Model save
Browse files
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 3 |
+
library_name: peft
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-dpo-qlora-uf6k-5e-6
|
| 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-dpo-qlora-uf6k-5e-6
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.6103
|
| 22 |
+
- Rewards/chosen: -0.2010
|
| 23 |
+
- Rewards/rejected: -0.4650
|
| 24 |
+
- Rewards/accuracies: 0.6860
|
| 25 |
+
- Rewards/margins: 0.2641
|
| 26 |
+
- Rewards/margins Max: 1.0353
|
| 27 |
+
- Rewards/margins Min: -0.4649
|
| 28 |
+
- Rewards/margins Std: 0.5062
|
| 29 |
+
- Logps/rejected: -305.0838
|
| 30 |
+
- Logps/chosen: -304.6903
|
| 31 |
+
- Logits/rejected: -2.7402
|
| 32 |
+
- Logits/chosen: -2.7722
|
| 33 |
+
|
| 34 |
+
## Model description
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Intended uses & limitations
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training and evaluation data
|
| 43 |
+
|
| 44 |
+
More information needed
|
| 45 |
+
|
| 46 |
+
## Training procedure
|
| 47 |
+
|
| 48 |
+
### Training hyperparameters
|
| 49 |
+
|
| 50 |
+
The following hyperparameters were used during training:
|
| 51 |
+
- learning_rate: 5e-06
|
| 52 |
+
- train_batch_size: 4
|
| 53 |
+
- eval_batch_size: 8
|
| 54 |
+
- seed: 42
|
| 55 |
+
- distributed_type: multi-GPU
|
| 56 |
+
- num_devices: 2
|
| 57 |
+
- gradient_accumulation_steps: 2
|
| 58 |
+
- total_train_batch_size: 16
|
| 59 |
+
- total_eval_batch_size: 16
|
| 60 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 61 |
+
- lr_scheduler_type: cosine
|
| 62 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 63 |
+
- num_epochs: 1
|
| 64 |
+
|
| 65 |
+
### Training results
|
| 66 |
+
|
| 67 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Rewards/margins Max | Rewards/margins Min | Rewards/margins Std | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 68 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:-------------------:|:-------------------:|:-------------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 69 |
+
| 0.6517 | 0.3 | 100 | 0.6560 | -0.0102 | -0.1055 | 0.6900 | 0.0954 | 0.4105 | -0.2027 | 0.2020 | -269.1304 | -285.6097 | -2.7441 | -2.7791 |
|
| 70 |
+
| 0.5795 | 0.61 | 200 | 0.6177 | -0.2139 | -0.4530 | 0.6820 | 0.2391 | 0.9738 | -0.4425 | 0.4777 | -303.8808 | -305.9859 | -2.7382 | -2.7701 |
|
| 71 |
+
| 0.5791 | 0.91 | 300 | 0.6103 | -0.2010 | -0.4650 | 0.6860 | 0.2641 | 1.0353 | -0.4649 | 0.5062 | -305.0838 | -304.6903 | -2.7402 | -2.7722 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- PEFT 0.7.1
|
| 77 |
+
- Transformers 4.39.0.dev0
|
| 78 |
+
- Pytorch 2.1.2+cu121
|
| 79 |
+
- Datasets 2.14.6
|
| 80 |
+
- Tokenizers 0.15.2
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c370009fea2832bc7b20b548fc845861beb748cbba4102c0f08fe4367248cce
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6308559574254741,
|
| 4 |
+
"train_runtime": 3887.3613,
|
| 5 |
+
"train_samples": 5263,
|
| 6 |
+
"train_samples_per_second": 1.354,
|
| 7 |
+
"train_steps_per_second": 0.085
|
| 8 |
+
}
|
runs/Jul25_08-24-43_notebook-deployment-48-7d9b6c99-khd85/events.out.tfevents.1721895982.notebook-deployment-48-7d9b6c99-khd85.3153937.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89c5ace5c905987827dc184ec523572923fd1f400e859aee18654dad310c7954
|
| 3 |
+
size 37307
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6308559574254741,
|
| 4 |
+
"train_runtime": 3887.3613,
|
| 5 |
+
"train_samples": 5263,
|
| 6 |
+
"train_samples_per_second": 1.354,
|
| 7 |
+
"train_steps_per_second": 0.085
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,681 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 329,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0,
|
| 13 |
+
"grad_norm": 2.09392729051305,
|
| 14 |
+
"learning_rate": 1.5151515151515152e-07,
|
| 15 |
+
"logits/chosen": -2.6820077896118164,
|
| 16 |
+
"logits/rejected": -2.6930205821990967,
|
| 17 |
+
"logps/chosen": -281.2528381347656,
|
| 18 |
+
"logps/rejected": -258.0622253417969,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/margins_max": 0.0,
|
| 24 |
+
"rewards/margins_min": 0.0,
|
| 25 |
+
"rewards/margins_std": 0.0,
|
| 26 |
+
"rewards/rejected": 0.0,
|
| 27 |
+
"step": 1
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 0.03,
|
| 31 |
+
"grad_norm": 2.0925129894368433,
|
| 32 |
+
"learning_rate": 1.5151515151515152e-06,
|
| 33 |
+
"logits/chosen": -2.7695705890655518,
|
| 34 |
+
"logits/rejected": -2.755211353302002,
|
| 35 |
+
"logps/chosen": -284.5762634277344,
|
| 36 |
+
"logps/rejected": -249.84103393554688,
|
| 37 |
+
"loss": 0.6932,
|
| 38 |
+
"rewards/accuracies": 0.4166666567325592,
|
| 39 |
+
"rewards/chosen": 0.00027878841501660645,
|
| 40 |
+
"rewards/margins": 0.00013094051973894238,
|
| 41 |
+
"rewards/margins_max": 0.0033118026331067085,
|
| 42 |
+
"rewards/margins_min": -0.002655874704942107,
|
| 43 |
+
"rewards/margins_std": 0.0026171421632170677,
|
| 44 |
+
"rewards/rejected": 0.0001478478661738336,
|
| 45 |
+
"step": 10
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"epoch": 0.06,
|
| 49 |
+
"grad_norm": 1.9815822382919812,
|
| 50 |
+
"learning_rate": 3.0303030303030305e-06,
|
| 51 |
+
"logits/chosen": -2.833542823791504,
|
| 52 |
+
"logits/rejected": -2.7807106971740723,
|
| 53 |
+
"logps/chosen": -291.4638977050781,
|
| 54 |
+
"logps/rejected": -270.51416015625,
|
| 55 |
+
"loss": 0.6925,
|
| 56 |
+
"rewards/accuracies": 0.625,
|
| 57 |
+
"rewards/chosen": 0.0006784057477489114,
|
| 58 |
+
"rewards/margins": 0.0018588269595056772,
|
| 59 |
+
"rewards/margins_max": 0.006385893560945988,
|
| 60 |
+
"rewards/margins_min": -0.0022843112237751484,
|
| 61 |
+
"rewards/margins_std": 0.0037722717970609665,
|
| 62 |
+
"rewards/rejected": -0.001180421095341444,
|
| 63 |
+
"step": 20
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"epoch": 0.09,
|
| 67 |
+
"grad_norm": 1.67492468935064,
|
| 68 |
+
"learning_rate": 4.5454545454545455e-06,
|
| 69 |
+
"logits/chosen": -2.860595226287842,
|
| 70 |
+
"logits/rejected": -2.8134350776672363,
|
| 71 |
+
"logps/chosen": -259.4276123046875,
|
| 72 |
+
"logps/rejected": -228.0315704345703,
|
| 73 |
+
"loss": 0.6913,
|
| 74 |
+
"rewards/accuracies": 0.637499988079071,
|
| 75 |
+
"rewards/chosen": -0.001823395723477006,
|
| 76 |
+
"rewards/margins": 0.004928162787109613,
|
| 77 |
+
"rewards/margins_max": 0.017859386280179024,
|
| 78 |
+
"rewards/margins_min": -0.005589204374700785,
|
| 79 |
+
"rewards/margins_std": 0.01042240671813488,
|
| 80 |
+
"rewards/rejected": -0.006751558743417263,
|
| 81 |
+
"step": 30
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 0.12,
|
| 85 |
+
"grad_norm": 1.5968086960496506,
|
| 86 |
+
"learning_rate": 4.993103596812269e-06,
|
| 87 |
+
"logits/chosen": -2.8253955841064453,
|
| 88 |
+
"logits/rejected": -2.762451648712158,
|
| 89 |
+
"logps/chosen": -317.4245910644531,
|
| 90 |
+
"logps/rejected": -226.83029174804688,
|
| 91 |
+
"loss": 0.6847,
|
| 92 |
+
"rewards/accuracies": 0.75,
|
| 93 |
+
"rewards/chosen": 0.001176043995656073,
|
| 94 |
+
"rewards/margins": 0.022631246596574783,
|
| 95 |
+
"rewards/margins_max": 0.05663127824664116,
|
| 96 |
+
"rewards/margins_min": -0.002739027375355363,
|
| 97 |
+
"rewards/margins_std": 0.027225341647863388,
|
| 98 |
+
"rewards/rejected": -0.021455200389027596,
|
| 99 |
+
"step": 40
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.15,
|
| 103 |
+
"grad_norm": 1.7771351450812785,
|
| 104 |
+
"learning_rate": 4.95941685833271e-06,
|
| 105 |
+
"logits/chosen": -2.7818782329559326,
|
| 106 |
+
"logits/rejected": -2.796020984649658,
|
| 107 |
+
"logps/chosen": -243.5937957763672,
|
| 108 |
+
"logps/rejected": -282.7391662597656,
|
| 109 |
+
"loss": 0.6835,
|
| 110 |
+
"rewards/accuracies": 0.737500011920929,
|
| 111 |
+
"rewards/chosen": -0.006589516066014767,
|
| 112 |
+
"rewards/margins": 0.02196229062974453,
|
| 113 |
+
"rewards/margins_max": 0.07266667485237122,
|
| 114 |
+
"rewards/margins_min": -0.01875842735171318,
|
| 115 |
+
"rewards/margins_std": 0.04057556018233299,
|
| 116 |
+
"rewards/rejected": -0.02855180762708187,
|
| 117 |
+
"step": 50
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 0.18,
|
| 121 |
+
"grad_norm": 1.962252962150773,
|
| 122 |
+
"learning_rate": 4.898051734555676e-06,
|
| 123 |
+
"logits/chosen": -2.818239688873291,
|
| 124 |
+
"logits/rejected": -2.8286662101745605,
|
| 125 |
+
"logps/chosen": -320.9726257324219,
|
| 126 |
+
"logps/rejected": -286.5067138671875,
|
| 127 |
+
"loss": 0.6723,
|
| 128 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 129 |
+
"rewards/chosen": 0.00948464497923851,
|
| 130 |
+
"rewards/margins": 0.042791008949279785,
|
| 131 |
+
"rewards/margins_max": 0.1431531012058258,
|
| 132 |
+
"rewards/margins_min": -0.06081182882189751,
|
| 133 |
+
"rewards/margins_std": 0.08912219107151031,
|
| 134 |
+
"rewards/rejected": -0.033306367695331573,
|
| 135 |
+
"step": 60
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.21,
|
| 139 |
+
"grad_norm": 1.8587103545341894,
|
| 140 |
+
"learning_rate": 4.809698831278217e-06,
|
| 141 |
+
"logits/chosen": -2.7206215858459473,
|
| 142 |
+
"logits/rejected": -2.7101528644561768,
|
| 143 |
+
"logps/chosen": -266.5228576660156,
|
| 144 |
+
"logps/rejected": -251.3375244140625,
|
| 145 |
+
"loss": 0.6729,
|
| 146 |
+
"rewards/accuracies": 0.699999988079071,
|
| 147 |
+
"rewards/chosen": -0.00045425668940879405,
|
| 148 |
+
"rewards/margins": 0.049175992608070374,
|
| 149 |
+
"rewards/margins_max": 0.13846850395202637,
|
| 150 |
+
"rewards/margins_min": -0.052676666527986526,
|
| 151 |
+
"rewards/margins_std": 0.08763651549816132,
|
| 152 |
+
"rewards/rejected": -0.04963024705648422,
|
| 153 |
+
"step": 70
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.24,
|
| 157 |
+
"grad_norm": 2.2136192637287415,
|
| 158 |
+
"learning_rate": 4.695352475952706e-06,
|
| 159 |
+
"logits/chosen": -2.807755947113037,
|
| 160 |
+
"logits/rejected": -2.7845520973205566,
|
| 161 |
+
"logps/chosen": -283.79571533203125,
|
| 162 |
+
"logps/rejected": -282.36761474609375,
|
| 163 |
+
"loss": 0.6639,
|
| 164 |
+
"rewards/accuracies": 0.737500011920929,
|
| 165 |
+
"rewards/chosen": -0.015222731046378613,
|
| 166 |
+
"rewards/margins": 0.05928920581936836,
|
| 167 |
+
"rewards/margins_max": 0.19641625881195068,
|
| 168 |
+
"rewards/margins_min": -0.08654795587062836,
|
| 169 |
+
"rewards/margins_std": 0.12829935550689697,
|
| 170 |
+
"rewards/rejected": -0.07451193034648895,
|
| 171 |
+
"step": 80
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"epoch": 0.27,
|
| 175 |
+
"grad_norm": 2.428092801188637,
|
| 176 |
+
"learning_rate": 4.556299527482029e-06,
|
| 177 |
+
"logits/chosen": -2.7647454738616943,
|
| 178 |
+
"logits/rejected": -2.7148241996765137,
|
| 179 |
+
"logps/chosen": -301.64178466796875,
|
| 180 |
+
"logps/rejected": -302.57427978515625,
|
| 181 |
+
"loss": 0.6646,
|
| 182 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 183 |
+
"rewards/chosen": -0.060897838324308395,
|
| 184 |
+
"rewards/margins": 0.049183908849954605,
|
| 185 |
+
"rewards/margins_max": 0.24967026710510254,
|
| 186 |
+
"rewards/margins_min": -0.15425434708595276,
|
| 187 |
+
"rewards/margins_std": 0.17774364352226257,
|
| 188 |
+
"rewards/rejected": -0.110081747174263,
|
| 189 |
+
"step": 90
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.3,
|
| 193 |
+
"grad_norm": 1.9682951507719486,
|
| 194 |
+
"learning_rate": 4.394104893853007e-06,
|
| 195 |
+
"logits/chosen": -2.860042095184326,
|
| 196 |
+
"logits/rejected": -2.823477268218994,
|
| 197 |
+
"logps/chosen": -277.8373718261719,
|
| 198 |
+
"logps/rejected": -274.130615234375,
|
| 199 |
+
"loss": 0.6517,
|
| 200 |
+
"rewards/accuracies": 0.762499988079071,
|
| 201 |
+
"rewards/chosen": -0.04375104233622551,
|
| 202 |
+
"rewards/margins": 0.1265198141336441,
|
| 203 |
+
"rewards/margins_max": 0.31651413440704346,
|
| 204 |
+
"rewards/margins_min": -0.057662349194288254,
|
| 205 |
+
"rewards/margins_std": 0.16801223158836365,
|
| 206 |
+
"rewards/rejected": -0.17027084529399872,
|
| 207 |
+
"step": 100
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.3,
|
| 211 |
+
"eval_logits/chosen": -2.7791192531585693,
|
| 212 |
+
"eval_logits/rejected": -2.744123697280884,
|
| 213 |
+
"eval_logps/chosen": -285.6097412109375,
|
| 214 |
+
"eval_logps/rejected": -269.1304016113281,
|
| 215 |
+
"eval_loss": 0.6559571623802185,
|
| 216 |
+
"eval_rewards/accuracies": 0.6899999976158142,
|
| 217 |
+
"eval_rewards/chosen": -0.0101632634177804,
|
| 218 |
+
"eval_rewards/margins": 0.0953514501452446,
|
| 219 |
+
"eval_rewards/margins_max": 0.410457581281662,
|
| 220 |
+
"eval_rewards/margins_min": -0.20267993211746216,
|
| 221 |
+
"eval_rewards/margins_std": 0.20197777450084686,
|
| 222 |
+
"eval_rewards/rejected": -0.10551470518112183,
|
| 223 |
+
"eval_runtime": 429.1012,
|
| 224 |
+
"eval_samples_per_second": 4.661,
|
| 225 |
+
"eval_steps_per_second": 0.291,
|
| 226 |
+
"step": 100
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.33,
|
| 230 |
+
"grad_norm": 2.460788298428497,
|
| 231 |
+
"learning_rate": 4.210593920593201e-06,
|
| 232 |
+
"logits/chosen": -2.7310755252838135,
|
| 233 |
+
"logits/rejected": -2.716763973236084,
|
| 234 |
+
"logps/chosen": -314.26788330078125,
|
| 235 |
+
"logps/rejected": -245.84921264648438,
|
| 236 |
+
"loss": 0.6531,
|
| 237 |
+
"rewards/accuracies": 0.637499988079071,
|
| 238 |
+
"rewards/chosen": -0.02511635422706604,
|
| 239 |
+
"rewards/margins": 0.07915838807821274,
|
| 240 |
+
"rewards/margins_max": 0.2661080062389374,
|
| 241 |
+
"rewards/margins_min": -0.1229986920952797,
|
| 242 |
+
"rewards/margins_std": 0.17364470660686493,
|
| 243 |
+
"rewards/rejected": -0.10427474975585938,
|
| 244 |
+
"step": 110
|
| 245 |
+
},
|
| 246 |
+
{
|
| 247 |
+
"epoch": 0.36,
|
| 248 |
+
"grad_norm": 2.4585744987342597,
|
| 249 |
+
"learning_rate": 4.007831848252212e-06,
|
| 250 |
+
"logits/chosen": -2.713639736175537,
|
| 251 |
+
"logits/rejected": -2.717987537384033,
|
| 252 |
+
"logps/chosen": -329.2998046875,
|
| 253 |
+
"logps/rejected": -291.00775146484375,
|
| 254 |
+
"loss": 0.6411,
|
| 255 |
+
"rewards/accuracies": 0.675000011920929,
|
| 256 |
+
"rewards/chosen": -0.05735304206609726,
|
| 257 |
+
"rewards/margins": 0.10872934013605118,
|
| 258 |
+
"rewards/margins_max": 0.3730432391166687,
|
| 259 |
+
"rewards/margins_min": -0.18109610676765442,
|
| 260 |
+
"rewards/margins_std": 0.2501201629638672,
|
| 261 |
+
"rewards/rejected": -0.16608235239982605,
|
| 262 |
+
"step": 120
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 0.4,
|
| 266 |
+
"grad_norm": 2.296742130510628,
|
| 267 |
+
"learning_rate": 3.7881005700938635e-06,
|
| 268 |
+
"logits/chosen": -2.784316301345825,
|
| 269 |
+
"logits/rejected": -2.794018507003784,
|
| 270 |
+
"logps/chosen": -270.2655334472656,
|
| 271 |
+
"logps/rejected": -249.17416381835938,
|
| 272 |
+
"loss": 0.6356,
|
| 273 |
+
"rewards/accuracies": 0.637499988079071,
|
| 274 |
+
"rewards/chosen": -0.0387195460498333,
|
| 275 |
+
"rewards/margins": 0.11184574663639069,
|
| 276 |
+
"rewards/margins_max": 0.4147607386112213,
|
| 277 |
+
"rewards/margins_min": -0.18675808608531952,
|
| 278 |
+
"rewards/margins_std": 0.26902341842651367,
|
| 279 |
+
"rewards/rejected": -0.15056531131267548,
|
| 280 |
+
"step": 130
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.43,
|
| 284 |
+
"grad_norm": 2.5520149273689916,
|
| 285 |
+
"learning_rate": 3.553872951569236e-06,
|
| 286 |
+
"logits/chosen": -2.7240898609161377,
|
| 287 |
+
"logits/rejected": -2.7166969776153564,
|
| 288 |
+
"logps/chosen": -307.4956970214844,
|
| 289 |
+
"logps/rejected": -302.90875244140625,
|
| 290 |
+
"loss": 0.6294,
|
| 291 |
+
"rewards/accuracies": 0.699999988079071,
|
| 292 |
+
"rewards/chosen": -0.13102450966835022,
|
| 293 |
+
"rewards/margins": 0.19427050650119781,
|
| 294 |
+
"rewards/margins_max": 0.5367092490196228,
|
| 295 |
+
"rewards/margins_min": -0.12927336990833282,
|
| 296 |
+
"rewards/margins_std": 0.2990504503250122,
|
| 297 |
+
"rewards/rejected": -0.32529503107070923,
|
| 298 |
+
"step": 140
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 0.46,
|
| 302 |
+
"grad_norm": 2.409073201436679,
|
| 303 |
+
"learning_rate": 3.307785000580313e-06,
|
| 304 |
+
"logits/chosen": -2.7703697681427,
|
| 305 |
+
"logits/rejected": -2.758200168609619,
|
| 306 |
+
"logps/chosen": -292.15093994140625,
|
| 307 |
+
"logps/rejected": -284.210693359375,
|
| 308 |
+
"loss": 0.6342,
|
| 309 |
+
"rewards/accuracies": 0.699999988079071,
|
| 310 |
+
"rewards/chosen": -0.21509556472301483,
|
| 311 |
+
"rewards/margins": 0.17580775916576385,
|
| 312 |
+
"rewards/margins_max": 0.5677517056465149,
|
| 313 |
+
"rewards/margins_min": -0.20906031131744385,
|
| 314 |
+
"rewards/margins_std": 0.3507745862007141,
|
| 315 |
+
"rewards/rejected": -0.3909032940864563,
|
| 316 |
+
"step": 150
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"epoch": 0.49,
|
| 320 |
+
"grad_norm": 3.016300738429993,
|
| 321 |
+
"learning_rate": 3.052606201731325e-06,
|
| 322 |
+
"logits/chosen": -2.735063076019287,
|
| 323 |
+
"logits/rejected": -2.718566417694092,
|
| 324 |
+
"logps/chosen": -261.0966796875,
|
| 325 |
+
"logps/rejected": -258.08099365234375,
|
| 326 |
+
"loss": 0.6538,
|
| 327 |
+
"rewards/accuracies": 0.625,
|
| 328 |
+
"rewards/chosen": -0.05592544749379158,
|
| 329 |
+
"rewards/margins": 0.11891476809978485,
|
| 330 |
+
"rewards/margins_max": 0.4770866930484772,
|
| 331 |
+
"rewards/margins_min": -0.23589050769805908,
|
| 332 |
+
"rewards/margins_std": 0.3210330605506897,
|
| 333 |
+
"rewards/rejected": -0.17484021186828613,
|
| 334 |
+
"step": 160
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"epoch": 0.52,
|
| 338 |
+
"grad_norm": 2.650519079076739,
|
| 339 |
+
"learning_rate": 2.7912083484274266e-06,
|
| 340 |
+
"logits/chosen": -2.747620105743408,
|
| 341 |
+
"logits/rejected": -2.6757848262786865,
|
| 342 |
+
"logps/chosen": -310.66448974609375,
|
| 343 |
+
"logps/rejected": -301.0819091796875,
|
| 344 |
+
"loss": 0.6027,
|
| 345 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 346 |
+
"rewards/chosen": -0.06945960223674774,
|
| 347 |
+
"rewards/margins": 0.2121526002883911,
|
| 348 |
+
"rewards/margins_max": 0.5673162937164307,
|
| 349 |
+
"rewards/margins_min": -0.14554527401924133,
|
| 350 |
+
"rewards/margins_std": 0.3131377100944519,
|
| 351 |
+
"rewards/rejected": -0.28161221742630005,
|
| 352 |
+
"step": 170
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.55,
|
| 356 |
+
"grad_norm": 2.9314631663166417,
|
| 357 |
+
"learning_rate": 2.526533223585641e-06,
|
| 358 |
+
"logits/chosen": -2.7748751640319824,
|
| 359 |
+
"logits/rejected": -2.718724489212036,
|
| 360 |
+
"logps/chosen": -268.07562255859375,
|
| 361 |
+
"logps/rejected": -257.9107971191406,
|
| 362 |
+
"loss": 0.6109,
|
| 363 |
+
"rewards/accuracies": 0.737500011920929,
|
| 364 |
+
"rewards/chosen": -0.11941893398761749,
|
| 365 |
+
"rewards/margins": 0.17082905769348145,
|
| 366 |
+
"rewards/margins_max": 0.5415267944335938,
|
| 367 |
+
"rewards/margins_min": -0.1771378219127655,
|
| 368 |
+
"rewards/margins_std": 0.31447431445121765,
|
| 369 |
+
"rewards/rejected": -0.29024797677993774,
|
| 370 |
+
"step": 180
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 0.58,
|
| 374 |
+
"grad_norm": 3.113744990751754,
|
| 375 |
+
"learning_rate": 2.2615594926807554e-06,
|
| 376 |
+
"logits/chosen": -2.7187552452087402,
|
| 377 |
+
"logits/rejected": -2.709470748901367,
|
| 378 |
+
"logps/chosen": -306.37664794921875,
|
| 379 |
+
"logps/rejected": -302.543701171875,
|
| 380 |
+
"loss": 0.598,
|
| 381 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 382 |
+
"rewards/chosen": -0.05795687437057495,
|
| 383 |
+
"rewards/margins": 0.2614460289478302,
|
| 384 |
+
"rewards/margins_max": 0.7187456488609314,
|
| 385 |
+
"rewards/margins_min": -0.1500331312417984,
|
| 386 |
+
"rewards/margins_std": 0.38883161544799805,
|
| 387 |
+
"rewards/rejected": -0.31940287351608276,
|
| 388 |
+
"step": 190
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.61,
|
| 392 |
+
"grad_norm": 4.46704837439659,
|
| 393 |
+
"learning_rate": 1.9992691817133025e-06,
|
| 394 |
+
"logits/chosen": -2.725795269012451,
|
| 395 |
+
"logits/rejected": -2.700993776321411,
|
| 396 |
+
"logps/chosen": -298.82855224609375,
|
| 397 |
+
"logps/rejected": -336.44140625,
|
| 398 |
+
"loss": 0.5795,
|
| 399 |
+
"rewards/accuracies": 0.699999988079071,
|
| 400 |
+
"rewards/chosen": -0.17227962613105774,
|
| 401 |
+
"rewards/margins": 0.30934515595436096,
|
| 402 |
+
"rewards/margins_max": 0.8330885171890259,
|
| 403 |
+
"rewards/margins_min": -0.24815456569194794,
|
| 404 |
+
"rewards/margins_std": 0.4841969907283783,
|
| 405 |
+
"rewards/rejected": -0.4816247820854187,
|
| 406 |
+
"step": 200
|
| 407 |
+
},
|
| 408 |
+
{
|
| 409 |
+
"epoch": 0.61,
|
| 410 |
+
"eval_logits/chosen": -2.770137310028076,
|
| 411 |
+
"eval_logits/rejected": -2.738154649734497,
|
| 412 |
+
"eval_logps/chosen": -305.9859313964844,
|
| 413 |
+
"eval_logps/rejected": -303.88079833984375,
|
| 414 |
+
"eval_loss": 0.6177265644073486,
|
| 415 |
+
"eval_rewards/accuracies": 0.6819999814033508,
|
| 416 |
+
"eval_rewards/chosen": -0.21392498910427094,
|
| 417 |
+
"eval_rewards/margins": 0.23909391462802887,
|
| 418 |
+
"eval_rewards/margins_max": 0.9738174080848694,
|
| 419 |
+
"eval_rewards/margins_min": -0.4424648880958557,
|
| 420 |
+
"eval_rewards/margins_std": 0.47772130370140076,
|
| 421 |
+
"eval_rewards/rejected": -0.4530189037322998,
|
| 422 |
+
"eval_runtime": 428.0295,
|
| 423 |
+
"eval_samples_per_second": 4.673,
|
| 424 |
+
"eval_steps_per_second": 0.292,
|
| 425 |
+
"step": 200
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 0.64,
|
| 429 |
+
"grad_norm": 3.3590225389683472,
|
| 430 |
+
"learning_rate": 1.742614117358029e-06,
|
| 431 |
+
"logits/chosen": -2.7592720985412598,
|
| 432 |
+
"logits/rejected": -2.724670886993408,
|
| 433 |
+
"logps/chosen": -328.0078430175781,
|
| 434 |
+
"logps/rejected": -345.1807556152344,
|
| 435 |
+
"loss": 0.5664,
|
| 436 |
+
"rewards/accuracies": 0.762499988079071,
|
| 437 |
+
"rewards/chosen": -0.22690674662590027,
|
| 438 |
+
"rewards/margins": 0.34196537733078003,
|
| 439 |
+
"rewards/margins_max": 0.8763746023178101,
|
| 440 |
+
"rewards/margins_min": -0.14572417736053467,
|
| 441 |
+
"rewards/margins_std": 0.465751975774765,
|
| 442 |
+
"rewards/rejected": -0.5688720941543579,
|
| 443 |
+
"step": 210
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.67,
|
| 447 |
+
"grad_norm": 5.608754753313119,
|
| 448 |
+
"learning_rate": 1.4944827069769125e-06,
|
| 449 |
+
"logits/chosen": -2.813154458999634,
|
| 450 |
+
"logits/rejected": -2.796706199645996,
|
| 451 |
+
"logps/chosen": -338.57867431640625,
|
| 452 |
+
"logps/rejected": -311.73773193359375,
|
| 453 |
+
"loss": 0.6278,
|
| 454 |
+
"rewards/accuracies": 0.6875,
|
| 455 |
+
"rewards/chosen": -0.25605496764183044,
|
| 456 |
+
"rewards/margins": 0.19726695120334625,
|
| 457 |
+
"rewards/margins_max": 0.692109227180481,
|
| 458 |
+
"rewards/margins_min": -0.2970424294471741,
|
| 459 |
+
"rewards/margins_std": 0.4508973956108093,
|
| 460 |
+
"rewards/rejected": -0.4533218741416931,
|
| 461 |
+
"step": 220
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.7,
|
| 465 |
+
"grad_norm": 5.028562977399784,
|
| 466 |
+
"learning_rate": 1.257667432355893e-06,
|
| 467 |
+
"logits/chosen": -2.784775972366333,
|
| 468 |
+
"logits/rejected": -2.8116180896759033,
|
| 469 |
+
"logps/chosen": -320.9100646972656,
|
| 470 |
+
"logps/rejected": -302.4966125488281,
|
| 471 |
+
"loss": 0.6612,
|
| 472 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 473 |
+
"rewards/chosen": -0.3226456046104431,
|
| 474 |
+
"rewards/margins": 0.07472896575927734,
|
| 475 |
+
"rewards/margins_max": 0.6551424264907837,
|
| 476 |
+
"rewards/margins_min": -0.5510781407356262,
|
| 477 |
+
"rewards/margins_std": 0.5290972590446472,
|
| 478 |
+
"rewards/rejected": -0.39737454056739807,
|
| 479 |
+
"step": 230
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.73,
|
| 483 |
+
"grad_norm": 8.214920370095605,
|
| 484 |
+
"learning_rate": 1.0348334229922677e-06,
|
| 485 |
+
"logits/chosen": -2.8330750465393066,
|
| 486 |
+
"logits/rejected": -2.793381452560425,
|
| 487 |
+
"logps/chosen": -310.33575439453125,
|
| 488 |
+
"logps/rejected": -318.92852783203125,
|
| 489 |
+
"loss": 0.623,
|
| 490 |
+
"rewards/accuracies": 0.675000011920929,
|
| 491 |
+
"rewards/chosen": -0.19318023324012756,
|
| 492 |
+
"rewards/margins": 0.24430537223815918,
|
| 493 |
+
"rewards/margins_max": 0.8275845646858215,
|
| 494 |
+
"rewards/margins_min": -0.34578949213027954,
|
| 495 |
+
"rewards/margins_std": 0.5300652384757996,
|
| 496 |
+
"rewards/rejected": -0.43748560547828674,
|
| 497 |
+
"step": 240
|
| 498 |
+
},
|
| 499 |
+
{
|
| 500 |
+
"epoch": 0.76,
|
| 501 |
+
"grad_norm": 3.4120571086616582,
|
| 502 |
+
"learning_rate": 8.284884626103165e-07,
|
| 503 |
+
"logits/chosen": -2.783628463745117,
|
| 504 |
+
"logits/rejected": -2.7553200721740723,
|
| 505 |
+
"logps/chosen": -310.9916687011719,
|
| 506 |
+
"logps/rejected": -346.27276611328125,
|
| 507 |
+
"loss": 0.6019,
|
| 508 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 509 |
+
"rewards/chosen": -0.09901777654886246,
|
| 510 |
+
"rewards/margins": 0.3181927502155304,
|
| 511 |
+
"rewards/margins_max": 0.9294508099555969,
|
| 512 |
+
"rewards/margins_min": -0.20651724934577942,
|
| 513 |
+
"rewards/margins_std": 0.5166513919830322,
|
| 514 |
+
"rewards/rejected": -0.41721048951148987,
|
| 515 |
+
"step": 250
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 0.79,
|
| 519 |
+
"grad_norm": 4.910453405014883,
|
| 520 |
+
"learning_rate": 6.409547664531734e-07,
|
| 521 |
+
"logits/chosen": -2.799520254135132,
|
| 522 |
+
"logits/rejected": -2.77909255027771,
|
| 523 |
+
"logps/chosen": -341.7703552246094,
|
| 524 |
+
"logps/rejected": -347.35455322265625,
|
| 525 |
+
"loss": 0.6077,
|
| 526 |
+
"rewards/accuracies": 0.762499988079071,
|
| 527 |
+
"rewards/chosen": -0.07788446545600891,
|
| 528 |
+
"rewards/margins": 0.2706037759780884,
|
| 529 |
+
"rewards/margins_max": 0.7577456831932068,
|
| 530 |
+
"rewards/margins_min": -0.09968160092830658,
|
| 531 |
+
"rewards/margins_std": 0.3826812505722046,
|
| 532 |
+
"rewards/rejected": -0.3484882414340973,
|
| 533 |
+
"step": 260
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.82,
|
| 537 |
+
"grad_norm": 5.466170918075581,
|
| 538 |
+
"learning_rate": 4.743428469705336e-07,
|
| 539 |
+
"logits/chosen": -2.751262903213501,
|
| 540 |
+
"logits/rejected": -2.7571630477905273,
|
| 541 |
+
"logps/chosen": -323.39410400390625,
|
| 542 |
+
"logps/rejected": -356.8424987792969,
|
| 543 |
+
"loss": 0.6027,
|
| 544 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 545 |
+
"rewards/chosen": -0.19586281478405,
|
| 546 |
+
"rewards/margins": 0.29274672269821167,
|
| 547 |
+
"rewards/margins_max": 0.7976349592208862,
|
| 548 |
+
"rewards/margins_min": -0.1788657009601593,
|
| 549 |
+
"rewards/margins_std": 0.4405595362186432,
|
| 550 |
+
"rewards/rejected": -0.4886094927787781,
|
| 551 |
+
"step": 270
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"epoch": 0.85,
|
| 555 |
+
"grad_norm": 3.1691510008843458,
|
| 556 |
+
"learning_rate": 3.3052776201888266e-07,
|
| 557 |
+
"logits/chosen": -2.800191879272461,
|
| 558 |
+
"logits/rejected": -2.795988082885742,
|
| 559 |
+
"logps/chosen": -346.5174255371094,
|
| 560 |
+
"logps/rejected": -325.9556579589844,
|
| 561 |
+
"loss": 0.5704,
|
| 562 |
+
"rewards/accuracies": 0.762499988079071,
|
| 563 |
+
"rewards/chosen": -0.2095174342393875,
|
| 564 |
+
"rewards/margins": 0.35230007767677307,
|
| 565 |
+
"rewards/margins_max": 0.9150822758674622,
|
| 566 |
+
"rewards/margins_min": -0.2514013946056366,
|
| 567 |
+
"rewards/margins_std": 0.5179048776626587,
|
| 568 |
+
"rewards/rejected": -0.561817467212677,
|
| 569 |
+
"step": 280
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.88,
|
| 573 |
+
"grad_norm": 5.145673782819818,
|
| 574 |
+
"learning_rate": 2.111280128780638e-07,
|
| 575 |
+
"logits/chosen": -2.7347121238708496,
|
| 576 |
+
"logits/rejected": -2.703662633895874,
|
| 577 |
+
"logps/chosen": -293.16033935546875,
|
| 578 |
+
"logps/rejected": -290.85107421875,
|
| 579 |
+
"loss": 0.6088,
|
| 580 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 581 |
+
"rewards/chosen": -0.1897042840719223,
|
| 582 |
+
"rewards/margins": 0.26360541582107544,
|
| 583 |
+
"rewards/margins_max": 0.8136995434761047,
|
| 584 |
+
"rewards/margins_min": -0.21997782588005066,
|
| 585 |
+
"rewards/margins_std": 0.4667910635471344,
|
| 586 |
+
"rewards/rejected": -0.45330971479415894,
|
| 587 |
+
"step": 290
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"epoch": 0.91,
|
| 591 |
+
"grad_norm": 6.333030325350309,
|
| 592 |
+
"learning_rate": 1.1748732956682023e-07,
|
| 593 |
+
"logits/chosen": -2.8414254188537598,
|
| 594 |
+
"logits/rejected": -2.791003704071045,
|
| 595 |
+
"logps/chosen": -348.19757080078125,
|
| 596 |
+
"logps/rejected": -340.9607849121094,
|
| 597 |
+
"loss": 0.5791,
|
| 598 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 599 |
+
"rewards/chosen": -0.24484387040138245,
|
| 600 |
+
"rewards/margins": 0.30887269973754883,
|
| 601 |
+
"rewards/margins_max": 0.9062288403511047,
|
| 602 |
+
"rewards/margins_min": -0.2699830234050751,
|
| 603 |
+
"rewards/margins_std": 0.5193021893501282,
|
| 604 |
+
"rewards/rejected": -0.5537165999412537,
|
| 605 |
+
"step": 300
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 0.91,
|
| 609 |
+
"eval_logits/chosen": -2.7722113132476807,
|
| 610 |
+
"eval_logits/rejected": -2.740241289138794,
|
| 611 |
+
"eval_logps/chosen": -304.6903381347656,
|
| 612 |
+
"eval_logps/rejected": -305.08380126953125,
|
| 613 |
+
"eval_loss": 0.6102678775787354,
|
| 614 |
+
"eval_rewards/accuracies": 0.6859999895095825,
|
| 615 |
+
"eval_rewards/chosen": -0.20096930861473083,
|
| 616 |
+
"eval_rewards/margins": 0.26407960057258606,
|
| 617 |
+
"eval_rewards/margins_max": 1.0352518558502197,
|
| 618 |
+
"eval_rewards/margins_min": -0.4649355411529541,
|
| 619 |
+
"eval_rewards/margins_std": 0.5062240362167358,
|
| 620 |
+
"eval_rewards/rejected": -0.4650489091873169,
|
| 621 |
+
"eval_runtime": 428.127,
|
| 622 |
+
"eval_samples_per_second": 4.672,
|
| 623 |
+
"eval_steps_per_second": 0.292,
|
| 624 |
+
"step": 300
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 0.94,
|
| 628 |
+
"grad_norm": 6.7576880294714785,
|
| 629 |
+
"learning_rate": 5.065954844616722e-08,
|
| 630 |
+
"logits/chosen": -2.7763688564300537,
|
| 631 |
+
"logits/rejected": -2.7473700046539307,
|
| 632 |
+
"logps/chosen": -290.21368408203125,
|
| 633 |
+
"logps/rejected": -329.4396667480469,
|
| 634 |
+
"loss": 0.5829,
|
| 635 |
+
"rewards/accuracies": 0.75,
|
| 636 |
+
"rewards/chosen": -0.13091912865638733,
|
| 637 |
+
"rewards/margins": 0.3528442680835724,
|
| 638 |
+
"rewards/margins_max": 0.8581522703170776,
|
| 639 |
+
"rewards/margins_min": -0.14479844272136688,
|
| 640 |
+
"rewards/margins_std": 0.44291505217552185,
|
| 641 |
+
"rewards/rejected": -0.4837633967399597,
|
| 642 |
+
"step": 310
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.97,
|
| 646 |
+
"grad_norm": 5.39765909106254,
|
| 647 |
+
"learning_rate": 1.1396752298723501e-08,
|
| 648 |
+
"logits/chosen": -2.820077896118164,
|
| 649 |
+
"logits/rejected": -2.7777419090270996,
|
| 650 |
+
"logps/chosen": -276.8497314453125,
|
| 651 |
+
"logps/rejected": -314.1188049316406,
|
| 652 |
+
"loss": 0.5654,
|
| 653 |
+
"rewards/accuracies": 0.737500011920929,
|
| 654 |
+
"rewards/chosen": -0.2790452241897583,
|
| 655 |
+
"rewards/margins": 0.28673863410949707,
|
| 656 |
+
"rewards/margins_max": 0.7317686080932617,
|
| 657 |
+
"rewards/margins_min": -0.12932872772216797,
|
| 658 |
+
"rewards/margins_std": 0.3887925446033478,
|
| 659 |
+
"rewards/rejected": -0.5657838582992554,
|
| 660 |
+
"step": 320
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 1.0,
|
| 664 |
+
"step": 329,
|
| 665 |
+
"total_flos": 0.0,
|
| 666 |
+
"train_loss": 0.6308559574254741,
|
| 667 |
+
"train_runtime": 3887.3613,
|
| 668 |
+
"train_samples_per_second": 1.354,
|
| 669 |
+
"train_steps_per_second": 0.085
|
| 670 |
+
}
|
| 671 |
+
],
|
| 672 |
+
"logging_steps": 10,
|
| 673 |
+
"max_steps": 329,
|
| 674 |
+
"num_input_tokens_seen": 0,
|
| 675 |
+
"num_train_epochs": 1,
|
| 676 |
+
"save_steps": 100,
|
| 677 |
+
"total_flos": 0.0,
|
| 678 |
+
"train_batch_size": 4,
|
| 679 |
+
"trial_name": null,
|
| 680 |
+
"trial_params": null
|
| 681 |
+
}
|