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-gpt_consistent-high-curriculum
|
| 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-gpt_consistent-high-curriculum
|
| 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: 0.4943
|
| 21 |
+
- Rewards/chosen: -0.9906
|
| 22 |
+
- Rewards/rejected: -1.9162
|
| 23 |
+
- Rewards/accuracies: 0.7328
|
| 24 |
+
- Rewards/margins: 0.9257
|
| 25 |
+
- Logps/rejected: -438.1469
|
| 26 |
+
- Logps/chosen: -384.1485
|
| 27 |
+
- Logits/rejected: 1.5457
|
| 28 |
+
- Logits/chosen: 0.3365
|
| 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 |
+
| 0.6617 | 0.1147 | 50 | 0.6429 | -0.0319 | -0.1521 | 0.6983 | 0.1202 | -261.7313 | -288.2777 | -2.4839 | -2.5612 |
|
| 66 |
+
| 0.5671 | 0.2294 | 100 | 0.5741 | -0.6263 | -1.0517 | 0.6897 | 0.4254 | -351.6960 | -347.7197 | -1.7192 | -1.9024 |
|
| 67 |
+
| 0.5328 | 0.3440 | 150 | 0.5237 | -0.5529 | -1.2131 | 0.7155 | 0.6602 | -367.8349 | -340.3821 | 0.6797 | -0.1069 |
|
| 68 |
+
| 0.5339 | 0.4587 | 200 | 0.5135 | -0.8694 | -1.6358 | 0.7284 | 0.7664 | -410.0984 | -372.0303 | 1.1593 | 0.1333 |
|
| 69 |
+
| 0.5206 | 0.5734 | 250 | 0.5051 | -1.1478 | -2.0209 | 0.7457 | 0.8731 | -448.6093 | -399.8708 | 2.1252 | 1.0597 |
|
| 70 |
+
| 0.5161 | 0.6881 | 300 | 0.4995 | -1.0692 | -1.9303 | 0.7414 | 0.8611 | -439.5535 | -392.0123 | 1.6077 | 0.4690 |
|
| 71 |
+
| 0.5113 | 0.8028 | 350 | 0.4953 | -0.9905 | -1.8979 | 0.7284 | 0.9073 | -436.3081 | -384.1431 | 1.4599 | 0.2827 |
|
| 72 |
+
| 0.5006 | 0.9174 | 400 | 0.4943 | -0.9906 | -1.9162 | 0.7328 | 0.9257 | -438.1469 | -384.1485 | 1.5457 | 0.3365 |
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5444534902178914,
|
| 5 |
+
"train_runtime": 11753.731,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.744,
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5444534902178914,
|
| 5 |
+
"train_runtime": 11753.731,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.744,
|
| 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": 1.0,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 436,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022935779816513763,
|
| 13 |
+
"grad_norm": 8.114700597292016,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.630662441253662,
|
| 16 |
+
"logits/rejected": -2.588312864303589,
|
| 17 |
+
"logps/chosen": -251.37826538085938,
|
| 18 |
+
"logps/rejected": -245.56118774414062,
|
| 19 |
+
"loss": 0.6932,
|
| 20 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 21 |
+
"rewards/chosen": 0.0003000783617608249,
|
| 22 |
+
"rewards/margins": 3.4782169677782804e-05,
|
| 23 |
+
"rewards/rejected": 0.00026529619935899973,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 9.2292553026982,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6437935829162598,
|
| 31 |
+
"logits/rejected": -2.5945396423339844,
|
| 32 |
+
"logps/chosen": -305.3244323730469,
|
| 33 |
+
"logps/rejected": -288.41082763671875,
|
| 34 |
+
"loss": 0.692,
|
| 35 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 36 |
+
"rewards/chosen": 0.0019361503655090928,
|
| 37 |
+
"rewards/margins": 0.0025582597590982914,
|
| 38 |
+
"rewards/rejected": -0.0006221095682121813,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 8.362484362577634,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6338467597961426,
|
| 46 |
+
"logits/rejected": -2.5890305042266846,
|
| 47 |
+
"logps/chosen": -290.099365234375,
|
| 48 |
+
"logps/rejected": -311.11444091796875,
|
| 49 |
+
"loss": 0.6875,
|
| 50 |
+
"rewards/accuracies": 0.706250011920929,
|
| 51 |
+
"rewards/chosen": 0.012083572335541248,
|
| 52 |
+
"rewards/margins": 0.01316638570278883,
|
| 53 |
+
"rewards/rejected": -0.0010828140657395124,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 8.520101675940221,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6183691024780273,
|
| 61 |
+
"logits/rejected": -2.586637258529663,
|
| 62 |
+
"logps/chosen": -266.4432067871094,
|
| 63 |
+
"logps/rejected": -251.4285125732422,
|
| 64 |
+
"loss": 0.6773,
|
| 65 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 66 |
+
"rewards/chosen": 0.022834882140159607,
|
| 67 |
+
"rewards/margins": 0.023432452231645584,
|
| 68 |
+
"rewards/rejected": -0.0005975713720545173,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 9.078671253466315,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.6266233921051025,
|
| 76 |
+
"logits/rejected": -2.5520858764648438,
|
| 77 |
+
"logps/chosen": -304.1348571777344,
|
| 78 |
+
"logps/rejected": -262.32061767578125,
|
| 79 |
+
"loss": 0.6617,
|
| 80 |
+
"rewards/accuracies": 0.643750011920929,
|
| 81 |
+
"rewards/chosen": 0.002302716951817274,
|
| 82 |
+
"rewards/margins": 0.0754440575838089,
|
| 83 |
+
"rewards/rejected": -0.07314133644104004,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.561208486557007,
|
| 89 |
+
"eval_logits/rejected": -2.4839041233062744,
|
| 90 |
+
"eval_logps/chosen": -288.2777099609375,
|
| 91 |
+
"eval_logps/rejected": -261.73126220703125,
|
| 92 |
+
"eval_loss": 0.6429266333580017,
|
| 93 |
+
"eval_rewards/accuracies": 0.6982758641242981,
|
| 94 |
+
"eval_rewards/chosen": -0.031876176595687866,
|
| 95 |
+
"eval_rewards/margins": 0.12021197378635406,
|
| 96 |
+
"eval_rewards/rejected": -0.15208815038204193,
|
| 97 |
+
"eval_runtime": 95.5992,
|
| 98 |
+
"eval_samples_per_second": 19.017,
|
| 99 |
+
"eval_steps_per_second": 0.303,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 11.596073380531864,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.547043561935425,
|
| 107 |
+
"logits/rejected": -2.4718384742736816,
|
| 108 |
+
"logps/chosen": -296.81402587890625,
|
| 109 |
+
"logps/rejected": -309.17828369140625,
|
| 110 |
+
"loss": 0.6408,
|
| 111 |
+
"rewards/accuracies": 0.625,
|
| 112 |
+
"rewards/chosen": -0.10427121073007584,
|
| 113 |
+
"rewards/margins": 0.10880544036626816,
|
| 114 |
+
"rewards/rejected": -0.2130766659975052,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 14.345685973653804,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.57369065284729,
|
| 122 |
+
"logits/rejected": -2.5374178886413574,
|
| 123 |
+
"logps/chosen": -286.39215087890625,
|
| 124 |
+
"logps/rejected": -286.45330810546875,
|
| 125 |
+
"loss": 0.6181,
|
| 126 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 127 |
+
"rewards/chosen": -0.11706429719924927,
|
| 128 |
+
"rewards/margins": 0.20492573082447052,
|
| 129 |
+
"rewards/rejected": -0.321990042924881,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 15.011490045096238,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.5018064975738525,
|
| 137 |
+
"logits/rejected": -2.4190878868103027,
|
| 138 |
+
"logps/chosen": -328.02960205078125,
|
| 139 |
+
"logps/rejected": -310.5951843261719,
|
| 140 |
+
"loss": 0.6089,
|
| 141 |
+
"rewards/accuracies": 0.71875,
|
| 142 |
+
"rewards/chosen": -0.22072939574718475,
|
| 143 |
+
"rewards/margins": 0.22173993289470673,
|
| 144 |
+
"rewards/rejected": -0.44246941804885864,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 12.064286648165384,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -2.316053628921509,
|
| 152 |
+
"logits/rejected": -2.2170588970184326,
|
| 153 |
+
"logps/chosen": -309.4535217285156,
|
| 154 |
+
"logps/rejected": -293.6477966308594,
|
| 155 |
+
"loss": 0.6033,
|
| 156 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 157 |
+
"rewards/chosen": -0.34011608362197876,
|
| 158 |
+
"rewards/margins": 0.24611127376556396,
|
| 159 |
+
"rewards/rejected": -0.5862273573875427,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 18.008124319426113,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -2.297001838684082,
|
| 167 |
+
"logits/rejected": -2.1643242835998535,
|
| 168 |
+
"logps/chosen": -329.64617919921875,
|
| 169 |
+
"logps/rejected": -325.0845947265625,
|
| 170 |
+
"loss": 0.5671,
|
| 171 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 172 |
+
"rewards/chosen": -0.3444996774196625,
|
| 173 |
+
"rewards/margins": 0.4821701645851135,
|
| 174 |
+
"rewards/rejected": -0.8266698718070984,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -1.9024137258529663,
|
| 180 |
+
"eval_logits/rejected": -1.719163417816162,
|
| 181 |
+
"eval_logps/chosen": -347.71966552734375,
|
| 182 |
+
"eval_logps/rejected": -351.6959533691406,
|
| 183 |
+
"eval_loss": 0.5741031765937805,
|
| 184 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 185 |
+
"eval_rewards/chosen": -0.6262954473495483,
|
| 186 |
+
"eval_rewards/margins": 0.4254392087459564,
|
| 187 |
+
"eval_rewards/rejected": -1.0517346858978271,
|
| 188 |
+
"eval_runtime": 94.9312,
|
| 189 |
+
"eval_samples_per_second": 19.151,
|
| 190 |
+
"eval_steps_per_second": 0.305,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 18.209810501109782,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -1.37660813331604,
|
| 198 |
+
"logits/rejected": -1.3109476566314697,
|
| 199 |
+
"logps/chosen": -306.15673828125,
|
| 200 |
+
"logps/rejected": -364.8440856933594,
|
| 201 |
+
"loss": 0.5638,
|
| 202 |
+
"rewards/accuracies": 0.65625,
|
| 203 |
+
"rewards/chosen": -0.5968034267425537,
|
| 204 |
+
"rewards/margins": 0.4451308250427246,
|
| 205 |
+
"rewards/rejected": -1.0419342517852783,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 30.619035412465948,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": -0.5667544007301331,
|
| 213 |
+
"logits/rejected": -0.11496512591838837,
|
| 214 |
+
"logps/chosen": -352.6363830566406,
|
| 215 |
+
"logps/rejected": -321.98138427734375,
|
| 216 |
+
"loss": 0.5429,
|
| 217 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 218 |
+
"rewards/chosen": -0.5322494506835938,
|
| 219 |
+
"rewards/margins": 0.4666845202445984,
|
| 220 |
+
"rewards/rejected": -0.9989339709281921,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 25.16692077643166,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": 0.31375107169151306,
|
| 228 |
+
"logits/rejected": 0.7534220814704895,
|
| 229 |
+
"logps/chosen": -333.3154602050781,
|
| 230 |
+
"logps/rejected": -383.91571044921875,
|
| 231 |
+
"loss": 0.5276,
|
| 232 |
+
"rewards/accuracies": 0.706250011920929,
|
| 233 |
+
"rewards/chosen": -0.756668746471405,
|
| 234 |
+
"rewards/margins": 0.7230855226516724,
|
| 235 |
+
"rewards/rejected": -1.479754090309143,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 21.8628335306802,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": 0.12007780373096466,
|
| 243 |
+
"logits/rejected": 0.32281678915023804,
|
| 244 |
+
"logps/chosen": -341.0147705078125,
|
| 245 |
+
"logps/rejected": -368.9652404785156,
|
| 246 |
+
"loss": 0.5337,
|
| 247 |
+
"rewards/accuracies": 0.71875,
|
| 248 |
+
"rewards/chosen": -0.6602333188056946,
|
| 249 |
+
"rewards/margins": 0.6314884424209595,
|
| 250 |
+
"rewards/rejected": -1.2917217016220093,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 26.496344548145927,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": 0.6530567407608032,
|
| 258 |
+
"logits/rejected": 1.3690438270568848,
|
| 259 |
+
"logps/chosen": -375.08856201171875,
|
| 260 |
+
"logps/rejected": -408.30181884765625,
|
| 261 |
+
"loss": 0.5328,
|
| 262 |
+
"rewards/accuracies": 0.731249988079071,
|
| 263 |
+
"rewards/chosen": -0.8673305511474609,
|
| 264 |
+
"rewards/margins": 0.7217624187469482,
|
| 265 |
+
"rewards/rejected": -1.5890929698944092,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": -0.10693416744470596,
|
| 271 |
+
"eval_logits/rejected": 0.6796554327011108,
|
| 272 |
+
"eval_logps/chosen": -340.38214111328125,
|
| 273 |
+
"eval_logps/rejected": -367.83489990234375,
|
| 274 |
+
"eval_loss": 0.5237244367599487,
|
| 275 |
+
"eval_rewards/accuracies": 0.7155172228813171,
|
| 276 |
+
"eval_rewards/chosen": -0.5529204607009888,
|
| 277 |
+
"eval_rewards/margins": 0.6602039933204651,
|
| 278 |
+
"eval_rewards/rejected": -1.2131245136260986,
|
| 279 |
+
"eval_runtime": 95.5215,
|
| 280 |
+
"eval_samples_per_second": 19.032,
|
| 281 |
+
"eval_steps_per_second": 0.304,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 22.090085860189564,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": -0.6242966651916504,
|
| 289 |
+
"logits/rejected": 0.1017322987318039,
|
| 290 |
+
"logps/chosen": -337.90997314453125,
|
| 291 |
+
"logps/rejected": -389.50592041015625,
|
| 292 |
+
"loss": 0.5376,
|
| 293 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 294 |
+
"rewards/chosen": -0.4958871901035309,
|
| 295 |
+
"rewards/margins": 0.6162235140800476,
|
| 296 |
+
"rewards/rejected": -1.1121107339859009,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 29.49090849875655,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": 0.3189530670642853,
|
| 304 |
+
"logits/rejected": 0.7786465287208557,
|
| 305 |
+
"logps/chosen": -349.08013916015625,
|
| 306 |
+
"logps/rejected": -422.4642639160156,
|
| 307 |
+
"loss": 0.5223,
|
| 308 |
+
"rewards/accuracies": 0.6875,
|
| 309 |
+
"rewards/chosen": -0.786384642124176,
|
| 310 |
+
"rewards/margins": 0.660692036151886,
|
| 311 |
+
"rewards/rejected": -1.4470767974853516,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 21.619610594454734,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": 0.5451809167861938,
|
| 319 |
+
"logits/rejected": 1.3867518901824951,
|
| 320 |
+
"logps/chosen": -417.93450927734375,
|
| 321 |
+
"logps/rejected": -453.64581298828125,
|
| 322 |
+
"loss": 0.5262,
|
| 323 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 324 |
+
"rewards/chosen": -1.0858782529830933,
|
| 325 |
+
"rewards/margins": 0.799088180065155,
|
| 326 |
+
"rewards/rejected": -1.8849666118621826,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 22.879012626203956,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": 0.8325613141059875,
|
| 334 |
+
"logits/rejected": 1.527114748954773,
|
| 335 |
+
"logps/chosen": -372.27227783203125,
|
| 336 |
+
"logps/rejected": -433.6874084472656,
|
| 337 |
+
"loss": 0.5396,
|
| 338 |
+
"rewards/accuracies": 0.71875,
|
| 339 |
+
"rewards/chosen": -0.9668887853622437,
|
| 340 |
+
"rewards/margins": 0.787070095539093,
|
| 341 |
+
"rewards/rejected": -1.7539589405059814,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 19.71033831580365,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": 0.25724777579307556,
|
| 349 |
+
"logits/rejected": 1.0539257526397705,
|
| 350 |
+
"logps/chosen": -339.7835388183594,
|
| 351 |
+
"logps/rejected": -395.4685363769531,
|
| 352 |
+
"loss": 0.5339,
|
| 353 |
+
"rewards/accuracies": 0.75,
|
| 354 |
+
"rewards/chosen": -0.7703949809074402,
|
| 355 |
+
"rewards/margins": 0.7419241070747375,
|
| 356 |
+
"rewards/rejected": -1.5123189687728882,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": 0.1333327293395996,
|
| 362 |
+
"eval_logits/rejected": 1.1592669486999512,
|
| 363 |
+
"eval_logps/chosen": -372.03033447265625,
|
| 364 |
+
"eval_logps/rejected": -410.098388671875,
|
| 365 |
+
"eval_loss": 0.5135313272476196,
|
| 366 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 367 |
+
"eval_rewards/chosen": -0.8694021701812744,
|
| 368 |
+
"eval_rewards/margins": 0.7663572430610657,
|
| 369 |
+
"eval_rewards/rejected": -1.6357594728469849,
|
| 370 |
+
"eval_runtime": 94.9597,
|
| 371 |
+
"eval_samples_per_second": 19.145,
|
| 372 |
+
"eval_steps_per_second": 0.305,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 24.166340722703165,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": -0.02151186764240265,
|
| 380 |
+
"logits/rejected": 0.7990398406982422,
|
| 381 |
+
"logps/chosen": -386.0631408691406,
|
| 382 |
+
"logps/rejected": -439.3416442871094,
|
| 383 |
+
"loss": 0.5165,
|
| 384 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 385 |
+
"rewards/chosen": -0.859545111656189,
|
| 386 |
+
"rewards/margins": 0.791865348815918,
|
| 387 |
+
"rewards/rejected": -1.651410460472107,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 29.831111320846208,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": 0.28069013357162476,
|
| 395 |
+
"logits/rejected": 0.9275471568107605,
|
| 396 |
+
"logps/chosen": -352.3778076171875,
|
| 397 |
+
"logps/rejected": -410.13238525390625,
|
| 398 |
+
"loss": 0.5119,
|
| 399 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 400 |
+
"rewards/chosen": -0.7737405300140381,
|
| 401 |
+
"rewards/margins": 0.7621678113937378,
|
| 402 |
+
"rewards/rejected": -1.5359084606170654,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 27.12612279862232,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": 0.31911998987197876,
|
| 410 |
+
"logits/rejected": 1.4982731342315674,
|
| 411 |
+
"logps/chosen": -380.7137145996094,
|
| 412 |
+
"logps/rejected": -428.25030517578125,
|
| 413 |
+
"loss": 0.5326,
|
| 414 |
+
"rewards/accuracies": 0.78125,
|
| 415 |
+
"rewards/chosen": -0.8266562223434448,
|
| 416 |
+
"rewards/margins": 0.8594551086425781,
|
| 417 |
+
"rewards/rejected": -1.6861114501953125,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 30.270434326846676,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": 0.08341093361377716,
|
| 425 |
+
"logits/rejected": 0.8582345843315125,
|
| 426 |
+
"logps/chosen": -345.74591064453125,
|
| 427 |
+
"logps/rejected": -403.52093505859375,
|
| 428 |
+
"loss": 0.5195,
|
| 429 |
+
"rewards/accuracies": 0.699999988079071,
|
| 430 |
+
"rewards/chosen": -0.8471673727035522,
|
| 431 |
+
"rewards/margins": 0.6442986726760864,
|
| 432 |
+
"rewards/rejected": -1.4914662837982178,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 24.62506692170475,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": 0.6792846918106079,
|
| 440 |
+
"logits/rejected": 1.6594598293304443,
|
| 441 |
+
"logps/chosen": -357.1456298828125,
|
| 442 |
+
"logps/rejected": -407.39794921875,
|
| 443 |
+
"loss": 0.5206,
|
| 444 |
+
"rewards/accuracies": 0.706250011920929,
|
| 445 |
+
"rewards/chosen": -0.9902165532112122,
|
| 446 |
+
"rewards/margins": 0.7551018595695496,
|
| 447 |
+
"rewards/rejected": -1.7453181743621826,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": 1.0596588850021362,
|
| 453 |
+
"eval_logits/rejected": 2.125216007232666,
|
| 454 |
+
"eval_logps/chosen": -399.8708190917969,
|
| 455 |
+
"eval_logps/rejected": -448.6092529296875,
|
| 456 |
+
"eval_loss": 0.5051079988479614,
|
| 457 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 458 |
+
"eval_rewards/chosen": -1.1478071212768555,
|
| 459 |
+
"eval_rewards/margins": 0.8730602860450745,
|
| 460 |
+
"eval_rewards/rejected": -2.020867347717285,
|
| 461 |
+
"eval_runtime": 95.375,
|
| 462 |
+
"eval_samples_per_second": 19.062,
|
| 463 |
+
"eval_steps_per_second": 0.304,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 25.512119180845332,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": 1.0334103107452393,
|
| 471 |
+
"logits/rejected": 1.9036356210708618,
|
| 472 |
+
"logps/chosen": -393.27923583984375,
|
| 473 |
+
"logps/rejected": -442.31634521484375,
|
| 474 |
+
"loss": 0.5225,
|
| 475 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 476 |
+
"rewards/chosen": -1.1329892873764038,
|
| 477 |
+
"rewards/margins": 0.6801810264587402,
|
| 478 |
+
"rewards/rejected": -1.8131701946258545,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 22.529726863175615,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": 0.06802092492580414,
|
| 486 |
+
"logits/rejected": 0.6059505343437195,
|
| 487 |
+
"logps/chosen": -379.96771240234375,
|
| 488 |
+
"logps/rejected": -420.07244873046875,
|
| 489 |
+
"loss": 0.5157,
|
| 490 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 491 |
+
"rewards/chosen": -0.902859091758728,
|
| 492 |
+
"rewards/margins": 0.6471339464187622,
|
| 493 |
+
"rewards/rejected": -1.5499929189682007,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 29.61705869960123,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": 0.36484724283218384,
|
| 501 |
+
"logits/rejected": 1.1159727573394775,
|
| 502 |
+
"logps/chosen": -378.2865905761719,
|
| 503 |
+
"logps/rejected": -446.976318359375,
|
| 504 |
+
"loss": 0.5034,
|
| 505 |
+
"rewards/accuracies": 0.6875,
|
| 506 |
+
"rewards/chosen": -0.9187177419662476,
|
| 507 |
+
"rewards/margins": 0.7557224035263062,
|
| 508 |
+
"rewards/rejected": -1.6744401454925537,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 32.55159666302868,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": 0.4420185089111328,
|
| 516 |
+
"logits/rejected": 1.4995023012161255,
|
| 517 |
+
"logps/chosen": -382.71185302734375,
|
| 518 |
+
"logps/rejected": -434.3760681152344,
|
| 519 |
+
"loss": 0.5136,
|
| 520 |
+
"rewards/accuracies": 0.75,
|
| 521 |
+
"rewards/chosen": -0.9455305933952332,
|
| 522 |
+
"rewards/margins": 0.9165040254592896,
|
| 523 |
+
"rewards/rejected": -1.862034559249878,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 23.09866796839295,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": 0.40113481879234314,
|
| 531 |
+
"logits/rejected": 1.5598547458648682,
|
| 532 |
+
"logps/chosen": -423.24615478515625,
|
| 533 |
+
"logps/rejected": -433.6683044433594,
|
| 534 |
+
"loss": 0.5161,
|
| 535 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 536 |
+
"rewards/chosen": -1.1671854257583618,
|
| 537 |
+
"rewards/margins": 0.6783391237258911,
|
| 538 |
+
"rewards/rejected": -1.845524549484253,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": 0.468977689743042,
|
| 544 |
+
"eval_logits/rejected": 1.6077337265014648,
|
| 545 |
+
"eval_logps/chosen": -392.01226806640625,
|
| 546 |
+
"eval_logps/rejected": -439.55352783203125,
|
| 547 |
+
"eval_loss": 0.4995412528514862,
|
| 548 |
+
"eval_rewards/accuracies": 0.7413793206214905,
|
| 549 |
+
"eval_rewards/chosen": -1.0692216157913208,
|
| 550 |
+
"eval_rewards/margins": 0.8610891103744507,
|
| 551 |
+
"eval_rewards/rejected": -1.9303104877471924,
|
| 552 |
+
"eval_runtime": 95.9852,
|
| 553 |
+
"eval_samples_per_second": 18.94,
|
| 554 |
+
"eval_steps_per_second": 0.302,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 23.93680772701649,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": 0.4794091284275055,
|
| 562 |
+
"logits/rejected": 1.1960773468017578,
|
| 563 |
+
"logps/chosen": -373.32708740234375,
|
| 564 |
+
"logps/rejected": -457.6255798339844,
|
| 565 |
+
"loss": 0.4872,
|
| 566 |
+
"rewards/accuracies": 0.768750011920929,
|
| 567 |
+
"rewards/chosen": -0.9246217012405396,
|
| 568 |
+
"rewards/margins": 0.8957231640815735,
|
| 569 |
+
"rewards/rejected": -1.8203446865081787,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 33.75818947226454,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": 0.50605309009552,
|
| 577 |
+
"logits/rejected": 1.3813035488128662,
|
| 578 |
+
"logps/chosen": -395.261474609375,
|
| 579 |
+
"logps/rejected": -462.6114196777344,
|
| 580 |
+
"loss": 0.4904,
|
| 581 |
+
"rewards/accuracies": 0.75,
|
| 582 |
+
"rewards/chosen": -1.1317824125289917,
|
| 583 |
+
"rewards/margins": 0.8041917085647583,
|
| 584 |
+
"rewards/rejected": -1.93597412109375,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 26.64078079133333,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": 0.7998399138450623,
|
| 592 |
+
"logits/rejected": 1.5950431823730469,
|
| 593 |
+
"logps/chosen": -358.0443420410156,
|
| 594 |
+
"logps/rejected": -450.14898681640625,
|
| 595 |
+
"loss": 0.4977,
|
| 596 |
+
"rewards/accuracies": 0.731249988079071,
|
| 597 |
+
"rewards/chosen": -1.0693919658660889,
|
| 598 |
+
"rewards/margins": 0.917527973651886,
|
| 599 |
+
"rewards/rejected": -1.9869201183319092,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 25.809036872331053,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": 0.4978507161140442,
|
| 607 |
+
"logits/rejected": 1.611681580543518,
|
| 608 |
+
"logps/chosen": -386.6365661621094,
|
| 609 |
+
"logps/rejected": -486.8155212402344,
|
| 610 |
+
"loss": 0.4861,
|
| 611 |
+
"rewards/accuracies": 0.800000011920929,
|
| 612 |
+
"rewards/chosen": -0.9396215677261353,
|
| 613 |
+
"rewards/margins": 1.1603189706802368,
|
| 614 |
+
"rewards/rejected": -2.099940538406372,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 25.432021542271517,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": 0.37401479482650757,
|
| 622 |
+
"logits/rejected": 1.426959753036499,
|
| 623 |
+
"logps/chosen": -393.84173583984375,
|
| 624 |
+
"logps/rejected": -441.216796875,
|
| 625 |
+
"loss": 0.5113,
|
| 626 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 627 |
+
"rewards/chosen": -1.0162795782089233,
|
| 628 |
+
"rewards/margins": 0.8663312792778015,
|
| 629 |
+
"rewards/rejected": -1.8826109170913696,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": 0.28273966908454895,
|
| 635 |
+
"eval_logits/rejected": 1.4599332809448242,
|
| 636 |
+
"eval_logps/chosen": -384.1430969238281,
|
| 637 |
+
"eval_logps/rejected": -436.3081359863281,
|
| 638 |
+
"eval_loss": 0.4952593147754669,
|
| 639 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 640 |
+
"eval_rewards/chosen": -0.9905301332473755,
|
| 641 |
+
"eval_rewards/margins": 0.9073269963264465,
|
| 642 |
+
"eval_rewards/rejected": -1.8978571891784668,
|
| 643 |
+
"eval_runtime": 98.0468,
|
| 644 |
+
"eval_samples_per_second": 18.542,
|
| 645 |
+
"eval_steps_per_second": 0.296,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 24.878023970166044,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": 0.011499330401420593,
|
| 653 |
+
"logits/rejected": 1.0976569652557373,
|
| 654 |
+
"logps/chosen": -381.95184326171875,
|
| 655 |
+
"logps/rejected": -447.99334716796875,
|
| 656 |
+
"loss": 0.4998,
|
| 657 |
+
"rewards/accuracies": 0.78125,
|
| 658 |
+
"rewards/chosen": -0.8815473318099976,
|
| 659 |
+
"rewards/margins": 0.9280586242675781,
|
| 660 |
+
"rewards/rejected": -1.8096059560775757,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 27.48602383662752,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": 0.2708785831928253,
|
| 668 |
+
"logits/rejected": 0.9552985429763794,
|
| 669 |
+
"logps/chosen": -397.5840148925781,
|
| 670 |
+
"logps/rejected": -466.92999267578125,
|
| 671 |
+
"loss": 0.4821,
|
| 672 |
+
"rewards/accuracies": 0.78125,
|
| 673 |
+
"rewards/chosen": -0.9426911473274231,
|
| 674 |
+
"rewards/margins": 0.8755933046340942,
|
| 675 |
+
"rewards/rejected": -1.8182843923568726,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 22.77493294832679,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": 0.3190244436264038,
|
| 683 |
+
"logits/rejected": 1.6189016103744507,
|
| 684 |
+
"logps/chosen": -393.47637939453125,
|
| 685 |
+
"logps/rejected": -460.20428466796875,
|
| 686 |
+
"loss": 0.5039,
|
| 687 |
+
"rewards/accuracies": 0.78125,
|
| 688 |
+
"rewards/chosen": -1.0399200916290283,
|
| 689 |
+
"rewards/margins": 0.9450514912605286,
|
| 690 |
+
"rewards/rejected": -1.9849714040756226,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 21.22789089972862,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": 0.4111763834953308,
|
| 698 |
+
"logits/rejected": 1.2173652648925781,
|
| 699 |
+
"logps/chosen": -375.0528869628906,
|
| 700 |
+
"logps/rejected": -447.7088317871094,
|
| 701 |
+
"loss": 0.5199,
|
| 702 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 703 |
+
"rewards/chosen": -0.9053140878677368,
|
| 704 |
+
"rewards/margins": 0.9649880528450012,
|
| 705 |
+
"rewards/rejected": -1.8703022003173828,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 25.945198405726252,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": 0.3489355742931366,
|
| 713 |
+
"logits/rejected": 1.733727216720581,
|
| 714 |
+
"logps/chosen": -385.9686584472656,
|
| 715 |
+
"logps/rejected": -407.83514404296875,
|
| 716 |
+
"loss": 0.5006,
|
| 717 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 718 |
+
"rewards/chosen": -0.9153203964233398,
|
| 719 |
+
"rewards/margins": 0.9544361233711243,
|
| 720 |
+
"rewards/rejected": -1.8697564601898193,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": 0.33653396368026733,
|
| 726 |
+
"eval_logits/rejected": 1.545719861984253,
|
| 727 |
+
"eval_logps/chosen": -384.1485290527344,
|
| 728 |
+
"eval_logps/rejected": -438.1469421386719,
|
| 729 |
+
"eval_loss": 0.49427667260169983,
|
| 730 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 731 |
+
"eval_rewards/chosen": -0.9905844330787659,
|
| 732 |
+
"eval_rewards/margins": 0.9256603717803955,
|
| 733 |
+
"eval_rewards/rejected": -1.9162448644638062,
|
| 734 |
+
"eval_runtime": 95.1799,
|
| 735 |
+
"eval_samples_per_second": 19.101,
|
| 736 |
+
"eval_steps_per_second": 0.305,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 29.414671515939542,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": 0.3652718663215637,
|
| 744 |
+
"logits/rejected": 1.4336962699890137,
|
| 745 |
+
"logps/chosen": -407.9559326171875,
|
| 746 |
+
"logps/rejected": -456.4903259277344,
|
| 747 |
+
"loss": 0.486,
|
| 748 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 749 |
+
"rewards/chosen": -0.9434130787849426,
|
| 750 |
+
"rewards/margins": 0.8941701650619507,
|
| 751 |
+
"rewards/rejected": -1.837583303451538,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 23.283884603445415,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": 0.2882634997367859,
|
| 759 |
+
"logits/rejected": 1.3258743286132812,
|
| 760 |
+
"logps/chosen": -375.098876953125,
|
| 761 |
+
"logps/rejected": -462.24639892578125,
|
| 762 |
+
"loss": 0.4823,
|
| 763 |
+
"rewards/accuracies": 0.731249988079071,
|
| 764 |
+
"rewards/chosen": -1.0388104915618896,
|
| 765 |
+
"rewards/margins": 0.9280673861503601,
|
| 766 |
+
"rewards/rejected": -1.9668779373168945,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 32.32447531053772,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": 0.6452657580375671,
|
| 774 |
+
"logits/rejected": 1.3607791662216187,
|
| 775 |
+
"logps/chosen": -378.0513916015625,
|
| 776 |
+
"logps/rejected": -468.52301025390625,
|
| 777 |
+
"loss": 0.504,
|
| 778 |
+
"rewards/accuracies": 0.75,
|
| 779 |
+
"rewards/chosen": -1.1194813251495361,
|
| 780 |
+
"rewards/margins": 0.9433539509773254,
|
| 781 |
+
"rewards/rejected": -2.062835216522217,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5444534902178914,
|
| 789 |
+
"train_runtime": 11753.731,
|
| 790 |
+
"train_samples_per_second": 4.744,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 436,
|
| 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 |
+
}
|