss": 0.5762364268302917,
|
| 275 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 276 |
+
"eval_rewards/chosen": -1.3757750988006592,
|
| 277 |
+
"eval_rewards/margins": 1.19010591506958,
|
| 278 |
+
"eval_rewards/rejected": -2.5658812522888184,
|
| 279 |
+
"eval_runtime": 90.6736,
|
| 280 |
+
"eval_samples_per_second": 20.193,
|
| 281 |
+
"eval_steps_per_second": 0.32,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 27.338096513767994,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": 0.46789199113845825,
|
| 289 |
+
"logits/rejected": 1.287762999534607,
|
| 290 |
+
"logps/chosen": -451.74456787109375,
|
| 291 |
+
"logps/rejected": -507.8985290527344,
|
| 292 |
+
"loss": 0.5691,
|
| 293 |
+
"rewards/accuracies": 0.6875,
|
| 294 |
+
"rewards/chosen": -1.5240428447723389,
|
| 295 |
+
"rewards/margins": 0.9985151290893555,
|
| 296 |
+
"rewards/rejected": -2.5225582122802734,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 20.374282543445702,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": 0.05013761669397354,
|
| 304 |
+
"logits/rejected": 0.7304227352142334,
|
| 305 |
+
"logps/chosen": -436.2395935058594,
|
| 306 |
+
"logps/rejected": -469.56884765625,
|
| 307 |
+
"loss": 0.5685,
|
| 308 |
+
"rewards/accuracies": 0.6875,
|
| 309 |
+
"rewards/chosen": -1.3213472366333008,
|
| 310 |
+
"rewards/margins": 0.7170308828353882,
|
| 311 |
+
"rewards/rejected": -2.0383780002593994,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 21.086569958558634,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": -0.4792841076850891,
|
| 319 |
+
"logits/rejected": 0.523780882358551,
|
| 320 |
+
"logps/chosen": -440.24420166015625,
|
| 321 |
+
"logps/rejected": -486.8812561035156,
|
| 322 |
+
"loss": 0.564,
|
| 323 |
+
"rewards/accuracies": 0.78125,
|
| 324 |
+
"rewards/chosen": -1.1006438732147217,
|
| 325 |
+
"rewards/margins": 1.168663501739502,
|
| 326 |
+
"rewards/rejected": -2.2693073749542236,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 19.913089361766232,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": -0.07349483668804169,
|
| 334 |
+
"logits/rejected": 0.8528989553451538,
|
| 335 |
+
"logps/chosen": -430.77471923828125,
|
| 336 |
+
"logps/rejected": -469.9756774902344,
|
| 337 |
+
"loss": 0.568,
|
| 338 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 339 |
+
"rewards/chosen": -1.5451713800430298,
|
| 340 |
+
"rewards/margins": 0.8096041679382324,
|
| 341 |
+
"rewards/rejected": -2.3547751903533936,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 16.58015899692404,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": -0.06599310785531998,
|
| 349 |
+
"logits/rejected": 1.1455082893371582,
|
| 350 |
+
"logps/chosen": -436.7762145996094,
|
| 351 |
+
"logps/rejected": -503.99127197265625,
|
| 352 |
+
"loss": 0.5607,
|
| 353 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 354 |
+
"rewards/chosen": -1.6843353509902954,
|
| 355 |
+
"rewards/margins": 1.066771388053894,
|
| 356 |
+
"rewards/rejected": -2.7511067390441895,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": -0.7141388654708862,
|
| 362 |
+
"eval_logits/rejected": 0.6491928100585938,
|
| 363 |
+
"eval_logps/chosen": -398.3775329589844,
|
| 364 |
+
"eval_logps/rejected": -464.0081481933594,
|
| 365 |
+
"eval_loss": 0.5560439229011536,
|
| 366 |
+
"eval_rewards/accuracies": 0.767241358757019,
|
| 367 |
+
"eval_rewards/chosen": -1.2277354001998901,
|
| 368 |
+
"eval_rewards/margins": 1.2215843200683594,
|
| 369 |
+
"eval_rewards/rejected": -2.449319839477539,
|
| 370 |
+
"eval_runtime": 90.9021,
|
| 371 |
+
"eval_samples_per_second": 20.143,
|
| 372 |
+
"eval_steps_per_second": 0.319,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 21.314455899070882,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": -0.3136857748031616,
|
| 380 |
+
"logits/rejected": 0.7702559232711792,
|
| 381 |
+
"logps/chosen": -407.5398254394531,
|
| 382 |
+
"logps/rejected": -480.2178649902344,
|
| 383 |
+
"loss": 0.5537,
|
| 384 |
+
"rewards/accuracies": 0.6875,
|
| 385 |
+
"rewards/chosen": -1.4408626556396484,
|
| 386 |
+
"rewards/margins": 0.908516526222229,
|
| 387 |
+
"rewards/rejected": -2.349379062652588,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 33.50940356009228,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": 0.8876802325248718,
|
| 395 |
+
"logits/rejected": 2.1045126914978027,
|
| 396 |
+
"logps/chosen": -468.2164001464844,
|
| 397 |
+
"logps/rejected": -526.6786499023438,
|
| 398 |
+
"loss": 0.5741,
|
| 399 |
+
"rewards/accuracies": 0.731249988079071,
|
| 400 |
+
"rewards/chosen": -1.8315035104751587,
|
| 401 |
+
"rewards/margins": 0.9734483957290649,
|
| 402 |
+
"rewards/rejected": -2.8049521446228027,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 30.59316545342991,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": 1.18466317653656,
|
| 410 |
+
"logits/rejected": 2.4997925758361816,
|
| 411 |
+
"logps/chosen": -467.6075744628906,
|
| 412 |
+
"logps/rejected": -524.8836669921875,
|
| 413 |
+
"loss": 0.5519,
|
| 414 |
+
"rewards/accuracies": 0.731249988079071,
|
| 415 |
+
"rewards/chosen": -1.8654839992523193,
|
| 416 |
+
"rewards/margins": 0.9904860258102417,
|
| 417 |
+
"rewards/rejected": -2.8559701442718506,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 20.516554684285598,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": 0.2621554434299469,
|
| 425 |
+
"logits/rejected": 1.582966923713684,
|
| 426 |
+
"logps/chosen": -450.7454528808594,
|
| 427 |
+
"logps/rejected": -559.3546142578125,
|
| 428 |
+
"loss": 0.5496,
|
| 429 |
+
"rewards/accuracies": 0.768750011920929,
|
| 430 |
+
"rewards/chosen": -1.5203564167022705,
|
| 431 |
+
"rewards/margins": 1.358971118927002,
|
| 432 |
+
"rewards/rejected": -2.8793272972106934,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 18.381009526967468,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": 0.046704571694135666,
|
| 440 |
+
"logits/rejected": 0.7480655908584595,
|
| 441 |
+
"logps/chosen": -425.90081787109375,
|
| 442 |
+
"logps/rejected": -522.4205322265625,
|
| 443 |
+
"loss": 0.5589,
|
| 444 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 445 |
+
"rewards/chosen": -1.4527461528778076,
|
| 446 |
+
"rewards/margins": 1.0694544315338135,
|
| 447 |
+
"rewards/rejected": -2.522200584411621,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": -0.24794088304042816,
|
| 453 |
+
"eval_logits/rejected": 1.3226882219314575,
|
| 454 |
+
"eval_logps/chosen": -408.1035461425781,
|
| 455 |
+
"eval_logps/rejected": -483.4844665527344,
|
| 456 |
+
"eval_loss": 0.5523289442062378,
|
| 457 |
+
"eval_rewards/accuracies": 0.7715517282485962,
|
| 458 |
+
"eval_rewards/chosen": -1.324995994567871,
|
| 459 |
+
"eval_rewards/margins": 1.319087028503418,
|
| 460 |
+
"eval_rewards/rejected": -2.644083023071289,
|
| 461 |
+
"eval_runtime": 92.8381,
|
| 462 |
+
"eval_samples_per_second": 19.723,
|
| 463 |
+
"eval_steps_per_second": 0.312,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 21.477935114375956,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": 0.25936073064804077,
|
| 471 |
+
"logits/rejected": 1.0254137516021729,
|
| 472 |
+
"logps/chosen": -452.327880859375,
|
| 473 |
+
"logps/rejected": -538.3406982421875,
|
| 474 |
+
"loss": 0.5539,
|
| 475 |
+
"rewards/accuracies": 0.699999988079071,
|
| 476 |
+
"rewards/chosen": -1.7254120111465454,
|
| 477 |
+
"rewards/margins": 0.924565315246582,
|
| 478 |
+
"rewards/rejected": -2.649976968765259,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 22.404068631830373,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": 0.9826221466064453,
|
| 486 |
+
"logits/rejected": 2.496065855026245,
|
| 487 |
+
"logps/chosen": -454.06201171875,
|
| 488 |
+
"logps/rejected": -520.5513916015625,
|
| 489 |
+
"loss": 0.5406,
|
| 490 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 491 |
+
"rewards/chosen": -1.8153613805770874,
|
| 492 |
+
"rewards/margins": 1.085078477859497,
|
| 493 |
+
"rewards/rejected": -2.900439739227295,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 24.981220974307632,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": 1.0183608531951904,
|
| 501 |
+
"logits/rejected": 2.100400924682617,
|
| 502 |
+
"logps/chosen": -443.5811462402344,
|
| 503 |
+
"logps/rejected": -558.9719848632812,
|
| 504 |
+
"loss": 0.5461,
|
| 505 |
+
"rewards/accuracies": 0.731249988079071,
|
| 506 |
+
"rewards/chosen": -1.7054297924041748,
|
| 507 |
+
"rewards/margins": 1.2532217502593994,
|
| 508 |
+
"rewards/rejected": -2.958651304244995,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 22.42240240560034,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": 1.1615647077560425,
|
| 516 |
+
"logits/rejected": 2.0770411491394043,
|
| 517 |
+
"logps/chosen": -479.4815979003906,
|
| 518 |
+
"logps/rejected": -552.2136840820312,
|
| 519 |
+
"loss": 0.5472,
|
| 520 |
+
"rewards/accuracies": 0.75,
|
| 521 |
+
"rewards/chosen": -1.7857574224472046,
|
| 522 |
+
"rewards/margins": 1.0808943510055542,
|
| 523 |
+
"rewards/rejected": -2.866652011871338,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 17.37829470785684,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": 0.5264058709144592,
|
| 531 |
+
"logits/rejected": 2.128044605255127,
|
| 532 |
+
"logps/chosen": -466.05633544921875,
|
| 533 |
+
"logps/rejected": -544.390869140625,
|
| 534 |
+
"loss": 0.5351,
|
| 535 |
+
"rewards/accuracies": 0.800000011920929,
|
| 536 |
+
"rewards/chosen": -1.641471266746521,
|
| 537 |
+
"rewards/margins": 1.2760158777236938,
|
| 538 |
+
"rewards/rejected": -2.9174869060516357,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": 0.27849671244621277,
|
| 544 |
+
"eval_logits/rejected": 1.952508568763733,
|
| 545 |
+
"eval_logps/chosen": -421.0749206542969,
|
| 546 |
+
"eval_logps/rejected": -503.1421203613281,
|
| 547 |
+
"eval_loss": 0.5429859757423401,
|
| 548 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 549 |
+
"eval_rewards/chosen": -1.4547101259231567,
|
| 550 |
+
"eval_rewards/margins": 1.3859494924545288,
|
| 551 |
+
"eval_rewards/rejected": -2.8406593799591064,
|
| 552 |
+
"eval_runtime": 90.64,
|
| 553 |
+
"eval_samples_per_second": 20.201,
|
| 554 |
+
"eval_steps_per_second": 0.32,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 20.02714298892451,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": 0.3929429054260254,
|
| 562 |
+
"logits/rejected": 1.4373176097869873,
|
| 563 |
+
"logps/chosen": -416.38616943359375,
|
| 564 |
+
"logps/rejected": -497.1776428222656,
|
| 565 |
+
"loss": 0.5427,
|
| 566 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 567 |
+
"rewards/chosen": -1.6008203029632568,
|
| 568 |
+
"rewards/margins": 0.9990032315254211,
|
| 569 |
+
"rewards/rejected": -2.599823474884033,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 17.430643463250824,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": 0.5950568318367004,
|
| 577 |
+
"logits/rejected": 1.5454418659210205,
|
| 578 |
+
"logps/chosen": -417.2794494628906,
|
| 579 |
+
"logps/rejected": -498.28594970703125,
|
| 580 |
+
"loss": 0.5494,
|
| 581 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 582 |
+
"rewards/chosen": -1.7299703359603882,
|
| 583 |
+
"rewards/margins": 0.8430070877075195,
|
| 584 |
+
"rewards/rejected": -2.5729775428771973,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 18.13700521728858,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": 0.24211910367012024,
|
| 592 |
+
"logits/rejected": 1.487108826637268,
|
| 593 |
+
"logps/chosen": -468.9916076660156,
|
| 594 |
+
"logps/rejected": -524.9107055664062,
|
| 595 |
+
"loss": 0.5477,
|
| 596 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 597 |
+
"rewards/chosen": -1.6977341175079346,
|
| 598 |
+
"rewards/margins": 1.0734937191009521,
|
| 599 |
+
"rewards/rejected": -2.7712278366088867,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 18.181148717866314,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": 0.9565867185592651,
|
| 607 |
+
"logits/rejected": 1.8202543258666992,
|
| 608 |
+
"logps/chosen": -449.505126953125,
|
| 609 |
+
"logps/rejected": -544.8054809570312,
|
| 610 |
+
"loss": 0.537,
|
| 611 |
+
"rewards/accuracies": 0.731249988079071,
|
| 612 |
+
"rewards/chosen": -1.8355414867401123,
|
| 613 |
+
"rewards/margins": 1.1922752857208252,
|
| 614 |
+
"rewards/rejected": -3.0278170108795166,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 22.57819934027554,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": 0.5353456139564514,
|
| 622 |
+
"logits/rejected": 1.7232787609100342,
|
| 623 |
+
"logps/chosen": -466.8003845214844,
|
| 624 |
+
"logps/rejected": -523.1851196289062,
|
| 625 |
+
"loss": 0.548,
|
| 626 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 627 |
+
"rewards/chosen": -1.7709314823150635,
|
| 628 |
+
"rewards/margins": 0.9179432988166809,
|
| 629 |
+
"rewards/rejected": -2.6888747215270996,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": 0.6372899413108826,
|
| 635 |
+
"eval_logits/rejected": 2.3928465843200684,
|
| 636 |
+
"eval_logps/chosen": -433.62774658203125,
|
| 637 |
+
"eval_logps/rejected": -518.405517578125,
|
| 638 |
+
"eval_loss": 0.5426377654075623,
|
| 639 |
+
"eval_rewards/accuracies": 0.7715517282485962,
|
| 640 |
+
"eval_rewards/chosen": -1.5802381038665771,
|
| 641 |
+
"eval_rewards/margins": 1.413055419921875,
|
| 642 |
+
"eval_rewards/rejected": -2.993293523788452,
|
| 643 |
+
"eval_runtime": 89.749,
|
| 644 |
+
"eval_samples_per_second": 20.401,
|
| 645 |
+
"eval_steps_per_second": 0.323,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 15.622675608360511,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": 0.3670940101146698,
|
| 653 |
+
"logits/rejected": 1.7792842388153076,
|
| 654 |
+
"logps/chosen": -446.5591735839844,
|
| 655 |
+
"logps/rejected": -531.0363159179688,
|
| 656 |
+
"loss": 0.5314,
|
| 657 |
+
"rewards/accuracies": 0.793749988079071,
|
| 658 |
+
"rewards/chosen": -1.6115410327911377,
|
| 659 |
+
"rewards/margins": 1.187956690788269,
|
| 660 |
+
"rewards/rejected": -2.799497604370117,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 19.069433933572203,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": 0.31656524538993835,
|
| 668 |
+
"logits/rejected": 1.853607416152954,
|
| 669 |
+
"logps/chosen": -419.215576171875,
|
| 670 |
+
"logps/rejected": -512.1829223632812,
|
| 671 |
+
"loss": 0.5406,
|
| 672 |
+
"rewards/accuracies": 0.78125,
|
| 673 |
+
"rewards/chosen": -1.481541395187378,
|
| 674 |
+
"rewards/margins": 1.2561315298080444,
|
| 675 |
+
"rewards/rejected": -2.737673282623291,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 23.53092447691508,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": 0.17315557599067688,
|
| 683 |
+
"logits/rejected": 1.6630465984344482,
|
| 684 |
+
"logps/chosen": -496.2525939941406,
|
| 685 |
+
"logps/rejected": -530.6498413085938,
|
| 686 |
+
"loss": 0.541,
|
| 687 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 688 |
+
"rewards/chosen": -1.7819627523422241,
|
| 689 |
+
"rewards/margins": 0.9505414962768555,
|
| 690 |
+
"rewards/rejected": -2.732504367828369,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 21.428741178294356,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": 0.45238789916038513,
|
| 698 |
+
"logits/rejected": 1.306971788406372,
|
| 699 |
+
"logps/chosen": -466.89324951171875,
|
| 700 |
+
"logps/rejected": -553.9127197265625,
|
| 701 |
+
"loss": 0.5468,
|
| 702 |
+
"rewards/accuracies": 0.71875,
|
| 703 |
+
"rewards/chosen": -1.5552027225494385,
|
| 704 |
+
"rewards/margins": 1.0361970663070679,
|
| 705 |
+
"rewards/rejected": -2.591399669647217,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 29.45545415193721,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": 0.8573619723320007,
|
| 713 |
+
"logits/rejected": 1.5074001550674438,
|
| 714 |
+
"logps/chosen": -456.8833923339844,
|
| 715 |
+
"logps/rejected": -535.3235473632812,
|
| 716 |
+
"loss": 0.5602,
|
| 717 |
+
"rewards/accuracies": 0.731249988079071,
|
| 718 |
+
"rewards/chosen": -1.781501054763794,
|
| 719 |
+
"rewards/margins": 0.9059982299804688,
|
| 720 |
+
"rewards/rejected": -2.6874990463256836,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": 0.7325018048286438,
|
| 726 |
+
"eval_logits/rejected": 2.4835281372070312,
|
| 727 |
+
"eval_logps/chosen": -434.1443176269531,
|
| 728 |
+
"eval_logps/rejected": -519.3194580078125,
|
| 729 |
+
"eval_loss": 0.5422444343566895,
|
| 730 |
+
"eval_rewards/accuracies": 0.767241358757019,
|
| 731 |
+
"eval_rewards/chosen": -1.585403323173523,
|
| 732 |
+
"eval_rewards/margins": 1.417029857635498,
|
| 733 |
+
"eval_rewards/rejected": -3.0024335384368896,
|
| 734 |
+
"eval_runtime": 90.5071,
|
| 735 |
+
"eval_samples_per_second": 20.23,
|
| 736 |
+
"eval_steps_per_second": 0.32,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 23.183003691416907,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": 0.6646889448165894,
|
| 744 |
+
"logits/rejected": 1.7684948444366455,
|
| 745 |
+
"logps/chosen": -474.71209716796875,
|
| 746 |
+
"logps/rejected": -539.7491455078125,
|
| 747 |
+
"loss": 0.5368,
|
| 748 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 749 |
+
"rewards/chosen": -1.8653221130371094,
|
| 750 |
+
"rewards/margins": 0.9348688125610352,
|
| 751 |
+
"rewards/rejected": -2.8001909255981445,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 20.16157572567649,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": 0.5646069645881653,
|
| 759 |
+
"logits/rejected": 2.1821658611297607,
|
| 760 |
+
"logps/chosen": -431.48663330078125,
|
| 761 |
+
"logps/rejected": -519.7050170898438,
|
| 762 |
+
"loss": 0.5427,
|
| 763 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 764 |
+
"rewards/chosen": -1.4873847961425781,
|
| 765 |
+
"rewards/margins": 1.369796633720398,
|
| 766 |
+
"rewards/rejected": -2.857182025909424,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 18.316684075654734,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": 0.780235230922699,
|
| 774 |
+
"logits/rejected": 1.9581851959228516,
|
| 775 |
+
"logps/chosen": -448.1625061035156,
|
| 776 |
+
"logps/rejected": -540.7398681640625,
|
| 777 |
+
"loss": 0.5411,
|
| 778 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 779 |
+
"rewards/chosen": -1.6593061685562134,
|
| 780 |
+
"rewards/margins": 1.1113975048065186,
|
| 781 |
+
"rewards/rejected": -2.7707035541534424,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5784502373134517,
|
| 789 |
+
"train_runtime": 11646.632,
|
| 790 |
+
"train_samples_per_second": 4.808,
|
| 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 |
+
}
|