Model save
Browse files- README.md +77 -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 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: Jessicahuang0018/zephyr-7b-sft-full
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-dpo-full
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-7b-dpo-full
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [Jessicahuang0018/zephyr-7b-sft-full](https://huggingface.co/Jessicahuang0018/zephyr-7b-sft-full) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5252
|
| 22 |
+
- Rewards/chosen: -0.6177
|
| 23 |
+
- Rewards/rejected: -1.4052
|
| 24 |
+
- Rewards/accuracies: 0.7773
|
| 25 |
+
- Rewards/margins: 0.7875
|
| 26 |
+
- Logps/rejected: -404.4023
|
| 27 |
+
- Logps/chosen: -326.9973
|
| 28 |
+
- Logits/rejected: -1.1062
|
| 29 |
+
- Logits/chosen: -1.2274
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-07
|
| 49 |
+
- train_batch_size: 8
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 8
|
| 54 |
+
- gradient_accumulation_steps: 2
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 64
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.6029 | 0.2092 | 100 | 0.6044 | -0.1356 | -0.4456 | 0.7031 | 0.3101 | -308.4451 | -278.7863 | -2.6772 | -2.7044 |
|
| 67 |
+
| 0.5641 | 0.4184 | 200 | 0.5478 | -0.4828 | -1.1391 | 0.7617 | 0.6563 | -377.7956 | -313.5140 | -1.6117 | -1.7057 |
|
| 68 |
+
| 0.5115 | 0.6276 | 300 | 0.5337 | -0.5489 | -1.2889 | 0.7656 | 0.7399 | -392.7693 | -320.1249 | -1.2038 | -1.3223 |
|
| 69 |
+
| 0.5248 | 0.8368 | 400 | 0.5252 | -0.6177 | -1.4052 | 0.7773 | 0.7875 | -404.4023 | -326.9973 | -1.1062 | -1.2274 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- Transformers 4.45.2
|
| 75 |
+
- Pytorch 2.5.1+rocm6.2
|
| 76 |
+
- Datasets 3.5.0
|
| 77 |
+
- Tokenizers 0.20.3
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5619526059058919,
|
| 5 |
+
"train_runtime": 17108.6731,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 3.573,
|
| 8 |
+
"train_steps_per_second": 0.028
|
| 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": "4.45.2"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5619526059058919,
|
| 5 |
+
"train_runtime": 17108.6731,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 3.573,
|
| 8 |
+
"train_steps_per_second": 0.028
|
| 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": 11.583866466472013,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.8379950523376465,
|
| 16 |
+
"logits/rejected": -2.8098106384277344,
|
| 17 |
+
"logps/chosen": -270.4354248046875,
|
| 18 |
+
"logps/rejected": -360.7401123046875,
|
| 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": 10.152186047261853,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.6558611392974854,
|
| 31 |
+
"logits/rejected": -2.6290385723114014,
|
| 32 |
+
"logps/chosen": -280.4989318847656,
|
| 33 |
+
"logps/rejected": -256.2272033691406,
|
| 34 |
+
"loss": 0.693,
|
| 35 |
+
"rewards/accuracies": 0.4861111044883728,
|
| 36 |
+
"rewards/chosen": 0.00041889649583026767,
|
| 37 |
+
"rewards/margins": 0.00031226000282913446,
|
| 38 |
+
"rewards/rejected": 0.00010663647844921798,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 10.23053474060914,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.712508201599121,
|
| 46 |
+
"logits/rejected": -2.6718993186950684,
|
| 47 |
+
"logps/chosen": -284.2615051269531,
|
| 48 |
+
"logps/rejected": -298.14898681640625,
|
| 49 |
+
"loss": 0.6926,
|
| 50 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 51 |
+
"rewards/chosen": 0.0035113892517983913,
|
| 52 |
+
"rewards/margins": 0.0010812545660883188,
|
| 53 |
+
"rewards/rejected": 0.002430134918540716,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 7.720120658846475,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.7409682273864746,
|
| 61 |
+
"logits/rejected": -2.6712632179260254,
|
| 62 |
+
"logps/chosen": -303.02618408203125,
|
| 63 |
+
"logps/rejected": -267.5990295410156,
|
| 64 |
+
"loss": 0.6897,
|
| 65 |
+
"rewards/accuracies": 0.625,
|
| 66 |
+
"rewards/chosen": 0.01898612640798092,
|
| 67 |
+
"rewards/margins": 0.00789168756455183,
|
| 68 |
+
"rewards/rejected": 0.011094437912106514,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 7.528467469118643,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.6612071990966797,
|
| 76 |
+
"logits/rejected": -2.6377806663513184,
|
| 77 |
+
"logps/chosen": -271.030029296875,
|
| 78 |
+
"logps/rejected": -244.3940887451172,
|
| 79 |
+
"loss": 0.6834,
|
| 80 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 81 |
+
"rewards/chosen": 0.050410717725753784,
|
| 82 |
+
"rewards/margins": 0.028806816786527634,
|
| 83 |
+
"rewards/rejected": 0.0216039028018713,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 10.208438942942724,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.5864808559417725,
|
| 91 |
+
"logits/rejected": -2.555248975753784,
|
| 92 |
+
"logps/chosen": -288.3023986816406,
|
| 93 |
+
"logps/rejected": -304.5211486816406,
|
| 94 |
+
"loss": 0.6709,
|
| 95 |
+
"rewards/accuracies": 0.65625,
|
| 96 |
+
"rewards/chosen": 0.033253263682127,
|
| 97 |
+
"rewards/margins": 0.04029017686843872,
|
| 98 |
+
"rewards/rejected": -0.007036912254989147,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 11.881528853183797,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.6687378883361816,
|
| 106 |
+
"logits/rejected": -2.618283748626709,
|
| 107 |
+
"logps/chosen": -273.0628967285156,
|
| 108 |
+
"logps/rejected": -309.696533203125,
|
| 109 |
+
"loss": 0.6523,
|
| 110 |
+
"rewards/accuracies": 0.6875,
|
| 111 |
+
"rewards/chosen": 0.006706203334033489,
|
| 112 |
+
"rewards/margins": 0.0958646684885025,
|
| 113 |
+
"rewards/rejected": -0.08915846049785614,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 14.04538333694439,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.6095685958862305,
|
| 121 |
+
"logits/rejected": -2.6213772296905518,
|
| 122 |
+
"logps/chosen": -278.0463562011719,
|
| 123 |
+
"logps/rejected": -278.8103332519531,
|
| 124 |
+
"loss": 0.6392,
|
| 125 |
+
"rewards/accuracies": 0.706250011920929,
|
| 126 |
+
"rewards/chosen": -0.06012658029794693,
|
| 127 |
+
"rewards/margins": 0.12167074531316757,
|
| 128 |
+
"rewards/rejected": -0.1817973107099533,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 19.703916951696367,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.731635093688965,
|
| 136 |
+
"logits/rejected": -2.696063995361328,
|
| 137 |
+
"logps/chosen": -329.30157470703125,
|
| 138 |
+
"logps/rejected": -283.0393371582031,
|
| 139 |
+
"loss": 0.6238,
|
| 140 |
+
"rewards/accuracies": 0.699999988079071,
|
| 141 |
+
"rewards/chosen": -0.01469563227146864,
|
| 142 |
+
"rewards/margins": 0.16161000728607178,
|
| 143 |
+
"rewards/rejected": -0.1763056218624115,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 13.22124745039195,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.6527259349823,
|
| 151 |
+
"logits/rejected": -2.6169190406799316,
|
| 152 |
+
"logps/chosen": -275.70172119140625,
|
| 153 |
+
"logps/rejected": -298.0541076660156,
|
| 154 |
+
"loss": 0.605,
|
| 155 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 156 |
+
"rewards/chosen": -0.15181758999824524,
|
| 157 |
+
"rewards/margins": 0.29173585772514343,
|
| 158 |
+
"rewards/rejected": -0.44355350732803345,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 11.139969526279701,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -2.6487300395965576,
|
| 166 |
+
"logits/rejected": -2.6412580013275146,
|
| 167 |
+
"logps/chosen": -277.281494140625,
|
| 168 |
+
"logps/rejected": -303.5362243652344,
|
| 169 |
+
"loss": 0.6029,
|
| 170 |
+
"rewards/accuracies": 0.762499988079071,
|
| 171 |
+
"rewards/chosen": -0.11172280460596085,
|
| 172 |
+
"rewards/margins": 0.34097820520401,
|
| 173 |
+
"rewards/rejected": -0.4527009427547455,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -2.704436779022217,
|
| 179 |
+
"eval_logits/rejected": -2.6772303581237793,
|
| 180 |
+
"eval_logps/chosen": -278.78631591796875,
|
| 181 |
+
"eval_logps/rejected": -308.44512939453125,
|
| 182 |
+
"eval_loss": 0.6044007539749146,
|
| 183 |
+
"eval_rewards/accuracies": 0.703125,
|
| 184 |
+
"eval_rewards/chosen": -0.13555391132831573,
|
| 185 |
+
"eval_rewards/margins": 0.3100648522377014,
|
| 186 |
+
"eval_rewards/rejected": -0.44561880826950073,
|
| 187 |
+
"eval_runtime": 165.4414,
|
| 188 |
+
"eval_samples_per_second": 12.089,
|
| 189 |
+
"eval_steps_per_second": 0.193,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 14.78935039360434,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -2.706756591796875,
|
| 197 |
+
"logits/rejected": -2.6693191528320312,
|
| 198 |
+
"logps/chosen": -335.3440246582031,
|
| 199 |
+
"logps/rejected": -314.557373046875,
|
| 200 |
+
"loss": 0.6021,
|
| 201 |
+
"rewards/accuracies": 0.6875,
|
| 202 |
+
"rewards/chosen": -0.2380789816379547,
|
| 203 |
+
"rewards/margins": 0.31388482451438904,
|
| 204 |
+
"rewards/rejected": -0.5519638061523438,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 13.768895735841397,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -2.615844249725342,
|
| 212 |
+
"logits/rejected": -2.613609790802002,
|
| 213 |
+
"logps/chosen": -320.2965087890625,
|
| 214 |
+
"logps/rejected": -336.942138671875,
|
| 215 |
+
"loss": 0.5929,
|
| 216 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 217 |
+
"rewards/chosen": -0.31260889768600464,
|
| 218 |
+
"rewards/margins": 0.3216027021408081,
|
| 219 |
+
"rewards/rejected": -0.634211540222168,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 20.230114151433302,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -2.6277177333831787,
|
| 227 |
+
"logits/rejected": -2.623868703842163,
|
| 228 |
+
"logps/chosen": -319.7072448730469,
|
| 229 |
+
"logps/rejected": -336.055908203125,
|
| 230 |
+
"loss": 0.5716,
|
| 231 |
+
"rewards/accuracies": 0.6875,
|
| 232 |
+
"rewards/chosen": -0.3001275658607483,
|
| 233 |
+
"rewards/margins": 0.34462252259254456,
|
| 234 |
+
"rewards/rejected": -0.6447501182556152,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 20.643526101049833,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -2.518947124481201,
|
| 242 |
+
"logits/rejected": -2.49458646774292,
|
| 243 |
+
"logps/chosen": -367.55987548828125,
|
| 244 |
+
"logps/rejected": -394.8923034667969,
|
| 245 |
+
"loss": 0.5739,
|
| 246 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 247 |
+
"rewards/chosen": -0.41105443239212036,
|
| 248 |
+
"rewards/margins": 0.49972137808799744,
|
| 249 |
+
"rewards/rejected": -0.9107757806777954,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 20.468259477792696,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": -2.2210803031921387,
|
| 257 |
+
"logits/rejected": -2.161284923553467,
|
| 258 |
+
"logps/chosen": -329.27447509765625,
|
| 259 |
+
"logps/rejected": -362.4422607421875,
|
| 260 |
+
"loss": 0.5794,
|
| 261 |
+
"rewards/accuracies": 0.668749988079071,
|
| 262 |
+
"rewards/chosen": -0.38728761672973633,
|
| 263 |
+
"rewards/margins": 0.4247269630432129,
|
| 264 |
+
"rewards/rejected": -0.8120145797729492,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 17.593512774728364,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": -2.1241800785064697,
|
| 272 |
+
"logits/rejected": -1.9932533502578735,
|
| 273 |
+
"logps/chosen": -323.16766357421875,
|
| 274 |
+
"logps/rejected": -362.8275451660156,
|
| 275 |
+
"loss": 0.5669,
|
| 276 |
+
"rewards/accuracies": 0.762499988079071,
|
| 277 |
+
"rewards/chosen": -0.4688601493835449,
|
| 278 |
+
"rewards/margins": 0.6374812126159668,
|
| 279 |
+
"rewards/rejected": -1.1063413619995117,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 20.383844588283708,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": -1.8940010070800781,
|
| 287 |
+
"logits/rejected": -1.801074743270874,
|
| 288 |
+
"logps/chosen": -311.9754333496094,
|
| 289 |
+
"logps/rejected": -341.491943359375,
|
| 290 |
+
"loss": 0.5668,
|
| 291 |
+
"rewards/accuracies": 0.668749988079071,
|
| 292 |
+
"rewards/chosen": -0.5860903263092041,
|
| 293 |
+
"rewards/margins": 0.46389254927635193,
|
| 294 |
+
"rewards/rejected": -1.0499829053878784,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 23.845354705688024,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": -1.5749564170837402,
|
| 302 |
+
"logits/rejected": -1.4734570980072021,
|
| 303 |
+
"logps/chosen": -318.86962890625,
|
| 304 |
+
"logps/rejected": -342.9805603027344,
|
| 305 |
+
"loss": 0.5325,
|
| 306 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 307 |
+
"rewards/chosen": -0.5563257336616516,
|
| 308 |
+
"rewards/margins": 0.5615246891975403,
|
| 309 |
+
"rewards/rejected": -1.1178505420684814,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 36.43353473183915,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": -1.584956169128418,
|
| 317 |
+
"logits/rejected": -1.515265703201294,
|
| 318 |
+
"logps/chosen": -337.5873107910156,
|
| 319 |
+
"logps/rejected": -378.912109375,
|
| 320 |
+
"loss": 0.5533,
|
| 321 |
+
"rewards/accuracies": 0.6875,
|
| 322 |
+
"rewards/chosen": -0.6411080360412598,
|
| 323 |
+
"rewards/margins": 0.527319073677063,
|
| 324 |
+
"rewards/rejected": -1.1684271097183228,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 26.532487463148613,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": -1.5711839199066162,
|
| 332 |
+
"logits/rejected": -1.502673625946045,
|
| 333 |
+
"logps/chosen": -323.0162048339844,
|
| 334 |
+
"logps/rejected": -371.4884338378906,
|
| 335 |
+
"loss": 0.5641,
|
| 336 |
+
"rewards/accuracies": 0.762499988079071,
|
| 337 |
+
"rewards/chosen": -0.5627155303955078,
|
| 338 |
+
"rewards/margins": 0.6155687570571899,
|
| 339 |
+
"rewards/rejected": -1.1782841682434082,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": -1.7056643962860107,
|
| 345 |
+
"eval_logits/rejected": -1.6116549968719482,
|
| 346 |
+
"eval_logps/chosen": -313.5140075683594,
|
| 347 |
+
"eval_logps/rejected": -377.7956237792969,
|
| 348 |
+
"eval_loss": 0.5477674007415771,
|
| 349 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 350 |
+
"eval_rewards/chosen": -0.48283088207244873,
|
| 351 |
+
"eval_rewards/margins": 0.6562928557395935,
|
| 352 |
+
"eval_rewards/rejected": -1.1391236782073975,
|
| 353 |
+
"eval_runtime": 206.3152,
|
| 354 |
+
"eval_samples_per_second": 9.694,
|
| 355 |
+
"eval_steps_per_second": 0.155,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 22.549163158159022,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": -1.5608060359954834,
|
| 363 |
+
"logits/rejected": -1.422942042350769,
|
| 364 |
+
"logps/chosen": -346.8080139160156,
|
| 365 |
+
"logps/rejected": -363.0429382324219,
|
| 366 |
+
"loss": 0.5543,
|
| 367 |
+
"rewards/accuracies": 0.731249988079071,
|
| 368 |
+
"rewards/chosen": -0.5096416473388672,
|
| 369 |
+
"rewards/margins": 0.5783377885818481,
|
| 370 |
+
"rewards/rejected": -1.0879795551300049,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 23.963564313390613,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": -1.545365571975708,
|
| 378 |
+
"logits/rejected": -1.441589593887329,
|
| 379 |
+
"logps/chosen": -333.80340576171875,
|
| 380 |
+
"logps/rejected": -361.42913818359375,
|
| 381 |
+
"loss": 0.5492,
|
| 382 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 383 |
+
"rewards/chosen": -0.5401382446289062,
|
| 384 |
+
"rewards/margins": 0.5902711153030396,
|
| 385 |
+
"rewards/rejected": -1.1304094791412354,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 21.314293618461466,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": -1.4857008457183838,
|
| 393 |
+
"logits/rejected": -1.408174991607666,
|
| 394 |
+
"logps/chosen": -331.1392517089844,
|
| 395 |
+
"logps/rejected": -392.140380859375,
|
| 396 |
+
"loss": 0.5637,
|
| 397 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 398 |
+
"rewards/chosen": -0.7635589838027954,
|
| 399 |
+
"rewards/margins": 0.6409043073654175,
|
| 400 |
+
"rewards/rejected": -1.4044631719589233,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 26.552602991742727,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": -1.5117828845977783,
|
| 408 |
+
"logits/rejected": -1.4116671085357666,
|
| 409 |
+
"logps/chosen": -328.97222900390625,
|
| 410 |
+
"logps/rejected": -379.9259948730469,
|
| 411 |
+
"loss": 0.5204,
|
| 412 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 413 |
+
"rewards/chosen": -0.6180346608161926,
|
| 414 |
+
"rewards/margins": 0.6596405506134033,
|
| 415 |
+
"rewards/rejected": -1.277674913406372,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 22.48166849269774,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": -1.2589095830917358,
|
| 423 |
+
"logits/rejected": -1.0939726829528809,
|
| 424 |
+
"logps/chosen": -382.38226318359375,
|
| 425 |
+
"logps/rejected": -413.89752197265625,
|
| 426 |
+
"loss": 0.539,
|
| 427 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 428 |
+
"rewards/chosen": -0.7211794257164001,
|
| 429 |
+
"rewards/margins": 0.7368528842926025,
|
| 430 |
+
"rewards/rejected": -1.458032250404358,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 31.572413268433298,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": -1.2480559349060059,
|
| 438 |
+
"logits/rejected": -1.1091158390045166,
|
| 439 |
+
"logps/chosen": -374.61962890625,
|
| 440 |
+
"logps/rejected": -404.3774108886719,
|
| 441 |
+
"loss": 0.5383,
|
| 442 |
+
"rewards/accuracies": 0.762499988079071,
|
| 443 |
+
"rewards/chosen": -0.7782480716705322,
|
| 444 |
+
"rewards/margins": 0.656974196434021,
|
| 445 |
+
"rewards/rejected": -1.4352221488952637,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 21.75156418734587,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": -1.2881791591644287,
|
| 453 |
+
"logits/rejected": -1.1539137363433838,
|
| 454 |
+
"logps/chosen": -343.4298400878906,
|
| 455 |
+
"logps/rejected": -379.86065673828125,
|
| 456 |
+
"loss": 0.543,
|
| 457 |
+
"rewards/accuracies": 0.675000011920929,
|
| 458 |
+
"rewards/chosen": -0.6388952732086182,
|
| 459 |
+
"rewards/margins": 0.5688951015472412,
|
| 460 |
+
"rewards/rejected": -1.207790493965149,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 23.061436094805646,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": -1.1816035509109497,
|
| 468 |
+
"logits/rejected": -1.0291564464569092,
|
| 469 |
+
"logps/chosen": -333.220458984375,
|
| 470 |
+
"logps/rejected": -370.09967041015625,
|
| 471 |
+
"loss": 0.5495,
|
| 472 |
+
"rewards/accuracies": 0.706250011920929,
|
| 473 |
+
"rewards/chosen": -0.6366706490516663,
|
| 474 |
+
"rewards/margins": 0.6222621202468872,
|
| 475 |
+
"rewards/rejected": -1.2589328289031982,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 20.823231284467255,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": -1.3357089757919312,
|
| 483 |
+
"logits/rejected": -1.2783232927322388,
|
| 484 |
+
"logps/chosen": -329.78460693359375,
|
| 485 |
+
"logps/rejected": -380.2730407714844,
|
| 486 |
+
"loss": 0.5386,
|
| 487 |
+
"rewards/accuracies": 0.731249988079071,
|
| 488 |
+
"rewards/chosen": -0.7694828510284424,
|
| 489 |
+
"rewards/margins": 0.5911151170730591,
|
| 490 |
+
"rewards/rejected": -1.3605979681015015,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 20.702370180538725,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": -1.3041666746139526,
|
| 498 |
+
"logits/rejected": -1.1101744174957275,
|
| 499 |
+
"logps/chosen": -373.60302734375,
|
| 500 |
+
"logps/rejected": -389.8649597167969,
|
| 501 |
+
"loss": 0.5115,
|
| 502 |
+
"rewards/accuracies": 0.706250011920929,
|
| 503 |
+
"rewards/chosen": -0.6723031997680664,
|
| 504 |
+
"rewards/margins": 0.6251998543739319,
|
| 505 |
+
"rewards/rejected": -1.2975029945373535,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": -1.322279453277588,
|
| 511 |
+
"eval_logits/rejected": -1.2038487195968628,
|
| 512 |
+
"eval_logps/chosen": -320.12493896484375,
|
| 513 |
+
"eval_logps/rejected": -392.7693176269531,
|
| 514 |
+
"eval_loss": 0.5337320566177368,
|
| 515 |
+
"eval_rewards/accuracies": 0.765625,
|
| 516 |
+
"eval_rewards/chosen": -0.548940122127533,
|
| 517 |
+
"eval_rewards/margins": 0.7399202585220337,
|
| 518 |
+
"eval_rewards/rejected": -1.2888603210449219,
|
| 519 |
+
"eval_runtime": 171.5093,
|
| 520 |
+
"eval_samples_per_second": 11.661,
|
| 521 |
+
"eval_steps_per_second": 0.187,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 23.91999554060663,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": -1.1424458026885986,
|
| 529 |
+
"logits/rejected": -1.046404480934143,
|
| 530 |
+
"logps/chosen": -371.3121032714844,
|
| 531 |
+
"logps/rejected": -396.40447998046875,
|
| 532 |
+
"loss": 0.5165,
|
| 533 |
+
"rewards/accuracies": 0.737500011920929,
|
| 534 |
+
"rewards/chosen": -0.6775599122047424,
|
| 535 |
+
"rewards/margins": 0.6300493478775024,
|
| 536 |
+
"rewards/rejected": -1.3076092004776,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 20.9212078859009,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": -1.1547837257385254,
|
| 544 |
+
"logits/rejected": -1.013450264930725,
|
| 545 |
+
"logps/chosen": -370.629638671875,
|
| 546 |
+
"logps/rejected": -387.43560791015625,
|
| 547 |
+
"loss": 0.5093,
|
| 548 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 549 |
+
"rewards/chosen": -0.7318499684333801,
|
| 550 |
+
"rewards/margins": 0.6536003947257996,
|
| 551 |
+
"rewards/rejected": -1.3854502439498901,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 24.136877828555647,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": -1.1623682975769043,
|
| 559 |
+
"logits/rejected": -1.0484459400177002,
|
| 560 |
+
"logps/chosen": -338.98895263671875,
|
| 561 |
+
"logps/rejected": -395.61602783203125,
|
| 562 |
+
"loss": 0.519,
|
| 563 |
+
"rewards/accuracies": 0.6875,
|
| 564 |
+
"rewards/chosen": -0.6542141437530518,
|
| 565 |
+
"rewards/margins": 0.6077429056167603,
|
| 566 |
+
"rewards/rejected": -1.2619569301605225,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 26.018359074294946,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": -1.2398521900177002,
|
| 574 |
+
"logits/rejected": -1.0911023616790771,
|
| 575 |
+
"logps/chosen": -355.0738220214844,
|
| 576 |
+
"logps/rejected": -431.7353515625,
|
| 577 |
+
"loss": 0.5227,
|
| 578 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 579 |
+
"rewards/chosen": -0.5684548020362854,
|
| 580 |
+
"rewards/margins": 0.8478826284408569,
|
| 581 |
+
"rewards/rejected": -1.4163373708724976,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 24.25963140644168,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": -1.1729114055633545,
|
| 589 |
+
"logits/rejected": -1.0101940631866455,
|
| 590 |
+
"logps/chosen": -335.6921081542969,
|
| 591 |
+
"logps/rejected": -396.78875732421875,
|
| 592 |
+
"loss": 0.5003,
|
| 593 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 594 |
+
"rewards/chosen": -0.657314658164978,
|
| 595 |
+
"rewards/margins": 0.7315484285354614,
|
| 596 |
+
"rewards/rejected": -1.3888630867004395,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 21.3034653324035,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": -1.201933741569519,
|
| 604 |
+
"logits/rejected": -0.9621894955635071,
|
| 605 |
+
"logps/chosen": -378.11016845703125,
|
| 606 |
+
"logps/rejected": -411.57177734375,
|
| 607 |
+
"loss": 0.5334,
|
| 608 |
+
"rewards/accuracies": 0.731249988079071,
|
| 609 |
+
"rewards/chosen": -0.7192701101303101,
|
| 610 |
+
"rewards/margins": 0.7363580465316772,
|
| 611 |
+
"rewards/rejected": -1.4556280374526978,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 23.05824742992241,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": -1.0932825803756714,
|
| 619 |
+
"logits/rejected": -0.9322390556335449,
|
| 620 |
+
"logps/chosen": -361.45855712890625,
|
| 621 |
+
"logps/rejected": -383.38348388671875,
|
| 622 |
+
"loss": 0.5287,
|
| 623 |
+
"rewards/accuracies": 0.71875,
|
| 624 |
+
"rewards/chosen": -0.7997485995292664,
|
| 625 |
+
"rewards/margins": 0.6485676765441895,
|
| 626 |
+
"rewards/rejected": -1.448316216468811,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 22.793937976155306,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": -0.9988774061203003,
|
| 634 |
+
"logits/rejected": -0.9683791399002075,
|
| 635 |
+
"logps/chosen": -346.8892822265625,
|
| 636 |
+
"logps/rejected": -440.4715881347656,
|
| 637 |
+
"loss": 0.516,
|
| 638 |
+
"rewards/accuracies": 0.71875,
|
| 639 |
+
"rewards/chosen": -0.9633440971374512,
|
| 640 |
+
"rewards/margins": 0.747096598148346,
|
| 641 |
+
"rewards/rejected": -1.7104406356811523,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 23.943540866299713,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": -1.1462968587875366,
|
| 649 |
+
"logits/rejected": -1.0620715618133545,
|
| 650 |
+
"logps/chosen": -383.8972473144531,
|
| 651 |
+
"logps/rejected": -434.6634216308594,
|
| 652 |
+
"loss": 0.511,
|
| 653 |
+
"rewards/accuracies": 0.731249988079071,
|
| 654 |
+
"rewards/chosen": -0.7514833211898804,
|
| 655 |
+
"rewards/margins": 0.6801032423973083,
|
| 656 |
+
"rewards/rejected": -1.431586503982544,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 27.044905725894765,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": -1.0648071765899658,
|
| 664 |
+
"logits/rejected": -0.924199104309082,
|
| 665 |
+
"logps/chosen": -353.1241149902344,
|
| 666 |
+
"logps/rejected": -417.89599609375,
|
| 667 |
+
"loss": 0.5248,
|
| 668 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 669 |
+
"rewards/chosen": -0.7453318238258362,
|
| 670 |
+
"rewards/margins": 0.7299187779426575,
|
| 671 |
+
"rewards/rejected": -1.4752506017684937,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": -1.2273505926132202,
|
| 677 |
+
"eval_logits/rejected": -1.1061500310897827,
|
| 678 |
+
"eval_logps/chosen": -326.99725341796875,
|
| 679 |
+
"eval_logps/rejected": -404.40228271484375,
|
| 680 |
+
"eval_loss": 0.5251822471618652,
|
| 681 |
+
"eval_rewards/accuracies": 0.77734375,
|
| 682 |
+
"eval_rewards/chosen": -0.6176633238792419,
|
| 683 |
+
"eval_rewards/margins": 0.7875270843505859,
|
| 684 |
+
"eval_rewards/rejected": -1.4051904678344727,
|
| 685 |
+
"eval_runtime": 164.4247,
|
| 686 |
+
"eval_samples_per_second": 12.164,
|
| 687 |
+
"eval_steps_per_second": 0.195,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 20.813895306539607,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": -1.1425502300262451,
|
| 695 |
+
"logits/rejected": -0.9420346021652222,
|
| 696 |
+
"logps/chosen": -370.9138488769531,
|
| 697 |
+
"logps/rejected": -416.3164978027344,
|
| 698 |
+
"loss": 0.5248,
|
| 699 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 700 |
+
"rewards/chosen": -0.6757674217224121,
|
| 701 |
+
"rewards/margins": 0.7509845495223999,
|
| 702 |
+
"rewards/rejected": -1.4267520904541016,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 27.427829452347144,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": -1.164847731590271,
|
| 710 |
+
"logits/rejected": -1.0420937538146973,
|
| 711 |
+
"logps/chosen": -371.6080017089844,
|
| 712 |
+
"logps/rejected": -413.12091064453125,
|
| 713 |
+
"loss": 0.5174,
|
| 714 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 715 |
+
"rewards/chosen": -0.7168834805488586,
|
| 716 |
+
"rewards/margins": 0.664655864238739,
|
| 717 |
+
"rewards/rejected": -1.3815393447875977,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 25.102555861952474,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": -1.0322716236114502,
|
| 725 |
+
"logits/rejected": -0.9212514758110046,
|
| 726 |
+
"logps/chosen": -359.2303771972656,
|
| 727 |
+
"logps/rejected": -423.27105712890625,
|
| 728 |
+
"loss": 0.5014,
|
| 729 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 730 |
+
"rewards/chosen": -0.7200824022293091,
|
| 731 |
+
"rewards/margins": 0.8008586764335632,
|
| 732 |
+
"rewards/rejected": -1.520941138267517,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 24.09733436821402,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": -1.1884571313858032,
|
| 740 |
+
"logits/rejected": -1.0310804843902588,
|
| 741 |
+
"logps/chosen": -350.55908203125,
|
| 742 |
+
"logps/rejected": -402.35052490234375,
|
| 743 |
+
"loss": 0.5188,
|
| 744 |
+
"rewards/accuracies": 0.768750011920929,
|
| 745 |
+
"rewards/chosen": -0.5973304510116577,
|
| 746 |
+
"rewards/margins": 0.6999295353889465,
|
| 747 |
+
"rewards/rejected": -1.297260046005249,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 21.830318518139613,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": -1.063053846359253,
|
| 755 |
+
"logits/rejected": -0.9926292300224304,
|
| 756 |
+
"logps/chosen": -374.13433837890625,
|
| 757 |
+
"logps/rejected": -451.5859375,
|
| 758 |
+
"loss": 0.5178,
|
| 759 |
+
"rewards/accuracies": 0.675000011920929,
|
| 760 |
+
"rewards/chosen": -0.7738043069839478,
|
| 761 |
+
"rewards/margins": 0.595235288143158,
|
| 762 |
+
"rewards/rejected": -1.36903977394104,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 23.470311120608795,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": -1.0980103015899658,
|
| 770 |
+
"logits/rejected": -0.9631898999214172,
|
| 771 |
+
"logps/chosen": -347.15081787109375,
|
| 772 |
+
"logps/rejected": -384.52691650390625,
|
| 773 |
+
"loss": 0.5214,
|
| 774 |
+
"rewards/accuracies": 0.75,
|
| 775 |
+
"rewards/chosen": -0.7243081331253052,
|
| 776 |
+
"rewards/margins": 0.6908310651779175,
|
| 777 |
+
"rewards/rejected": -1.4151391983032227,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 20.563815183603257,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": -1.1189404726028442,
|
| 785 |
+
"logits/rejected": -0.9989368319511414,
|
| 786 |
+
"logps/chosen": -355.0899353027344,
|
| 787 |
+
"logps/rejected": -417.033935546875,
|
| 788 |
+
"loss": 0.5035,
|
| 789 |
+
"rewards/accuracies": 0.699999988079071,
|
| 790 |
+
"rewards/chosen": -0.6831948757171631,
|
| 791 |
+
"rewards/margins": 0.6488174200057983,
|
| 792 |
+
"rewards/rejected": -1.332012414932251,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5619526059058919,
|
| 800 |
+
"train_runtime": 17108.6731,
|
| 801 |
+
"train_samples_per_second": 3.573,
|
| 802 |
+
"train_steps_per_second": 0.028
|
| 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 |
+
}
|