Commit
•
f4df214
1
Parent(s):
75f04dd
Model save
Browse files- README.md +76 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on an unknown dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.5055
|
| 21 |
+
- Rewards/chosen: -0.9459
|
| 22 |
+
- Rewards/rejected: -1.9033
|
| 23 |
+
- Rewards/accuracies: 0.7812
|
| 24 |
+
- Rewards/margins: 0.9573
|
| 25 |
+
- Logps/rejected: -452.9886
|
| 26 |
+
- Logps/chosen: -357.2234
|
| 27 |
+
- Logits/rejected: 1.2877
|
| 28 |
+
- Logits/chosen: 0.5168
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 8
|
| 50 |
+
- seed: 42
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
+
- total_eval_batch_size: 64
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.5659 | 0.2092 | 100 | 0.5701 | -0.7175 | -1.2998 | 0.7148 | 0.5824 | -392.6456 | -334.3749 | -0.7859 | -0.9821 |
|
| 66 |
+
| 0.5493 | 0.4184 | 200 | 0.5264 | -0.7175 | -1.5459 | 0.7734 | 0.8284 | -417.2501 | -334.3752 | 0.2805 | -0.3142 |
|
| 67 |
+
| 0.493 | 0.6276 | 300 | 0.5119 | -0.9435 | -1.8186 | 0.7617 | 0.8751 | -444.5215 | -356.9808 | 1.2687 | 0.5290 |
|
| 68 |
+
| 0.5014 | 0.8368 | 400 | 0.5055 | -0.9459 | -1.9033 | 0.7812 | 0.9573 | -452.9886 | -357.2234 | 1.2877 | 0.5168 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- Transformers 4.43.1
|
| 74 |
+
- Pytorch 2.1.2+cu121
|
| 75 |
+
- Datasets 2.20.0
|
| 76 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5398832953624646,
|
| 5 |
+
"train_runtime": 10917.977,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 5.599,
|
| 8 |
+
"train_steps_per_second": 0.044
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.43.1"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5398832953624646,
|
| 5 |
+
"train_runtime": 10917.977,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 5.599,
|
| 8 |
+
"train_steps_per_second": 0.044
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0020920502092050207,
|
| 13 |
+
"grad_norm": 9.16837234811457,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.7655434608459473,
|
| 16 |
+
"logits/rejected": -2.7158141136169434,
|
| 17 |
+
"logps/chosen": -269.8756103515625,
|
| 18 |
+
"logps/rejected": -360.6216125488281,
|
| 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": 0.02092050209205021,
|
| 28 |
+
"grad_norm": 9.188352376621864,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.5924127101898193,
|
| 31 |
+
"logits/rejected": -2.562790632247925,
|
| 32 |
+
"logps/chosen": -264.7654724121094,
|
| 33 |
+
"logps/rejected": -251.53240966796875,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.4375,
|
| 36 |
+
"rewards/chosen": 0.00027513509849086404,
|
| 37 |
+
"rewards/margins": -9.524480992695317e-05,
|
| 38 |
+
"rewards/rejected": 0.00037037988658994436,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 8.405862905672732,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.6532669067382812,
|
| 46 |
+
"logits/rejected": -2.605334997177124,
|
| 47 |
+
"logps/chosen": -281.41314697265625,
|
| 48 |
+
"logps/rejected": -296.80389404296875,
|
| 49 |
+
"loss": 0.6924,
|
| 50 |
+
"rewards/accuracies": 0.518750011920929,
|
| 51 |
+
"rewards/chosen": 0.0009061357122845948,
|
| 52 |
+
"rewards/margins": 0.0009388698963448405,
|
| 53 |
+
"rewards/rejected": -3.2734242267906666e-05,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 8.298184796125266,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.6643333435058594,
|
| 61 |
+
"logits/rejected": -2.5918784141540527,
|
| 62 |
+
"logps/chosen": -299.40911865234375,
|
| 63 |
+
"logps/rejected": -263.7054748535156,
|
| 64 |
+
"loss": 0.6884,
|
| 65 |
+
"rewards/accuracies": 0.643750011920929,
|
| 66 |
+
"rewards/chosen": 0.012471171095967293,
|
| 67 |
+
"rewards/margins": 0.010278983041644096,
|
| 68 |
+
"rewards/rejected": 0.002192187588661909,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 8.153488217762229,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.556988477706909,
|
| 76 |
+
"logits/rejected": -2.5218937397003174,
|
| 77 |
+
"logps/chosen": -263.11932373046875,
|
| 78 |
+
"logps/rejected": -245.16162109375,
|
| 79 |
+
"loss": 0.6774,
|
| 80 |
+
"rewards/accuracies": 0.706250011920929,
|
| 81 |
+
"rewards/chosen": 0.039762504398822784,
|
| 82 |
+
"rewards/margins": 0.047586839646101,
|
| 83 |
+
"rewards/rejected": -0.007824338972568512,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 8.472043005988137,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.497899055480957,
|
| 91 |
+
"logits/rejected": -2.457843780517578,
|
| 92 |
+
"logps/chosen": -279.7335205078125,
|
| 93 |
+
"logps/rejected": -297.64471435546875,
|
| 94 |
+
"loss": 0.6611,
|
| 95 |
+
"rewards/accuracies": 0.612500011920929,
|
| 96 |
+
"rewards/chosen": 6.177672185003757e-05,
|
| 97 |
+
"rewards/margins": 0.04885866492986679,
|
| 98 |
+
"rewards/rejected": -0.0487968847155571,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 11.234854782109899,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.5127627849578857,
|
| 106 |
+
"logits/rejected": -2.4409728050231934,
|
| 107 |
+
"logps/chosen": -273.99835205078125,
|
| 108 |
+
"logps/rejected": -313.34991455078125,
|
| 109 |
+
"loss": 0.6384,
|
| 110 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 111 |
+
"rewards/chosen": -0.02334701642394066,
|
| 112 |
+
"rewards/margins": 0.1286955624818802,
|
| 113 |
+
"rewards/rejected": -0.15204259753227234,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 10.848666524907674,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.360724449157715,
|
| 121 |
+
"logits/rejected": -2.3582727909088135,
|
| 122 |
+
"logps/chosen": -271.3426513671875,
|
| 123 |
+
"logps/rejected": -283.6495666503906,
|
| 124 |
+
"loss": 0.6157,
|
| 125 |
+
"rewards/accuracies": 0.768750011920929,
|
| 126 |
+
"rewards/chosen": -0.05673746392130852,
|
| 127 |
+
"rewards/margins": 0.21231010556221008,
|
| 128 |
+
"rewards/rejected": -0.2690475583076477,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 19.300722351787282,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.557610273361206,
|
| 136 |
+
"logits/rejected": -2.4776089191436768,
|
| 137 |
+
"logps/chosen": -337.05694580078125,
|
| 138 |
+
"logps/rejected": -305.25653076171875,
|
| 139 |
+
"loss": 0.6024,
|
| 140 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 141 |
+
"rewards/chosen": -0.1427958607673645,
|
| 142 |
+
"rewards/margins": 0.26870062947273254,
|
| 143 |
+
"rewards/rejected": -0.41149649024009705,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 20.757494940175373,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.0054640769958496,
|
| 151 |
+
"logits/rejected": -1.9451837539672852,
|
| 152 |
+
"logps/chosen": -310.46173095703125,
|
| 153 |
+
"logps/rejected": -353.8874816894531,
|
| 154 |
+
"loss": 0.5716,
|
| 155 |
+
"rewards/accuracies": 0.706250011920929,
|
| 156 |
+
"rewards/chosen": -0.5131224393844604,
|
| 157 |
+
"rewards/margins": 0.5017130970954895,
|
| 158 |
+
"rewards/rejected": -1.0148354768753052,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 18.659168826580355,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -1.321213960647583,
|
| 166 |
+
"logits/rejected": -1.129945158958435,
|
| 167 |
+
"logps/chosen": -307.0208435058594,
|
| 168 |
+
"logps/rejected": -361.28350830078125,
|
| 169 |
+
"loss": 0.5659,
|
| 170 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 171 |
+
"rewards/chosen": -0.466524213552475,
|
| 172 |
+
"rewards/margins": 0.5754273533821106,
|
| 173 |
+
"rewards/rejected": -1.0419515371322632,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -0.9821491241455078,
|
| 179 |
+
"eval_logits/rejected": -0.7858694791793823,
|
| 180 |
+
"eval_logps/chosen": -334.3749084472656,
|
| 181 |
+
"eval_logps/rejected": -392.64556884765625,
|
| 182 |
+
"eval_loss": 0.5701435804367065,
|
| 183 |
+
"eval_rewards/accuracies": 0.71484375,
|
| 184 |
+
"eval_rewards/chosen": -0.7174500226974487,
|
| 185 |
+
"eval_rewards/margins": 0.5823848843574524,
|
| 186 |
+
"eval_rewards/rejected": -1.299834966659546,
|
| 187 |
+
"eval_runtime": 99.7714,
|
| 188 |
+
"eval_samples_per_second": 20.046,
|
| 189 |
+
"eval_steps_per_second": 0.321,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 17.55736534565649,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -1.0365899801254272,
|
| 197 |
+
"logits/rejected": -0.571634829044342,
|
| 198 |
+
"logps/chosen": -372.1576843261719,
|
| 199 |
+
"logps/rejected": -384.3282165527344,
|
| 200 |
+
"loss": 0.57,
|
| 201 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 202 |
+
"rewards/chosen": -0.676292896270752,
|
| 203 |
+
"rewards/margins": 0.5884906053543091,
|
| 204 |
+
"rewards/rejected": -1.2647836208343506,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 20.044772578276955,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -0.6921079754829407,
|
| 212 |
+
"logits/rejected": -0.4483819603919983,
|
| 213 |
+
"logps/chosen": -354.0092468261719,
|
| 214 |
+
"logps/rejected": -391.1457824707031,
|
| 215 |
+
"loss": 0.5521,
|
| 216 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 217 |
+
"rewards/chosen": -0.7282991409301758,
|
| 218 |
+
"rewards/margins": 0.5842112898826599,
|
| 219 |
+
"rewards/rejected": -1.31251060962677,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 20.68475231351549,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -0.3410087823867798,
|
| 227 |
+
"logits/rejected": -0.04661134257912636,
|
| 228 |
+
"logps/chosen": -377.7615966796875,
|
| 229 |
+
"logps/rejected": -426.6678771972656,
|
| 230 |
+
"loss": 0.5404,
|
| 231 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 232 |
+
"rewards/chosen": -0.9577744603157043,
|
| 233 |
+
"rewards/margins": 0.6243709325790405,
|
| 234 |
+
"rewards/rejected": -1.5821453332901,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 17.900468089200906,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -0.7499442100524902,
|
| 242 |
+
"logits/rejected": -0.48287004232406616,
|
| 243 |
+
"logps/chosen": -378.3758850097656,
|
| 244 |
+
"logps/rejected": -425.58148193359375,
|
| 245 |
+
"loss": 0.5392,
|
| 246 |
+
"rewards/accuracies": 0.75,
|
| 247 |
+
"rewards/chosen": -0.6240967512130737,
|
| 248 |
+
"rewards/margins": 0.6252509355545044,
|
| 249 |
+
"rewards/rejected": -1.2493478059768677,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 28.14974497379761,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": -0.2267804592847824,
|
| 257 |
+
"logits/rejected": 0.15516111254692078,
|
| 258 |
+
"logps/chosen": -373.9515380859375,
|
| 259 |
+
"logps/rejected": -421.302001953125,
|
| 260 |
+
"loss": 0.5406,
|
| 261 |
+
"rewards/accuracies": 0.706250011920929,
|
| 262 |
+
"rewards/chosen": -0.9071490168571472,
|
| 263 |
+
"rewards/margins": 0.5438954830169678,
|
| 264 |
+
"rewards/rejected": -1.4510444402694702,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 26.21920743231775,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": -0.5155479907989502,
|
| 272 |
+
"logits/rejected": 0.20259642601013184,
|
| 273 |
+
"logps/chosen": -330.3095703125,
|
| 274 |
+
"logps/rejected": -401.3794250488281,
|
| 275 |
+
"loss": 0.5368,
|
| 276 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 277 |
+
"rewards/chosen": -0.5726653337478638,
|
| 278 |
+
"rewards/margins": 0.8669228553771973,
|
| 279 |
+
"rewards/rejected": -1.439588189125061,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 29.878459650293532,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": -0.2268867939710617,
|
| 287 |
+
"logits/rejected": 0.22177407145500183,
|
| 288 |
+
"logps/chosen": -337.85345458984375,
|
| 289 |
+
"logps/rejected": -385.3772888183594,
|
| 290 |
+
"loss": 0.5435,
|
| 291 |
+
"rewards/accuracies": 0.699999988079071,
|
| 292 |
+
"rewards/chosen": -0.8853822946548462,
|
| 293 |
+
"rewards/margins": 0.6570433974266052,
|
| 294 |
+
"rewards/rejected": -1.542425513267517,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 23.677554496063618,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": 0.14835922420024872,
|
| 302 |
+
"logits/rejected": 0.698757529258728,
|
| 303 |
+
"logps/chosen": -335.9161682128906,
|
| 304 |
+
"logps/rejected": -375.7064514160156,
|
| 305 |
+
"loss": 0.504,
|
| 306 |
+
"rewards/accuracies": 0.706250011920929,
|
| 307 |
+
"rewards/chosen": -0.7958806753158569,
|
| 308 |
+
"rewards/margins": 0.6854288578033447,
|
| 309 |
+
"rewards/rejected": -1.4813096523284912,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 23.203690544454616,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": 0.024832600727677345,
|
| 317 |
+
"logits/rejected": 0.5344244837760925,
|
| 318 |
+
"logps/chosen": -348.7262268066406,
|
| 319 |
+
"logps/rejected": -404.1893005371094,
|
| 320 |
+
"loss": 0.5174,
|
| 321 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 322 |
+
"rewards/chosen": -0.7743078470230103,
|
| 323 |
+
"rewards/margins": 0.6698081493377686,
|
| 324 |
+
"rewards/rejected": -1.4441161155700684,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 50.04702866429734,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": 0.0846758633852005,
|
| 332 |
+
"logits/rejected": 0.5746791958808899,
|
| 333 |
+
"logps/chosen": -342.2915954589844,
|
| 334 |
+
"logps/rejected": -408.9627685546875,
|
| 335 |
+
"loss": 0.5493,
|
| 336 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 337 |
+
"rewards/chosen": -0.8239612579345703,
|
| 338 |
+
"rewards/margins": 0.7452048659324646,
|
| 339 |
+
"rewards/rejected": -1.5691661834716797,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": -0.31423458456993103,
|
| 345 |
+
"eval_logits/rejected": 0.28053638339042664,
|
| 346 |
+
"eval_logps/chosen": -334.3751525878906,
|
| 347 |
+
"eval_logps/rejected": -417.25006103515625,
|
| 348 |
+
"eval_loss": 0.5263724327087402,
|
| 349 |
+
"eval_rewards/accuracies": 0.7734375,
|
| 350 |
+
"eval_rewards/chosen": -0.717452347278595,
|
| 351 |
+
"eval_rewards/margins": 0.8284271955490112,
|
| 352 |
+
"eval_rewards/rejected": -1.545879602432251,
|
| 353 |
+
"eval_runtime": 97.9594,
|
| 354 |
+
"eval_samples_per_second": 20.417,
|
| 355 |
+
"eval_steps_per_second": 0.327,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 24.06066475347068,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": -0.26085931062698364,
|
| 363 |
+
"logits/rejected": 0.5198220014572144,
|
| 364 |
+
"logps/chosen": -365.29010009765625,
|
| 365 |
+
"logps/rejected": -394.47723388671875,
|
| 366 |
+
"loss": 0.5396,
|
| 367 |
+
"rewards/accuracies": 0.731249988079071,
|
| 368 |
+
"rewards/chosen": -0.7562572360038757,
|
| 369 |
+
"rewards/margins": 0.675930380821228,
|
| 370 |
+
"rewards/rejected": -1.4321876764297485,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 28.79899654759585,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": 0.5355015397071838,
|
| 378 |
+
"logits/rejected": 1.2820476293563843,
|
| 379 |
+
"logps/chosen": -367.3331604003906,
|
| 380 |
+
"logps/rejected": -419.95538330078125,
|
| 381 |
+
"loss": 0.5283,
|
| 382 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 383 |
+
"rewards/chosen": -1.0195591449737549,
|
| 384 |
+
"rewards/margins": 0.7230145335197449,
|
| 385 |
+
"rewards/rejected": -1.7425734996795654,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 21.063139682190887,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": 0.5966787338256836,
|
| 393 |
+
"logits/rejected": 1.3136788606643677,
|
| 394 |
+
"logps/chosen": -338.0008239746094,
|
| 395 |
+
"logps/rejected": -426.60162353515625,
|
| 396 |
+
"loss": 0.5352,
|
| 397 |
+
"rewards/accuracies": 0.737500011920929,
|
| 398 |
+
"rewards/chosen": -0.8904770612716675,
|
| 399 |
+
"rewards/margins": 0.8978098630905151,
|
| 400 |
+
"rewards/rejected": -1.7882869243621826,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 22.340209509704692,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": 0.36957186460494995,
|
| 408 |
+
"logits/rejected": 0.9800383448600769,
|
| 409 |
+
"logps/chosen": -344.3677062988281,
|
| 410 |
+
"logps/rejected": -398.9412536621094,
|
| 411 |
+
"loss": 0.5049,
|
| 412 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 413 |
+
"rewards/chosen": -0.8063153028488159,
|
| 414 |
+
"rewards/margins": 0.6925672292709351,
|
| 415 |
+
"rewards/rejected": -1.498882532119751,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 25.32395453596937,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": 0.6675643920898438,
|
| 423 |
+
"logits/rejected": 1.5727179050445557,
|
| 424 |
+
"logps/chosen": -386.45343017578125,
|
| 425 |
+
"logps/rejected": -431.385986328125,
|
| 426 |
+
"loss": 0.5053,
|
| 427 |
+
"rewards/accuracies": 0.768750011920929,
|
| 428 |
+
"rewards/chosen": -0.8825046420097351,
|
| 429 |
+
"rewards/margins": 0.835895836353302,
|
| 430 |
+
"rewards/rejected": -1.7184005975723267,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 32.04785964174811,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": 0.7512863874435425,
|
| 438 |
+
"logits/rejected": 1.690630316734314,
|
| 439 |
+
"logps/chosen": -394.2430114746094,
|
| 440 |
+
"logps/rejected": -439.67205810546875,
|
| 441 |
+
"loss": 0.5095,
|
| 442 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 443 |
+
"rewards/chosen": -0.9994521141052246,
|
| 444 |
+
"rewards/margins": 0.8580014109611511,
|
| 445 |
+
"rewards/rejected": -1.8574535846710205,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 24.904083578928375,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": 0.7148577570915222,
|
| 453 |
+
"logits/rejected": 1.4790935516357422,
|
| 454 |
+
"logps/chosen": -385.87371826171875,
|
| 455 |
+
"logps/rejected": -443.53839111328125,
|
| 456 |
+
"loss": 0.5171,
|
| 457 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 458 |
+
"rewards/chosen": -1.0889328718185425,
|
| 459 |
+
"rewards/margins": 0.76996910572052,
|
| 460 |
+
"rewards/rejected": -1.8589019775390625,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 25.263400006388935,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": 0.6911333799362183,
|
| 468 |
+
"logits/rejected": 1.6273902654647827,
|
| 469 |
+
"logps/chosen": -368.5780334472656,
|
| 470 |
+
"logps/rejected": -429.0997009277344,
|
| 471 |
+
"loss": 0.5258,
|
| 472 |
+
"rewards/accuracies": 0.699999988079071,
|
| 473 |
+
"rewards/chosen": -1.0685292482376099,
|
| 474 |
+
"rewards/margins": 0.8003263473510742,
|
| 475 |
+
"rewards/rejected": -1.8688554763793945,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 26.846743184768357,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": 0.6784590482711792,
|
| 483 |
+
"logits/rejected": 1.4638246297836304,
|
| 484 |
+
"logps/chosen": -365.30206298828125,
|
| 485 |
+
"logps/rejected": -430.20660400390625,
|
| 486 |
+
"loss": 0.5113,
|
| 487 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 488 |
+
"rewards/chosen": -1.1421018838882446,
|
| 489 |
+
"rewards/margins": 0.7403939366340637,
|
| 490 |
+
"rewards/rejected": -1.8824958801269531,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 29.645083802417286,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": 0.45342984795570374,
|
| 498 |
+
"logits/rejected": 1.6018460988998413,
|
| 499 |
+
"logps/chosen": -403.8066711425781,
|
| 500 |
+
"logps/rejected": -440.07098388671875,
|
| 501 |
+
"loss": 0.493,
|
| 502 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 503 |
+
"rewards/chosen": -1.0277760028839111,
|
| 504 |
+
"rewards/margins": 0.7721818685531616,
|
| 505 |
+
"rewards/rejected": -1.7999578714370728,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": 0.5289936065673828,
|
| 511 |
+
"eval_logits/rejected": 1.2686660289764404,
|
| 512 |
+
"eval_logps/chosen": -356.9808044433594,
|
| 513 |
+
"eval_logps/rejected": -444.52154541015625,
|
| 514 |
+
"eval_loss": 0.5118551850318909,
|
| 515 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 516 |
+
"eval_rewards/chosen": -0.9435087442398071,
|
| 517 |
+
"eval_rewards/margins": 0.8750853538513184,
|
| 518 |
+
"eval_rewards/rejected": -1.8185943365097046,
|
| 519 |
+
"eval_runtime": 97.9113,
|
| 520 |
+
"eval_samples_per_second": 20.427,
|
| 521 |
+
"eval_steps_per_second": 0.327,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 26.615301380561615,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": 0.6738792657852173,
|
| 529 |
+
"logits/rejected": 1.4009501934051514,
|
| 530 |
+
"logps/chosen": -398.75958251953125,
|
| 531 |
+
"logps/rejected": -437.30633544921875,
|
| 532 |
+
"loss": 0.4932,
|
| 533 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 534 |
+
"rewards/chosen": -1.0345656871795654,
|
| 535 |
+
"rewards/margins": 0.753504753112793,
|
| 536 |
+
"rewards/rejected": -1.7880703210830688,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 32.234049319731376,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": 0.8933159112930298,
|
| 544 |
+
"logits/rejected": 1.8882709741592407,
|
| 545 |
+
"logps/chosen": -401.068115234375,
|
| 546 |
+
"logps/rejected": -431.6943359375,
|
| 547 |
+
"loss": 0.4861,
|
| 548 |
+
"rewards/accuracies": 0.762499988079071,
|
| 549 |
+
"rewards/chosen": -1.0735524892807007,
|
| 550 |
+
"rewards/margins": 0.7948076128959656,
|
| 551 |
+
"rewards/rejected": -1.868360161781311,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 34.15386792642874,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": 0.35191336274147034,
|
| 559 |
+
"logits/rejected": 1.2871559858322144,
|
| 560 |
+
"logps/chosen": -360.3251037597656,
|
| 561 |
+
"logps/rejected": -431.16534423828125,
|
| 562 |
+
"loss": 0.496,
|
| 563 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 564 |
+
"rewards/chosen": -0.9808635711669922,
|
| 565 |
+
"rewards/margins": 0.7355011701583862,
|
| 566 |
+
"rewards/rejected": -1.7163646221160889,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 36.00524525047531,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": 0.36807161569595337,
|
| 574 |
+
"logits/rejected": 1.3383424282073975,
|
| 575 |
+
"logps/chosen": -374.64703369140625,
|
| 576 |
+
"logps/rejected": -469.4212951660156,
|
| 577 |
+
"loss": 0.5019,
|
| 578 |
+
"rewards/accuracies": 0.800000011920929,
|
| 579 |
+
"rewards/chosen": -0.8456922769546509,
|
| 580 |
+
"rewards/margins": 1.047476053237915,
|
| 581 |
+
"rewards/rejected": -1.8931680917739868,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 28.054514541029086,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": 0.6923775672912598,
|
| 589 |
+
"logits/rejected": 1.5702488422393799,
|
| 590 |
+
"logps/chosen": -356.87725830078125,
|
| 591 |
+
"logps/rejected": -427.8857421875,
|
| 592 |
+
"loss": 0.4787,
|
| 593 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 594 |
+
"rewards/chosen": -0.9867467880249023,
|
| 595 |
+
"rewards/margins": 0.8851081728935242,
|
| 596 |
+
"rewards/rejected": -1.8718547821044922,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 24.593262334348687,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": 0.7163609266281128,
|
| 604 |
+
"logits/rejected": 1.9110931158065796,
|
| 605 |
+
"logps/chosen": -414.01123046875,
|
| 606 |
+
"logps/rejected": -465.447998046875,
|
| 607 |
+
"loss": 0.515,
|
| 608 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 609 |
+
"rewards/chosen": -1.1131541728973389,
|
| 610 |
+
"rewards/margins": 0.928746223449707,
|
| 611 |
+
"rewards/rejected": -2.041900157928467,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 24.621281841757344,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": 1.1743080615997314,
|
| 619 |
+
"logits/rejected": 1.871708631515503,
|
| 620 |
+
"logps/chosen": -397.70159912109375,
|
| 621 |
+
"logps/rejected": -431.24530029296875,
|
| 622 |
+
"loss": 0.5066,
|
| 623 |
+
"rewards/accuracies": 0.737500011920929,
|
| 624 |
+
"rewards/chosen": -1.2236586809158325,
|
| 625 |
+
"rewards/margins": 0.7059282064437866,
|
| 626 |
+
"rewards/rejected": -1.9295867681503296,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 26.620813748068915,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": 1.1752651929855347,
|
| 634 |
+
"logits/rejected": 1.894327163696289,
|
| 635 |
+
"logps/chosen": -367.64251708984375,
|
| 636 |
+
"logps/rejected": -479.59814453125,
|
| 637 |
+
"loss": 0.505,
|
| 638 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 639 |
+
"rewards/chosen": -1.281599760055542,
|
| 640 |
+
"rewards/margins": 0.9278450012207031,
|
| 641 |
+
"rewards/rejected": -2.209444522857666,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 24.66450917199271,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": 0.6577990651130676,
|
| 649 |
+
"logits/rejected": 1.2567031383514404,
|
| 650 |
+
"logps/chosen": -414.7494201660156,
|
| 651 |
+
"logps/rejected": -473.61859130859375,
|
| 652 |
+
"loss": 0.4898,
|
| 653 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 654 |
+
"rewards/chosen": -1.091284990310669,
|
| 655 |
+
"rewards/margins": 0.8050578832626343,
|
| 656 |
+
"rewards/rejected": -1.8963428735733032,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 25.534386282276298,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": 0.7974405884742737,
|
| 664 |
+
"logits/rejected": 1.8116296529769897,
|
| 665 |
+
"logps/chosen": -378.2085876464844,
|
| 666 |
+
"logps/rejected": -467.15380859375,
|
| 667 |
+
"loss": 0.5014,
|
| 668 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 669 |
+
"rewards/chosen": -1.076610803604126,
|
| 670 |
+
"rewards/margins": 0.9254143834114075,
|
| 671 |
+
"rewards/rejected": -2.0020251274108887,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": 0.5168243050575256,
|
| 677 |
+
"eval_logits/rejected": 1.2876626253128052,
|
| 678 |
+
"eval_logps/chosen": -357.22344970703125,
|
| 679 |
+
"eval_logps/rejected": -452.98858642578125,
|
| 680 |
+
"eval_loss": 0.5055292248725891,
|
| 681 |
+
"eval_rewards/accuracies": 0.78125,
|
| 682 |
+
"eval_rewards/chosen": -0.9459351897239685,
|
| 683 |
+
"eval_rewards/margins": 0.957329511642456,
|
| 684 |
+
"eval_rewards/rejected": -1.9032648801803589,
|
| 685 |
+
"eval_runtime": 98.0802,
|
| 686 |
+
"eval_samples_per_second": 20.391,
|
| 687 |
+
"eval_steps_per_second": 0.326,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 26.77180405179113,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": 0.6059648990631104,
|
| 695 |
+
"logits/rejected": 1.6763797998428345,
|
| 696 |
+
"logps/chosen": -400.90740966796875,
|
| 697 |
+
"logps/rejected": -458.0049743652344,
|
| 698 |
+
"loss": 0.5081,
|
| 699 |
+
"rewards/accuracies": 0.768750011920929,
|
| 700 |
+
"rewards/chosen": -1.019736409187317,
|
| 701 |
+
"rewards/margins": 0.8720607757568359,
|
| 702 |
+
"rewards/rejected": -1.8917968273162842,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 27.43056085728676,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": 0.3962455987930298,
|
| 710 |
+
"logits/rejected": 1.2819584608078003,
|
| 711 |
+
"logps/chosen": -398.86578369140625,
|
| 712 |
+
"logps/rejected": -457.84796142578125,
|
| 713 |
+
"loss": 0.495,
|
| 714 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 715 |
+
"rewards/chosen": -1.0178868770599365,
|
| 716 |
+
"rewards/margins": 0.8161880373954773,
|
| 717 |
+
"rewards/rejected": -1.8340747356414795,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 28.218529380366814,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": 0.9617902636528015,
|
| 725 |
+
"logits/rejected": 1.7479569911956787,
|
| 726 |
+
"logps/chosen": -389.6866760253906,
|
| 727 |
+
"logps/rejected": -455.4461975097656,
|
| 728 |
+
"loss": 0.4738,
|
| 729 |
+
"rewards/accuracies": 0.78125,
|
| 730 |
+
"rewards/chosen": -1.1064532995224,
|
| 731 |
+
"rewards/margins": 0.8703316450119019,
|
| 732 |
+
"rewards/rejected": -1.9767850637435913,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 31.300447561945898,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": 0.5907198786735535,
|
| 740 |
+
"logits/rejected": 1.4784181118011475,
|
| 741 |
+
"logps/chosen": -381.84088134765625,
|
| 742 |
+
"logps/rejected": -447.17059326171875,
|
| 743 |
+
"loss": 0.5009,
|
| 744 |
+
"rewards/accuracies": 0.75,
|
| 745 |
+
"rewards/chosen": -0.9751038551330566,
|
| 746 |
+
"rewards/margins": 0.8055868148803711,
|
| 747 |
+
"rewards/rejected": -1.7806905508041382,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 24.39105540939906,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": 0.6014220714569092,
|
| 755 |
+
"logits/rejected": 1.414150357246399,
|
| 756 |
+
"logps/chosen": -400.45184326171875,
|
| 757 |
+
"logps/rejected": -494.45965576171875,
|
| 758 |
+
"loss": 0.4975,
|
| 759 |
+
"rewards/accuracies": 0.706250011920929,
|
| 760 |
+
"rewards/chosen": -1.0672663450241089,
|
| 761 |
+
"rewards/margins": 0.806206226348877,
|
| 762 |
+
"rewards/rejected": -1.8734725713729858,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 27.720726174975443,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": 0.7587317824363708,
|
| 770 |
+
"logits/rejected": 1.6242033243179321,
|
| 771 |
+
"logps/chosen": -371.478759765625,
|
| 772 |
+
"logps/rejected": -423.82684326171875,
|
| 773 |
+
"loss": 0.5041,
|
| 774 |
+
"rewards/accuracies": 0.706250011920929,
|
| 775 |
+
"rewards/chosen": -0.994895339012146,
|
| 776 |
+
"rewards/margins": 0.8169215321540833,
|
| 777 |
+
"rewards/rejected": -1.811816930770874,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 26.198352977438653,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": 0.5299074053764343,
|
| 785 |
+
"logits/rejected": 1.6109470129013062,
|
| 786 |
+
"logps/chosen": -373.5680847167969,
|
| 787 |
+
"logps/rejected": -463.97283935546875,
|
| 788 |
+
"loss": 0.4888,
|
| 789 |
+
"rewards/accuracies": 0.762499988079071,
|
| 790 |
+
"rewards/chosen": -0.9487025141716003,
|
| 791 |
+
"rewards/margins": 0.8672690391540527,
|
| 792 |
+
"rewards/rejected": -1.8159716129302979,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5398832953624646,
|
| 800 |
+
"train_runtime": 10917.977,
|
| 801 |
+
"train_samples_per_second": 5.599,
|
| 802 |
+
"train_steps_per_second": 0.044
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 478,
|
| 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": 8,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|