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_j-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_j-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: 1.4447
|
| 22 |
+
- Rewards/chosen: -16.0215
|
| 23 |
+
- Rewards/rejected: -16.8041
|
| 24 |
+
- Rewards/accuracies: 0.6261
|
| 25 |
+
- Rewards/margins: 0.7826
|
| 26 |
+
- Logps/rejected: -168.0413
|
| 27 |
+
- Logps/chosen: -160.2150
|
| 28 |
+
- Logits/rejected: -0.3389
|
| 29 |
+
- Logits/chosen: -0.3606
|
| 30 |
+
- Nll Loss: 0.2798
|
| 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 |
+
| 1.7833 | 0.2311 | 100 | 1.6389 | -15.3939 | -15.7792 | 0.5783 | 0.3853 | -157.7921 | -153.9390 | -0.3065 | -0.3333 | 0.2678 |
|
| 68 |
+
| 1.5321 | 0.4623 | 200 | 1.5242 | -15.8988 | -16.5121 | 0.5978 | 0.6132 | -165.1206 | -158.9884 | -0.4244 | -0.4423 | 0.2764 |
|
| 69 |
+
| 1.4722 | 0.6934 | 300 | 1.4633 | -16.0803 | -16.8141 | 0.6217 | 0.7338 | -168.1411 | -160.8031 | -0.3641 | -0.3856 | 0.2790 |
|
| 70 |
+
| 1.4589 | 0.9246 | 400 | 1.4447 | -16.0215 | -16.8041 | 0.6261 | 0.7826 | -168.0413 | -160.2150 | -0.3389 | -0.3606 | 0.2798 |
|
| 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": 1.6635627029118714,
|
| 5 |
+
"train_runtime": 10783.4889,
|
| 6 |
+
"train_samples": 55376,
|
| 7 |
+
"train_samples_per_second": 5.135,
|
| 8 |
+
"train_steps_per_second": 0.04
|
| 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": 1.6635627029118714,
|
| 5 |
+
"train_runtime": 10783.4889,
|
| 6 |
+
"train_samples": 55376,
|
| 7 |
+
"train_samples_per_second": 5.135,
|
| 8 |
+
"train_steps_per_second": 0.04
|
| 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": 67.55843353271484,
|
| 14 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 15 |
+
"logits/chosen": -0.33515310287475586,
|
| 16 |
+
"logits/rejected": -0.31493520736694336,
|
| 17 |
+
"logps/chosen": -269.29345703125,
|
| 18 |
+
"logps/rejected": -267.6085205078125,
|
| 19 |
+
"loss": 2.9222,
|
| 20 |
+
"nll_loss": 1.0535928010940552,
|
| 21 |
+
"rewards/accuracies": 0.4906249940395355,
|
| 22 |
+
"rewards/chosen": -26.929346084594727,
|
| 23 |
+
"rewards/margins": -0.16849184036254883,
|
| 24 |
+
"rewards/rejected": -26.760854721069336,
|
| 25 |
+
"step": 10
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.04622941346431667,
|
| 29 |
+
"grad_norm": 57.353782653808594,
|
| 30 |
+
"learning_rate": 4.545454545454545e-07,
|
| 31 |
+
"logits/chosen": -0.3512418866157532,
|
| 32 |
+
"logits/rejected": -0.3330627977848053,
|
| 33 |
+
"logps/chosen": -260.02117919921875,
|
| 34 |
+
"logps/rejected": -266.5718078613281,
|
| 35 |
+
"loss": 2.8619,
|
| 36 |
+
"nll_loss": 0.9778251647949219,
|
| 37 |
+
"rewards/accuracies": 0.565625011920929,
|
| 38 |
+
"rewards/chosen": -26.002117156982422,
|
| 39 |
+
"rewards/margins": 0.6550623178482056,
|
| 40 |
+
"rewards/rejected": -26.657180786132812,
|
| 41 |
+
"step": 20
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.06934412019647501,
|
| 45 |
+
"grad_norm": 57.872154235839844,
|
| 46 |
+
"learning_rate": 6.818181818181817e-07,
|
| 47 |
+
"logits/chosen": -0.38214609026908875,
|
| 48 |
+
"logits/rejected": -0.36934739351272583,
|
| 49 |
+
"logps/chosen": -243.16397094726562,
|
| 50 |
+
"logps/rejected": -246.39990234375,
|
| 51 |
+
"loss": 2.6585,
|
| 52 |
+
"nll_loss": 1.018157958984375,
|
| 53 |
+
"rewards/accuracies": 0.512499988079071,
|
| 54 |
+
"rewards/chosen": -24.316394805908203,
|
| 55 |
+
"rewards/margins": 0.3235955238342285,
|
| 56 |
+
"rewards/rejected": -24.639989852905273,
|
| 57 |
+
"step": 30
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.09245882692863334,
|
| 61 |
+
"grad_norm": 51.343589782714844,
|
| 62 |
+
"learning_rate": 9.09090909090909e-07,
|
| 63 |
+
"logits/chosen": -0.6919909119606018,
|
| 64 |
+
"logits/rejected": -0.6750722527503967,
|
| 65 |
+
"logps/chosen": -202.0379180908203,
|
| 66 |
+
"logps/rejected": -203.7409210205078,
|
| 67 |
+
"loss": 2.385,
|
| 68 |
+
"nll_loss": 0.866622805595398,
|
| 69 |
+
"rewards/accuracies": 0.5218750238418579,
|
| 70 |
+
"rewards/chosen": -20.203792572021484,
|
| 71 |
+
"rewards/margins": 0.17029908299446106,
|
| 72 |
+
"rewards/rejected": -20.37409019470215,
|
| 73 |
+
"step": 40
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.11557353366079168,
|
| 77 |
+
"grad_norm": 45.946136474609375,
|
| 78 |
+
"learning_rate": 9.845360824742267e-07,
|
| 79 |
+
"logits/chosen": -0.8110788464546204,
|
| 80 |
+
"logits/rejected": -0.7866124510765076,
|
| 81 |
+
"logps/chosen": -176.52212524414062,
|
| 82 |
+
"logps/rejected": -175.50418090820312,
|
| 83 |
+
"loss": 2.1668,
|
| 84 |
+
"nll_loss": 0.4587995409965515,
|
| 85 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 86 |
+
"rewards/chosen": -17.65221405029297,
|
| 87 |
+
"rewards/margins": -0.10179616510868073,
|
| 88 |
+
"rewards/rejected": -17.550418853759766,
|
| 89 |
+
"step": 50
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.13868824039295002,
|
| 93 |
+
"grad_norm": 55.89133834838867,
|
| 94 |
+
"learning_rate": 9.587628865979382e-07,
|
| 95 |
+
"logits/chosen": -0.635937511920929,
|
| 96 |
+
"logits/rejected": -0.6386198997497559,
|
| 97 |
+
"logps/chosen": -158.75927734375,
|
| 98 |
+
"logps/rejected": -159.04640197753906,
|
| 99 |
+
"loss": 1.9369,
|
| 100 |
+
"nll_loss": 0.41199779510498047,
|
| 101 |
+
"rewards/accuracies": 0.528124988079071,
|
| 102 |
+
"rewards/chosen": -15.87592887878418,
|
| 103 |
+
"rewards/margins": 0.028712665662169456,
|
| 104 |
+
"rewards/rejected": -15.904638290405273,
|
| 105 |
+
"step": 60
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.16180294712510834,
|
| 109 |
+
"grad_norm": 51.20368957519531,
|
| 110 |
+
"learning_rate": 9.329896907216495e-07,
|
| 111 |
+
"logits/chosen": -0.46112680435180664,
|
| 112 |
+
"logits/rejected": -0.4352455139160156,
|
| 113 |
+
"logps/chosen": -154.1770782470703,
|
| 114 |
+
"logps/rejected": -156.22377014160156,
|
| 115 |
+
"loss": 1.8871,
|
| 116 |
+
"nll_loss": 0.3339909613132477,
|
| 117 |
+
"rewards/accuracies": 0.5218750238418579,
|
| 118 |
+
"rewards/chosen": -15.417707443237305,
|
| 119 |
+
"rewards/margins": 0.20467153191566467,
|
| 120 |
+
"rewards/rejected": -15.622377395629883,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.1849176538572667,
|
| 125 |
+
"grad_norm": 46.16290283203125,
|
| 126 |
+
"learning_rate": 9.072164948453608e-07,
|
| 127 |
+
"logits/chosen": -0.36660072207450867,
|
| 128 |
+
"logits/rejected": -0.3421391248703003,
|
| 129 |
+
"logps/chosen": -158.54318237304688,
|
| 130 |
+
"logps/rejected": -161.6597442626953,
|
| 131 |
+
"loss": 1.7235,
|
| 132 |
+
"nll_loss": 0.2865411043167114,
|
| 133 |
+
"rewards/accuracies": 0.5093749761581421,
|
| 134 |
+
"rewards/chosen": -15.85431957244873,
|
| 135 |
+
"rewards/margins": 0.311655193567276,
|
| 136 |
+
"rewards/rejected": -16.165973663330078,
|
| 137 |
+
"step": 80
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.208032360589425,
|
| 141 |
+
"grad_norm": 49.6414680480957,
|
| 142 |
+
"learning_rate": 8.814432989690721e-07,
|
| 143 |
+
"logits/chosen": -0.3728088140487671,
|
| 144 |
+
"logits/rejected": -0.3610958456993103,
|
| 145 |
+
"logps/chosen": -153.40731811523438,
|
| 146 |
+
"logps/rejected": -161.5341796875,
|
| 147 |
+
"loss": 1.6191,
|
| 148 |
+
"nll_loss": 0.255443274974823,
|
| 149 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 150 |
+
"rewards/chosen": -15.340731620788574,
|
| 151 |
+
"rewards/margins": 0.812686562538147,
|
| 152 |
+
"rewards/rejected": -16.153419494628906,
|
| 153 |
+
"step": 90
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.23114706732158335,
|
| 157 |
+
"grad_norm": 45.195945739746094,
|
| 158 |
+
"learning_rate": 8.556701030927834e-07,
|
| 159 |
+
"logits/chosen": -0.34290799498558044,
|
| 160 |
+
"logits/rejected": -0.3276888430118561,
|
| 161 |
+
"logps/chosen": -156.65350341796875,
|
| 162 |
+
"logps/rejected": -159.73312377929688,
|
| 163 |
+
"loss": 1.7833,
|
| 164 |
+
"nll_loss": 0.2992493510246277,
|
| 165 |
+
"rewards/accuracies": 0.546875,
|
| 166 |
+
"rewards/chosen": -15.665351867675781,
|
| 167 |
+
"rewards/margins": 0.30796217918395996,
|
| 168 |
+
"rewards/rejected": -15.973315238952637,
|
| 169 |
+
"step": 100
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.23114706732158335,
|
| 173 |
+
"eval_logits/chosen": -0.3332868814468384,
|
| 174 |
+
"eval_logits/rejected": -0.30650395154953003,
|
| 175 |
+
"eval_logps/chosen": -153.93896484375,
|
| 176 |
+
"eval_logps/rejected": -157.79205322265625,
|
| 177 |
+
"eval_loss": 1.6389189958572388,
|
| 178 |
+
"eval_nll_loss": 0.26784786581993103,
|
| 179 |
+
"eval_rewards/accuracies": 0.5782608985900879,
|
| 180 |
+
"eval_rewards/chosen": -15.393896102905273,
|
| 181 |
+
"eval_rewards/margins": 0.38530847430229187,
|
| 182 |
+
"eval_rewards/rejected": -15.779205322265625,
|
| 183 |
+
"eval_runtime": 77.324,
|
| 184 |
+
"eval_samples_per_second": 23.615,
|
| 185 |
+
"eval_steps_per_second": 1.487,
|
| 186 |
+
"step": 100
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 0.2542617740537417,
|
| 190 |
+
"grad_norm": 46.2556037902832,
|
| 191 |
+
"learning_rate": 8.298969072164948e-07,
|
| 192 |
+
"logits/chosen": -0.28695493936538696,
|
| 193 |
+
"logits/rejected": -0.2580588459968567,
|
| 194 |
+
"logps/chosen": -150.8438262939453,
|
| 195 |
+
"logps/rejected": -155.1436767578125,
|
| 196 |
+
"loss": 1.5662,
|
| 197 |
+
"nll_loss": 0.2598631978034973,
|
| 198 |
+
"rewards/accuracies": 0.5625,
|
| 199 |
+
"rewards/chosen": -15.084381103515625,
|
| 200 |
+
"rewards/margins": 0.4299860894680023,
|
| 201 |
+
"rewards/rejected": -15.514368057250977,
|
| 202 |
+
"step": 110
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 0.27737648078590005,
|
| 206 |
+
"grad_norm": 44.54212188720703,
|
| 207 |
+
"learning_rate": 8.041237113402062e-07,
|
| 208 |
+
"logits/chosen": -0.3080243170261383,
|
| 209 |
+
"logits/rejected": -0.2822437286376953,
|
| 210 |
+
"logps/chosen": -157.50094604492188,
|
| 211 |
+
"logps/rejected": -159.069580078125,
|
| 212 |
+
"loss": 1.6159,
|
| 213 |
+
"nll_loss": 0.30583077669143677,
|
| 214 |
+
"rewards/accuracies": 0.5406249761581421,
|
| 215 |
+
"rewards/chosen": -15.750096321105957,
|
| 216 |
+
"rewards/margins": 0.1568617820739746,
|
| 217 |
+
"rewards/rejected": -15.906957626342773,
|
| 218 |
+
"step": 120
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 0.30049118751805837,
|
| 222 |
+
"grad_norm": 55.04152297973633,
|
| 223 |
+
"learning_rate": 7.783505154639175e-07,
|
| 224 |
+
"logits/chosen": -0.30607444047927856,
|
| 225 |
+
"logits/rejected": -0.2829833924770355,
|
| 226 |
+
"logps/chosen": -157.97000122070312,
|
| 227 |
+
"logps/rejected": -166.240478515625,
|
| 228 |
+
"loss": 1.6775,
|
| 229 |
+
"nll_loss": 0.2982478737831116,
|
| 230 |
+
"rewards/accuracies": 0.596875011920929,
|
| 231 |
+
"rewards/chosen": -15.79699993133545,
|
| 232 |
+
"rewards/margins": 0.8270493745803833,
|
| 233 |
+
"rewards/rejected": -16.624048233032227,
|
| 234 |
+
"step": 130
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.3236058942502167,
|
| 238 |
+
"grad_norm": 68.56768035888672,
|
| 239 |
+
"learning_rate": 7.525773195876288e-07,
|
| 240 |
+
"logits/chosen": -0.38547760248184204,
|
| 241 |
+
"logits/rejected": -0.3799718916416168,
|
| 242 |
+
"logps/chosen": -148.39646911621094,
|
| 243 |
+
"logps/rejected": -153.57923889160156,
|
| 244 |
+
"loss": 1.6639,
|
| 245 |
+
"nll_loss": 0.2873283624649048,
|
| 246 |
+
"rewards/accuracies": 0.581250011920929,
|
| 247 |
+
"rewards/chosen": -14.83964729309082,
|
| 248 |
+
"rewards/margins": 0.5182766318321228,
|
| 249 |
+
"rewards/rejected": -15.35792350769043,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.34672060098237506,
|
| 254 |
+
"grad_norm": 61.17930603027344,
|
| 255 |
+
"learning_rate": 7.268041237113402e-07,
|
| 256 |
+
"logits/chosen": -0.4227396845817566,
|
| 257 |
+
"logits/rejected": -0.41468995809555054,
|
| 258 |
+
"logps/chosen": -147.5921173095703,
|
| 259 |
+
"logps/rejected": -153.60935974121094,
|
| 260 |
+
"loss": 1.7192,
|
| 261 |
+
"nll_loss": 0.28114941716194153,
|
| 262 |
+
"rewards/accuracies": 0.609375,
|
| 263 |
+
"rewards/chosen": -14.759210586547852,
|
| 264 |
+
"rewards/margins": 0.6017246246337891,
|
| 265 |
+
"rewards/rejected": -15.360937118530273,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3698353077145334,
|
| 270 |
+
"grad_norm": 41.435455322265625,
|
| 271 |
+
"learning_rate": 7.010309278350515e-07,
|
| 272 |
+
"logits/chosen": -0.45233017206192017,
|
| 273 |
+
"logits/rejected": -0.4274457097053528,
|
| 274 |
+
"logps/chosen": -162.49929809570312,
|
| 275 |
+
"logps/rejected": -166.28240966796875,
|
| 276 |
+
"loss": 1.5534,
|
| 277 |
+
"nll_loss": 0.28693827986717224,
|
| 278 |
+
"rewards/accuracies": 0.5625,
|
| 279 |
+
"rewards/chosen": -16.249929428100586,
|
| 280 |
+
"rewards/margins": 0.3783140480518341,
|
| 281 |
+
"rewards/rejected": -16.62824249267578,
|
| 282 |
+
"step": 160
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3929500144466917,
|
| 286 |
+
"grad_norm": 47.89544677734375,
|
| 287 |
+
"learning_rate": 6.752577319587629e-07,
|
| 288 |
+
"logits/chosen": -0.36802831292152405,
|
| 289 |
+
"logits/rejected": -0.35963720083236694,
|
| 290 |
+
"logps/chosen": -158.8728790283203,
|
| 291 |
+
"logps/rejected": -166.11460876464844,
|
| 292 |
+
"loss": 1.4425,
|
| 293 |
+
"nll_loss": 0.2846711277961731,
|
| 294 |
+
"rewards/accuracies": 0.578125,
|
| 295 |
+
"rewards/chosen": -15.887288093566895,
|
| 296 |
+
"rewards/margins": 0.7241734862327576,
|
| 297 |
+
"rewards/rejected": -16.611461639404297,
|
| 298 |
+
"step": 170
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 0.41606472117885,
|
| 302 |
+
"grad_norm": 47.56120681762695,
|
| 303 |
+
"learning_rate": 6.494845360824742e-07,
|
| 304 |
+
"logits/chosen": -0.4111746847629547,
|
| 305 |
+
"logits/rejected": -0.4110477566719055,
|
| 306 |
+
"logps/chosen": -152.79598999023438,
|
| 307 |
+
"logps/rejected": -159.20770263671875,
|
| 308 |
+
"loss": 1.4742,
|
| 309 |
+
"nll_loss": 0.2701548635959625,
|
| 310 |
+
"rewards/accuracies": 0.5531250238418579,
|
| 311 |
+
"rewards/chosen": -15.279600143432617,
|
| 312 |
+
"rewards/margins": 0.6411706805229187,
|
| 313 |
+
"rewards/rejected": -15.920770645141602,
|
| 314 |
+
"step": 180
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 0.4391794279110084,
|
| 318 |
+
"grad_norm": 51.05569076538086,
|
| 319 |
+
"learning_rate": 6.237113402061855e-07,
|
| 320 |
+
"logits/chosen": -0.4827125668525696,
|
| 321 |
+
"logits/rejected": -0.4709135591983795,
|
| 322 |
+
"logps/chosen": -161.96566772460938,
|
| 323 |
+
"logps/rejected": -168.30032348632812,
|
| 324 |
+
"loss": 1.546,
|
| 325 |
+
"nll_loss": 0.29454725980758667,
|
| 326 |
+
"rewards/accuracies": 0.5843750238418579,
|
| 327 |
+
"rewards/chosen": -16.19656753540039,
|
| 328 |
+
"rewards/margins": 0.63346266746521,
|
| 329 |
+
"rewards/rejected": -16.83003044128418,
|
| 330 |
+
"step": 190
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"epoch": 0.4622941346431667,
|
| 334 |
+
"grad_norm": 44.95571517944336,
|
| 335 |
+
"learning_rate": 5.979381443298969e-07,
|
| 336 |
+
"logits/chosen": -0.426901638507843,
|
| 337 |
+
"logits/rejected": -0.42319783568382263,
|
| 338 |
+
"logps/chosen": -160.36892700195312,
|
| 339 |
+
"logps/rejected": -166.8142547607422,
|
| 340 |
+
"loss": 1.5321,
|
| 341 |
+
"nll_loss": 0.29501739144325256,
|
| 342 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 343 |
+
"rewards/chosen": -16.03689193725586,
|
| 344 |
+
"rewards/margins": 0.644534707069397,
|
| 345 |
+
"rewards/rejected": -16.681427001953125,
|
| 346 |
+
"step": 200
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"epoch": 0.4622941346431667,
|
| 350 |
+
"eval_logits/chosen": -0.4423229396343231,
|
| 351 |
+
"eval_logits/rejected": -0.42438310384750366,
|
| 352 |
+
"eval_logps/chosen": -158.98841857910156,
|
| 353 |
+
"eval_logps/rejected": -165.12060546875,
|
| 354 |
+
"eval_loss": 1.524242877960205,
|
| 355 |
+
"eval_nll_loss": 0.2763634920120239,
|
| 356 |
+
"eval_rewards/accuracies": 0.5978260636329651,
|
| 357 |
+
"eval_rewards/chosen": -15.898841857910156,
|
| 358 |
+
"eval_rewards/margins": 0.6132183074951172,
|
| 359 |
+
"eval_rewards/rejected": -16.512060165405273,
|
| 360 |
+
"eval_runtime": 77.1561,
|
| 361 |
+
"eval_samples_per_second": 23.666,
|
| 362 |
+
"eval_steps_per_second": 1.49,
|
| 363 |
+
"step": 200
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.48540884137532503,
|
| 367 |
+
"grad_norm": 52.92426300048828,
|
| 368 |
+
"learning_rate": 5.721649484536082e-07,
|
| 369 |
+
"logits/chosen": -0.45326417684555054,
|
| 370 |
+
"logits/rejected": -0.42941370606422424,
|
| 371 |
+
"logps/chosen": -156.1173553466797,
|
| 372 |
+
"logps/rejected": -158.58352661132812,
|
| 373 |
+
"loss": 1.5104,
|
| 374 |
+
"nll_loss": 0.28336626291275024,
|
| 375 |
+
"rewards/accuracies": 0.5531250238418579,
|
| 376 |
+
"rewards/chosen": -15.611738204956055,
|
| 377 |
+
"rewards/margins": 0.246616929769516,
|
| 378 |
+
"rewards/rejected": -15.858355522155762,
|
| 379 |
+
"step": 210
|
| 380 |
+
},
|
| 381 |
+
{
|
| 382 |
+
"epoch": 0.5085235481074833,
|
| 383 |
+
"grad_norm": 50.266258239746094,
|
| 384 |
+
"learning_rate": 5.463917525773195e-07,
|
| 385 |
+
"logits/chosen": -0.471625953912735,
|
| 386 |
+
"logits/rejected": -0.4460209906101227,
|
| 387 |
+
"logps/chosen": -168.0143585205078,
|
| 388 |
+
"logps/rejected": -173.90097045898438,
|
| 389 |
+
"loss": 1.569,
|
| 390 |
+
"nll_loss": 0.28898459672927856,
|
| 391 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 392 |
+
"rewards/chosen": -16.801433563232422,
|
| 393 |
+
"rewards/margins": 0.5886625051498413,
|
| 394 |
+
"rewards/rejected": -17.390098571777344,
|
| 395 |
+
"step": 220
|
| 396 |
+
},
|
| 397 |
+
{
|
| 398 |
+
"epoch": 0.5316382548396418,
|
| 399 |
+
"grad_norm": 55.33750534057617,
|
| 400 |
+
"learning_rate": 5.20618556701031e-07,
|
| 401 |
+
"logits/chosen": -0.42029595375061035,
|
| 402 |
+
"logits/rejected": -0.4011649191379547,
|
| 403 |
+
"logps/chosen": -167.03567504882812,
|
| 404 |
+
"logps/rejected": -173.90760803222656,
|
| 405 |
+
"loss": 1.5782,
|
| 406 |
+
"nll_loss": 0.29872751235961914,
|
| 407 |
+
"rewards/accuracies": 0.6031249761581421,
|
| 408 |
+
"rewards/chosen": -16.703567504882812,
|
| 409 |
+
"rewards/margins": 0.687191903591156,
|
| 410 |
+
"rewards/rejected": -17.39076042175293,
|
| 411 |
+
"step": 230
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 0.5547529615718001,
|
| 415 |
+
"grad_norm": 48.64881896972656,
|
| 416 |
+
"learning_rate": 4.948453608247422e-07,
|
| 417 |
+
"logits/chosen": -0.37133297324180603,
|
| 418 |
+
"logits/rejected": -0.3616218566894531,
|
| 419 |
+
"logps/chosen": -166.56610107421875,
|
| 420 |
+
"logps/rejected": -171.88787841796875,
|
| 421 |
+
"loss": 1.4277,
|
| 422 |
+
"nll_loss": 0.2906336784362793,
|
| 423 |
+
"rewards/accuracies": 0.559374988079071,
|
| 424 |
+
"rewards/chosen": -16.6566104888916,
|
| 425 |
+
"rewards/margins": 0.532177746295929,
|
| 426 |
+
"rewards/rejected": -17.18878746032715,
|
| 427 |
+
"step": 240
|
| 428 |
+
},
|
| 429 |
+
{
|
| 430 |
+
"epoch": 0.5778676683039584,
|
| 431 |
+
"grad_norm": 53.73638916015625,
|
| 432 |
+
"learning_rate": 4.6907216494845357e-07,
|
| 433 |
+
"logits/chosen": -0.3499279022216797,
|
| 434 |
+
"logits/rejected": -0.3348368704319,
|
| 435 |
+
"logps/chosen": -161.66017150878906,
|
| 436 |
+
"logps/rejected": -165.3099822998047,
|
| 437 |
+
"loss": 1.5715,
|
| 438 |
+
"nll_loss": 0.24935810267925262,
|
| 439 |
+
"rewards/accuracies": 0.550000011920929,
|
| 440 |
+
"rewards/chosen": -16.166015625,
|
| 441 |
+
"rewards/margins": 0.3649832606315613,
|
| 442 |
+
"rewards/rejected": -16.5310001373291,
|
| 443 |
+
"step": 250
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.6009823750361167,
|
| 447 |
+
"grad_norm": 51.90835952758789,
|
| 448 |
+
"learning_rate": 4.432989690721649e-07,
|
| 449 |
+
"logits/chosen": -0.46265679597854614,
|
| 450 |
+
"logits/rejected": -0.4500146508216858,
|
| 451 |
+
"logps/chosen": -158.95870971679688,
|
| 452 |
+
"logps/rejected": -167.43258666992188,
|
| 453 |
+
"loss": 1.4266,
|
| 454 |
+
"nll_loss": 0.2810400724411011,
|
| 455 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 456 |
+
"rewards/chosen": -15.895869255065918,
|
| 457 |
+
"rewards/margins": 0.8473905324935913,
|
| 458 |
+
"rewards/rejected": -16.743261337280273,
|
| 459 |
+
"step": 260
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 0.624097081768275,
|
| 463 |
+
"grad_norm": 43.483646392822266,
|
| 464 |
+
"learning_rate": 4.175257731958763e-07,
|
| 465 |
+
"logits/chosen": -0.4157690107822418,
|
| 466 |
+
"logits/rejected": -0.41007503867149353,
|
| 467 |
+
"logps/chosen": -158.64816284179688,
|
| 468 |
+
"logps/rejected": -165.8525848388672,
|
| 469 |
+
"loss": 1.4656,
|
| 470 |
+
"nll_loss": 0.2797052264213562,
|
| 471 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 472 |
+
"rewards/chosen": -15.864816665649414,
|
| 473 |
+
"rewards/margins": 0.7204429507255554,
|
| 474 |
+
"rewards/rejected": -16.58526039123535,
|
| 475 |
+
"step": 270
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.6472117885004334,
|
| 479 |
+
"grad_norm": 55.48883819580078,
|
| 480 |
+
"learning_rate": 3.917525773195876e-07,
|
| 481 |
+
"logits/chosen": -0.34205523133277893,
|
| 482 |
+
"logits/rejected": -0.3176972568035126,
|
| 483 |
+
"logps/chosen": -161.7351531982422,
|
| 484 |
+
"logps/rejected": -171.16500854492188,
|
| 485 |
+
"loss": 1.4658,
|
| 486 |
+
"nll_loss": 0.29218602180480957,
|
| 487 |
+
"rewards/accuracies": 0.625,
|
| 488 |
+
"rewards/chosen": -16.173513412475586,
|
| 489 |
+
"rewards/margins": 0.9429864883422852,
|
| 490 |
+
"rewards/rejected": -17.116500854492188,
|
| 491 |
+
"step": 280
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6703264952325917,
|
| 495 |
+
"grad_norm": 58.60624694824219,
|
| 496 |
+
"learning_rate": 3.659793814432989e-07,
|
| 497 |
+
"logits/chosen": -0.3818144202232361,
|
| 498 |
+
"logits/rejected": -0.35421329736709595,
|
| 499 |
+
"logps/chosen": -165.26620483398438,
|
| 500 |
+
"logps/rejected": -171.03085327148438,
|
| 501 |
+
"loss": 1.4217,
|
| 502 |
+
"nll_loss": 0.29271894693374634,
|
| 503 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 504 |
+
"rewards/chosen": -16.526620864868164,
|
| 505 |
+
"rewards/margins": 0.5764636397361755,
|
| 506 |
+
"rewards/rejected": -17.103084564208984,
|
| 507 |
+
"step": 290
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 0.6934412019647501,
|
| 511 |
+
"grad_norm": 43.18596267700195,
|
| 512 |
+
"learning_rate": 3.402061855670103e-07,
|
| 513 |
+
"logits/chosen": -0.44146934151649475,
|
| 514 |
+
"logits/rejected": -0.4326046109199524,
|
| 515 |
+
"logps/chosen": -168.69503784179688,
|
| 516 |
+
"logps/rejected": -178.41549682617188,
|
| 517 |
+
"loss": 1.4722,
|
| 518 |
+
"nll_loss": 0.3029026389122009,
|
| 519 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 520 |
+
"rewards/chosen": -16.869503021240234,
|
| 521 |
+
"rewards/margins": 0.9720472097396851,
|
| 522 |
+
"rewards/rejected": -17.841550827026367,
|
| 523 |
+
"step": 300
|
| 524 |
+
},
|
| 525 |
+
{
|
| 526 |
+
"epoch": 0.6934412019647501,
|
| 527 |
+
"eval_logits/chosen": -0.3856337070465088,
|
| 528 |
+
"eval_logits/rejected": -0.36413437128067017,
|
| 529 |
+
"eval_logps/chosen": -160.80313110351562,
|
| 530 |
+
"eval_logps/rejected": -168.14109802246094,
|
| 531 |
+
"eval_loss": 1.463273048400879,
|
| 532 |
+
"eval_nll_loss": 0.27904874086380005,
|
| 533 |
+
"eval_rewards/accuracies": 0.6217391490936279,
|
| 534 |
+
"eval_rewards/chosen": -16.080310821533203,
|
| 535 |
+
"eval_rewards/margins": 0.7337984442710876,
|
| 536 |
+
"eval_rewards/rejected": -16.814109802246094,
|
| 537 |
+
"eval_runtime": 76.8594,
|
| 538 |
+
"eval_samples_per_second": 23.758,
|
| 539 |
+
"eval_steps_per_second": 1.496,
|
| 540 |
+
"step": 300
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"epoch": 0.7165559086969084,
|
| 544 |
+
"grad_norm": 48.37858200073242,
|
| 545 |
+
"learning_rate": 3.1443298969072163e-07,
|
| 546 |
+
"logits/chosen": -0.43834060430526733,
|
| 547 |
+
"logits/rejected": -0.4327595829963684,
|
| 548 |
+
"logps/chosen": -170.76321411132812,
|
| 549 |
+
"logps/rejected": -174.53939819335938,
|
| 550 |
+
"loss": 1.5353,
|
| 551 |
+
"nll_loss": 0.30162352323532104,
|
| 552 |
+
"rewards/accuracies": 0.559374988079071,
|
| 553 |
+
"rewards/chosen": -17.07632064819336,
|
| 554 |
+
"rewards/margins": 0.37761738896369934,
|
| 555 |
+
"rewards/rejected": -17.45393943786621,
|
| 556 |
+
"step": 310
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"epoch": 0.7396706154290668,
|
| 560 |
+
"grad_norm": 45.166900634765625,
|
| 561 |
+
"learning_rate": 2.8865979381443296e-07,
|
| 562 |
+
"logits/chosen": -0.4189019799232483,
|
| 563 |
+
"logits/rejected": -0.41374340653419495,
|
| 564 |
+
"logps/chosen": -159.76388549804688,
|
| 565 |
+
"logps/rejected": -168.59652709960938,
|
| 566 |
+
"loss": 1.521,
|
| 567 |
+
"nll_loss": 0.28271549940109253,
|
| 568 |
+
"rewards/accuracies": 0.628125011920929,
|
| 569 |
+
"rewards/chosen": -15.976388931274414,
|
| 570 |
+
"rewards/margins": 0.8832640647888184,
|
| 571 |
+
"rewards/rejected": -16.85965347290039,
|
| 572 |
+
"step": 320
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 0.7627853221612251,
|
| 576 |
+
"grad_norm": 46.85552215576172,
|
| 577 |
+
"learning_rate": 2.6288659793814435e-07,
|
| 578 |
+
"logits/chosen": -0.4116067886352539,
|
| 579 |
+
"logits/rejected": -0.4161633551120758,
|
| 580 |
+
"logps/chosen": -159.32391357421875,
|
| 581 |
+
"logps/rejected": -166.09951782226562,
|
| 582 |
+
"loss": 1.4975,
|
| 583 |
+
"nll_loss": 0.299803227186203,
|
| 584 |
+
"rewards/accuracies": 0.628125011920929,
|
| 585 |
+
"rewards/chosen": -15.932391166687012,
|
| 586 |
+
"rewards/margins": 0.6775625944137573,
|
| 587 |
+
"rewards/rejected": -16.609954833984375,
|
| 588 |
+
"step": 330
|
| 589 |
+
},
|
| 590 |
+
{
|
| 591 |
+
"epoch": 0.7859000288933834,
|
| 592 |
+
"grad_norm": 47.712646484375,
|
| 593 |
+
"learning_rate": 2.3711340206185566e-07,
|
| 594 |
+
"logits/chosen": -0.4130411148071289,
|
| 595 |
+
"logits/rejected": -0.40607109665870667,
|
| 596 |
+
"logps/chosen": -162.58676147460938,
|
| 597 |
+
"logps/rejected": -167.87673950195312,
|
| 598 |
+
"loss": 1.3896,
|
| 599 |
+
"nll_loss": 0.27490609884262085,
|
| 600 |
+
"rewards/accuracies": 0.581250011920929,
|
| 601 |
+
"rewards/chosen": -16.258676528930664,
|
| 602 |
+
"rewards/margins": 0.5289959907531738,
|
| 603 |
+
"rewards/rejected": -16.78767204284668,
|
| 604 |
+
"step": 340
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"epoch": 0.8090147356255417,
|
| 608 |
+
"grad_norm": 45.12944412231445,
|
| 609 |
+
"learning_rate": 2.11340206185567e-07,
|
| 610 |
+
"logits/chosen": -0.40707340836524963,
|
| 611 |
+
"logits/rejected": -0.4076949656009674,
|
| 612 |
+
"logps/chosen": -164.3544921875,
|
| 613 |
+
"logps/rejected": -173.22579956054688,
|
| 614 |
+
"loss": 1.4861,
|
| 615 |
+
"nll_loss": 0.279820054769516,
|
| 616 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 617 |
+
"rewards/chosen": -16.435449600219727,
|
| 618 |
+
"rewards/margins": 0.8871291279792786,
|
| 619 |
+
"rewards/rejected": -17.32257843017578,
|
| 620 |
+
"step": 350
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 0.8321294423577,
|
| 624 |
+
"grad_norm": 55.76411437988281,
|
| 625 |
+
"learning_rate": 1.8556701030927835e-07,
|
| 626 |
+
"logits/chosen": -0.41376179456710815,
|
| 627 |
+
"logits/rejected": -0.4020335078239441,
|
| 628 |
+
"logps/chosen": -167.02381896972656,
|
| 629 |
+
"logps/rejected": -172.5863800048828,
|
| 630 |
+
"loss": 1.5281,
|
| 631 |
+
"nll_loss": 0.2744566798210144,
|
| 632 |
+
"rewards/accuracies": 0.596875011920929,
|
| 633 |
+
"rewards/chosen": -16.702383041381836,
|
| 634 |
+
"rewards/margins": 0.5562567114830017,
|
| 635 |
+
"rewards/rejected": -17.258638381958008,
|
| 636 |
+
"step": 360
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"epoch": 0.8552441490898585,
|
| 640 |
+
"grad_norm": 48.47395706176758,
|
| 641 |
+
"learning_rate": 1.5979381443298966e-07,
|
| 642 |
+
"logits/chosen": -0.4272083342075348,
|
| 643 |
+
"logits/rejected": -0.4035620093345642,
|
| 644 |
+
"logps/chosen": -161.60487365722656,
|
| 645 |
+
"logps/rejected": -171.45762634277344,
|
| 646 |
+
"loss": 1.3967,
|
| 647 |
+
"nll_loss": 0.2558661103248596,
|
| 648 |
+
"rewards/accuracies": 0.65625,
|
| 649 |
+
"rewards/chosen": -16.160486221313477,
|
| 650 |
+
"rewards/margins": 0.9852760434150696,
|
| 651 |
+
"rewards/rejected": -17.145763397216797,
|
| 652 |
+
"step": 370
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"epoch": 0.8783588558220168,
|
| 656 |
+
"grad_norm": 50.784507751464844,
|
| 657 |
+
"learning_rate": 1.3402061855670102e-07,
|
| 658 |
+
"logits/chosen": -0.4153415262699127,
|
| 659 |
+
"logits/rejected": -0.403484582901001,
|
| 660 |
+
"logps/chosen": -159.54061889648438,
|
| 661 |
+
"logps/rejected": -166.57180786132812,
|
| 662 |
+
"loss": 1.414,
|
| 663 |
+
"nll_loss": 0.2998126149177551,
|
| 664 |
+
"rewards/accuracies": 0.612500011920929,
|
| 665 |
+
"rewards/chosen": -15.954061508178711,
|
| 666 |
+
"rewards/margins": 0.7031179666519165,
|
| 667 |
+
"rewards/rejected": -16.657180786132812,
|
| 668 |
+
"step": 380
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 0.9014735625541751,
|
| 672 |
+
"grad_norm": 55.96210861206055,
|
| 673 |
+
"learning_rate": 1.0824742268041237e-07,
|
| 674 |
+
"logits/chosen": -0.4493354856967926,
|
| 675 |
+
"logits/rejected": -0.43461060523986816,
|
| 676 |
+
"logps/chosen": -163.90878295898438,
|
| 677 |
+
"logps/rejected": -174.5413055419922,
|
| 678 |
+
"loss": 1.4429,
|
| 679 |
+
"nll_loss": 0.27769935131073,
|
| 680 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 681 |
+
"rewards/chosen": -16.39087677001953,
|
| 682 |
+
"rewards/margins": 1.063251256942749,
|
| 683 |
+
"rewards/rejected": -17.45412826538086,
|
| 684 |
+
"step": 390
|
| 685 |
+
},
|
| 686 |
+
{
|
| 687 |
+
"epoch": 0.9245882692863334,
|
| 688 |
+
"grad_norm": 49.90566635131836,
|
| 689 |
+
"learning_rate": 8.24742268041237e-08,
|
| 690 |
+
"logits/chosen": -0.3982897698879242,
|
| 691 |
+
"logits/rejected": -0.37326449155807495,
|
| 692 |
+
"logps/chosen": -166.83291625976562,
|
| 693 |
+
"logps/rejected": -172.27056884765625,
|
| 694 |
+
"loss": 1.4589,
|
| 695 |
+
"nll_loss": 0.3044116795063019,
|
| 696 |
+
"rewards/accuracies": 0.578125,
|
| 697 |
+
"rewards/chosen": -16.683292388916016,
|
| 698 |
+
"rewards/margins": 0.5437662601470947,
|
| 699 |
+
"rewards/rejected": -17.227060317993164,
|
| 700 |
+
"step": 400
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"epoch": 0.9245882692863334,
|
| 704 |
+
"eval_logits/chosen": -0.360569566488266,
|
| 705 |
+
"eval_logits/rejected": -0.33890673518180847,
|
| 706 |
+
"eval_logps/chosen": -160.21502685546875,
|
| 707 |
+
"eval_logps/rejected": -168.04132080078125,
|
| 708 |
+
"eval_loss": 1.4446684122085571,
|
| 709 |
+
"eval_nll_loss": 0.2798323929309845,
|
| 710 |
+
"eval_rewards/accuracies": 0.626086950302124,
|
| 711 |
+
"eval_rewards/chosen": -16.021503448486328,
|
| 712 |
+
"eval_rewards/margins": 0.7826284766197205,
|
| 713 |
+
"eval_rewards/rejected": -16.80413246154785,
|
| 714 |
+
"eval_runtime": 76.9997,
|
| 715 |
+
"eval_samples_per_second": 23.714,
|
| 716 |
+
"eval_steps_per_second": 1.494,
|
| 717 |
+
"step": 400
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.9477029760184917,
|
| 721 |
+
"grad_norm": 50.10642623901367,
|
| 722 |
+
"learning_rate": 5.670103092783505e-08,
|
| 723 |
+
"logits/chosen": -0.3168153762817383,
|
| 724 |
+
"logits/rejected": -0.3060024678707123,
|
| 725 |
+
"logps/chosen": -152.92799377441406,
|
| 726 |
+
"logps/rejected": -162.26162719726562,
|
| 727 |
+
"loss": 1.4003,
|
| 728 |
+
"nll_loss": 0.2777474522590637,
|
| 729 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 730 |
+
"rewards/chosen": -15.29279899597168,
|
| 731 |
+
"rewards/margins": 0.9333623051643372,
|
| 732 |
+
"rewards/rejected": -16.22616195678711,
|
| 733 |
+
"step": 410
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9708176827506501,
|
| 737 |
+
"grad_norm": 46.16789245605469,
|
| 738 |
+
"learning_rate": 3.092783505154639e-08,
|
| 739 |
+
"logits/chosen": -0.3549926280975342,
|
| 740 |
+
"logits/rejected": -0.33413809537887573,
|
| 741 |
+
"logps/chosen": -155.63693237304688,
|
| 742 |
+
"logps/rejected": -161.7698211669922,
|
| 743 |
+
"loss": 1.4104,
|
| 744 |
+
"nll_loss": 0.28400248289108276,
|
| 745 |
+
"rewards/accuracies": 0.596875011920929,
|
| 746 |
+
"rewards/chosen": -15.563693046569824,
|
| 747 |
+
"rewards/margins": 0.6132909059524536,
|
| 748 |
+
"rewards/rejected": -16.176982879638672,
|
| 749 |
+
"step": 420
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 0.9939323894828085,
|
| 753 |
+
"grad_norm": 47.918479919433594,
|
| 754 |
+
"learning_rate": 5.154639175257731e-09,
|
| 755 |
+
"logits/chosen": -0.37286701798439026,
|
| 756 |
+
"logits/rejected": -0.38741737604141235,
|
| 757 |
+
"logps/chosen": -163.7307586669922,
|
| 758 |
+
"logps/rejected": -171.1845245361328,
|
| 759 |
+
"loss": 1.4847,
|
| 760 |
+
"nll_loss": 0.26778078079223633,
|
| 761 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 762 |
+
"rewards/chosen": -16.373075485229492,
|
| 763 |
+
"rewards/margins": 0.7453770041465759,
|
| 764 |
+
"rewards/rejected": -17.118452072143555,
|
| 765 |
+
"step": 430
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"epoch": 0.9985553308292401,
|
| 769 |
+
"step": 432,
|
| 770 |
+
"total_flos": 0.0,
|
| 771 |
+
"train_loss": 1.6635627029118714,
|
| 772 |
+
"train_runtime": 10783.4889,
|
| 773 |
+
"train_samples_per_second": 5.135,
|
| 774 |
+
"train_steps_per_second": 0.04
|
| 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 |
+
}
|