Commit
•
a5b0816
1
Parent(s):
528477b
Model save
Browse files- README.md +78 -0
- all_results.json +9 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: peft
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: mistralai/Mistral-7B-v0.1
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-dpo-qlora
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-7b-dpo-qlora
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5035
|
| 22 |
+
- Rewards/chosen: -2.0213
|
| 23 |
+
- Rewards/rejected: -3.0170
|
| 24 |
+
- Rewards/accuracies: 0.7656
|
| 25 |
+
- Rewards/margins: 0.9957
|
| 26 |
+
- Logps/rejected: -549.2363
|
| 27 |
+
- Logps/chosen: -448.5603
|
| 28 |
+
- Logits/rejected: -1.1850
|
| 29 |
+
- Logits/chosen: -1.2569
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-06
|
| 49 |
+
- train_batch_size: 4
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 8
|
| 54 |
+
- gradient_accumulation_steps: 4
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 64
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.5646 | 0.2093 | 100 | 0.5739 | -0.9253 | -1.4816 | 0.7188 | 0.5564 | -395.6964 | -338.9565 | -1.9267 | -1.9878 |
|
| 67 |
+
| 0.5524 | 0.4186 | 200 | 0.5318 | -0.8476 | -1.5395 | 0.7617 | 0.6919 | -401.4810 | -331.1845 | -1.5104 | -1.5801 |
|
| 68 |
+
| 0.4977 | 0.6279 | 300 | 0.5100 | -1.8821 | -2.8383 | 0.7773 | 0.9562 | -531.3586 | -434.6388 | -1.1156 | -1.1878 |
|
| 69 |
+
| 0.5096 | 0.8373 | 400 | 0.5035 | -2.0213 | -3.0170 | 0.7656 | 0.9957 | -549.2363 | -448.5603 | -1.1850 | -1.2569 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- PEFT 0.13.2
|
| 75 |
+
- Transformers 4.45.2
|
| 76 |
+
- Pytorch 2.1.2+cu121
|
| 77 |
+
- Datasets 3.0.1
|
| 78 |
+
- Tokenizers 0.20.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9984301412872841,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5421361258444796,
|
| 5 |
+
"train_runtime": 7660.497,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 7.98,
|
| 8 |
+
"train_steps_per_second": 0.062
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9984301412872841,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5421361258444796,
|
| 5 |
+
"train_runtime": 7660.497,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 7.98,
|
| 8 |
+
"train_steps_per_second": 0.062
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9984301412872841,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 477,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0020931449502878076,
|
| 13 |
+
"grad_norm": 0.8892470002174377,
|
| 14 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 15 |
+
"logits/chosen": -2.5115435123443604,
|
| 16 |
+
"logits/rejected": -2.418633460998535,
|
| 17 |
+
"logps/chosen": -296.6790771484375,
|
| 18 |
+
"logps/rejected": -343.70672607421875,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.020931449502878074,
|
| 28 |
+
"grad_norm": 0.8503940105438232,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-06,
|
| 30 |
+
"logits/chosen": -2.2245330810546875,
|
| 31 |
+
"logits/rejected": -2.1629724502563477,
|
| 32 |
+
"logps/chosen": -283.80975341796875,
|
| 33 |
+
"logps/rejected": -255.17117309570312,
|
| 34 |
+
"loss": 0.6926,
|
| 35 |
+
"rewards/accuracies": 0.4861111044883728,
|
| 36 |
+
"rewards/chosen": 0.008690375834703445,
|
| 37 |
+
"rewards/margins": 0.0015612257411703467,
|
| 38 |
+
"rewards/rejected": 0.0071291509084403515,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04186289900575615,
|
| 43 |
+
"grad_norm": 0.8254271149635315,
|
| 44 |
+
"learning_rate": 2.0833333333333334e-06,
|
| 45 |
+
"logits/chosen": -2.335510015487671,
|
| 46 |
+
"logits/rejected": -2.22532057762146,
|
| 47 |
+
"logps/chosen": -286.57855224609375,
|
| 48 |
+
"logps/rejected": -261.207763671875,
|
| 49 |
+
"loss": 0.6874,
|
| 50 |
+
"rewards/accuracies": 0.65625,
|
| 51 |
+
"rewards/chosen": 0.03880625218153,
|
| 52 |
+
"rewards/margins": 0.011353857815265656,
|
| 53 |
+
"rewards/rejected": 0.027452394366264343,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06279434850863422,
|
| 58 |
+
"grad_norm": 0.8504378795623779,
|
| 59 |
+
"learning_rate": 3.125e-06,
|
| 60 |
+
"logits/chosen": -2.2984628677368164,
|
| 61 |
+
"logits/rejected": -2.184295892715454,
|
| 62 |
+
"logps/chosen": -261.04315185546875,
|
| 63 |
+
"logps/rejected": -227.68020629882812,
|
| 64 |
+
"loss": 0.6805,
|
| 65 |
+
"rewards/accuracies": 0.699999988079071,
|
| 66 |
+
"rewards/chosen": 0.06454791873693466,
|
| 67 |
+
"rewards/margins": 0.028396595269441605,
|
| 68 |
+
"rewards/rejected": 0.03615132346749306,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.0837257980115123,
|
| 73 |
+
"grad_norm": 0.9619700908660889,
|
| 74 |
+
"learning_rate": 4.166666666666667e-06,
|
| 75 |
+
"logits/chosen": -2.1841602325439453,
|
| 76 |
+
"logits/rejected": -2.1042842864990234,
|
| 77 |
+
"logps/chosen": -244.4292449951172,
|
| 78 |
+
"logps/rejected": -248.12057495117188,
|
| 79 |
+
"loss": 0.664,
|
| 80 |
+
"rewards/accuracies": 0.6875,
|
| 81 |
+
"rewards/chosen": 0.03205280378460884,
|
| 82 |
+
"rewards/margins": 0.06447319686412811,
|
| 83 |
+
"rewards/rejected": -0.03242039680480957,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10465724751439037,
|
| 88 |
+
"grad_norm": 1.198764443397522,
|
| 89 |
+
"learning_rate": 4.999731868769027e-06,
|
| 90 |
+
"logits/chosen": -2.2160000801086426,
|
| 91 |
+
"logits/rejected": -2.09934663772583,
|
| 92 |
+
"logps/chosen": -269.432861328125,
|
| 93 |
+
"logps/rejected": -274.974365234375,
|
| 94 |
+
"loss": 0.6454,
|
| 95 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 96 |
+
"rewards/chosen": -0.08960571140050888,
|
| 97 |
+
"rewards/margins": 0.10663382709026337,
|
| 98 |
+
"rewards/rejected": -0.19623954594135284,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12558869701726844,
|
| 103 |
+
"grad_norm": 1.695876121520996,
|
| 104 |
+
"learning_rate": 4.9903533134293035e-06,
|
| 105 |
+
"logits/chosen": -2.236078977584839,
|
| 106 |
+
"logits/rejected": -2.1389355659484863,
|
| 107 |
+
"logps/chosen": -262.9930725097656,
|
| 108 |
+
"logps/rejected": -276.64984130859375,
|
| 109 |
+
"loss": 0.6264,
|
| 110 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 111 |
+
"rewards/chosen": -0.264637291431427,
|
| 112 |
+
"rewards/margins": 0.2180159091949463,
|
| 113 |
+
"rewards/rejected": -0.4826532006263733,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14652014652014653,
|
| 118 |
+
"grad_norm": 2.115797519683838,
|
| 119 |
+
"learning_rate": 4.967625656594782e-06,
|
| 120 |
+
"logits/chosen": -2.1270174980163574,
|
| 121 |
+
"logits/rejected": -2.063481569290161,
|
| 122 |
+
"logps/chosen": -319.60888671875,
|
| 123 |
+
"logps/rejected": -334.22705078125,
|
| 124 |
+
"loss": 0.607,
|
| 125 |
+
"rewards/accuracies": 0.737500011920929,
|
| 126 |
+
"rewards/chosen": -0.4378367066383362,
|
| 127 |
+
"rewards/margins": 0.2902434766292572,
|
| 128 |
+
"rewards/rejected": -0.728080153465271,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.1674515960230246,
|
| 133 |
+
"grad_norm": 2.7295830249786377,
|
| 134 |
+
"learning_rate": 4.93167072587771e-06,
|
| 135 |
+
"logits/chosen": -2.275411605834961,
|
| 136 |
+
"logits/rejected": -2.0854580402374268,
|
| 137 |
+
"logps/chosen": -358.58154296875,
|
| 138 |
+
"logps/rejected": -320.63385009765625,
|
| 139 |
+
"loss": 0.6069,
|
| 140 |
+
"rewards/accuracies": 0.768750011920929,
|
| 141 |
+
"rewards/chosen": -0.49991363286972046,
|
| 142 |
+
"rewards/margins": 0.31041082739830017,
|
| 143 |
+
"rewards/rejected": -0.8103245496749878,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18838304552590268,
|
| 148 |
+
"grad_norm": 2.6313765048980713,
|
| 149 |
+
"learning_rate": 4.882681251368549e-06,
|
| 150 |
+
"logits/chosen": -2.1408169269561768,
|
| 151 |
+
"logits/rejected": -2.067002058029175,
|
| 152 |
+
"logps/chosen": -289.2994079589844,
|
| 153 |
+
"logps/rejected": -331.0093994140625,
|
| 154 |
+
"loss": 0.576,
|
| 155 |
+
"rewards/accuracies": 0.699999988079071,
|
| 156 |
+
"rewards/chosen": -0.5190187692642212,
|
| 157 |
+
"rewards/margins": 0.3863602876663208,
|
| 158 |
+
"rewards/rejected": -0.9053791165351868,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20931449502878074,
|
| 163 |
+
"grad_norm": 3.2287309169769287,
|
| 164 |
+
"learning_rate": 4.8209198325401815e-06,
|
| 165 |
+
"logits/chosen": -2.0671892166137695,
|
| 166 |
+
"logits/rejected": -1.958622694015503,
|
| 167 |
+
"logps/chosen": -370.7698059082031,
|
| 168 |
+
"logps/rejected": -413.80120849609375,
|
| 169 |
+
"loss": 0.5646,
|
| 170 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 171 |
+
"rewards/chosen": -0.933598518371582,
|
| 172 |
+
"rewards/margins": 0.5866175889968872,
|
| 173 |
+
"rewards/rejected": -1.5202162265777588,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20931449502878074,
|
| 178 |
+
"eval_logits/chosen": -1.9878015518188477,
|
| 179 |
+
"eval_logits/rejected": -1.9267165660858154,
|
| 180 |
+
"eval_logps/chosen": -338.9564514160156,
|
| 181 |
+
"eval_logps/rejected": -395.69635009765625,
|
| 182 |
+
"eval_loss": 0.573878288269043,
|
| 183 |
+
"eval_rewards/accuracies": 0.71875,
|
| 184 |
+
"eval_rewards/chosen": -0.9252709150314331,
|
| 185 |
+
"eval_rewards/margins": 0.5563679933547974,
|
| 186 |
+
"eval_rewards/rejected": -1.4816389083862305,
|
| 187 |
+
"eval_runtime": 110.707,
|
| 188 |
+
"eval_samples_per_second": 18.066,
|
| 189 |
+
"eval_steps_per_second": 0.289,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2302459445316588,
|
| 194 |
+
"grad_norm": 3.1588852405548096,
|
| 195 |
+
"learning_rate": 4.746717530629565e-06,
|
| 196 |
+
"logits/chosen": -2.0595319271087646,
|
| 197 |
+
"logits/rejected": -1.9567897319793701,
|
| 198 |
+
"logps/chosen": -360.9635009765625,
|
| 199 |
+
"logps/rejected": -388.3221130371094,
|
| 200 |
+
"loss": 0.5789,
|
| 201 |
+
"rewards/accuracies": 0.65625,
|
| 202 |
+
"rewards/chosen": -0.860044002532959,
|
| 203 |
+
"rewards/margins": 0.4338661730289459,
|
| 204 |
+
"rewards/rejected": -1.2939101457595825,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.25117739403453687,
|
| 209 |
+
"grad_norm": 2.8980185985565186,
|
| 210 |
+
"learning_rate": 4.660472094042121e-06,
|
| 211 |
+
"logits/chosen": -1.8865253925323486,
|
| 212 |
+
"logits/rejected": -1.8495429754257202,
|
| 213 |
+
"logps/chosen": -350.013916015625,
|
| 214 |
+
"logps/rejected": -394.4359130859375,
|
| 215 |
+
"loss": 0.5592,
|
| 216 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 217 |
+
"rewards/chosen": -0.9938249588012695,
|
| 218 |
+
"rewards/margins": 0.5278288125991821,
|
| 219 |
+
"rewards/rejected": -1.5216538906097412,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.272108843537415,
|
| 224 |
+
"grad_norm": 3.334541082382202,
|
| 225 |
+
"learning_rate": 4.5626458262912745e-06,
|
| 226 |
+
"logits/chosen": -1.8878917694091797,
|
| 227 |
+
"logits/rejected": -1.8218252658843994,
|
| 228 |
+
"logps/chosen": -396.32659912109375,
|
| 229 |
+
"logps/rejected": -439.26513671875,
|
| 230 |
+
"loss": 0.5373,
|
| 231 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 232 |
+
"rewards/chosen": -1.3216043710708618,
|
| 233 |
+
"rewards/margins": 0.5591233968734741,
|
| 234 |
+
"rewards/rejected": -1.880727767944336,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.29304029304029305,
|
| 239 |
+
"grad_norm": 3.480584144592285,
|
| 240 |
+
"learning_rate": 4.453763107901676e-06,
|
| 241 |
+
"logits/chosen": -1.8928722143173218,
|
| 242 |
+
"logits/rejected": -1.7833354473114014,
|
| 243 |
+
"logps/chosen": -396.819091796875,
|
| 244 |
+
"logps/rejected": -430.72674560546875,
|
| 245 |
+
"loss": 0.549,
|
| 246 |
+
"rewards/accuracies": 0.75,
|
| 247 |
+
"rewards/chosen": -1.0090841054916382,
|
| 248 |
+
"rewards/margins": 0.7491954565048218,
|
| 249 |
+
"rewards/rejected": -1.75827956199646,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3139717425431711,
|
| 254 |
+
"grad_norm": 3.2153046131134033,
|
| 255 |
+
"learning_rate": 4.33440758555951e-06,
|
| 256 |
+
"logits/chosen": -1.8725719451904297,
|
| 257 |
+
"logits/rejected": -1.7752315998077393,
|
| 258 |
+
"logps/chosen": -358.0649719238281,
|
| 259 |
+
"logps/rejected": -379.94464111328125,
|
| 260 |
+
"loss": 0.5479,
|
| 261 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 262 |
+
"rewards/chosen": -0.9543037414550781,
|
| 263 |
+
"rewards/margins": 0.5243933796882629,
|
| 264 |
+
"rewards/rejected": -1.4786970615386963,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.3349031920460492,
|
| 269 |
+
"grad_norm": 3.572585344314575,
|
| 270 |
+
"learning_rate": 4.205219043576955e-06,
|
| 271 |
+
"logits/chosen": -1.8374134302139282,
|
| 272 |
+
"logits/rejected": -1.6999791860580444,
|
| 273 |
+
"logps/chosen": -347.92791748046875,
|
| 274 |
+
"logps/rejected": -397.92236328125,
|
| 275 |
+
"loss": 0.5582,
|
| 276 |
+
"rewards/accuracies": 0.737500011920929,
|
| 277 |
+
"rewards/chosen": -0.8851444125175476,
|
| 278 |
+
"rewards/margins": 0.7162662744522095,
|
| 279 |
+
"rewards/rejected": -1.6014108657836914,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35583464154892724,
|
| 284 |
+
"grad_norm": 4.003506183624268,
|
| 285 |
+
"learning_rate": 4.066889974440757e-06,
|
| 286 |
+
"logits/chosen": -1.4849274158477783,
|
| 287 |
+
"logits/rejected": -1.420831561088562,
|
| 288 |
+
"logps/chosen": -422.58807373046875,
|
| 289 |
+
"logps/rejected": -454.4142150878906,
|
| 290 |
+
"loss": 0.5563,
|
| 291 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 292 |
+
"rewards/chosen": -1.7905734777450562,
|
| 293 |
+
"rewards/margins": 0.5325244665145874,
|
| 294 |
+
"rewards/rejected": -2.3230979442596436,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37676609105180536,
|
| 299 |
+
"grad_norm": 3.9669156074523926,
|
| 300 |
+
"learning_rate": 3.92016186682789e-06,
|
| 301 |
+
"logits/chosen": -1.6295009851455688,
|
| 302 |
+
"logits/rejected": -1.550588846206665,
|
| 303 |
+
"logps/chosen": -370.55133056640625,
|
| 304 |
+
"logps/rejected": -407.09735107421875,
|
| 305 |
+
"loss": 0.5112,
|
| 306 |
+
"rewards/accuracies": 0.731249988079071,
|
| 307 |
+
"rewards/chosen": -1.166177749633789,
|
| 308 |
+
"rewards/margins": 0.6077830195426941,
|
| 309 |
+
"rewards/rejected": -1.773960828781128,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.3976975405546834,
|
| 314 |
+
"grad_norm": 4.69191312789917,
|
| 315 |
+
"learning_rate": 3.7658212309857576e-06,
|
| 316 |
+
"logits/chosen": -1.5681777000427246,
|
| 317 |
+
"logits/rejected": -1.5323166847229004,
|
| 318 |
+
"logps/chosen": -348.3360900878906,
|
| 319 |
+
"logps/rejected": -402.2555847167969,
|
| 320 |
+
"loss": 0.5366,
|
| 321 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 322 |
+
"rewards/chosen": -1.0483043193817139,
|
| 323 |
+
"rewards/margins": 0.5953764915466309,
|
| 324 |
+
"rewards/rejected": -1.6436808109283447,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.4186289900575615,
|
| 329 |
+
"grad_norm": 3.6216626167297363,
|
| 330 |
+
"learning_rate": 3.604695382782159e-06,
|
| 331 |
+
"logits/chosen": -1.514801263809204,
|
| 332 |
+
"logits/rejected": -1.5004067420959473,
|
| 333 |
+
"logps/chosen": -343.3397216796875,
|
| 334 |
+
"logps/rejected": -422.3468322753906,
|
| 335 |
+
"loss": 0.5524,
|
| 336 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 337 |
+
"rewards/chosen": -1.099816918373108,
|
| 338 |
+
"rewards/margins": 0.5951756834983826,
|
| 339 |
+
"rewards/rejected": -1.6949926614761353,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.4186289900575615,
|
| 344 |
+
"eval_logits/chosen": -1.5800749063491821,
|
| 345 |
+
"eval_logits/rejected": -1.5103871822357178,
|
| 346 |
+
"eval_logps/chosen": -331.1845397949219,
|
| 347 |
+
"eval_logps/rejected": -401.48101806640625,
|
| 348 |
+
"eval_loss": 0.5318315029144287,
|
| 349 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 350 |
+
"eval_rewards/chosen": -0.8475518226623535,
|
| 351 |
+
"eval_rewards/margins": 0.6919333338737488,
|
| 352 |
+
"eval_rewards/rejected": -1.5394850969314575,
|
| 353 |
+
"eval_runtime": 110.7567,
|
| 354 |
+
"eval_samples_per_second": 18.058,
|
| 355 |
+
"eval_steps_per_second": 0.289,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.43956043956043955,
|
| 360 |
+
"grad_norm": 4.658122539520264,
|
| 361 |
+
"learning_rate": 3.437648009023905e-06,
|
| 362 |
+
"logits/chosen": -1.5987634658813477,
|
| 363 |
+
"logits/rejected": -1.4259040355682373,
|
| 364 |
+
"logps/chosen": -384.1799011230469,
|
| 365 |
+
"logps/rejected": -397.27679443359375,
|
| 366 |
+
"loss": 0.5351,
|
| 367 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 368 |
+
"rewards/chosen": -0.9452459216117859,
|
| 369 |
+
"rewards/margins": 0.6696128845214844,
|
| 370 |
+
"rewards/rejected": -1.614858865737915,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4604918890633176,
|
| 375 |
+
"grad_norm": 4.130067348480225,
|
| 376 |
+
"learning_rate": 3.265574537815398e-06,
|
| 377 |
+
"logits/chosen": -1.387083649635315,
|
| 378 |
+
"logits/rejected": -1.3616085052490234,
|
| 379 |
+
"logps/chosen": -346.65240478515625,
|
| 380 |
+
"logps/rejected": -429.26007080078125,
|
| 381 |
+
"loss": 0.5306,
|
| 382 |
+
"rewards/accuracies": 0.731249988079071,
|
| 383 |
+
"rewards/chosen": -1.253117561340332,
|
| 384 |
+
"rewards/margins": 0.7737840414047241,
|
| 385 |
+
"rewards/rejected": -2.0269014835357666,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.48142333856619574,
|
| 390 |
+
"grad_norm": 3.835031032562256,
|
| 391 |
+
"learning_rate": 3.089397338773569e-06,
|
| 392 |
+
"logits/chosen": -1.5156991481781006,
|
| 393 |
+
"logits/rejected": -1.389380693435669,
|
| 394 |
+
"logps/chosen": -364.519287109375,
|
| 395 |
+
"logps/rejected": -443.21649169921875,
|
| 396 |
+
"loss": 0.5296,
|
| 397 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 398 |
+
"rewards/chosen": -1.2139475345611572,
|
| 399 |
+
"rewards/margins": 0.9700647592544556,
|
| 400 |
+
"rewards/rejected": -2.1840121746063232,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.5023547880690737,
|
| 405 |
+
"grad_norm": 3.9292032718658447,
|
| 406 |
+
"learning_rate": 2.9100607788275547e-06,
|
| 407 |
+
"logits/chosen": -1.4860846996307373,
|
| 408 |
+
"logits/rejected": -1.3620604276657104,
|
| 409 |
+
"logps/chosen": -396.4831848144531,
|
| 410 |
+
"logps/rejected": -455.0039978027344,
|
| 411 |
+
"loss": 0.4992,
|
| 412 |
+
"rewards/accuracies": 0.731249988079071,
|
| 413 |
+
"rewards/chosen": -1.2688108682632446,
|
| 414 |
+
"rewards/margins": 0.8345935940742493,
|
| 415 |
+
"rewards/rejected": -2.1034045219421387,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5232862375719518,
|
| 420 |
+
"grad_norm": 3.9835152626037598,
|
| 421 |
+
"learning_rate": 2.72852616010567e-06,
|
| 422 |
+
"logits/chosen": -1.3683382272720337,
|
| 423 |
+
"logits/rejected": -1.203263282775879,
|
| 424 |
+
"logps/chosen": -439.71734619140625,
|
| 425 |
+
"logps/rejected": -500.26995849609375,
|
| 426 |
+
"loss": 0.5193,
|
| 427 |
+
"rewards/accuracies": 0.762499988079071,
|
| 428 |
+
"rewards/chosen": -1.8103727102279663,
|
| 429 |
+
"rewards/margins": 0.9737499356269836,
|
| 430 |
+
"rewards/rejected": -2.7841227054595947,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.54421768707483,
|
| 435 |
+
"grad_norm": 4.053774356842041,
|
| 436 |
+
"learning_rate": 2.5457665670441937e-06,
|
| 437 |
+
"logits/chosen": -1.3506557941436768,
|
| 438 |
+
"logits/rejected": -1.3202815055847168,
|
| 439 |
+
"logps/chosen": -425.1206970214844,
|
| 440 |
+
"logps/rejected": -482.37060546875,
|
| 441 |
+
"loss": 0.5157,
|
| 442 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 443 |
+
"rewards/chosen": -1.7851899862289429,
|
| 444 |
+
"rewards/margins": 0.7175502181053162,
|
| 445 |
+
"rewards/rejected": -2.5027403831481934,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.565149136577708,
|
| 450 |
+
"grad_norm": 3.4300835132598877,
|
| 451 |
+
"learning_rate": 2.3627616503391813e-06,
|
| 452 |
+
"logits/chosen": -1.3187822103500366,
|
| 453 |
+
"logits/rejected": -1.2432146072387695,
|
| 454 |
+
"logps/chosen": -435.5782775878906,
|
| 455 |
+
"logps/rejected": -497.5511169433594,
|
| 456 |
+
"loss": 0.5274,
|
| 457 |
+
"rewards/accuracies": 0.737500011920929,
|
| 458 |
+
"rewards/chosen": -1.6222509145736694,
|
| 459 |
+
"rewards/margins": 0.8602613210678101,
|
| 460 |
+
"rewards/rejected": -2.4825127124786377,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5860805860805861,
|
| 465 |
+
"grad_norm": 4.350644588470459,
|
| 466 |
+
"learning_rate": 2.1804923757009885e-06,
|
| 467 |
+
"logits/chosen": -1.173998236656189,
|
| 468 |
+
"logits/rejected": -1.0935077667236328,
|
| 469 |
+
"logps/chosen": -396.106201171875,
|
| 470 |
+
"logps/rejected": -447.62359619140625,
|
| 471 |
+
"loss": 0.5269,
|
| 472 |
+
"rewards/accuracies": 0.706250011920929,
|
| 473 |
+
"rewards/chosen": -1.5709511041641235,
|
| 474 |
+
"rewards/margins": 0.7292364835739136,
|
| 475 |
+
"rewards/rejected": -2.300187587738037,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.6070120355834642,
|
| 480 |
+
"grad_norm": 3.7417469024658203,
|
| 481 |
+
"learning_rate": 1.9999357655598894e-06,
|
| 482 |
+
"logits/chosen": -1.29354727268219,
|
| 483 |
+
"logits/rejected": -1.1772781610488892,
|
| 484 |
+
"logps/chosen": -411.03814697265625,
|
| 485 |
+
"logps/rejected": -484.0711975097656,
|
| 486 |
+
"loss": 0.5207,
|
| 487 |
+
"rewards/accuracies": 0.737500011920929,
|
| 488 |
+
"rewards/chosen": -1.7882314920425415,
|
| 489 |
+
"rewards/margins": 0.787774920463562,
|
| 490 |
+
"rewards/rejected": -2.5760064125061035,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6279434850863422,
|
| 495 |
+
"grad_norm": 5.170950889587402,
|
| 496 |
+
"learning_rate": 1.8220596619089576e-06,
|
| 497 |
+
"logits/chosen": -1.2801265716552734,
|
| 498 |
+
"logits/rejected": -1.1654409170150757,
|
| 499 |
+
"logps/chosen": -476.93023681640625,
|
| 500 |
+
"logps/rejected": -508.8740234375,
|
| 501 |
+
"loss": 0.4977,
|
| 502 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 503 |
+
"rewards/chosen": -1.7797939777374268,
|
| 504 |
+
"rewards/margins": 0.7282718420028687,
|
| 505 |
+
"rewards/rejected": -2.508065700531006,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6279434850863422,
|
| 510 |
+
"eval_logits/chosen": -1.187766194343567,
|
| 511 |
+
"eval_logits/rejected": -1.1155990362167358,
|
| 512 |
+
"eval_logps/chosen": -434.6387939453125,
|
| 513 |
+
"eval_logps/rejected": -531.358642578125,
|
| 514 |
+
"eval_loss": 0.5100404620170593,
|
| 515 |
+
"eval_rewards/accuracies": 0.77734375,
|
| 516 |
+
"eval_rewards/chosen": -1.882094383239746,
|
| 517 |
+
"eval_rewards/margins": 0.9561675786972046,
|
| 518 |
+
"eval_rewards/rejected": -2.838261842727661,
|
| 519 |
+
"eval_runtime": 110.6981,
|
| 520 |
+
"eval_samples_per_second": 18.067,
|
| 521 |
+
"eval_steps_per_second": 0.289,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6488749345892203,
|
| 526 |
+
"grad_norm": 4.2551469802856445,
|
| 527 |
+
"learning_rate": 1.647817538357072e-06,
|
| 528 |
+
"logits/chosen": -1.2920277118682861,
|
| 529 |
+
"logits/rejected": -1.1881086826324463,
|
| 530 |
+
"logps/chosen": -491.37213134765625,
|
| 531 |
+
"logps/rejected": -541.3331298828125,
|
| 532 |
+
"loss": 0.4999,
|
| 533 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 534 |
+
"rewards/chosen": -2.065791130065918,
|
| 535 |
+
"rewards/margins": 1.0087900161743164,
|
| 536 |
+
"rewards/rejected": -3.0745811462402344,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6698063840920984,
|
| 541 |
+
"grad_norm": 5.16890811920166,
|
| 542 |
+
"learning_rate": 1.4781433892011132e-06,
|
| 543 |
+
"logits/chosen": -1.2393107414245605,
|
| 544 |
+
"logits/rejected": -1.1026521921157837,
|
| 545 |
+
"logps/chosen": -445.18408203125,
|
| 546 |
+
"logps/rejected": -523.7984619140625,
|
| 547 |
+
"loss": 0.4901,
|
| 548 |
+
"rewards/accuracies": 0.800000011920929,
|
| 549 |
+
"rewards/chosen": -2.0319957733154297,
|
| 550 |
+
"rewards/margins": 1.0875332355499268,
|
| 551 |
+
"rewards/rejected": -3.1195292472839355,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6907378335949764,
|
| 556 |
+
"grad_norm": 4.174256324768066,
|
| 557 |
+
"learning_rate": 1.3139467229135999e-06,
|
| 558 |
+
"logits/chosen": -1.3440932035446167,
|
| 559 |
+
"logits/rejected": -1.3159363269805908,
|
| 560 |
+
"logps/chosen": -418.71905517578125,
|
| 561 |
+
"logps/rejected": -499.96612548828125,
|
| 562 |
+
"loss": 0.5117,
|
| 563 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 564 |
+
"rewards/chosen": -1.6654974222183228,
|
| 565 |
+
"rewards/margins": 0.8459057807922363,
|
| 566 |
+
"rewards/rejected": -2.5114030838012695,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7116692830978545,
|
| 571 |
+
"grad_norm": 4.793274402618408,
|
| 572 |
+
"learning_rate": 1.1561076868822756e-06,
|
| 573 |
+
"logits/chosen": -1.2274696826934814,
|
| 574 |
+
"logits/rejected": -1.1634485721588135,
|
| 575 |
+
"logps/chosen": -472.03167724609375,
|
| 576 |
+
"logps/rejected": -538.4515380859375,
|
| 577 |
+
"loss": 0.5043,
|
| 578 |
+
"rewards/accuracies": 0.731249988079071,
|
| 579 |
+
"rewards/chosen": -2.003690719604492,
|
| 580 |
+
"rewards/margins": 0.8766867518424988,
|
| 581 |
+
"rewards/rejected": -2.8803775310516357,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7326007326007326,
|
| 586 |
+
"grad_norm": 4.15632963180542,
|
| 587 |
+
"learning_rate": 1.0054723495346484e-06,
|
| 588 |
+
"logits/chosen": -1.2941699028015137,
|
| 589 |
+
"logits/rejected": -1.1815686225891113,
|
| 590 |
+
"logps/chosen": -464.3369140625,
|
| 591 |
+
"logps/rejected": -537.1565551757812,
|
| 592 |
+
"loss": 0.467,
|
| 593 |
+
"rewards/accuracies": 0.793749988079071,
|
| 594 |
+
"rewards/chosen": -2.2074246406555176,
|
| 595 |
+
"rewards/margins": 0.9847939610481262,
|
| 596 |
+
"rewards/rejected": -3.192218542098999,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7535321821036107,
|
| 601 |
+
"grad_norm": 3.9081270694732666,
|
| 602 |
+
"learning_rate": 8.628481651367876e-07,
|
| 603 |
+
"logits/chosen": -1.2783113718032837,
|
| 604 |
+
"logits/rejected": -1.1381876468658447,
|
| 605 |
+
"logps/chosen": -522.1004028320312,
|
| 606 |
+
"logps/rejected": -580.3312377929688,
|
| 607 |
+
"loss": 0.5229,
|
| 608 |
+
"rewards/accuracies": 0.762499988079071,
|
| 609 |
+
"rewards/chosen": -2.310837745666504,
|
| 610 |
+
"rewards/margins": 1.0041909217834473,
|
| 611 |
+
"rewards/rejected": -3.315028429031372,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7744636316064888,
|
| 616 |
+
"grad_norm": 4.175972938537598,
|
| 617 |
+
"learning_rate": 7.289996455765749e-07,
|
| 618 |
+
"logits/chosen": -1.305639386177063,
|
| 619 |
+
"logits/rejected": -1.1913295984268188,
|
| 620 |
+
"logps/chosen": -472.50103759765625,
|
| 621 |
+
"logps/rejected": -539.5281372070312,
|
| 622 |
+
"loss": 0.515,
|
| 623 |
+
"rewards/accuracies": 0.762499988079071,
|
| 624 |
+
"rewards/chosen": -2.2103240489959717,
|
| 625 |
+
"rewards/margins": 0.9552016258239746,
|
| 626 |
+
"rewards/rejected": -3.1655256748199463,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7953950811093669,
|
| 631 |
+
"grad_norm": 4.144151210784912,
|
| 632 |
+
"learning_rate": 6.046442623320145e-07,
|
| 633 |
+
"logits/chosen": -1.1470415592193604,
|
| 634 |
+
"logits/rejected": -1.1134142875671387,
|
| 635 |
+
"logps/chosen": -463.9144592285156,
|
| 636 |
+
"logps/rejected": -589.644287109375,
|
| 637 |
+
"loss": 0.4996,
|
| 638 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 639 |
+
"rewards/chosen": -2.329439401626587,
|
| 640 |
+
"rewards/margins": 1.0926482677459717,
|
| 641 |
+
"rewards/rejected": -3.4220874309539795,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8163265306122449,
|
| 646 |
+
"grad_norm": 3.9941041469573975,
|
| 647 |
+
"learning_rate": 4.904486005914027e-07,
|
| 648 |
+
"logits/chosen": -1.3255889415740967,
|
| 649 |
+
"logits/rejected": -1.2115201950073242,
|
| 650 |
+
"logps/chosen": -541.8289794921875,
|
| 651 |
+
"logps/rejected": -589.1190185546875,
|
| 652 |
+
"loss": 0.4868,
|
| 653 |
+
"rewards/accuracies": 0.71875,
|
| 654 |
+
"rewards/chosen": -2.216815710067749,
|
| 655 |
+
"rewards/margins": 0.8514670133590698,
|
| 656 |
+
"rewards/rejected": -3.0682828426361084,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.837257980115123,
|
| 661 |
+
"grad_norm": 4.9559149742126465,
|
| 662 |
+
"learning_rate": 3.8702478614051353e-07,
|
| 663 |
+
"logits/chosen": -1.2322378158569336,
|
| 664 |
+
"logits/rejected": -1.1362693309783936,
|
| 665 |
+
"logps/chosen": -454.14727783203125,
|
| 666 |
+
"logps/rejected": -534.5038452148438,
|
| 667 |
+
"loss": 0.5096,
|
| 668 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 669 |
+
"rewards/chosen": -2.1453731060028076,
|
| 670 |
+
"rewards/margins": 0.9312618374824524,
|
| 671 |
+
"rewards/rejected": -3.0766348838806152,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.837257980115123,
|
| 676 |
+
"eval_logits/chosen": -1.2569077014923096,
|
| 677 |
+
"eval_logits/rejected": -1.1850428581237793,
|
| 678 |
+
"eval_logps/chosen": -448.560302734375,
|
| 679 |
+
"eval_logps/rejected": -549.236328125,
|
| 680 |
+
"eval_loss": 0.5035256743431091,
|
| 681 |
+
"eval_rewards/accuracies": 0.765625,
|
| 682 |
+
"eval_rewards/chosen": -2.0213088989257812,
|
| 683 |
+
"eval_rewards/margins": 0.9957298040390015,
|
| 684 |
+
"eval_rewards/rejected": -3.017038583755493,
|
| 685 |
+
"eval_runtime": 110.7002,
|
| 686 |
+
"eval_samples_per_second": 18.067,
|
| 687 |
+
"eval_steps_per_second": 0.289,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.858189429618001,
|
| 692 |
+
"grad_norm": 4.206878185272217,
|
| 693 |
+
"learning_rate": 2.9492720416985004e-07,
|
| 694 |
+
"logits/chosen": -1.3825483322143555,
|
| 695 |
+
"logits/rejected": -1.2790623903274536,
|
| 696 |
+
"logps/chosen": -492.615234375,
|
| 697 |
+
"logps/rejected": -567.6419067382812,
|
| 698 |
+
"loss": 0.5101,
|
| 699 |
+
"rewards/accuracies": 0.768750011920929,
|
| 700 |
+
"rewards/chosen": -2.1133949756622314,
|
| 701 |
+
"rewards/margins": 1.0342962741851807,
|
| 702 |
+
"rewards/rejected": -3.147691249847412,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8791208791208791,
|
| 707 |
+
"grad_norm": 4.6710124015808105,
|
| 708 |
+
"learning_rate": 2.1464952759020857e-07,
|
| 709 |
+
"logits/chosen": -1.2218900918960571,
|
| 710 |
+
"logits/rejected": -1.1765925884246826,
|
| 711 |
+
"logps/chosen": -447.31451416015625,
|
| 712 |
+
"logps/rejected": -551.3411865234375,
|
| 713 |
+
"loss": 0.5055,
|
| 714 |
+
"rewards/accuracies": 0.737500011920929,
|
| 715 |
+
"rewards/chosen": -2.0722880363464355,
|
| 716 |
+
"rewards/margins": 0.9839819073677063,
|
| 717 |
+
"rewards/rejected": -3.056270122528076,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.9000523286237572,
|
| 722 |
+
"grad_norm": 3.861823081970215,
|
| 723 |
+
"learning_rate": 1.4662207078575685e-07,
|
| 724 |
+
"logits/chosen": -1.2683221101760864,
|
| 725 |
+
"logits/rejected": -1.1284939050674438,
|
| 726 |
+
"logps/chosen": -473.9938049316406,
|
| 727 |
+
"logps/rejected": -547.7625122070312,
|
| 728 |
+
"loss": 0.478,
|
| 729 |
+
"rewards/accuracies": 0.768750011920929,
|
| 730 |
+
"rewards/chosen": -1.9979991912841797,
|
| 731 |
+
"rewards/margins": 0.9907487630844116,
|
| 732 |
+
"rewards/rejected": -2.988748073577881,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9209837781266352,
|
| 737 |
+
"grad_norm": 4.48820686340332,
|
| 738 |
+
"learning_rate": 9.120948298936422e-08,
|
| 739 |
+
"logits/chosen": -1.4051973819732666,
|
| 740 |
+
"logits/rejected": -1.2804924249649048,
|
| 741 |
+
"logps/chosen": -447.3819274902344,
|
| 742 |
+
"logps/rejected": -505.71221923828125,
|
| 743 |
+
"loss": 0.4913,
|
| 744 |
+
"rewards/accuracies": 0.699999988079071,
|
| 745 |
+
"rewards/chosen": -2.0068001747131348,
|
| 746 |
+
"rewards/margins": 0.9424559473991394,
|
| 747 |
+
"rewards/rejected": -2.94925594329834,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9419152276295133,
|
| 752 |
+
"grad_norm": 4.315399169921875,
|
| 753 |
+
"learning_rate": 4.870879364444109e-08,
|
| 754 |
+
"logits/chosen": -1.1320288181304932,
|
| 755 |
+
"logits/rejected": -1.0678789615631104,
|
| 756 |
+
"logps/chosen": -466.1886291503906,
|
| 757 |
+
"logps/rejected": -553.8982543945312,
|
| 758 |
+
"loss": 0.4949,
|
| 759 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 760 |
+
"rewards/chosen": -2.0924811363220215,
|
| 761 |
+
"rewards/margins": 0.9840418696403503,
|
| 762 |
+
"rewards/rejected": -3.0765230655670166,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9628466771323915,
|
| 767 |
+
"grad_norm": 4.0728840827941895,
|
| 768 |
+
"learning_rate": 1.93478202307823e-08,
|
| 769 |
+
"logits/chosen": -1.290541410446167,
|
| 770 |
+
"logits/rejected": -1.133880376815796,
|
| 771 |
+
"logps/chosen": -460.96710205078125,
|
| 772 |
+
"logps/rejected": -513.3797607421875,
|
| 773 |
+
"loss": 0.5042,
|
| 774 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 775 |
+
"rewards/chosen": -2.039780855178833,
|
| 776 |
+
"rewards/margins": 0.937634289264679,
|
| 777 |
+
"rewards/rejected": -2.977414608001709,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9837781266352695,
|
| 782 |
+
"grad_norm": 3.986497640609741,
|
| 783 |
+
"learning_rate": 3.283947088983663e-09,
|
| 784 |
+
"logits/chosen": -1.2861273288726807,
|
| 785 |
+
"logits/rejected": -1.1986534595489502,
|
| 786 |
+
"logps/chosen": -484.498046875,
|
| 787 |
+
"logps/rejected": -541.9881591796875,
|
| 788 |
+
"loss": 0.4914,
|
| 789 |
+
"rewards/accuracies": 0.6875,
|
| 790 |
+
"rewards/chosen": -2.0398573875427246,
|
| 791 |
+
"rewards/margins": 0.8533897399902344,
|
| 792 |
+
"rewards/rejected": -2.893247127532959,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.9984301412872841,
|
| 797 |
+
"step": 477,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5421361258444796,
|
| 800 |
+
"train_runtime": 7660.497,
|
| 801 |
+
"train_samples_per_second": 7.98,
|
| 802 |
+
"train_steps_per_second": 0.062
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 477,
|
| 807 |
+
"num_input_tokens_seen": 0,
|
| 808 |
+
"num_train_epochs": 1,
|
| 809 |
+
"save_steps": 100,
|
| 810 |
+
"stateful_callbacks": {
|
| 811 |
+
"TrainerControl": {
|
| 812 |
+
"args": {
|
| 813 |
+
"should_epoch_stop": false,
|
| 814 |
+
"should_evaluate": false,
|
| 815 |
+
"should_log": false,
|
| 816 |
+
"should_save": true,
|
| 817 |
+
"should_training_stop": true
|
| 818 |
+
},
|
| 819 |
+
"attributes": {}
|
| 820 |
+
}
|
| 821 |
+
},
|
| 822 |
+
"total_flos": 0.0,
|
| 823 |
+
"train_batch_size": 4,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|