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: 0.5044
|
| 21 |
+
- Rewards/chosen: -0.9256
|
| 22 |
+
- Rewards/rejected: -1.8971
|
| 23 |
+
- Rewards/accuracies: 0.7852
|
| 24 |
+
- Rewards/margins: 0.9715
|
| 25 |
+
- Logps/rejected: -452.3751
|
| 26 |
+
- Logps/chosen: -355.1891
|
| 27 |
+
- Logits/rejected: 1.7396
|
| 28 |
+
- Logits/chosen: 0.8151
|
| 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 |
+
| 0.5652 | 0.2092 | 100 | 0.5692 | -0.6767 | -1.2533 | 0.7070 | 0.5765 | -387.9908 | -330.3037 | -0.3959 | -0.6102 |
|
| 66 |
+
| 0.5456 | 0.4184 | 200 | 0.5314 | -0.7136 | -1.4550 | 0.7656 | 0.7414 | -408.1622 | -333.9880 | 0.0197 | -0.6245 |
|
| 67 |
+
| 0.494 | 0.6276 | 300 | 0.5094 | -0.8720 | -1.8005 | 0.7852 | 0.9285 | -442.7074 | -349.8286 | 1.1371 | 0.2961 |
|
| 68 |
+
| 0.5019 | 0.8368 | 400 | 0.5044 | -0.9256 | -1.8971 | 0.7852 | 0.9715 | -452.3751 | -355.1891 | 1.7396 | 0.8151 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- Transformers 4.44.0
|
| 74 |
+
- Pytorch 2.1.2+cu121
|
| 75 |
+
- Datasets 2.21.0
|
| 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": 0.5400737299579955,
|
| 5 |
+
"train_runtime": 7416.7822,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 8.243,
|
| 8 |
+
"train_steps_per_second": 0.064
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5400737299579955,
|
| 5 |
+
"train_runtime": 7416.7822,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 8.243,
|
| 8 |
+
"train_steps_per_second": 0.064
|
| 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": 9.168965006424878,
|
| 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": 9.12449411192981,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.59224796295166,
|
| 31 |
+
"logits/rejected": -2.562605381011963,
|
| 32 |
+
"logps/chosen": -264.8179016113281,
|
| 33 |
+
"logps/rejected": -251.55332946777344,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.4305555522441864,
|
| 36 |
+
"rewards/chosen": -0.00024893987574614584,
|
| 37 |
+
"rewards/margins": -0.0004101950326003134,
|
| 38 |
+
"rewards/rejected": 0.00016125504043884575,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 8.402884792192983,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.653071880340576,
|
| 46 |
+
"logits/rejected": -2.6053452491760254,
|
| 47 |
+
"logps/chosen": -281.4222106933594,
|
| 48 |
+
"logps/rejected": -296.7785949707031,
|
| 49 |
+
"loss": 0.6924,
|
| 50 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 51 |
+
"rewards/chosen": 0.000815415580291301,
|
| 52 |
+
"rewards/margins": 0.0005955949891358614,
|
| 53 |
+
"rewards/rejected": 0.00021982062025927007,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 8.331266596707612,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.664562940597534,
|
| 61 |
+
"logits/rejected": -2.59194278717041,
|
| 62 |
+
"logps/chosen": -299.4393615722656,
|
| 63 |
+
"logps/rejected": -263.73602294921875,
|
| 64 |
+
"loss": 0.6885,
|
| 65 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 66 |
+
"rewards/chosen": 0.01216866448521614,
|
| 67 |
+
"rewards/margins": 0.010281884111464024,
|
| 68 |
+
"rewards/rejected": 0.0018867806065827608,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 8.163671353448509,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.556959867477417,
|
| 76 |
+
"logits/rejected": -2.5217204093933105,
|
| 77 |
+
"logps/chosen": -263.1289978027344,
|
| 78 |
+
"logps/rejected": -245.099609375,
|
| 79 |
+
"loss": 0.6777,
|
| 80 |
+
"rewards/accuracies": 0.699999988079071,
|
| 81 |
+
"rewards/chosen": 0.039665959775447845,
|
| 82 |
+
"rewards/margins": 0.04687044769525528,
|
| 83 |
+
"rewards/rejected": -0.0072044916450977325,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 8.415963996114856,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.498046875,
|
| 91 |
+
"logits/rejected": -2.4585351943969727,
|
| 92 |
+
"logps/chosen": -279.64141845703125,
|
| 93 |
+
"logps/rejected": -297.4805603027344,
|
| 94 |
+
"loss": 0.6615,
|
| 95 |
+
"rewards/accuracies": 0.612500011920929,
|
| 96 |
+
"rewards/chosen": 0.0009827238973230124,
|
| 97 |
+
"rewards/margins": 0.04813830927014351,
|
| 98 |
+
"rewards/rejected": -0.04715558886528015,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 11.37240796902324,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.513566255569458,
|
| 106 |
+
"logits/rejected": -2.4418997764587402,
|
| 107 |
+
"logps/chosen": -273.7225036621094,
|
| 108 |
+
"logps/rejected": -312.88446044921875,
|
| 109 |
+
"loss": 0.639,
|
| 110 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 111 |
+
"rewards/chosen": -0.020588582381606102,
|
| 112 |
+
"rewards/margins": 0.1267995536327362,
|
| 113 |
+
"rewards/rejected": -0.14738816022872925,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 10.780449834370726,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.3635454177856445,
|
| 121 |
+
"logits/rejected": -2.3618197441101074,
|
| 122 |
+
"logps/chosen": -270.852783203125,
|
| 123 |
+
"logps/rejected": -283.3929138183594,
|
| 124 |
+
"loss": 0.6158,
|
| 125 |
+
"rewards/accuracies": 0.762499988079071,
|
| 126 |
+
"rewards/chosen": -0.05183873698115349,
|
| 127 |
+
"rewards/margins": 0.21464291214942932,
|
| 128 |
+
"rewards/rejected": -0.2664816379547119,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 18.854218139571227,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.5561511516571045,
|
| 136 |
+
"logits/rejected": -2.474567413330078,
|
| 137 |
+
"logps/chosen": -336.8047790527344,
|
| 138 |
+
"logps/rejected": -304.82586669921875,
|
| 139 |
+
"loss": 0.6023,
|
| 140 |
+
"rewards/accuracies": 0.668749988079071,
|
| 141 |
+
"rewards/chosen": -0.14027419686317444,
|
| 142 |
+
"rewards/margins": 0.2669159471988678,
|
| 143 |
+
"rewards/rejected": -0.40719014406204224,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 14.738392316661798,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.082120180130005,
|
| 151 |
+
"logits/rejected": -2.03467059135437,
|
| 152 |
+
"logps/chosen": -295.71002197265625,
|
| 153 |
+
"logps/rejected": -333.998046875,
|
| 154 |
+
"loss": 0.5716,
|
| 155 |
+
"rewards/accuracies": 0.762499988079071,
|
| 156 |
+
"rewards/chosen": -0.3656051754951477,
|
| 157 |
+
"rewards/margins": 0.4503363072872162,
|
| 158 |
+
"rewards/rejected": -0.8159415125846863,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 19.0889825670233,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -1.0137593746185303,
|
| 166 |
+
"logits/rejected": -0.8453589677810669,
|
| 167 |
+
"logps/chosen": -314.6478271484375,
|
| 168 |
+
"logps/rejected": -370.57684326171875,
|
| 169 |
+
"loss": 0.5652,
|
| 170 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 171 |
+
"rewards/chosen": -0.5427938103675842,
|
| 172 |
+
"rewards/margins": 0.5920915007591248,
|
| 173 |
+
"rewards/rejected": -1.1348851919174194,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -0.6102238297462463,
|
| 179 |
+
"eval_logits/rejected": -0.3958854675292969,
|
| 180 |
+
"eval_logps/chosen": -330.3037414550781,
|
| 181 |
+
"eval_logps/rejected": -387.99078369140625,
|
| 182 |
+
"eval_loss": 0.5692113041877747,
|
| 183 |
+
"eval_rewards/accuracies": 0.70703125,
|
| 184 |
+
"eval_rewards/chosen": -0.6767381429672241,
|
| 185 |
+
"eval_rewards/margins": 0.5765491724014282,
|
| 186 |
+
"eval_rewards/rejected": -1.2532871961593628,
|
| 187 |
+
"eval_runtime": 92.7565,
|
| 188 |
+
"eval_samples_per_second": 21.562,
|
| 189 |
+
"eval_steps_per_second": 0.345,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 18.680147346638737,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -0.5662704110145569,
|
| 197 |
+
"logits/rejected": -0.05761973187327385,
|
| 198 |
+
"logps/chosen": -377.0027770996094,
|
| 199 |
+
"logps/rejected": -387.5830383300781,
|
| 200 |
+
"loss": 0.5651,
|
| 201 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 202 |
+
"rewards/chosen": -0.7247437238693237,
|
| 203 |
+
"rewards/margins": 0.5725878477096558,
|
| 204 |
+
"rewards/rejected": -1.297331690788269,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 21.74920149487147,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -0.38402482867240906,
|
| 212 |
+
"logits/rejected": -0.08361969143152237,
|
| 213 |
+
"logps/chosen": -353.30731201171875,
|
| 214 |
+
"logps/rejected": -394.52532958984375,
|
| 215 |
+
"loss": 0.551,
|
| 216 |
+
"rewards/accuracies": 0.731249988079071,
|
| 217 |
+
"rewards/chosen": -0.7212796211242676,
|
| 218 |
+
"rewards/margins": 0.6250264048576355,
|
| 219 |
+
"rewards/rejected": -1.3463060855865479,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 18.13991460993244,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -0.39832931756973267,
|
| 227 |
+
"logits/rejected": -0.04852822422981262,
|
| 228 |
+
"logps/chosen": -357.71588134765625,
|
| 229 |
+
"logps/rejected": -406.4501953125,
|
| 230 |
+
"loss": 0.5346,
|
| 231 |
+
"rewards/accuracies": 0.737500011920929,
|
| 232 |
+
"rewards/chosen": -0.7573168873786926,
|
| 233 |
+
"rewards/margins": 0.6226516962051392,
|
| 234 |
+
"rewards/rejected": -1.3799686431884766,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 19.080428474602492,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -0.10298992693424225,
|
| 242 |
+
"logits/rejected": 0.23123951256275177,
|
| 243 |
+
"logps/chosen": -394.36370849609375,
|
| 244 |
+
"logps/rejected": -445.19732666015625,
|
| 245 |
+
"loss": 0.5329,
|
| 246 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 247 |
+
"rewards/chosen": -0.7839752435684204,
|
| 248 |
+
"rewards/margins": 0.6615303754806519,
|
| 249 |
+
"rewards/rejected": -1.4455056190490723,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 24.41518003453127,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": 0.23929806053638458,
|
| 257 |
+
"logits/rejected": 0.663245677947998,
|
| 258 |
+
"logps/chosen": -381.1101989746094,
|
| 259 |
+
"logps/rejected": -428.31121826171875,
|
| 260 |
+
"loss": 0.5412,
|
| 261 |
+
"rewards/accuracies": 0.668749988079071,
|
| 262 |
+
"rewards/chosen": -0.9787357449531555,
|
| 263 |
+
"rewards/margins": 0.5424004793167114,
|
| 264 |
+
"rewards/rejected": -1.5211361646652222,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 30.00328915671681,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": 0.18379352986812592,
|
| 272 |
+
"logits/rejected": 0.9613186717033386,
|
| 273 |
+
"logps/chosen": -354.5564270019531,
|
| 274 |
+
"logps/rejected": -423.7784118652344,
|
| 275 |
+
"loss": 0.5338,
|
| 276 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 277 |
+
"rewards/chosen": -0.8151340484619141,
|
| 278 |
+
"rewards/margins": 0.8484436869621277,
|
| 279 |
+
"rewards/rejected": -1.6635777950286865,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 32.90887166557451,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": 0.6914592385292053,
|
| 287 |
+
"logits/rejected": 1.118165373802185,
|
| 288 |
+
"logps/chosen": -350.73297119140625,
|
| 289 |
+
"logps/rejected": -398.38079833984375,
|
| 290 |
+
"loss": 0.5397,
|
| 291 |
+
"rewards/accuracies": 0.699999988079071,
|
| 292 |
+
"rewards/chosen": -1.014177918434143,
|
| 293 |
+
"rewards/margins": 0.6582826375961304,
|
| 294 |
+
"rewards/rejected": -1.6724605560302734,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 22.287968378600482,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": 0.37205350399017334,
|
| 302 |
+
"logits/rejected": 1.014420509338379,
|
| 303 |
+
"logps/chosen": -331.5960998535156,
|
| 304 |
+
"logps/rejected": -370.4664611816406,
|
| 305 |
+
"loss": 0.5103,
|
| 306 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 307 |
+
"rewards/chosen": -0.7526803612709045,
|
| 308 |
+
"rewards/margins": 0.6762296557426453,
|
| 309 |
+
"rewards/rejected": -1.4289101362228394,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 25.06102572422969,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": 0.3608325123786926,
|
| 317 |
+
"logits/rejected": 1.0164767503738403,
|
| 318 |
+
"logps/chosen": -363.13262939453125,
|
| 319 |
+
"logps/rejected": -417.058349609375,
|
| 320 |
+
"loss": 0.5144,
|
| 321 |
+
"rewards/accuracies": 0.71875,
|
| 322 |
+
"rewards/chosen": -0.9183723330497742,
|
| 323 |
+
"rewards/margins": 0.6544338464736938,
|
| 324 |
+
"rewards/rejected": -1.5728062391281128,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 56.1449128192664,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": 0.13207800686359406,
|
| 332 |
+
"logits/rejected": 0.5271329879760742,
|
| 333 |
+
"logps/chosen": -351.56219482421875,
|
| 334 |
+
"logps/rejected": -415.3944396972656,
|
| 335 |
+
"loss": 0.5456,
|
| 336 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 337 |
+
"rewards/chosen": -0.9166672825813293,
|
| 338 |
+
"rewards/margins": 0.7168152332305908,
|
| 339 |
+
"rewards/rejected": -1.6334823369979858,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": -0.6244610548019409,
|
| 345 |
+
"eval_logits/rejected": 0.01971178501844406,
|
| 346 |
+
"eval_logps/chosen": -333.98797607421875,
|
| 347 |
+
"eval_logps/rejected": -408.1622314453125,
|
| 348 |
+
"eval_loss": 0.5313767194747925,
|
| 349 |
+
"eval_rewards/accuracies": 0.765625,
|
| 350 |
+
"eval_rewards/chosen": -0.7135804295539856,
|
| 351 |
+
"eval_rewards/margins": 0.7414206862449646,
|
| 352 |
+
"eval_rewards/rejected": -1.4550011157989502,
|
| 353 |
+
"eval_runtime": 92.967,
|
| 354 |
+
"eval_samples_per_second": 21.513,
|
| 355 |
+
"eval_steps_per_second": 0.344,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 28.205388358273062,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": -0.4052346348762512,
|
| 363 |
+
"logits/rejected": 0.4885997772216797,
|
| 364 |
+
"logps/chosen": -368.7042236328125,
|
| 365 |
+
"logps/rejected": -401.01116943359375,
|
| 366 |
+
"loss": 0.5385,
|
| 367 |
+
"rewards/accuracies": 0.75,
|
| 368 |
+
"rewards/chosen": -0.7903989553451538,
|
| 369 |
+
"rewards/margins": 0.7071281671524048,
|
| 370 |
+
"rewards/rejected": -1.4975273609161377,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 31.469903246953805,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": 0.48687195777893066,
|
| 378 |
+
"logits/rejected": 1.3709402084350586,
|
| 379 |
+
"logps/chosen": -373.0572204589844,
|
| 380 |
+
"logps/rejected": -430.284423828125,
|
| 381 |
+
"loss": 0.5311,
|
| 382 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 383 |
+
"rewards/chosen": -1.0767996311187744,
|
| 384 |
+
"rewards/margins": 0.7690645456314087,
|
| 385 |
+
"rewards/rejected": -1.8458640575408936,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 21.508720745531846,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": 0.6478261947631836,
|
| 393 |
+
"logits/rejected": 1.4287010431289673,
|
| 394 |
+
"logps/chosen": -348.21307373046875,
|
| 395 |
+
"logps/rejected": -428.8194885253906,
|
| 396 |
+
"loss": 0.5366,
|
| 397 |
+
"rewards/accuracies": 0.737500011920929,
|
| 398 |
+
"rewards/chosen": -0.9925994873046875,
|
| 399 |
+
"rewards/margins": 0.8178664445877075,
|
| 400 |
+
"rewards/rejected": -1.8104660511016846,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 23.741045093534996,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": 0.6465433835983276,
|
| 408 |
+
"logits/rejected": 1.3921277523040771,
|
| 409 |
+
"logps/chosen": -352.5728454589844,
|
| 410 |
+
"logps/rejected": -413.220947265625,
|
| 411 |
+
"loss": 0.5045,
|
| 412 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 413 |
+
"rewards/chosen": -0.8883668184280396,
|
| 414 |
+
"rewards/margins": 0.7533131837844849,
|
| 415 |
+
"rewards/rejected": -1.6416800022125244,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 24.18519529429216,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": 0.7936254143714905,
|
| 423 |
+
"logits/rejected": 1.7098544836044312,
|
| 424 |
+
"logps/chosen": -389.4216003417969,
|
| 425 |
+
"logps/rejected": -437.39141845703125,
|
| 426 |
+
"loss": 0.5098,
|
| 427 |
+
"rewards/accuracies": 0.768750011920929,
|
| 428 |
+
"rewards/chosen": -0.9121864438056946,
|
| 429 |
+
"rewards/margins": 0.8662682771682739,
|
| 430 |
+
"rewards/rejected": -1.7784545421600342,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 30.73692690145265,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": 0.7679815292358398,
|
| 438 |
+
"logits/rejected": 1.6959028244018555,
|
| 439 |
+
"logps/chosen": -401.0132751464844,
|
| 440 |
+
"logps/rejected": -448.0699768066406,
|
| 441 |
+
"loss": 0.509,
|
| 442 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 443 |
+
"rewards/chosen": -1.0671550035476685,
|
| 444 |
+
"rewards/margins": 0.8742769956588745,
|
| 445 |
+
"rewards/rejected": -1.941432237625122,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 21.173165688823637,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": 0.5351714491844177,
|
| 453 |
+
"logits/rejected": 1.2392023801803589,
|
| 454 |
+
"logps/chosen": -379.1543884277344,
|
| 455 |
+
"logps/rejected": -429.35150146484375,
|
| 456 |
+
"loss": 0.5231,
|
| 457 |
+
"rewards/accuracies": 0.675000011920929,
|
| 458 |
+
"rewards/chosen": -1.0217397212982178,
|
| 459 |
+
"rewards/margins": 0.6952935457229614,
|
| 460 |
+
"rewards/rejected": -1.7170331478118896,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 28.68227240987258,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": 0.8288900256156921,
|
| 468 |
+
"logits/rejected": 1.7930643558502197,
|
| 469 |
+
"logps/chosen": -372.87237548828125,
|
| 470 |
+
"logps/rejected": -432.2880859375,
|
| 471 |
+
"loss": 0.5307,
|
| 472 |
+
"rewards/accuracies": 0.71875,
|
| 473 |
+
"rewards/chosen": -1.1114728450775146,
|
| 474 |
+
"rewards/margins": 0.7892667055130005,
|
| 475 |
+
"rewards/rejected": -1.9007394313812256,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 24.56550896170595,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": 0.7682640552520752,
|
| 483 |
+
"logits/rejected": 1.5038535594940186,
|
| 484 |
+
"logps/chosen": -370.6294860839844,
|
| 485 |
+
"logps/rejected": -434.703125,
|
| 486 |
+
"loss": 0.5206,
|
| 487 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 488 |
+
"rewards/chosen": -1.1953760385513306,
|
| 489 |
+
"rewards/margins": 0.7320847511291504,
|
| 490 |
+
"rewards/rejected": -1.92746102809906,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 24.820945259716815,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": 0.2875315546989441,
|
| 498 |
+
"logits/rejected": 1.408119559288025,
|
| 499 |
+
"logps/chosen": -400.6693420410156,
|
| 500 |
+
"logps/rejected": -436.7262268066406,
|
| 501 |
+
"loss": 0.494,
|
| 502 |
+
"rewards/accuracies": 0.71875,
|
| 503 |
+
"rewards/chosen": -0.9964024424552917,
|
| 504 |
+
"rewards/margins": 0.7701075077056885,
|
| 505 |
+
"rewards/rejected": -1.766510009765625,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": 0.29611948132514954,
|
| 511 |
+
"eval_logits/rejected": 1.1370961666107178,
|
| 512 |
+
"eval_logps/chosen": -349.82861328125,
|
| 513 |
+
"eval_logps/rejected": -442.7073669433594,
|
| 514 |
+
"eval_loss": 0.5094121098518372,
|
| 515 |
+
"eval_rewards/accuracies": 0.78515625,
|
| 516 |
+
"eval_rewards/chosen": -0.871986985206604,
|
| 517 |
+
"eval_rewards/margins": 0.9284656643867493,
|
| 518 |
+
"eval_rewards/rejected": -1.8004525899887085,
|
| 519 |
+
"eval_runtime": 93.1533,
|
| 520 |
+
"eval_samples_per_second": 21.47,
|
| 521 |
+
"eval_steps_per_second": 0.344,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 26.748349709062442,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": 0.5810573101043701,
|
| 529 |
+
"logits/rejected": 1.3016048669815063,
|
| 530 |
+
"logps/chosen": -399.5757751464844,
|
| 531 |
+
"logps/rejected": -436.1788024902344,
|
| 532 |
+
"loss": 0.4938,
|
| 533 |
+
"rewards/accuracies": 0.731249988079071,
|
| 534 |
+
"rewards/chosen": -1.0427275896072388,
|
| 535 |
+
"rewards/margins": 0.7340670228004456,
|
| 536 |
+
"rewards/rejected": -1.776794672012329,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 31.96747246205928,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": 1.1102755069732666,
|
| 544 |
+
"logits/rejected": 2.059471368789673,
|
| 545 |
+
"logps/chosen": -419.55902099609375,
|
| 546 |
+
"logps/rejected": -446.55548095703125,
|
| 547 |
+
"loss": 0.4909,
|
| 548 |
+
"rewards/accuracies": 0.75,
|
| 549 |
+
"rewards/chosen": -1.2584612369537354,
|
| 550 |
+
"rewards/margins": 0.7585103511810303,
|
| 551 |
+
"rewards/rejected": -2.0169715881347656,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 31.34877880671993,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": 0.4595663547515869,
|
| 559 |
+
"logits/rejected": 1.4255298376083374,
|
| 560 |
+
"logps/chosen": -364.82562255859375,
|
| 561 |
+
"logps/rejected": -434.95831298828125,
|
| 562 |
+
"loss": 0.4957,
|
| 563 |
+
"rewards/accuracies": 0.699999988079071,
|
| 564 |
+
"rewards/chosen": -1.025868535041809,
|
| 565 |
+
"rewards/margins": 0.7284258008003235,
|
| 566 |
+
"rewards/rejected": -1.7542946338653564,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 36.534964414934784,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": 0.5645917654037476,
|
| 574 |
+
"logits/rejected": 1.5419148206710815,
|
| 575 |
+
"logps/chosen": -382.41680908203125,
|
| 576 |
+
"logps/rejected": -477.4661560058594,
|
| 577 |
+
"loss": 0.5004,
|
| 578 |
+
"rewards/accuracies": 0.793749988079071,
|
| 579 |
+
"rewards/chosen": -0.9233899116516113,
|
| 580 |
+
"rewards/margins": 1.0502269268035889,
|
| 581 |
+
"rewards/rejected": -1.9736168384552002,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 23.351076956086505,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": 0.7957900762557983,
|
| 589 |
+
"logits/rejected": 1.681909203529358,
|
| 590 |
+
"logps/chosen": -360.25079345703125,
|
| 591 |
+
"logps/rejected": -430.71343994140625,
|
| 592 |
+
"loss": 0.4759,
|
| 593 |
+
"rewards/accuracies": 0.762499988079071,
|
| 594 |
+
"rewards/chosen": -1.0204814672470093,
|
| 595 |
+
"rewards/margins": 0.8796509504318237,
|
| 596 |
+
"rewards/rejected": -1.900132417678833,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 24.72192771367193,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": 0.7120740413665771,
|
| 604 |
+
"logits/rejected": 1.9698877334594727,
|
| 605 |
+
"logps/chosen": -409.35009765625,
|
| 606 |
+
"logps/rejected": -462.37213134765625,
|
| 607 |
+
"loss": 0.5113,
|
| 608 |
+
"rewards/accuracies": 0.762499988079071,
|
| 609 |
+
"rewards/chosen": -1.066542387008667,
|
| 610 |
+
"rewards/margins": 0.944599449634552,
|
| 611 |
+
"rewards/rejected": -2.0111420154571533,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 25.85275095494242,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": 1.2129358053207397,
|
| 619 |
+
"logits/rejected": 2.045804738998413,
|
| 620 |
+
"logps/chosen": -397.6884460449219,
|
| 621 |
+
"logps/rejected": -434.45635986328125,
|
| 622 |
+
"loss": 0.5097,
|
| 623 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 624 |
+
"rewards/chosen": -1.2235265970230103,
|
| 625 |
+
"rewards/margins": 0.7381704449653625,
|
| 626 |
+
"rewards/rejected": -1.961696982383728,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 26.772769459890974,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": 1.6128628253936768,
|
| 634 |
+
"logits/rejected": 2.2228505611419678,
|
| 635 |
+
"logps/chosen": -372.97344970703125,
|
| 636 |
+
"logps/rejected": -485.552734375,
|
| 637 |
+
"loss": 0.5079,
|
| 638 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 639 |
+
"rewards/chosen": -1.3349088430404663,
|
| 640 |
+
"rewards/margins": 0.9340818524360657,
|
| 641 |
+
"rewards/rejected": -2.2689907550811768,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 24.093673944544452,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": 1.175621509552002,
|
| 649 |
+
"logits/rejected": 1.7995212078094482,
|
| 650 |
+
"logps/chosen": -421.20068359375,
|
| 651 |
+
"logps/rejected": -482.9652404785156,
|
| 652 |
+
"loss": 0.4904,
|
| 653 |
+
"rewards/accuracies": 0.731249988079071,
|
| 654 |
+
"rewards/chosen": -1.1557972431182861,
|
| 655 |
+
"rewards/margins": 0.8340117335319519,
|
| 656 |
+
"rewards/rejected": -1.9898090362548828,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 25.903427781490553,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": 1.2189610004425049,
|
| 664 |
+
"logits/rejected": 2.2791786193847656,
|
| 665 |
+
"logps/chosen": -381.3320007324219,
|
| 666 |
+
"logps/rejected": -468.74359130859375,
|
| 667 |
+
"loss": 0.5019,
|
| 668 |
+
"rewards/accuracies": 0.762499988079071,
|
| 669 |
+
"rewards/chosen": -1.1078449487686157,
|
| 670 |
+
"rewards/margins": 0.9100783467292786,
|
| 671 |
+
"rewards/rejected": -2.01792311668396,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": 0.8151233792304993,
|
| 677 |
+
"eval_logits/rejected": 1.7396329641342163,
|
| 678 |
+
"eval_logps/chosen": -355.1890869140625,
|
| 679 |
+
"eval_logps/rejected": -452.37506103515625,
|
| 680 |
+
"eval_loss": 0.5043635368347168,
|
| 681 |
+
"eval_rewards/accuracies": 0.78515625,
|
| 682 |
+
"eval_rewards/chosen": -0.9255913496017456,
|
| 683 |
+
"eval_rewards/margins": 0.9715384244918823,
|
| 684 |
+
"eval_rewards/rejected": -1.8971298933029175,
|
| 685 |
+
"eval_runtime": 93.4716,
|
| 686 |
+
"eval_samples_per_second": 21.397,
|
| 687 |
+
"eval_steps_per_second": 0.342,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 28.39031028613065,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": 0.9069422483444214,
|
| 695 |
+
"logits/rejected": 2.2471582889556885,
|
| 696 |
+
"logps/chosen": -399.8828430175781,
|
| 697 |
+
"logps/rejected": -458.23358154296875,
|
| 698 |
+
"loss": 0.5082,
|
| 699 |
+
"rewards/accuracies": 0.768750011920929,
|
| 700 |
+
"rewards/chosen": -1.009491205215454,
|
| 701 |
+
"rewards/margins": 0.8845922350883484,
|
| 702 |
+
"rewards/rejected": -1.8940832614898682,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 26.704848418881845,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": 0.633719801902771,
|
| 710 |
+
"logits/rejected": 1.6843284368515015,
|
| 711 |
+
"logps/chosen": -395.73773193359375,
|
| 712 |
+
"logps/rejected": -453.5660095214844,
|
| 713 |
+
"loss": 0.4959,
|
| 714 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 715 |
+
"rewards/chosen": -0.9866059422492981,
|
| 716 |
+
"rewards/margins": 0.8046499490737915,
|
| 717 |
+
"rewards/rejected": -1.7912559509277344,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 27.04072698273277,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": 1.1087729930877686,
|
| 725 |
+
"logits/rejected": 2.018793821334839,
|
| 726 |
+
"logps/chosen": -387.2413635253906,
|
| 727 |
+
"logps/rejected": -453.595947265625,
|
| 728 |
+
"loss": 0.4749,
|
| 729 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 730 |
+
"rewards/chosen": -1.0820001363754272,
|
| 731 |
+
"rewards/margins": 0.8762818574905396,
|
| 732 |
+
"rewards/rejected": -1.9582821130752563,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 31.988161638127448,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": 0.8210460543632507,
|
| 740 |
+
"logits/rejected": 1.825097680091858,
|
| 741 |
+
"logps/chosen": -378.8327331542969,
|
| 742 |
+
"logps/rejected": -445.6421813964844,
|
| 743 |
+
"loss": 0.5011,
|
| 744 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 745 |
+
"rewards/chosen": -0.945022463798523,
|
| 746 |
+
"rewards/margins": 0.8203840255737305,
|
| 747 |
+
"rewards/rejected": -1.765406608581543,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 24.964570138956866,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": 0.8392779231071472,
|
| 755 |
+
"logits/rejected": 1.7072703838348389,
|
| 756 |
+
"logps/chosen": -397.63751220703125,
|
| 757 |
+
"logps/rejected": -497.10888671875,
|
| 758 |
+
"loss": 0.4977,
|
| 759 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 760 |
+
"rewards/chosen": -1.0391228199005127,
|
| 761 |
+
"rewards/margins": 0.8608413934707642,
|
| 762 |
+
"rewards/rejected": -1.8999640941619873,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 27.14765135099825,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": 0.9813226461410522,
|
| 770 |
+
"logits/rejected": 1.978114366531372,
|
| 771 |
+
"logps/chosen": -371.80206298828125,
|
| 772 |
+
"logps/rejected": -422.38787841796875,
|
| 773 |
+
"loss": 0.5027,
|
| 774 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 775 |
+
"rewards/chosen": -0.9981285929679871,
|
| 776 |
+
"rewards/margins": 0.7992982864379883,
|
| 777 |
+
"rewards/rejected": -1.7974268198013306,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 25.51430496068903,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": 0.8643745183944702,
|
| 785 |
+
"logits/rejected": 1.9723937511444092,
|
| 786 |
+
"logps/chosen": -376.01934814453125,
|
| 787 |
+
"logps/rejected": -464.6620178222656,
|
| 788 |
+
"loss": 0.4911,
|
| 789 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 790 |
+
"rewards/chosen": -0.9732152819633484,
|
| 791 |
+
"rewards/margins": 0.8496482968330383,
|
| 792 |
+
"rewards/rejected": -1.8228635787963867,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5400737299579955,
|
| 800 |
+
"train_runtime": 7416.7822,
|
| 801 |
+
"train_samples_per_second": 8.243,
|
| 802 |
+
"train_steps_per_second": 0.064
|
| 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 |
+
}
|