Model save
Browse files- README.md +76 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +751 -0
README.md
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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-3
|
| 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-3
|
| 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: 0.5175
|
| 21 |
+
- Rewards/chosen: -0.8697
|
| 22 |
+
- Rewards/rejected: -1.8604
|
| 23 |
+
- Rewards/accuracies: 0.7716
|
| 24 |
+
- Rewards/margins: 0.9907
|
| 25 |
+
- Logps/rejected: -405.1198
|
| 26 |
+
- Logps/chosen: -362.5746
|
| 27 |
+
- Logits/rejected: 1.7998
|
| 28 |
+
- Logits/chosen: 0.2407
|
| 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 |
+
| 0.5982 | 0.2286 | 100 | 0.5954 | -0.1704 | -0.5853 | 0.7155 | 0.4149 | -277.6054 | -292.6412 | -2.2902 | -2.4282 |
|
| 66 |
+
| 0.5439 | 0.4571 | 200 | 0.5353 | -0.8124 | -1.6747 | 0.7586 | 0.8623 | -386.5413 | -356.8392 | -0.0415 | -1.0380 |
|
| 67 |
+
| 0.5109 | 0.6857 | 300 | 0.5190 | -0.6985 | -1.6462 | 0.7716 | 0.9478 | -383.7008 | -345.4529 | 1.1016 | -0.3942 |
|
| 68 |
+
| 0.5457 | 0.9143 | 400 | 0.5175 | -0.8697 | -1.8604 | 0.7716 | 0.9907 | -405.1198 | -362.5746 | 1.7998 | 0.2407 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- Transformers 4.44.0.dev0
|
| 74 |
+
- Pytorch 2.1.2
|
| 75 |
+
- Datasets 2.20.0
|
| 76 |
+
- 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": 0.5586249282758351,
|
| 5 |
+
"train_runtime": 11006.8714,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 5.088,
|
| 8 |
+
"train_steps_per_second": 0.04
|
| 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": 0.5586249282758351,
|
| 5 |
+
"train_runtime": 11006.8714,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 5.088,
|
| 8 |
+
"train_steps_per_second": 0.04
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,751 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9988571428571429,
|
| 5 |
+
"eval_steps": 100,
|
| 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": 9.822878263417993,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.700991630554199,
|
| 16 |
+
"logits/rejected": -2.625051975250244,
|
| 17 |
+
"logps/chosen": -301.26373291015625,
|
| 18 |
+
"logps/rejected": -281.7487487792969,
|
| 19 |
+
"loss": 0.693,
|
| 20 |
+
"rewards/accuracies": 0.40625,
|
| 21 |
+
"rewards/chosen": 0.00022422037727665156,
|
| 22 |
+
"rewards/margins": 0.0002149858046323061,
|
| 23 |
+
"rewards/rejected": 9.234552635462023e-06,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 7.790697115653487,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6405527591705322,
|
| 31 |
+
"logits/rejected": -2.6053428649902344,
|
| 32 |
+
"logps/chosen": -278.93023681640625,
|
| 33 |
+
"logps/rejected": -254.70883178710938,
|
| 34 |
+
"loss": 0.6923,
|
| 35 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 36 |
+
"rewards/chosen": 0.002409698674455285,
|
| 37 |
+
"rewards/margins": 0.0018069592770189047,
|
| 38 |
+
"rewards/rejected": 0.0006027388735674322,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 8.37723659427994,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.637849807739258,
|
| 46 |
+
"logits/rejected": -2.616870403289795,
|
| 47 |
+
"logps/chosen": -263.4436950683594,
|
| 48 |
+
"logps/rejected": -263.60205078125,
|
| 49 |
+
"loss": 0.6871,
|
| 50 |
+
"rewards/accuracies": 0.612500011920929,
|
| 51 |
+
"rewards/chosen": 0.012393913231790066,
|
| 52 |
+
"rewards/margins": 0.00848361849784851,
|
| 53 |
+
"rewards/rejected": 0.003910294268280268,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 9.39434892949384,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.64766263961792,
|
| 61 |
+
"logits/rejected": -2.585386276245117,
|
| 62 |
+
"logps/chosen": -290.52587890625,
|
| 63 |
+
"logps/rejected": -268.34405517578125,
|
| 64 |
+
"loss": 0.6756,
|
| 65 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 66 |
+
"rewards/chosen": 0.03471367806196213,
|
| 67 |
+
"rewards/margins": 0.041840970516204834,
|
| 68 |
+
"rewards/rejected": -0.007127293851226568,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 13.922334228887747,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.676335573196411,
|
| 76 |
+
"logits/rejected": -2.5971837043762207,
|
| 77 |
+
"logps/chosen": -294.0897216796875,
|
| 78 |
+
"logps/rejected": -253.9338836669922,
|
| 79 |
+
"loss": 0.6628,
|
| 80 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 81 |
+
"rewards/chosen": 0.022205885499715805,
|
| 82 |
+
"rewards/margins": 0.09290142357349396,
|
| 83 |
+
"rewards/rejected": -0.07069554179906845,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.13714285714285715,
|
| 88 |
+
"grad_norm": 11.361649612146103,
|
| 89 |
+
"learning_rate": 4.979579212164186e-07,
|
| 90 |
+
"logits/chosen": -2.5759811401367188,
|
| 91 |
+
"logits/rejected": -2.472777843475342,
|
| 92 |
+
"logps/chosen": -290.1505432128906,
|
| 93 |
+
"logps/rejected": -270.30242919921875,
|
| 94 |
+
"loss": 0.6374,
|
| 95 |
+
"rewards/accuracies": 0.65625,
|
| 96 |
+
"rewards/chosen": -0.09644584357738495,
|
| 97 |
+
"rewards/margins": 0.1210569366812706,
|
| 98 |
+
"rewards/rejected": -0.21750275790691376,
|
| 99 |
+
"step": 60
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.16,
|
| 103 |
+
"grad_norm": 14.922787112175543,
|
| 104 |
+
"learning_rate": 4.946196886175515e-07,
|
| 105 |
+
"logits/chosen": -2.5715270042419434,
|
| 106 |
+
"logits/rejected": -2.5131115913391113,
|
| 107 |
+
"logps/chosen": -284.8800964355469,
|
| 108 |
+
"logps/rejected": -291.2018737792969,
|
| 109 |
+
"loss": 0.6102,
|
| 110 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 111 |
+
"rewards/chosen": -0.09101851284503937,
|
| 112 |
+
"rewards/margins": 0.22473247349262238,
|
| 113 |
+
"rewards/rejected": -0.31575098633766174,
|
| 114 |
+
"step": 70
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.18285714285714286,
|
| 118 |
+
"grad_norm": 11.602482470889782,
|
| 119 |
+
"learning_rate": 4.897191188239667e-07,
|
| 120 |
+
"logits/chosen": -2.5607805252075195,
|
| 121 |
+
"logits/rejected": -2.50270938873291,
|
| 122 |
+
"logps/chosen": -291.697998046875,
|
| 123 |
+
"logps/rejected": -307.26983642578125,
|
| 124 |
+
"loss": 0.6078,
|
| 125 |
+
"rewards/accuracies": 0.706250011920929,
|
| 126 |
+
"rewards/chosen": -0.2466896027326584,
|
| 127 |
+
"rewards/margins": 0.27291998267173767,
|
| 128 |
+
"rewards/rejected": -0.5196096301078796,
|
| 129 |
+
"step": 80
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.2057142857142857,
|
| 133 |
+
"grad_norm": 17.533063770924706,
|
| 134 |
+
"learning_rate": 4.832875107981763e-07,
|
| 135 |
+
"logits/chosen": -2.6349377632141113,
|
| 136 |
+
"logits/rejected": -2.5697710514068604,
|
| 137 |
+
"logps/chosen": -295.2223815917969,
|
| 138 |
+
"logps/rejected": -310.2930603027344,
|
| 139 |
+
"loss": 0.6076,
|
| 140 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 141 |
+
"rewards/chosen": -0.18937139213085175,
|
| 142 |
+
"rewards/margins": 0.3580796718597412,
|
| 143 |
+
"rewards/rejected": -0.5474510192871094,
|
| 144 |
+
"step": 90
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.22857142857142856,
|
| 148 |
+
"grad_norm": 16.194420237420413,
|
| 149 |
+
"learning_rate": 4.753659419387223e-07,
|
| 150 |
+
"logits/chosen": -2.6082775592803955,
|
| 151 |
+
"logits/rejected": -2.501986026763916,
|
| 152 |
+
"logps/chosen": -317.41748046875,
|
| 153 |
+
"logps/rejected": -301.41156005859375,
|
| 154 |
+
"loss": 0.5982,
|
| 155 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 156 |
+
"rewards/chosen": -0.31031134724617004,
|
| 157 |
+
"rewards/margins": 0.3692900538444519,
|
| 158 |
+
"rewards/rejected": -0.6796014904975891,
|
| 159 |
+
"step": 100
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.22857142857142856,
|
| 163 |
+
"eval_logits/chosen": -2.428199529647827,
|
| 164 |
+
"eval_logits/rejected": -2.290175437927246,
|
| 165 |
+
"eval_logps/chosen": -292.6412048339844,
|
| 166 |
+
"eval_logps/rejected": -277.6054382324219,
|
| 167 |
+
"eval_loss": 0.5954251885414124,
|
| 168 |
+
"eval_rewards/accuracies": 0.7155172228813171,
|
| 169 |
+
"eval_rewards/chosen": -0.17037244141101837,
|
| 170 |
+
"eval_rewards/margins": 0.41492074728012085,
|
| 171 |
+
"eval_rewards/rejected": -0.5852932333946228,
|
| 172 |
+
"eval_runtime": 92.2211,
|
| 173 |
+
"eval_samples_per_second": 19.854,
|
| 174 |
+
"eval_steps_per_second": 0.314,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.25142857142857145,
|
| 179 |
+
"grad_norm": 17.063109893518277,
|
| 180 |
+
"learning_rate": 4.660050057270191e-07,
|
| 181 |
+
"logits/chosen": -2.2657978534698486,
|
| 182 |
+
"logits/rejected": -2.190037965774536,
|
| 183 |
+
"logps/chosen": -358.4273376464844,
|
| 184 |
+
"logps/rejected": -371.21563720703125,
|
| 185 |
+
"loss": 0.591,
|
| 186 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 187 |
+
"rewards/chosen": -0.4212372303009033,
|
| 188 |
+
"rewards/margins": 0.31182247400283813,
|
| 189 |
+
"rewards/rejected": -0.7330597639083862,
|
| 190 |
+
"step": 110
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2742857142857143,
|
| 194 |
+
"grad_norm": 18.104917147072054,
|
| 195 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 196 |
+
"logits/chosen": -1.961683988571167,
|
| 197 |
+
"logits/rejected": -1.7426531314849854,
|
| 198 |
+
"logps/chosen": -373.987060546875,
|
| 199 |
+
"logps/rejected": -336.79095458984375,
|
| 200 |
+
"loss": 0.5711,
|
| 201 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 202 |
+
"rewards/chosen": -0.6308258175849915,
|
| 203 |
+
"rewards/margins": 0.4199417233467102,
|
| 204 |
+
"rewards/rejected": -1.0507675409317017,
|
| 205 |
+
"step": 120
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.29714285714285715,
|
| 209 |
+
"grad_norm": 23.635801722109512,
|
| 210 |
+
"learning_rate": 4.432129880904388e-07,
|
| 211 |
+
"logits/chosen": -1.0558098554611206,
|
| 212 |
+
"logits/rejected": -0.7218537330627441,
|
| 213 |
+
"logps/chosen": -395.3338317871094,
|
| 214 |
+
"logps/rejected": -383.4067077636719,
|
| 215 |
+
"loss": 0.551,
|
| 216 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 217 |
+
"rewards/chosen": -0.901779055595398,
|
| 218 |
+
"rewards/margins": 0.42758530378341675,
|
| 219 |
+
"rewards/rejected": -1.3293644189834595,
|
| 220 |
+
"step": 130
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.32,
|
| 224 |
+
"grad_norm": 23.352536101734486,
|
| 225 |
+
"learning_rate": 4.299274747394055e-07,
|
| 226 |
+
"logits/chosen": -0.8035767674446106,
|
| 227 |
+
"logits/rejected": -0.4614217281341553,
|
| 228 |
+
"logps/chosen": -391.3183288574219,
|
| 229 |
+
"logps/rejected": -408.2334289550781,
|
| 230 |
+
"loss": 0.5542,
|
| 231 |
+
"rewards/accuracies": 0.731249988079071,
|
| 232 |
+
"rewards/chosen": -0.7587070465087891,
|
| 233 |
+
"rewards/margins": 0.6045126914978027,
|
| 234 |
+
"rewards/rejected": -1.3632197380065918,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.34285714285714286,
|
| 239 |
+
"grad_norm": 20.925582545841166,
|
| 240 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 241 |
+
"logits/chosen": -0.7857555150985718,
|
| 242 |
+
"logits/rejected": -0.3248134255409241,
|
| 243 |
+
"logps/chosen": -371.37091064453125,
|
| 244 |
+
"logps/rejected": -420.19720458984375,
|
| 245 |
+
"loss": 0.5294,
|
| 246 |
+
"rewards/accuracies": 0.706250011920929,
|
| 247 |
+
"rewards/chosen": -0.8438502550125122,
|
| 248 |
+
"rewards/margins": 0.5951322317123413,
|
| 249 |
+
"rewards/rejected": -1.438982367515564,
|
| 250 |
+
"step": 150
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3657142857142857,
|
| 254 |
+
"grad_norm": 24.471284191884987,
|
| 255 |
+
"learning_rate": 4.000011566683401e-07,
|
| 256 |
+
"logits/chosen": -0.6328128576278687,
|
| 257 |
+
"logits/rejected": -0.07759309560060501,
|
| 258 |
+
"logps/chosen": -406.115966796875,
|
| 259 |
+
"logps/rejected": -437.8179626464844,
|
| 260 |
+
"loss": 0.5424,
|
| 261 |
+
"rewards/accuracies": 0.737500011920929,
|
| 262 |
+
"rewards/chosen": -1.0677564144134521,
|
| 263 |
+
"rewards/margins": 0.7539954781532288,
|
| 264 |
+
"rewards/rejected": -1.8217518329620361,
|
| 265 |
+
"step": 160
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.38857142857142857,
|
| 269 |
+
"grad_norm": 24.225836916738423,
|
| 270 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 271 |
+
"logits/chosen": -1.5064557790756226,
|
| 272 |
+
"logits/rejected": -1.0156046152114868,
|
| 273 |
+
"logps/chosen": -380.225830078125,
|
| 274 |
+
"logps/rejected": -391.9797668457031,
|
| 275 |
+
"loss": 0.5359,
|
| 276 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 277 |
+
"rewards/chosen": -0.7612096667289734,
|
| 278 |
+
"rewards/margins": 0.5012774467468262,
|
| 279 |
+
"rewards/rejected": -1.2624871730804443,
|
| 280 |
+
"step": 170
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.4114285714285714,
|
| 284 |
+
"grad_norm": 21.58523060279369,
|
| 285 |
+
"learning_rate": 3.662488473675315e-07,
|
| 286 |
+
"logits/chosen": -1.3468199968338013,
|
| 287 |
+
"logits/rejected": -0.5501200556755066,
|
| 288 |
+
"logps/chosen": -415.23846435546875,
|
| 289 |
+
"logps/rejected": -432.3148498535156,
|
| 290 |
+
"loss": 0.5511,
|
| 291 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 292 |
+
"rewards/chosen": -0.8505870699882507,
|
| 293 |
+
"rewards/margins": 0.873055636882782,
|
| 294 |
+
"rewards/rejected": -1.7236427068710327,
|
| 295 |
+
"step": 180
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.4342857142857143,
|
| 299 |
+
"grad_norm": 22.105812847980097,
|
| 300 |
+
"learning_rate": 3.48203751140067e-07,
|
| 301 |
+
"logits/chosen": -1.2279399633407593,
|
| 302 |
+
"logits/rejected": -0.5397592782974243,
|
| 303 |
+
"logps/chosen": -367.1853942871094,
|
| 304 |
+
"logps/rejected": -383.40606689453125,
|
| 305 |
+
"loss": 0.5454,
|
| 306 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 307 |
+
"rewards/chosen": -0.9092782735824585,
|
| 308 |
+
"rewards/margins": 0.5798004269599915,
|
| 309 |
+
"rewards/rejected": -1.4890786409378052,
|
| 310 |
+
"step": 190
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.45714285714285713,
|
| 314 |
+
"grad_norm": 18.556629832547383,
|
| 315 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 316 |
+
"logits/chosen": -1.407790184020996,
|
| 317 |
+
"logits/rejected": -0.6917704343795776,
|
| 318 |
+
"logps/chosen": -349.32354736328125,
|
| 319 |
+
"logps/rejected": -378.60113525390625,
|
| 320 |
+
"loss": 0.5439,
|
| 321 |
+
"rewards/accuracies": 0.71875,
|
| 322 |
+
"rewards/chosen": -0.8098081350326538,
|
| 323 |
+
"rewards/margins": 0.687397301197052,
|
| 324 |
+
"rewards/rejected": -1.4972054958343506,
|
| 325 |
+
"step": 200
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.45714285714285713,
|
| 329 |
+
"eval_logits/chosen": -1.037967562675476,
|
| 330 |
+
"eval_logits/rejected": -0.04150526225566864,
|
| 331 |
+
"eval_logps/chosen": -356.8392333984375,
|
| 332 |
+
"eval_logps/rejected": -386.5412902832031,
|
| 333 |
+
"eval_loss": 0.5353143215179443,
|
| 334 |
+
"eval_rewards/accuracies": 0.7586206793785095,
|
| 335 |
+
"eval_rewards/chosen": -0.8123530149459839,
|
| 336 |
+
"eval_rewards/margins": 0.8622987270355225,
|
| 337 |
+
"eval_rewards/rejected": -1.674651861190796,
|
| 338 |
+
"eval_runtime": 90.9527,
|
| 339 |
+
"eval_samples_per_second": 20.131,
|
| 340 |
+
"eval_steps_per_second": 0.319,
|
| 341 |
+
"step": 200
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"epoch": 0.48,
|
| 345 |
+
"grad_norm": 20.355837183026274,
|
| 346 |
+
"learning_rate": 3.103511916141658e-07,
|
| 347 |
+
"logits/chosen": -0.7288470268249512,
|
| 348 |
+
"logits/rejected": -0.05566452816128731,
|
| 349 |
+
"logps/chosen": -357.4866943359375,
|
| 350 |
+
"logps/rejected": -403.9605407714844,
|
| 351 |
+
"loss": 0.5321,
|
| 352 |
+
"rewards/accuracies": 0.706250011920929,
|
| 353 |
+
"rewards/chosen": -0.9403315782546997,
|
| 354 |
+
"rewards/margins": 0.6464741230010986,
|
| 355 |
+
"rewards/rejected": -1.5868055820465088,
|
| 356 |
+
"step": 210
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.5028571428571429,
|
| 360 |
+
"grad_norm": 22.193503671674048,
|
| 361 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 362 |
+
"logits/chosen": -0.47817331552505493,
|
| 363 |
+
"logits/rejected": 0.3535853326320648,
|
| 364 |
+
"logps/chosen": -388.7688293457031,
|
| 365 |
+
"logps/rejected": -414.6498107910156,
|
| 366 |
+
"loss": 0.5487,
|
| 367 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 368 |
+
"rewards/chosen": -1.0370280742645264,
|
| 369 |
+
"rewards/margins": 0.6476354002952576,
|
| 370 |
+
"rewards/rejected": -1.6846634149551392,
|
| 371 |
+
"step": 220
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.5257142857142857,
|
| 375 |
+
"grad_norm": 20.666846900072,
|
| 376 |
+
"learning_rate": 2.709592897595191e-07,
|
| 377 |
+
"logits/chosen": -0.1568877398967743,
|
| 378 |
+
"logits/rejected": 0.7903480529785156,
|
| 379 |
+
"logps/chosen": -379.3858642578125,
|
| 380 |
+
"logps/rejected": -402.08740234375,
|
| 381 |
+
"loss": 0.531,
|
| 382 |
+
"rewards/accuracies": 0.731249988079071,
|
| 383 |
+
"rewards/chosen": -0.983267605304718,
|
| 384 |
+
"rewards/margins": 0.6447398662567139,
|
| 385 |
+
"rewards/rejected": -1.6280075311660767,
|
| 386 |
+
"step": 230
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.5485714285714286,
|
| 390 |
+
"grad_norm": 26.65066022502421,
|
| 391 |
+
"learning_rate": 2.509992316440332e-07,
|
| 392 |
+
"logits/chosen": -0.4197590947151184,
|
| 393 |
+
"logits/rejected": 0.6013392210006714,
|
| 394 |
+
"logps/chosen": -385.48638916015625,
|
| 395 |
+
"logps/rejected": -448.73175048828125,
|
| 396 |
+
"loss": 0.5245,
|
| 397 |
+
"rewards/accuracies": 0.78125,
|
| 398 |
+
"rewards/chosen": -0.8677653074264526,
|
| 399 |
+
"rewards/margins": 0.9053336977958679,
|
| 400 |
+
"rewards/rejected": -1.7730991840362549,
|
| 401 |
+
"step": 240
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.5714285714285714,
|
| 405 |
+
"grad_norm": 21.858480788312963,
|
| 406 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 407 |
+
"logits/chosen": -0.17110395431518555,
|
| 408 |
+
"logits/rejected": 0.49592718482017517,
|
| 409 |
+
"logps/chosen": -368.26263427734375,
|
| 410 |
+
"logps/rejected": -430.48468017578125,
|
| 411 |
+
"loss": 0.5351,
|
| 412 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 413 |
+
"rewards/chosen": -0.8763637542724609,
|
| 414 |
+
"rewards/margins": 0.7264786958694458,
|
| 415 |
+
"rewards/rejected": -1.6028425693511963,
|
| 416 |
+
"step": 250
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5942857142857143,
|
| 420 |
+
"grad_norm": 23.021393487481784,
|
| 421 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 422 |
+
"logits/chosen": -0.3194964528083801,
|
| 423 |
+
"logits/rejected": 0.2826068699359894,
|
| 424 |
+
"logps/chosen": -373.82830810546875,
|
| 425 |
+
"logps/rejected": -423.619384765625,
|
| 426 |
+
"loss": 0.5387,
|
| 427 |
+
"rewards/accuracies": 0.706250011920929,
|
| 428 |
+
"rewards/chosen": -0.9404159784317017,
|
| 429 |
+
"rewards/margins": 0.5623483657836914,
|
| 430 |
+
"rewards/rejected": -1.502764344215393,
|
| 431 |
+
"step": 260
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.6171428571428571,
|
| 435 |
+
"grad_norm": 28.56553316601309,
|
| 436 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 437 |
+
"logits/chosen": 0.10294933617115021,
|
| 438 |
+
"logits/rejected": 1.3178448677062988,
|
| 439 |
+
"logps/chosen": -370.7629089355469,
|
| 440 |
+
"logps/rejected": -405.086181640625,
|
| 441 |
+
"loss": 0.5159,
|
| 442 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 443 |
+
"rewards/chosen": -0.9823705554008484,
|
| 444 |
+
"rewards/margins": 0.7634164690971375,
|
| 445 |
+
"rewards/rejected": -1.7457869052886963,
|
| 446 |
+
"step": 270
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.64,
|
| 450 |
+
"grad_norm": 24.30460588422491,
|
| 451 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 452 |
+
"logits/chosen": 0.1169511079788208,
|
| 453 |
+
"logits/rejected": 1.0327664613723755,
|
| 454 |
+
"logps/chosen": -367.6756591796875,
|
| 455 |
+
"logps/rejected": -442.3148498535156,
|
| 456 |
+
"loss": 0.5251,
|
| 457 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 458 |
+
"rewards/chosen": -0.9463755488395691,
|
| 459 |
+
"rewards/margins": 0.8457048535346985,
|
| 460 |
+
"rewards/rejected": -1.792080283164978,
|
| 461 |
+
"step": 280
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.6628571428571428,
|
| 465 |
+
"grad_norm": 28.120774422152078,
|
| 466 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 467 |
+
"logits/chosen": 0.10812599956989288,
|
| 468 |
+
"logits/rejected": 1.0026741027832031,
|
| 469 |
+
"logps/chosen": -394.53759765625,
|
| 470 |
+
"logps/rejected": -435.6455078125,
|
| 471 |
+
"loss": 0.5215,
|
| 472 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 473 |
+
"rewards/chosen": -0.9363173246383667,
|
| 474 |
+
"rewards/margins": 0.764653205871582,
|
| 475 |
+
"rewards/rejected": -1.7009706497192383,
|
| 476 |
+
"step": 290
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.6857142857142857,
|
| 480 |
+
"grad_norm": 20.243539038486684,
|
| 481 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 482 |
+
"logits/chosen": -0.2318529188632965,
|
| 483 |
+
"logits/rejected": 1.2198255062103271,
|
| 484 |
+
"logps/chosen": -389.1872863769531,
|
| 485 |
+
"logps/rejected": -424.034912109375,
|
| 486 |
+
"loss": 0.5109,
|
| 487 |
+
"rewards/accuracies": 0.762499988079071,
|
| 488 |
+
"rewards/chosen": -0.8727809190750122,
|
| 489 |
+
"rewards/margins": 0.8411470651626587,
|
| 490 |
+
"rewards/rejected": -1.713927984237671,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6857142857142857,
|
| 495 |
+
"eval_logits/chosen": -0.3942234218120575,
|
| 496 |
+
"eval_logits/rejected": 1.1016443967819214,
|
| 497 |
+
"eval_logps/chosen": -345.452880859375,
|
| 498 |
+
"eval_logps/rejected": -383.7008361816406,
|
| 499 |
+
"eval_loss": 0.5189641118049622,
|
| 500 |
+
"eval_rewards/accuracies": 0.7715517282485962,
|
| 501 |
+
"eval_rewards/chosen": -0.6984895467758179,
|
| 502 |
+
"eval_rewards/margins": 0.9477577805519104,
|
| 503 |
+
"eval_rewards/rejected": -1.6462475061416626,
|
| 504 |
+
"eval_runtime": 91.163,
|
| 505 |
+
"eval_samples_per_second": 20.085,
|
| 506 |
+
"eval_steps_per_second": 0.318,
|
| 507 |
+
"step": 300
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 0.7085714285714285,
|
| 511 |
+
"grad_norm": 21.712761933059525,
|
| 512 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 513 |
+
"logits/chosen": -0.18514932692050934,
|
| 514 |
+
"logits/rejected": 0.6901147961616516,
|
| 515 |
+
"logps/chosen": -342.0721740722656,
|
| 516 |
+
"logps/rejected": -390.0445861816406,
|
| 517 |
+
"loss": 0.5324,
|
| 518 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 519 |
+
"rewards/chosen": -0.8576802015304565,
|
| 520 |
+
"rewards/margins": 0.6708122491836548,
|
| 521 |
+
"rewards/rejected": -1.5284925699234009,
|
| 522 |
+
"step": 310
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.7314285714285714,
|
| 526 |
+
"grad_norm": 22.89579586434415,
|
| 527 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 528 |
+
"logits/chosen": 0.14465376734733582,
|
| 529 |
+
"logits/rejected": 0.9812980890274048,
|
| 530 |
+
"logps/chosen": -342.1743469238281,
|
| 531 |
+
"logps/rejected": -399.4386901855469,
|
| 532 |
+
"loss": 0.5259,
|
| 533 |
+
"rewards/accuracies": 0.675000011920929,
|
| 534 |
+
"rewards/chosen": -0.9789196848869324,
|
| 535 |
+
"rewards/margins": 0.6055852174758911,
|
| 536 |
+
"rewards/rejected": -1.5845048427581787,
|
| 537 |
+
"step": 320
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.7542857142857143,
|
| 541 |
+
"grad_norm": 21.156935890608,
|
| 542 |
+
"learning_rate": 8.601038193139438e-08,
|
| 543 |
+
"logits/chosen": -0.23074492812156677,
|
| 544 |
+
"logits/rejected": 0.9273271560668945,
|
| 545 |
+
"logps/chosen": -383.02691650390625,
|
| 546 |
+
"logps/rejected": -410.26519775390625,
|
| 547 |
+
"loss": 0.5203,
|
| 548 |
+
"rewards/accuracies": 0.78125,
|
| 549 |
+
"rewards/chosen": -0.8380874395370483,
|
| 550 |
+
"rewards/margins": 0.7866854071617126,
|
| 551 |
+
"rewards/rejected": -1.6247727870941162,
|
| 552 |
+
"step": 330
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.7771428571428571,
|
| 556 |
+
"grad_norm": 21.402992880624122,
|
| 557 |
+
"learning_rate": 7.146574594727572e-08,
|
| 558 |
+
"logits/chosen": 0.4683713912963867,
|
| 559 |
+
"logits/rejected": 1.2558484077453613,
|
| 560 |
+
"logps/chosen": -365.50213623046875,
|
| 561 |
+
"logps/rejected": -427.18682861328125,
|
| 562 |
+
"loss": 0.5174,
|
| 563 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 564 |
+
"rewards/chosen": -0.9955114126205444,
|
| 565 |
+
"rewards/margins": 0.8561191558837891,
|
| 566 |
+
"rewards/rejected": -1.8516305685043335,
|
| 567 |
+
"step": 340
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.8,
|
| 571 |
+
"grad_norm": 23.419375804789023,
|
| 572 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 573 |
+
"logits/chosen": -0.03433036804199219,
|
| 574 |
+
"logits/rejected": 1.043505072593689,
|
| 575 |
+
"logps/chosen": -382.8368225097656,
|
| 576 |
+
"logps/rejected": -416.30145263671875,
|
| 577 |
+
"loss": 0.5252,
|
| 578 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 579 |
+
"rewards/chosen": -0.9312959909439087,
|
| 580 |
+
"rewards/margins": 0.6887425780296326,
|
| 581 |
+
"rewards/rejected": -1.620038628578186,
|
| 582 |
+
"step": 350
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.8228571428571428,
|
| 586 |
+
"grad_norm": 21.8208411590094,
|
| 587 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 588 |
+
"logits/chosen": -0.20763058960437775,
|
| 589 |
+
"logits/rejected": 1.1433674097061157,
|
| 590 |
+
"logps/chosen": -370.57281494140625,
|
| 591 |
+
"logps/rejected": -423.702392578125,
|
| 592 |
+
"loss": 0.5099,
|
| 593 |
+
"rewards/accuracies": 0.793749988079071,
|
| 594 |
+
"rewards/chosen": -0.8516775369644165,
|
| 595 |
+
"rewards/margins": 0.8744813799858093,
|
| 596 |
+
"rewards/rejected": -1.726158857345581,
|
| 597 |
+
"step": 360
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.8457142857142858,
|
| 601 |
+
"grad_norm": 21.479692671582267,
|
| 602 |
+
"learning_rate": 3.500802900154412e-08,
|
| 603 |
+
"logits/chosen": -0.02174173854291439,
|
| 604 |
+
"logits/rejected": 1.3574512004852295,
|
| 605 |
+
"logps/chosen": -351.3383483886719,
|
| 606 |
+
"logps/rejected": -407.7884826660156,
|
| 607 |
+
"loss": 0.515,
|
| 608 |
+
"rewards/accuracies": 0.762499988079071,
|
| 609 |
+
"rewards/chosen": -0.8027693033218384,
|
| 610 |
+
"rewards/margins": 0.8909598588943481,
|
| 611 |
+
"rewards/rejected": -1.6937291622161865,
|
| 612 |
+
"step": 370
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.8685714285714285,
|
| 616 |
+
"grad_norm": 23.836354879990434,
|
| 617 |
+
"learning_rate": 2.550629574310309e-08,
|
| 618 |
+
"logits/chosen": -0.2868313491344452,
|
| 619 |
+
"logits/rejected": 1.0226128101348877,
|
| 620 |
+
"logps/chosen": -414.9786682128906,
|
| 621 |
+
"logps/rejected": -421.4689025878906,
|
| 622 |
+
"loss": 0.5176,
|
| 623 |
+
"rewards/accuracies": 0.71875,
|
| 624 |
+
"rewards/chosen": -0.969223141670227,
|
| 625 |
+
"rewards/margins": 0.6714719533920288,
|
| 626 |
+
"rewards/rejected": -1.6406952142715454,
|
| 627 |
+
"step": 380
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.8914285714285715,
|
| 631 |
+
"grad_norm": 24.654106172641406,
|
| 632 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 633 |
+
"logits/chosen": 0.04682580381631851,
|
| 634 |
+
"logits/rejected": 0.8462721109390259,
|
| 635 |
+
"logps/chosen": -395.02130126953125,
|
| 636 |
+
"logps/rejected": -450.59375,
|
| 637 |
+
"loss": 0.525,
|
| 638 |
+
"rewards/accuracies": 0.71875,
|
| 639 |
+
"rewards/chosen": -0.836483359336853,
|
| 640 |
+
"rewards/margins": 0.7217278480529785,
|
| 641 |
+
"rewards/rejected": -1.5582112073898315,
|
| 642 |
+
"step": 390
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.9142857142857143,
|
| 646 |
+
"grad_norm": 35.84442600986327,
|
| 647 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 648 |
+
"logits/chosen": 0.2935855984687805,
|
| 649 |
+
"logits/rejected": 0.8280227780342102,
|
| 650 |
+
"logps/chosen": -379.0578918457031,
|
| 651 |
+
"logps/rejected": -430.62261962890625,
|
| 652 |
+
"loss": 0.5457,
|
| 653 |
+
"rewards/accuracies": 0.731249988079071,
|
| 654 |
+
"rewards/chosen": -1.0032463073730469,
|
| 655 |
+
"rewards/margins": 0.6372436881065369,
|
| 656 |
+
"rewards/rejected": -1.640489935874939,
|
| 657 |
+
"step": 400
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.9142857142857143,
|
| 661 |
+
"eval_logits/chosen": 0.24073126912117004,
|
| 662 |
+
"eval_logits/rejected": 1.7997794151306152,
|
| 663 |
+
"eval_logps/chosen": -362.5745849609375,
|
| 664 |
+
"eval_logps/rejected": -405.11981201171875,
|
| 665 |
+
"eval_loss": 0.5175118446350098,
|
| 666 |
+
"eval_rewards/accuracies": 0.7715517282485962,
|
| 667 |
+
"eval_rewards/chosen": -0.8697066307067871,
|
| 668 |
+
"eval_rewards/margins": 0.9907302856445312,
|
| 669 |
+
"eval_rewards/rejected": -1.8604369163513184,
|
| 670 |
+
"eval_runtime": 90.8335,
|
| 671 |
+
"eval_samples_per_second": 20.158,
|
| 672 |
+
"eval_steps_per_second": 0.319,
|
| 673 |
+
"step": 400
|
| 674 |
+
},
|
| 675 |
+
{
|
| 676 |
+
"epoch": 0.9371428571428572,
|
| 677 |
+
"grad_norm": 27.47931239793458,
|
| 678 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 679 |
+
"logits/chosen": 0.10173013061285019,
|
| 680 |
+
"logits/rejected": 1.089665174484253,
|
| 681 |
+
"logps/chosen": -389.411865234375,
|
| 682 |
+
"logps/rejected": -432.0257873535156,
|
| 683 |
+
"loss": 0.5148,
|
| 684 |
+
"rewards/accuracies": 0.731249988079071,
|
| 685 |
+
"rewards/chosen": -1.0123193264007568,
|
| 686 |
+
"rewards/margins": 0.710638165473938,
|
| 687 |
+
"rewards/rejected": -1.7229573726654053,
|
| 688 |
+
"step": 410
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.96,
|
| 692 |
+
"grad_norm": 21.72575201877371,
|
| 693 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 694 |
+
"logits/chosen": -0.029790764674544334,
|
| 695 |
+
"logits/rejected": 1.4268020391464233,
|
| 696 |
+
"logps/chosen": -357.87603759765625,
|
| 697 |
+
"logps/rejected": -413.14227294921875,
|
| 698 |
+
"loss": 0.5143,
|
| 699 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 700 |
+
"rewards/chosen": -0.7512796521186829,
|
| 701 |
+
"rewards/margins": 1.0402743816375732,
|
| 702 |
+
"rewards/rejected": -1.7915542125701904,
|
| 703 |
+
"step": 420
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.9828571428571429,
|
| 707 |
+
"grad_norm": 22.231222051131574,
|
| 708 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 709 |
+
"logits/chosen": 0.29859107732772827,
|
| 710 |
+
"logits/rejected": 1.3065626621246338,
|
| 711 |
+
"logps/chosen": -373.48834228515625,
|
| 712 |
+
"logps/rejected": -435.72515869140625,
|
| 713 |
+
"loss": 0.5136,
|
| 714 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 715 |
+
"rewards/chosen": -0.9125651121139526,
|
| 716 |
+
"rewards/margins": 0.8079904317855835,
|
| 717 |
+
"rewards/rejected": -1.7205555438995361,
|
| 718 |
+
"step": 430
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.9988571428571429,
|
| 722 |
+
"step": 437,
|
| 723 |
+
"total_flos": 0.0,
|
| 724 |
+
"train_loss": 0.5586249282758351,
|
| 725 |
+
"train_runtime": 11006.8714,
|
| 726 |
+
"train_samples_per_second": 5.088,
|
| 727 |
+
"train_steps_per_second": 0.04
|
| 728 |
+
}
|
| 729 |
+
],
|
| 730 |
+
"logging_steps": 10,
|
| 731 |
+
"max_steps": 437,
|
| 732 |
+
"num_input_tokens_seen": 0,
|
| 733 |
+
"num_train_epochs": 1,
|
| 734 |
+
"save_steps": 100,
|
| 735 |
+
"stateful_callbacks": {
|
| 736 |
+
"TrainerControl": {
|
| 737 |
+
"args": {
|
| 738 |
+
"should_epoch_stop": false,
|
| 739 |
+
"should_evaluate": false,
|
| 740 |
+
"should_log": false,
|
| 741 |
+
"should_save": true,
|
| 742 |
+
"should_training_stop": true
|
| 743 |
+
},
|
| 744 |
+
"attributes": {}
|
| 745 |
+
}
|
| 746 |
+
},
|
| 747 |
+
"total_flos": 0.0,
|
| 748 |
+
"train_batch_size": 8,
|
| 749 |
+
"trial_name": null,
|
| 750 |
+
"trial_params": null
|
| 751 |
+
}
|