Model save
Browse files- README.md +78 -0
- all_results.json +9 -0
- generation_config.json +12 -0
- train_results.json +9 -0
- trainer_state.json +798 -0
README.md
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: llama3.1
|
| 4 |
+
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- cpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: llama3.1-cpo-full-0913
|
| 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 |
+
# llama3.1-cpo-full-0913
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 1.5947
|
| 22 |
+
- Rewards/chosen: -15.5964
|
| 23 |
+
- Rewards/rejected: -16.3155
|
| 24 |
+
- Rewards/accuracies: 0.6261
|
| 25 |
+
- Rewards/margins: 0.7192
|
| 26 |
+
- Logps/rejected: -163.1553
|
| 27 |
+
- Logps/chosen: -155.9637
|
| 28 |
+
- Logits/rejected: -0.4910
|
| 29 |
+
- Logits/chosen: -0.5144
|
| 30 |
+
- Nll Loss: 0.4262
|
| 31 |
+
|
| 32 |
+
## Model description
|
| 33 |
+
|
| 34 |
+
More information needed
|
| 35 |
+
|
| 36 |
+
## Intended uses & limitations
|
| 37 |
+
|
| 38 |
+
More information needed
|
| 39 |
+
|
| 40 |
+
## Training and evaluation data
|
| 41 |
+
|
| 42 |
+
More information needed
|
| 43 |
+
|
| 44 |
+
## Training procedure
|
| 45 |
+
|
| 46 |
+
### Training hyperparameters
|
| 47 |
+
|
| 48 |
+
The following hyperparameters were used during training:
|
| 49 |
+
- learning_rate: 1e-06
|
| 50 |
+
- train_batch_size: 4
|
| 51 |
+
- eval_batch_size: 4
|
| 52 |
+
- seed: 42
|
| 53 |
+
- distributed_type: multi-GPU
|
| 54 |
+
- num_devices: 4
|
| 55 |
+
- gradient_accumulation_steps: 8
|
| 56 |
+
- total_train_batch_size: 128
|
| 57 |
+
- total_eval_batch_size: 16
|
| 58 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 59 |
+
- lr_scheduler_type: linear
|
| 60 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 61 |
+
- num_epochs: 1
|
| 62 |
+
|
| 63 |
+
### Training results
|
| 64 |
+
|
| 65 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Nll Loss |
|
| 66 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:--------:|
|
| 67 |
+
| 1.9304 | 0.2311 | 100 | 1.7873 | -14.9945 | -15.3576 | 0.5804 | 0.3632 | -153.5762 | -149.9445 | -0.3649 | -0.3854 | 0.4085 |
|
| 68 |
+
| 1.6908 | 0.4623 | 200 | 1.6702 | -15.6437 | -16.2439 | 0.5978 | 0.6002 | -162.4385 | -156.4369 | -0.3777 | -0.4014 | 0.4252 |
|
| 69 |
+
| 1.6317 | 0.6934 | 300 | 1.6162 | -15.4682 | -16.1519 | 0.6152 | 0.6837 | -161.5185 | -154.6818 | -0.4753 | -0.4948 | 0.4202 |
|
| 70 |
+
| 1.62 | 0.9246 | 400 | 1.5947 | -15.5964 | -16.3155 | 0.6261 | 0.7192 | -163.1553 | -155.9637 | -0.4910 | -0.5144 | 0.4262 |
|
| 71 |
+
|
| 72 |
+
|
| 73 |
+
### Framework versions
|
| 74 |
+
|
| 75 |
+
- Transformers 4.44.2
|
| 76 |
+
- Pytorch 2.3.1
|
| 77 |
+
- Datasets 2.21.0
|
| 78 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9985553308292401,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 1.7731637126869626,
|
| 5 |
+
"train_runtime": 10231.9294,
|
| 6 |
+
"train_samples": 55376,
|
| 7 |
+
"train_samples_per_second": 5.412,
|
| 8 |
+
"train_steps_per_second": 0.042
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 128000,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
128001,
|
| 6 |
+
128008,
|
| 7 |
+
128009
|
| 8 |
+
],
|
| 9 |
+
"temperature": 0.6,
|
| 10 |
+
"top_p": 0.9,
|
| 11 |
+
"transformers_version": "4.44.2"
|
| 12 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9985553308292401,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 1.7731637126869626,
|
| 5 |
+
"train_runtime": 10231.9294,
|
| 6 |
+
"train_samples": 55376,
|
| 7 |
+
"train_samples_per_second": 5.412,
|
| 8 |
+
"train_steps_per_second": 0.042
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,798 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9985553308292401,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 432,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.023114706732158336,
|
| 13 |
+
"grad_norm": 66.92803955078125,
|
| 14 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 15 |
+
"logits/chosen": -0.33564168214797974,
|
| 16 |
+
"logits/rejected": -0.3153206706047058,
|
| 17 |
+
"logps/chosen": -269.33428955078125,
|
| 18 |
+
"logps/rejected": -267.60894775390625,
|
| 19 |
+
"loss": 2.6157,
|
| 20 |
+
"nll_loss": 0.741317629814148,
|
| 21 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 22 |
+
"rewards/chosen": -26.933429718017578,
|
| 23 |
+
"rewards/margins": -0.17253029346466064,
|
| 24 |
+
"rewards/rejected": -26.76089859008789,
|
| 25 |
+
"step": 10
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.04622941346431667,
|
| 29 |
+
"grad_norm": 54.904842376708984,
|
| 30 |
+
"learning_rate": 4.545454545454545e-07,
|
| 31 |
+
"logits/chosen": -0.3472834527492523,
|
| 32 |
+
"logits/rejected": -0.3292314112186432,
|
| 33 |
+
"logps/chosen": -260.78680419921875,
|
| 34 |
+
"logps/rejected": -267.32977294921875,
|
| 35 |
+
"loss": 2.5223,
|
| 36 |
+
"nll_loss": 0.7186762094497681,
|
| 37 |
+
"rewards/accuracies": 0.565625011920929,
|
| 38 |
+
"rewards/chosen": -26.078683853149414,
|
| 39 |
+
"rewards/margins": 0.6542952060699463,
|
| 40 |
+
"rewards/rejected": -26.732980728149414,
|
| 41 |
+
"step": 20
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.06934412019647501,
|
| 45 |
+
"grad_norm": 57.42607498168945,
|
| 46 |
+
"learning_rate": 6.818181818181817e-07,
|
| 47 |
+
"logits/chosen": -0.3462437689304352,
|
| 48 |
+
"logits/rejected": -0.334714412689209,
|
| 49 |
+
"logps/chosen": -247.49801635742188,
|
| 50 |
+
"logps/rejected": -250.79483032226562,
|
| 51 |
+
"loss": 2.3549,
|
| 52 |
+
"nll_loss": 0.7035976052284241,
|
| 53 |
+
"rewards/accuracies": 0.515625,
|
| 54 |
+
"rewards/chosen": -24.749801635742188,
|
| 55 |
+
"rewards/margins": 0.3296825885772705,
|
| 56 |
+
"rewards/rejected": -25.079483032226562,
|
| 57 |
+
"step": 30
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.09245882692863334,
|
| 61 |
+
"grad_norm": 46.98875427246094,
|
| 62 |
+
"learning_rate": 9.09090909090909e-07,
|
| 63 |
+
"logits/chosen": -0.5729629993438721,
|
| 64 |
+
"logits/rejected": -0.5595733523368835,
|
| 65 |
+
"logps/chosen": -215.0082244873047,
|
| 66 |
+
"logps/rejected": -216.9280548095703,
|
| 67 |
+
"loss": 2.1725,
|
| 68 |
+
"nll_loss": 0.6498099565505981,
|
| 69 |
+
"rewards/accuracies": 0.518750011920929,
|
| 70 |
+
"rewards/chosen": -21.500822067260742,
|
| 71 |
+
"rewards/margins": 0.19198258221149445,
|
| 72 |
+
"rewards/rejected": -21.692806243896484,
|
| 73 |
+
"step": 40
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.11557353366079168,
|
| 77 |
+
"grad_norm": 47.18745803833008,
|
| 78 |
+
"learning_rate": 9.845360824742267e-07,
|
| 79 |
+
"logits/chosen": -0.8356858491897583,
|
| 80 |
+
"logits/rejected": -0.8102104067802429,
|
| 81 |
+
"logps/chosen": -196.44061279296875,
|
| 82 |
+
"logps/rejected": -195.4991455078125,
|
| 83 |
+
"loss": 2.1857,
|
| 84 |
+
"nll_loss": 0.5283800959587097,
|
| 85 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 86 |
+
"rewards/chosen": -19.644062042236328,
|
| 87 |
+
"rewards/margins": -0.09414808452129364,
|
| 88 |
+
"rewards/rejected": -19.549915313720703,
|
| 89 |
+
"step": 50
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.13868824039295002,
|
| 93 |
+
"grad_norm": 55.287715911865234,
|
| 94 |
+
"learning_rate": 9.587628865979382e-07,
|
| 95 |
+
"logits/chosen": -0.6823571920394897,
|
| 96 |
+
"logits/rejected": -0.68729567527771,
|
| 97 |
+
"logps/chosen": -164.70936584472656,
|
| 98 |
+
"logps/rejected": -164.9478302001953,
|
| 99 |
+
"loss": 2.0037,
|
| 100 |
+
"nll_loss": 0.4647987484931946,
|
| 101 |
+
"rewards/accuracies": 0.534375011920929,
|
| 102 |
+
"rewards/chosen": -16.470935821533203,
|
| 103 |
+
"rewards/margins": 0.02384711429476738,
|
| 104 |
+
"rewards/rejected": -16.494781494140625,
|
| 105 |
+
"step": 60
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.16180294712510834,
|
| 109 |
+
"grad_norm": 51.4761962890625,
|
| 110 |
+
"learning_rate": 9.329896907216495e-07,
|
| 111 |
+
"logits/chosen": -0.5190773010253906,
|
| 112 |
+
"logits/rejected": -0.49602770805358887,
|
| 113 |
+
"logps/chosen": -155.39878845214844,
|
| 114 |
+
"logps/rejected": -157.3554229736328,
|
| 115 |
+
"loss": 1.9943,
|
| 116 |
+
"nll_loss": 0.4453979432582855,
|
| 117 |
+
"rewards/accuracies": 0.518750011920929,
|
| 118 |
+
"rewards/chosen": -15.539876937866211,
|
| 119 |
+
"rewards/margins": 0.195662721991539,
|
| 120 |
+
"rewards/rejected": -15.735540390014648,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.1849176538572667,
|
| 125 |
+
"grad_norm": 46.25474166870117,
|
| 126 |
+
"learning_rate": 9.072164948453608e-07,
|
| 127 |
+
"logits/chosen": -0.4949778914451599,
|
| 128 |
+
"logits/rejected": -0.47060757875442505,
|
| 129 |
+
"logps/chosen": -158.2879180908203,
|
| 130 |
+
"logps/rejected": -161.47145080566406,
|
| 131 |
+
"loss": 1.8594,
|
| 132 |
+
"nll_loss": 0.4298928380012512,
|
| 133 |
+
"rewards/accuracies": 0.5093749761581421,
|
| 134 |
+
"rewards/chosen": -15.828791618347168,
|
| 135 |
+
"rewards/margins": 0.3183526396751404,
|
| 136 |
+
"rewards/rejected": -16.147144317626953,
|
| 137 |
+
"step": 80
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.208032360589425,
|
| 141 |
+
"grad_norm": 49.99284744262695,
|
| 142 |
+
"learning_rate": 8.814432989690721e-07,
|
| 143 |
+
"logits/chosen": -0.37455958127975464,
|
| 144 |
+
"logits/rejected": -0.3673579692840576,
|
| 145 |
+
"logps/chosen": -153.99154663085938,
|
| 146 |
+
"logps/rejected": -161.9925079345703,
|
| 147 |
+
"loss": 1.7748,
|
| 148 |
+
"nll_loss": 0.42108353972435,
|
| 149 |
+
"rewards/accuracies": 0.59375,
|
| 150 |
+
"rewards/chosen": -15.39915657043457,
|
| 151 |
+
"rewards/margins": 0.8000966310501099,
|
| 152 |
+
"rewards/rejected": -16.19925308227539,
|
| 153 |
+
"step": 90
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.23114706732158335,
|
| 157 |
+
"grad_norm": 47.385650634765625,
|
| 158 |
+
"learning_rate": 8.556701030927834e-07,
|
| 159 |
+
"logits/chosen": -0.3636273443698883,
|
| 160 |
+
"logits/rejected": -0.3537142872810364,
|
| 161 |
+
"logps/chosen": -153.48483276367188,
|
| 162 |
+
"logps/rejected": -156.58370971679688,
|
| 163 |
+
"loss": 1.9304,
|
| 164 |
+
"nll_loss": 0.4187200665473938,
|
| 165 |
+
"rewards/accuracies": 0.5406249761581421,
|
| 166 |
+
"rewards/chosen": -15.348485946655273,
|
| 167 |
+
"rewards/margins": 0.3098832070827484,
|
| 168 |
+
"rewards/rejected": -15.658369064331055,
|
| 169 |
+
"step": 100
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.23114706732158335,
|
| 173 |
+
"eval_logits/chosen": -0.3853626549243927,
|
| 174 |
+
"eval_logits/rejected": -0.36488524079322815,
|
| 175 |
+
"eval_logps/chosen": -149.94454956054688,
|
| 176 |
+
"eval_logps/rejected": -153.576171875,
|
| 177 |
+
"eval_loss": 1.7872967720031738,
|
| 178 |
+
"eval_nll_loss": 0.4085230827331543,
|
| 179 |
+
"eval_rewards/accuracies": 0.5804347991943359,
|
| 180 |
+
"eval_rewards/chosen": -14.994454383850098,
|
| 181 |
+
"eval_rewards/margins": 0.36316320300102234,
|
| 182 |
+
"eval_rewards/rejected": -15.357619285583496,
|
| 183 |
+
"eval_runtime": 73.8265,
|
| 184 |
+
"eval_samples_per_second": 24.734,
|
| 185 |
+
"eval_steps_per_second": 1.558,
|
| 186 |
+
"step": 100
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 0.2542617740537417,
|
| 190 |
+
"grad_norm": 47.490840911865234,
|
| 191 |
+
"learning_rate": 8.298969072164948e-07,
|
| 192 |
+
"logits/chosen": -0.3666607141494751,
|
| 193 |
+
"logits/rejected": -0.3409901261329651,
|
| 194 |
+
"logps/chosen": -147.7798309326172,
|
| 195 |
+
"logps/rejected": -152.14883422851562,
|
| 196 |
+
"loss": 1.7164,
|
| 197 |
+
"nll_loss": 0.40307506918907166,
|
| 198 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 199 |
+
"rewards/chosen": -14.777984619140625,
|
| 200 |
+
"rewards/margins": 0.4368988573551178,
|
| 201 |
+
"rewards/rejected": -15.214881896972656,
|
| 202 |
+
"step": 110
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 0.27737648078590005,
|
| 206 |
+
"grad_norm": 45.41544723510742,
|
| 207 |
+
"learning_rate": 8.041237113402062e-07,
|
| 208 |
+
"logits/chosen": -0.36932340264320374,
|
| 209 |
+
"logits/rejected": -0.3461097776889801,
|
| 210 |
+
"logps/chosen": -155.2689208984375,
|
| 211 |
+
"logps/rejected": -156.53543090820312,
|
| 212 |
+
"loss": 1.7549,
|
| 213 |
+
"nll_loss": 0.42143669724464417,
|
| 214 |
+
"rewards/accuracies": 0.534375011920929,
|
| 215 |
+
"rewards/chosen": -15.526891708374023,
|
| 216 |
+
"rewards/margins": 0.12664994597434998,
|
| 217 |
+
"rewards/rejected": -15.653543472290039,
|
| 218 |
+
"step": 120
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 0.30049118751805837,
|
| 222 |
+
"grad_norm": 55.190162658691406,
|
| 223 |
+
"learning_rate": 7.783505154639175e-07,
|
| 224 |
+
"logits/chosen": -0.36812376976013184,
|
| 225 |
+
"logits/rejected": -0.3466120660305023,
|
| 226 |
+
"logps/chosen": -153.25393676757812,
|
| 227 |
+
"logps/rejected": -161.2375030517578,
|
| 228 |
+
"loss": 1.8234,
|
| 229 |
+
"nll_loss": 0.4233035147190094,
|
| 230 |
+
"rewards/accuracies": 0.59375,
|
| 231 |
+
"rewards/chosen": -15.325393676757812,
|
| 232 |
+
"rewards/margins": 0.7983576655387878,
|
| 233 |
+
"rewards/rejected": -16.123750686645508,
|
| 234 |
+
"step": 130
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.3236058942502167,
|
| 238 |
+
"grad_norm": 58.641231536865234,
|
| 239 |
+
"learning_rate": 7.525773195876288e-07,
|
| 240 |
+
"logits/chosen": -0.42011794447898865,
|
| 241 |
+
"logits/rejected": -0.41268259286880493,
|
| 242 |
+
"logps/chosen": -144.4340362548828,
|
| 243 |
+
"logps/rejected": -149.6342010498047,
|
| 244 |
+
"loss": 1.7976,
|
| 245 |
+
"nll_loss": 0.4138007164001465,
|
| 246 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 247 |
+
"rewards/chosen": -14.443403244018555,
|
| 248 |
+
"rewards/margins": 0.520018458366394,
|
| 249 |
+
"rewards/rejected": -14.963422775268555,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.34672060098237506,
|
| 254 |
+
"grad_norm": 59.50709533691406,
|
| 255 |
+
"learning_rate": 7.268041237113402e-07,
|
| 256 |
+
"logits/chosen": -0.45735687017440796,
|
| 257 |
+
"logits/rejected": -0.44332581758499146,
|
| 258 |
+
"logps/chosen": -141.5747528076172,
|
| 259 |
+
"logps/rejected": -147.01950073242188,
|
| 260 |
+
"loss": 1.8755,
|
| 261 |
+
"nll_loss": 0.41054767370224,
|
| 262 |
+
"rewards/accuracies": 0.578125,
|
| 263 |
+
"rewards/chosen": -14.157475471496582,
|
| 264 |
+
"rewards/margins": 0.5444743037223816,
|
| 265 |
+
"rewards/rejected": -14.701950073242188,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3698353077145334,
|
| 270 |
+
"grad_norm": 40.405235290527344,
|
| 271 |
+
"learning_rate": 7.010309278350515e-07,
|
| 272 |
+
"logits/chosen": -0.4439857602119446,
|
| 273 |
+
"logits/rejected": -0.41624826192855835,
|
| 274 |
+
"logps/chosen": -156.5723876953125,
|
| 275 |
+
"logps/rejected": -160.45640563964844,
|
| 276 |
+
"loss": 1.7024,
|
| 277 |
+
"nll_loss": 0.4132766127586365,
|
| 278 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 279 |
+
"rewards/chosen": -15.65723991394043,
|
| 280 |
+
"rewards/margins": 0.38840025663375854,
|
| 281 |
+
"rewards/rejected": -16.045639038085938,
|
| 282 |
+
"step": 160
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3929500144466917,
|
| 286 |
+
"grad_norm": 51.05006790161133,
|
| 287 |
+
"learning_rate": 6.752577319587629e-07,
|
| 288 |
+
"logits/chosen": -0.39237576723098755,
|
| 289 |
+
"logits/rejected": -0.3862777650356293,
|
| 290 |
+
"logps/chosen": -152.97549438476562,
|
| 291 |
+
"logps/rejected": -160.07443237304688,
|
| 292 |
+
"loss": 1.599,
|
| 293 |
+
"nll_loss": 0.41846928000450134,
|
| 294 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 295 |
+
"rewards/chosen": -15.2975492477417,
|
| 296 |
+
"rewards/margins": 0.7098936438560486,
|
| 297 |
+
"rewards/rejected": -16.007442474365234,
|
| 298 |
+
"step": 170
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 0.41606472117885,
|
| 302 |
+
"grad_norm": 48.777130126953125,
|
| 303 |
+
"learning_rate": 6.494845360824742e-07,
|
| 304 |
+
"logits/chosen": -0.4046599864959717,
|
| 305 |
+
"logits/rejected": -0.4020842909812927,
|
| 306 |
+
"logps/chosen": -148.8914794921875,
|
| 307 |
+
"logps/rejected": -155.31307983398438,
|
| 308 |
+
"loss": 1.6521,
|
| 309 |
+
"nll_loss": 0.4291691780090332,
|
| 310 |
+
"rewards/accuracies": 0.565625011920929,
|
| 311 |
+
"rewards/chosen": -14.889147758483887,
|
| 312 |
+
"rewards/margins": 0.6421611309051514,
|
| 313 |
+
"rewards/rejected": -15.531309127807617,
|
| 314 |
+
"step": 180
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 0.4391794279110084,
|
| 318 |
+
"grad_norm": 51.11661911010742,
|
| 319 |
+
"learning_rate": 6.237113402061855e-07,
|
| 320 |
+
"logits/chosen": -0.42425212264060974,
|
| 321 |
+
"logits/rejected": -0.4136783480644226,
|
| 322 |
+
"logps/chosen": -156.27357482910156,
|
| 323 |
+
"logps/rejected": -162.0409393310547,
|
| 324 |
+
"loss": 1.7189,
|
| 325 |
+
"nll_loss": 0.42820248007774353,
|
| 326 |
+
"rewards/accuracies": 0.5843750238418579,
|
| 327 |
+
"rewards/chosen": -15.627357482910156,
|
| 328 |
+
"rewards/margins": 0.5767360329627991,
|
| 329 |
+
"rewards/rejected": -16.2040958404541,
|
| 330 |
+
"step": 190
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"epoch": 0.4622941346431667,
|
| 334 |
+
"grad_norm": 45.06605529785156,
|
| 335 |
+
"learning_rate": 5.979381443298969e-07,
|
| 336 |
+
"logits/chosen": -0.3681766986846924,
|
| 337 |
+
"logits/rejected": -0.36281704902648926,
|
| 338 |
+
"logps/chosen": -157.00155639648438,
|
| 339 |
+
"logps/rejected": -163.3647003173828,
|
| 340 |
+
"loss": 1.6908,
|
| 341 |
+
"nll_loss": 0.4334492087364197,
|
| 342 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 343 |
+
"rewards/chosen": -15.700152397155762,
|
| 344 |
+
"rewards/margins": 0.6363152265548706,
|
| 345 |
+
"rewards/rejected": -16.336469650268555,
|
| 346 |
+
"step": 200
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"epoch": 0.4622941346431667,
|
| 350 |
+
"eval_logits/chosen": -0.4014091491699219,
|
| 351 |
+
"eval_logits/rejected": -0.3777381181716919,
|
| 352 |
+
"eval_logps/chosen": -156.43685913085938,
|
| 353 |
+
"eval_logps/rejected": -162.43849182128906,
|
| 354 |
+
"eval_loss": 1.6701573133468628,
|
| 355 |
+
"eval_nll_loss": 0.42524340748786926,
|
| 356 |
+
"eval_rewards/accuracies": 0.5978260636329651,
|
| 357 |
+
"eval_rewards/chosen": -15.643685340881348,
|
| 358 |
+
"eval_rewards/margins": 0.6001652479171753,
|
| 359 |
+
"eval_rewards/rejected": -16.243852615356445,
|
| 360 |
+
"eval_runtime": 73.8669,
|
| 361 |
+
"eval_samples_per_second": 24.72,
|
| 362 |
+
"eval_steps_per_second": 1.557,
|
| 363 |
+
"step": 200
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.48540884137532503,
|
| 367 |
+
"grad_norm": 52.39714813232422,
|
| 368 |
+
"learning_rate": 5.721649484536082e-07,
|
| 369 |
+
"logits/chosen": -0.4187684953212738,
|
| 370 |
+
"logits/rejected": -0.39357370138168335,
|
| 371 |
+
"logps/chosen": -151.34727478027344,
|
| 372 |
+
"logps/rejected": -153.41477966308594,
|
| 373 |
+
"loss": 1.6691,
|
| 374 |
+
"nll_loss": 0.42135825753211975,
|
| 375 |
+
"rewards/accuracies": 0.546875,
|
| 376 |
+
"rewards/chosen": -15.134727478027344,
|
| 377 |
+
"rewards/margins": 0.2067503184080124,
|
| 378 |
+
"rewards/rejected": -15.34147834777832,
|
| 379 |
+
"step": 210
|
| 380 |
+
},
|
| 381 |
+
{
|
| 382 |
+
"epoch": 0.5085235481074833,
|
| 383 |
+
"grad_norm": 48.871177673339844,
|
| 384 |
+
"learning_rate": 5.463917525773195e-07,
|
| 385 |
+
"logits/chosen": -0.46896496415138245,
|
| 386 |
+
"logits/rejected": -0.4411422312259674,
|
| 387 |
+
"logps/chosen": -160.63766479492188,
|
| 388 |
+
"logps/rejected": -166.47012329101562,
|
| 389 |
+
"loss": 1.7268,
|
| 390 |
+
"nll_loss": 0.43222665786743164,
|
| 391 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 392 |
+
"rewards/chosen": -16.063764572143555,
|
| 393 |
+
"rewards/margins": 0.5832474231719971,
|
| 394 |
+
"rewards/rejected": -16.64701271057129,
|
| 395 |
+
"step": 220
|
| 396 |
+
},
|
| 397 |
+
{
|
| 398 |
+
"epoch": 0.5316382548396418,
|
| 399 |
+
"grad_norm": 46.8618278503418,
|
| 400 |
+
"learning_rate": 5.20618556701031e-07,
|
| 401 |
+
"logits/chosen": -0.4806763529777527,
|
| 402 |
+
"logits/rejected": -0.4655001163482666,
|
| 403 |
+
"logps/chosen": -164.1845245361328,
|
| 404 |
+
"logps/rejected": -170.3590087890625,
|
| 405 |
+
"loss": 1.753,
|
| 406 |
+
"nll_loss": 0.4462898373603821,
|
| 407 |
+
"rewards/accuracies": 0.5843750238418579,
|
| 408 |
+
"rewards/chosen": -16.4184513092041,
|
| 409 |
+
"rewards/margins": 0.6174517869949341,
|
| 410 |
+
"rewards/rejected": -17.03590202331543,
|
| 411 |
+
"step": 230
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 0.5547529615718001,
|
| 415 |
+
"grad_norm": 48.733802795410156,
|
| 416 |
+
"learning_rate": 4.948453608247422e-07,
|
| 417 |
+
"logits/chosen": -0.42387205362319946,
|
| 418 |
+
"logits/rejected": -0.4110351502895355,
|
| 419 |
+
"logps/chosen": -159.4880828857422,
|
| 420 |
+
"logps/rejected": -164.54312133789062,
|
| 421 |
+
"loss": 1.5882,
|
| 422 |
+
"nll_loss": 0.4322621822357178,
|
| 423 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 424 |
+
"rewards/chosen": -15.948808670043945,
|
| 425 |
+
"rewards/margins": 0.5055034756660461,
|
| 426 |
+
"rewards/rejected": -16.45431137084961,
|
| 427 |
+
"step": 240
|
| 428 |
+
},
|
| 429 |
+
{
|
| 430 |
+
"epoch": 0.5778676683039584,
|
| 431 |
+
"grad_norm": 53.66923141479492,
|
| 432 |
+
"learning_rate": 4.6907216494845357e-07,
|
| 433 |
+
"logits/chosen": -0.45519933104515076,
|
| 434 |
+
"logits/rejected": -0.44147634506225586,
|
| 435 |
+
"logps/chosen": -156.13479614257812,
|
| 436 |
+
"logps/rejected": -159.08212280273438,
|
| 437 |
+
"loss": 1.7404,
|
| 438 |
+
"nll_loss": 0.4152873158454895,
|
| 439 |
+
"rewards/accuracies": 0.5406249761581421,
|
| 440 |
+
"rewards/chosen": -15.613479614257812,
|
| 441 |
+
"rewards/margins": 0.29473432898521423,
|
| 442 |
+
"rewards/rejected": -15.908210754394531,
|
| 443 |
+
"step": 250
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.6009823750361167,
|
| 447 |
+
"grad_norm": 50.42735290527344,
|
| 448 |
+
"learning_rate": 4.432989690721649e-07,
|
| 449 |
+
"logits/chosen": -0.5512745976448059,
|
| 450 |
+
"logits/rejected": -0.5374751091003418,
|
| 451 |
+
"logps/chosen": -152.5029296875,
|
| 452 |
+
"logps/rejected": -160.7849884033203,
|
| 453 |
+
"loss": 1.5895,
|
| 454 |
+
"nll_loss": 0.4251280725002289,
|
| 455 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 456 |
+
"rewards/chosen": -15.250292778015137,
|
| 457 |
+
"rewards/margins": 0.8282074928283691,
|
| 458 |
+
"rewards/rejected": -16.07849884033203,
|
| 459 |
+
"step": 260
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 0.624097081768275,
|
| 463 |
+
"grad_norm": 42.25716781616211,
|
| 464 |
+
"learning_rate": 4.175257731958763e-07,
|
| 465 |
+
"logits/chosen": -0.5728802680969238,
|
| 466 |
+
"logits/rejected": -0.5652969479560852,
|
| 467 |
+
"logps/chosen": -153.87515258789062,
|
| 468 |
+
"logps/rejected": -160.92990112304688,
|
| 469 |
+
"loss": 1.622,
|
| 470 |
+
"nll_loss": 0.42097169160842896,
|
| 471 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 472 |
+
"rewards/chosen": -15.3875150680542,
|
| 473 |
+
"rewards/margins": 0.7054744362831116,
|
| 474 |
+
"rewards/rejected": -16.09299087524414,
|
| 475 |
+
"step": 270
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.6472117885004334,
|
| 479 |
+
"grad_norm": 52.24223327636719,
|
| 480 |
+
"learning_rate": 3.917525773195876e-07,
|
| 481 |
+
"logits/chosen": -0.5287462472915649,
|
| 482 |
+
"logits/rejected": -0.5079108476638794,
|
| 483 |
+
"logps/chosen": -155.820068359375,
|
| 484 |
+
"logps/rejected": -165.27578735351562,
|
| 485 |
+
"loss": 1.6145,
|
| 486 |
+
"nll_loss": 0.42984142899513245,
|
| 487 |
+
"rewards/accuracies": 0.643750011920929,
|
| 488 |
+
"rewards/chosen": -15.582005500793457,
|
| 489 |
+
"rewards/margins": 0.9455726742744446,
|
| 490 |
+
"rewards/rejected": -16.527578353881836,
|
| 491 |
+
"step": 280
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6703264952325917,
|
| 495 |
+
"grad_norm": 54.0615119934082,
|
| 496 |
+
"learning_rate": 3.659793814432989e-07,
|
| 497 |
+
"logits/chosen": -0.5829291343688965,
|
| 498 |
+
"logits/rejected": -0.5588080883026123,
|
| 499 |
+
"logps/chosen": -160.86459350585938,
|
| 500 |
+
"logps/rejected": -166.34828186035156,
|
| 501 |
+
"loss": 1.5808,
|
| 502 |
+
"nll_loss": 0.4233360290527344,
|
| 503 |
+
"rewards/accuracies": 0.609375,
|
| 504 |
+
"rewards/chosen": -16.08646011352539,
|
| 505 |
+
"rewards/margins": 0.5483680963516235,
|
| 506 |
+
"rewards/rejected": -16.63482666015625,
|
| 507 |
+
"step": 290
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 0.6934412019647501,
|
| 511 |
+
"grad_norm": 43.99635696411133,
|
| 512 |
+
"learning_rate": 3.402061855670103e-07,
|
| 513 |
+
"logits/chosen": -0.5929441452026367,
|
| 514 |
+
"logits/rejected": -0.585766613483429,
|
| 515 |
+
"logps/chosen": -164.1534423828125,
|
| 516 |
+
"logps/rejected": -173.26889038085938,
|
| 517 |
+
"loss": 1.6317,
|
| 518 |
+
"nll_loss": 0.43874359130859375,
|
| 519 |
+
"rewards/accuracies": 0.609375,
|
| 520 |
+
"rewards/chosen": -16.415346145629883,
|
| 521 |
+
"rewards/margins": 0.911544144153595,
|
| 522 |
+
"rewards/rejected": -17.326889038085938,
|
| 523 |
+
"step": 300
|
| 524 |
+
},
|
| 525 |
+
{
|
| 526 |
+
"epoch": 0.6934412019647501,
|
| 527 |
+
"eval_logits/chosen": -0.49481379985809326,
|
| 528 |
+
"eval_logits/rejected": -0.4753292500972748,
|
| 529 |
+
"eval_logps/chosen": -154.68177795410156,
|
| 530 |
+
"eval_logps/rejected": -161.5185089111328,
|
| 531 |
+
"eval_loss": 1.6161738634109497,
|
| 532 |
+
"eval_nll_loss": 0.42018982768058777,
|
| 533 |
+
"eval_rewards/accuracies": 0.615217387676239,
|
| 534 |
+
"eval_rewards/chosen": -15.468178749084473,
|
| 535 |
+
"eval_rewards/margins": 0.6836734414100647,
|
| 536 |
+
"eval_rewards/rejected": -16.151851654052734,
|
| 537 |
+
"eval_runtime": 74.0087,
|
| 538 |
+
"eval_samples_per_second": 24.673,
|
| 539 |
+
"eval_steps_per_second": 1.554,
|
| 540 |
+
"step": 300
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"epoch": 0.7165559086969084,
|
| 544 |
+
"grad_norm": 47.26166534423828,
|
| 545 |
+
"learning_rate": 3.1443298969072163e-07,
|
| 546 |
+
"logits/chosen": -0.5505023002624512,
|
| 547 |
+
"logits/rejected": -0.5488861799240112,
|
| 548 |
+
"logps/chosen": -164.3349609375,
|
| 549 |
+
"logps/rejected": -167.97909545898438,
|
| 550 |
+
"loss": 1.6871,
|
| 551 |
+
"nll_loss": 0.42464059591293335,
|
| 552 |
+
"rewards/accuracies": 0.550000011920929,
|
| 553 |
+
"rewards/chosen": -16.433496475219727,
|
| 554 |
+
"rewards/margins": 0.36441320180892944,
|
| 555 |
+
"rewards/rejected": -16.797908782958984,
|
| 556 |
+
"step": 310
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"epoch": 0.7396706154290668,
|
| 560 |
+
"grad_norm": 44.56684875488281,
|
| 561 |
+
"learning_rate": 2.8865979381443296e-07,
|
| 562 |
+
"logits/chosen": -0.5739372372627258,
|
| 563 |
+
"logits/rejected": -0.5621416568756104,
|
| 564 |
+
"logps/chosen": -154.22021484375,
|
| 565 |
+
"logps/rejected": -162.76797485351562,
|
| 566 |
+
"loss": 1.6709,
|
| 567 |
+
"nll_loss": 0.42187362909317017,
|
| 568 |
+
"rewards/accuracies": 0.6343749761581421,
|
| 569 |
+
"rewards/chosen": -15.422021865844727,
|
| 570 |
+
"rewards/margins": 0.8547781109809875,
|
| 571 |
+
"rewards/rejected": -16.276798248291016,
|
| 572 |
+
"step": 320
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 0.7627853221612251,
|
| 576 |
+
"grad_norm": 48.82611083984375,
|
| 577 |
+
"learning_rate": 2.6288659793814435e-07,
|
| 578 |
+
"logits/chosen": -0.584633469581604,
|
| 579 |
+
"logits/rejected": -0.5864993333816528,
|
| 580 |
+
"logps/chosen": -154.85739135742188,
|
| 581 |
+
"logps/rejected": -161.60858154296875,
|
| 582 |
+
"loss": 1.6549,
|
| 583 |
+
"nll_loss": 0.43096083402633667,
|
| 584 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 585 |
+
"rewards/chosen": -15.485738754272461,
|
| 586 |
+
"rewards/margins": 0.6751174330711365,
|
| 587 |
+
"rewards/rejected": -16.160858154296875,
|
| 588 |
+
"step": 330
|
| 589 |
+
},
|
| 590 |
+
{
|
| 591 |
+
"epoch": 0.7859000288933834,
|
| 592 |
+
"grad_norm": 46.865325927734375,
|
| 593 |
+
"learning_rate": 2.3711340206185566e-07,
|
| 594 |
+
"logits/chosen": -0.5921510457992554,
|
| 595 |
+
"logits/rejected": -0.5866528153419495,
|
| 596 |
+
"logps/chosen": -157.51187133789062,
|
| 597 |
+
"logps/rejected": -162.7019500732422,
|
| 598 |
+
"loss": 1.5515,
|
| 599 |
+
"nll_loss": 0.4305364489555359,
|
| 600 |
+
"rewards/accuracies": 0.5718749761581421,
|
| 601 |
+
"rewards/chosen": -15.751187324523926,
|
| 602 |
+
"rewards/margins": 0.5190097093582153,
|
| 603 |
+
"rewards/rejected": -16.27019500732422,
|
| 604 |
+
"step": 340
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"epoch": 0.8090147356255417,
|
| 608 |
+
"grad_norm": 45.32524871826172,
|
| 609 |
+
"learning_rate": 2.11340206185567e-07,
|
| 610 |
+
"logits/chosen": -0.574793815612793,
|
| 611 |
+
"logits/rejected": -0.575833261013031,
|
| 612 |
+
"logps/chosen": -158.30972290039062,
|
| 613 |
+
"logps/rejected": -167.1160888671875,
|
| 614 |
+
"loss": 1.6559,
|
| 615 |
+
"nll_loss": 0.4133908152580261,
|
| 616 |
+
"rewards/accuracies": 0.6031249761581421,
|
| 617 |
+
"rewards/chosen": -15.830973625183105,
|
| 618 |
+
"rewards/margins": 0.8806363940238953,
|
| 619 |
+
"rewards/rejected": -16.71160888671875,
|
| 620 |
+
"step": 350
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 0.8321294423577,
|
| 624 |
+
"grad_norm": 53.514156341552734,
|
| 625 |
+
"learning_rate": 1.8556701030927835e-07,
|
| 626 |
+
"logits/chosen": -0.5572192668914795,
|
| 627 |
+
"logits/rejected": -0.5455694794654846,
|
| 628 |
+
"logps/chosen": -161.35818481445312,
|
| 629 |
+
"logps/rejected": -166.35061645507812,
|
| 630 |
+
"loss": 1.7006,
|
| 631 |
+
"nll_loss": 0.4319698214530945,
|
| 632 |
+
"rewards/accuracies": 0.606249988079071,
|
| 633 |
+
"rewards/chosen": -16.135818481445312,
|
| 634 |
+
"rewards/margins": 0.4992440342903137,
|
| 635 |
+
"rewards/rejected": -16.635061264038086,
|
| 636 |
+
"step": 360
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"epoch": 0.8552441490898585,
|
| 640 |
+
"grad_norm": 47.664703369140625,
|
| 641 |
+
"learning_rate": 1.5979381443298966e-07,
|
| 642 |
+
"logits/chosen": -0.5517255067825317,
|
| 643 |
+
"logits/rejected": -0.5260570049285889,
|
| 644 |
+
"logps/chosen": -155.55886840820312,
|
| 645 |
+
"logps/rejected": -165.2543487548828,
|
| 646 |
+
"loss": 1.5564,
|
| 647 |
+
"nll_loss": 0.42134684324264526,
|
| 648 |
+
"rewards/accuracies": 0.659375011920929,
|
| 649 |
+
"rewards/chosen": -15.555887222290039,
|
| 650 |
+
"rewards/margins": 0.9695472717285156,
|
| 651 |
+
"rewards/rejected": -16.525434494018555,
|
| 652 |
+
"step": 370
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"epoch": 0.8783588558220168,
|
| 656 |
+
"grad_norm": 48.99809646606445,
|
| 657 |
+
"learning_rate": 1.3402061855670102e-07,
|
| 658 |
+
"logits/chosen": -0.5434561967849731,
|
| 659 |
+
"logits/rejected": -0.5287705063819885,
|
| 660 |
+
"logps/chosen": -154.3162841796875,
|
| 661 |
+
"logps/rejected": -161.20472717285156,
|
| 662 |
+
"loss": 1.5817,
|
| 663 |
+
"nll_loss": 0.431587278842926,
|
| 664 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 665 |
+
"rewards/chosen": -15.431628227233887,
|
| 666 |
+
"rewards/margins": 0.6888439655303955,
|
| 667 |
+
"rewards/rejected": -16.120471954345703,
|
| 668 |
+
"step": 380
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 0.9014735625541751,
|
| 672 |
+
"grad_norm": 52.46923065185547,
|
| 673 |
+
"learning_rate": 1.0824742268041237e-07,
|
| 674 |
+
"logits/chosen": -0.5911422967910767,
|
| 675 |
+
"logits/rejected": -0.5752480626106262,
|
| 676 |
+
"logps/chosen": -158.2704620361328,
|
| 677 |
+
"logps/rejected": -168.7931671142578,
|
| 678 |
+
"loss": 1.6055,
|
| 679 |
+
"nll_loss": 0.42453208565711975,
|
| 680 |
+
"rewards/accuracies": 0.625,
|
| 681 |
+
"rewards/chosen": -15.827044486999512,
|
| 682 |
+
"rewards/margins": 1.0522701740264893,
|
| 683 |
+
"rewards/rejected": -16.879314422607422,
|
| 684 |
+
"step": 390
|
| 685 |
+
},
|
| 686 |
+
{
|
| 687 |
+
"epoch": 0.9245882692863334,
|
| 688 |
+
"grad_norm": 50.29128646850586,
|
| 689 |
+
"learning_rate": 8.24742268041237e-08,
|
| 690 |
+
"logits/chosen": -0.5491029024124146,
|
| 691 |
+
"logits/rejected": -0.5335959196090698,
|
| 692 |
+
"logps/chosen": -162.01458740234375,
|
| 693 |
+
"logps/rejected": -167.11669921875,
|
| 694 |
+
"loss": 1.62,
|
| 695 |
+
"nll_loss": 0.44402360916137695,
|
| 696 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 697 |
+
"rewards/chosen": -16.201457977294922,
|
| 698 |
+
"rewards/margins": 0.5102119445800781,
|
| 699 |
+
"rewards/rejected": -16.711669921875,
|
| 700 |
+
"step": 400
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"epoch": 0.9245882692863334,
|
| 704 |
+
"eval_logits/chosen": -0.514378547668457,
|
| 705 |
+
"eval_logits/rejected": -0.49102360010147095,
|
| 706 |
+
"eval_logps/chosen": -155.9636688232422,
|
| 707 |
+
"eval_logps/rejected": -163.15530395507812,
|
| 708 |
+
"eval_loss": 1.5946580171585083,
|
| 709 |
+
"eval_nll_loss": 0.4261849522590637,
|
| 710 |
+
"eval_rewards/accuracies": 0.626086950302124,
|
| 711 |
+
"eval_rewards/chosen": -15.596366882324219,
|
| 712 |
+
"eval_rewards/margins": 0.7191624045372009,
|
| 713 |
+
"eval_rewards/rejected": -16.315528869628906,
|
| 714 |
+
"eval_runtime": 73.9588,
|
| 715 |
+
"eval_samples_per_second": 24.689,
|
| 716 |
+
"eval_steps_per_second": 1.555,
|
| 717 |
+
"step": 400
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.9477029760184917,
|
| 721 |
+
"grad_norm": 49.76476287841797,
|
| 722 |
+
"learning_rate": 5.670103092783505e-08,
|
| 723 |
+
"logits/chosen": -0.47304850816726685,
|
| 724 |
+
"logits/rejected": -0.46452435851097107,
|
| 725 |
+
"logps/chosen": -148.42233276367188,
|
| 726 |
+
"logps/rejected": -157.0984649658203,
|
| 727 |
+
"loss": 1.5582,
|
| 728 |
+
"nll_loss": 0.42344313859939575,
|
| 729 |
+
"rewards/accuracies": 0.6031249761581421,
|
| 730 |
+
"rewards/chosen": -14.842233657836914,
|
| 731 |
+
"rewards/margins": 0.8676150441169739,
|
| 732 |
+
"rewards/rejected": -15.709848403930664,
|
| 733 |
+
"step": 410
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9708176827506501,
|
| 737 |
+
"grad_norm": 45.05620574951172,
|
| 738 |
+
"learning_rate": 3.092783505154639e-08,
|
| 739 |
+
"logits/chosen": -0.5102072954177856,
|
| 740 |
+
"logits/rejected": -0.4865845739841461,
|
| 741 |
+
"logps/chosen": -150.34286499023438,
|
| 742 |
+
"logps/rejected": -156.001708984375,
|
| 743 |
+
"loss": 1.5674,
|
| 744 |
+
"nll_loss": 0.4283737242221832,
|
| 745 |
+
"rewards/accuracies": 0.590624988079071,
|
| 746 |
+
"rewards/chosen": -15.03428840637207,
|
| 747 |
+
"rewards/margins": 0.5658840537071228,
|
| 748 |
+
"rewards/rejected": -15.600171089172363,
|
| 749 |
+
"step": 420
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 0.9939323894828085,
|
| 753 |
+
"grad_norm": 46.765380859375,
|
| 754 |
+
"learning_rate": 5.154639175257731e-09,
|
| 755 |
+
"logits/chosen": -0.5253250002861023,
|
| 756 |
+
"logits/rejected": -0.534714937210083,
|
| 757 |
+
"logps/chosen": -157.79779052734375,
|
| 758 |
+
"logps/rejected": -164.88742065429688,
|
| 759 |
+
"loss": 1.645,
|
| 760 |
+
"nll_loss": 0.42819744348526,
|
| 761 |
+
"rewards/accuracies": 0.6031249761581421,
|
| 762 |
+
"rewards/chosen": -15.779779434204102,
|
| 763 |
+
"rewards/margins": 0.7089639902114868,
|
| 764 |
+
"rewards/rejected": -16.48874282836914,
|
| 765 |
+
"step": 430
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"epoch": 0.9985553308292401,
|
| 769 |
+
"step": 432,
|
| 770 |
+
"total_flos": 0.0,
|
| 771 |
+
"train_loss": 1.7731637126869626,
|
| 772 |
+
"train_runtime": 10231.9294,
|
| 773 |
+
"train_samples_per_second": 5.412,
|
| 774 |
+
"train_steps_per_second": 0.042
|
| 775 |
+
}
|
| 776 |
+
],
|
| 777 |
+
"logging_steps": 10,
|
| 778 |
+
"max_steps": 432,
|
| 779 |
+
"num_input_tokens_seen": 0,
|
| 780 |
+
"num_train_epochs": 1,
|
| 781 |
+
"save_steps": 100,
|
| 782 |
+
"stateful_callbacks": {
|
| 783 |
+
"TrainerControl": {
|
| 784 |
+
"args": {
|
| 785 |
+
"should_epoch_stop": false,
|
| 786 |
+
"should_evaluate": false,
|
| 787 |
+
"should_log": false,
|
| 788 |
+
"should_save": true,
|
| 789 |
+
"should_training_stop": true
|
| 790 |
+
},
|
| 791 |
+
"attributes": {}
|
| 792 |
+
}
|
| 793 |
+
},
|
| 794 |
+
"total_flos": 0.0,
|
| 795 |
+
"train_batch_size": 4,
|
| 796 |
+
"trial_name": null,
|
| 797 |
+
"trial_params": null
|
| 798 |
+
}
|