Model save
Browse files- README.md +78 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +603 -0
README.md
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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-magpi-low-margin-3-epochs
|
| 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-magpi-low-margin-3-epochs
|
| 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: 0.0010
|
| 21 |
+
- Rewards/chosen: -3.9690
|
| 22 |
+
- Rewards/rejected: -58.1646
|
| 23 |
+
- Rewards/accuracies: 0.9980
|
| 24 |
+
- Rewards/margins: 54.1956
|
| 25 |
+
- Logps/rejected: -6457.2515
|
| 26 |
+
- Logps/chosen: -763.8833
|
| 27 |
+
- Logits/rejected: 4.2613
|
| 28 |
+
- Logits/chosen: 1.2520
|
| 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: 55
|
| 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: 3
|
| 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 |
+
| 0.009 | 0.4739 | 50 | 0.0026 | -0.6343 | -34.9234 | 0.9960 | 34.2890 | -4133.1240 | -430.4127 | -2.6406 | -2.9323 |
|
| 66 |
+
| 0.0158 | 0.9479 | 100 | 0.0003 | -1.2700 | -43.4290 | 1.0 | 42.1589 | -4983.6846 | -493.9866 | 3.0564 | -0.5346 |
|
| 67 |
+
| 0.0007 | 1.4218 | 150 | 0.0005 | -2.9913 | -52.6639 | 0.9980 | 49.6726 | -5907.1768 | -666.1114 | 4.1711 | 0.7924 |
|
| 68 |
+
| 0.0015 | 1.8957 | 200 | 0.0006 | -3.5933 | -54.2858 | 0.9980 | 50.6925 | -6069.3657 | -726.3069 | 4.0713 | 1.0487 |
|
| 69 |
+
| 0.0 | 2.3697 | 250 | 0.0009 | -3.9372 | -57.6402 | 0.9980 | 53.7030 | -6404.8037 | -760.6977 | 4.2488 | 1.2599 |
|
| 70 |
+
| 0.0001 | 2.8436 | 300 | 0.0010 | -3.9690 | -58.1646 | 0.9980 | 54.1956 | -6457.2515 | -763.8833 | 4.2613 | 1.2520 |
|
| 71 |
+
|
| 72 |
+
|
| 73 |
+
### Framework versions
|
| 74 |
+
|
| 75 |
+
- Transformers 4.44.0.dev0
|
| 76 |
+
- Pytorch 2.1.2
|
| 77 |
+
- Datasets 2.20.0
|
| 78 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.985781990521327,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.04501082229597535,
|
| 5 |
+
"train_runtime": 9068.4154,
|
| 6 |
+
"train_samples": 13500,
|
| 7 |
+
"train_samples_per_second": 4.466,
|
| 8 |
+
"train_steps_per_second": 0.035
|
| 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.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.985781990521327,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.04501082229597535,
|
| 5 |
+
"train_runtime": 9068.4154,
|
| 6 |
+
"train_samples": 13500,
|
| 7 |
+
"train_samples_per_second": 4.466,
|
| 8 |
+
"train_steps_per_second": 0.035
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,603 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.985781990521327,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 315,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0947867298578199,
|
| 13 |
+
"grad_norm": 45.73923520041355,
|
| 14 |
+
"learning_rate": 1.5624999999999999e-07,
|
| 15 |
+
"logits/chosen": -2.8428008556365967,
|
| 16 |
+
"logits/rejected": -2.6068577766418457,
|
| 17 |
+
"logps/chosen": -368.2735290527344,
|
| 18 |
+
"logps/rejected": -498.0081481933594,
|
| 19 |
+
"loss": 0.6856,
|
| 20 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 21 |
+
"rewards/chosen": 0.0027675286401063204,
|
| 22 |
+
"rewards/margins": 0.014056036248803139,
|
| 23 |
+
"rewards/rejected": -0.011288506910204887,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.1895734597156398,
|
| 28 |
+
"grad_norm": 16.393339414177866,
|
| 29 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 30 |
+
"logits/chosen": -2.853433132171631,
|
| 31 |
+
"logits/rejected": -2.6817729473114014,
|
| 32 |
+
"logps/chosen": -370.88690185546875,
|
| 33 |
+
"logps/rejected": -610.00048828125,
|
| 34 |
+
"loss": 0.473,
|
| 35 |
+
"rewards/accuracies": 1.0,
|
| 36 |
+
"rewards/chosen": 0.09298940747976303,
|
| 37 |
+
"rewards/margins": 0.63857501745224,
|
| 38 |
+
"rewards/rejected": -0.5455856323242188,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.2843601895734597,
|
| 43 |
+
"grad_norm": 4.343562060345208,
|
| 44 |
+
"learning_rate": 4.6874999999999996e-07,
|
| 45 |
+
"logits/chosen": -2.9111626148223877,
|
| 46 |
+
"logits/rejected": -2.6944823265075684,
|
| 47 |
+
"logps/chosen": -330.89703369140625,
|
| 48 |
+
"logps/rejected": -913.5744018554688,
|
| 49 |
+
"loss": 0.1487,
|
| 50 |
+
"rewards/accuracies": 1.0,
|
| 51 |
+
"rewards/chosen": 0.38449177145957947,
|
| 52 |
+
"rewards/margins": 4.493498802185059,
|
| 53 |
+
"rewards/rejected": -4.109007358551025,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.3791469194312796,
|
| 58 |
+
"grad_norm": 1.5210747026486233,
|
| 59 |
+
"learning_rate": 4.990147841143461e-07,
|
| 60 |
+
"logits/chosen": -2.918667793273926,
|
| 61 |
+
"logits/rejected": -2.744393825531006,
|
| 62 |
+
"logps/chosen": -352.75640869140625,
|
| 63 |
+
"logps/rejected": -1823.7945556640625,
|
| 64 |
+
"loss": 0.0266,
|
| 65 |
+
"rewards/accuracies": 1.0,
|
| 66 |
+
"rewards/chosen": 0.18624749779701233,
|
| 67 |
+
"rewards/margins": 13.044224739074707,
|
| 68 |
+
"rewards/rejected": -12.857976913452148,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.47393364928909953,
|
| 73 |
+
"grad_norm": 3.5117794451703537,
|
| 74 |
+
"learning_rate": 4.950256493879794e-07,
|
| 75 |
+
"logits/chosen": -2.8872039318084717,
|
| 76 |
+
"logits/rejected": -2.6208763122558594,
|
| 77 |
+
"logps/chosen": -432.4601135253906,
|
| 78 |
+
"logps/rejected": -3069.329345703125,
|
| 79 |
+
"loss": 0.009,
|
| 80 |
+
"rewards/accuracies": 1.0,
|
| 81 |
+
"rewards/chosen": -0.5042247176170349,
|
| 82 |
+
"rewards/margins": 25.241214752197266,
|
| 83 |
+
"rewards/rejected": -25.745441436767578,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.47393364928909953,
|
| 88 |
+
"eval_logits/chosen": -2.932264804840088,
|
| 89 |
+
"eval_logits/rejected": -2.640563726425171,
|
| 90 |
+
"eval_logps/chosen": -430.4127197265625,
|
| 91 |
+
"eval_logps/rejected": -4133.1240234375,
|
| 92 |
+
"eval_loss": 0.002647999208420515,
|
| 93 |
+
"eval_rewards/accuracies": 0.9959677457809448,
|
| 94 |
+
"eval_rewards/chosen": -0.6343097686767578,
|
| 95 |
+
"eval_rewards/margins": 34.28904724121094,
|
| 96 |
+
"eval_rewards/rejected": -34.92335510253906,
|
| 97 |
+
"eval_runtime": 193.1739,
|
| 98 |
+
"eval_samples_per_second": 20.215,
|
| 99 |
+
"eval_steps_per_second": 0.321,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.5687203791469194,
|
| 104 |
+
"grad_norm": 2.067910467176069,
|
| 105 |
+
"learning_rate": 4.88020090697132e-07,
|
| 106 |
+
"logits/chosen": -2.941565990447998,
|
| 107 |
+
"logits/rejected": -2.5026259422302246,
|
| 108 |
+
"logps/chosen": -496.873046875,
|
| 109 |
+
"logps/rejected": -3218.004638671875,
|
| 110 |
+
"loss": 0.0218,
|
| 111 |
+
"rewards/accuracies": 0.987500011920929,
|
| 112 |
+
"rewards/chosen": -1.0619714260101318,
|
| 113 |
+
"rewards/margins": 25.82952880859375,
|
| 114 |
+
"rewards/rejected": -26.89150047302246,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.6635071090047393,
|
| 119 |
+
"grad_norm": 27.786761703670376,
|
| 120 |
+
"learning_rate": 4.780843509929904e-07,
|
| 121 |
+
"logits/chosen": -1.2228466272354126,
|
| 122 |
+
"logits/rejected": 0.15184922516345978,
|
| 123 |
+
"logps/chosen": -490.0467834472656,
|
| 124 |
+
"logps/rejected": -3536.85302734375,
|
| 125 |
+
"loss": 0.0127,
|
| 126 |
+
"rewards/accuracies": 0.9937499761581421,
|
| 127 |
+
"rewards/chosen": -1.3228168487548828,
|
| 128 |
+
"rewards/margins": 29.06142807006836,
|
| 129 |
+
"rewards/rejected": -30.384246826171875,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.7582938388625592,
|
| 134 |
+
"grad_norm": 1.926082688082676,
|
| 135 |
+
"learning_rate": 4.6534074564712217e-07,
|
| 136 |
+
"logits/chosen": -0.27225083112716675,
|
| 137 |
+
"logits/rejected": 1.7922433614730835,
|
| 138 |
+
"logps/chosen": -459.4078063964844,
|
| 139 |
+
"logps/rejected": -3558.192626953125,
|
| 140 |
+
"loss": 0.0027,
|
| 141 |
+
"rewards/accuracies": 1.0,
|
| 142 |
+
"rewards/chosen": -1.0403556823730469,
|
| 143 |
+
"rewards/margins": 29.467926025390625,
|
| 144 |
+
"rewards/rejected": -30.508281707763672,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.8530805687203792,
|
| 149 |
+
"grad_norm": 1.213238062829923,
|
| 150 |
+
"learning_rate": 4.4994615667026846e-07,
|
| 151 |
+
"logits/chosen": -0.5399178266525269,
|
| 152 |
+
"logits/rejected": 2.844102382659912,
|
| 153 |
+
"logps/chosen": -458.8783264160156,
|
| 154 |
+
"logps/rejected": -3624.69921875,
|
| 155 |
+
"loss": 0.0043,
|
| 156 |
+
"rewards/accuracies": 1.0,
|
| 157 |
+
"rewards/chosen": -0.9510987997055054,
|
| 158 |
+
"rewards/margins": 30.12051773071289,
|
| 159 |
+
"rewards/rejected": -31.071613311767578,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.9478672985781991,
|
| 164 |
+
"grad_norm": 2.2220593339773402,
|
| 165 |
+
"learning_rate": 4.320901013934887e-07,
|
| 166 |
+
"logits/chosen": 0.34000855684280396,
|
| 167 |
+
"logits/rejected": 3.4551188945770264,
|
| 168 |
+
"logps/chosen": -497.6631774902344,
|
| 169 |
+
"logps/rejected": -4114.6044921875,
|
| 170 |
+
"loss": 0.0158,
|
| 171 |
+
"rewards/accuracies": 1.0,
|
| 172 |
+
"rewards/chosen": -1.247301459312439,
|
| 173 |
+
"rewards/margins": 34.767173767089844,
|
| 174 |
+
"rewards/rejected": -36.01448059082031,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.9478672985781991,
|
| 179 |
+
"eval_logits/chosen": -0.5345638394355774,
|
| 180 |
+
"eval_logits/rejected": 3.056351661682129,
|
| 181 |
+
"eval_logps/chosen": -493.986572265625,
|
| 182 |
+
"eval_logps/rejected": -4983.6845703125,
|
| 183 |
+
"eval_loss": 0.0003322066040709615,
|
| 184 |
+
"eval_rewards/accuracies": 1.0,
|
| 185 |
+
"eval_rewards/chosen": -1.2700488567352295,
|
| 186 |
+
"eval_rewards/margins": 42.158912658691406,
|
| 187 |
+
"eval_rewards/rejected": -43.42896270751953,
|
| 188 |
+
"eval_runtime": 192.4704,
|
| 189 |
+
"eval_samples_per_second": 20.289,
|
| 190 |
+
"eval_steps_per_second": 0.322,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 1.042654028436019,
|
| 195 |
+
"grad_norm": 0.22005560080220604,
|
| 196 |
+
"learning_rate": 4.119923993874379e-07,
|
| 197 |
+
"logits/chosen": -0.7820493578910828,
|
| 198 |
+
"logits/rejected": 3.120879650115967,
|
| 199 |
+
"logps/chosen": -517.9307861328125,
|
| 200 |
+
"logps/rejected": -3928.251953125,
|
| 201 |
+
"loss": 0.006,
|
| 202 |
+
"rewards/accuracies": 1.0,
|
| 203 |
+
"rewards/chosen": -1.4943805932998657,
|
| 204 |
+
"rewards/margins": 32.6540641784668,
|
| 205 |
+
"rewards/rejected": -34.14844512939453,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 1.1374407582938388,
|
| 210 |
+
"grad_norm": 0.12162920408221319,
|
| 211 |
+
"learning_rate": 3.899004663415083e-07,
|
| 212 |
+
"logits/chosen": -0.34208816289901733,
|
| 213 |
+
"logits/rejected": 3.0167593955993652,
|
| 214 |
+
"logps/chosen": -538.0582275390625,
|
| 215 |
+
"logps/rejected": -4409.0302734375,
|
| 216 |
+
"loss": 0.0011,
|
| 217 |
+
"rewards/accuracies": 1.0,
|
| 218 |
+
"rewards/chosen": -1.8046563863754272,
|
| 219 |
+
"rewards/margins": 36.82363510131836,
|
| 220 |
+
"rewards/rejected": -38.62828826904297,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.2322274881516588,
|
| 225 |
+
"grad_norm": 0.09818677325286322,
|
| 226 |
+
"learning_rate": 3.6608626821692824e-07,
|
| 227 |
+
"logits/chosen": 0.6416308879852295,
|
| 228 |
+
"logits/rejected": 3.524348735809326,
|
| 229 |
+
"logps/chosen": -605.2796630859375,
|
| 230 |
+
"logps/rejected": -4457.169921875,
|
| 231 |
+
"loss": 0.0005,
|
| 232 |
+
"rewards/accuracies": 1.0,
|
| 233 |
+
"rewards/chosen": -2.267808437347412,
|
| 234 |
+
"rewards/margins": 37.17861557006836,
|
| 235 |
+
"rewards/rejected": -39.4464225769043,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 1.3270142180094786,
|
| 240 |
+
"grad_norm": 0.4137355125115123,
|
| 241 |
+
"learning_rate": 3.408429731701635e-07,
|
| 242 |
+
"logits/chosen": 1.477095603942871,
|
| 243 |
+
"logits/rejected": 4.739525318145752,
|
| 244 |
+
"logps/chosen": -659.0659790039062,
|
| 245 |
+
"logps/rejected": -4712.701171875,
|
| 246 |
+
"loss": 0.0002,
|
| 247 |
+
"rewards/accuracies": 1.0,
|
| 248 |
+
"rewards/chosen": -2.851088762283325,
|
| 249 |
+
"rewards/margins": 39.070613861083984,
|
| 250 |
+
"rewards/rejected": -41.92170333862305,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.4218009478672986,
|
| 255 |
+
"grad_norm": 0.038412138392737674,
|
| 256 |
+
"learning_rate": 3.144813424636031e-07,
|
| 257 |
+
"logits/chosen": 1.158630132675171,
|
| 258 |
+
"logits/rejected": 4.032771110534668,
|
| 259 |
+
"logps/chosen": -702.2916259765625,
|
| 260 |
+
"logps/rejected": -4613.56201171875,
|
| 261 |
+
"loss": 0.0007,
|
| 262 |
+
"rewards/accuracies": 1.0,
|
| 263 |
+
"rewards/chosen": -3.405531406402588,
|
| 264 |
+
"rewards/margins": 37.08515930175781,
|
| 265 |
+
"rewards/rejected": -40.490692138671875,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.4218009478672986,
|
| 270 |
+
"eval_logits/chosen": 0.7923636436462402,
|
| 271 |
+
"eval_logits/rejected": 4.171131134033203,
|
| 272 |
+
"eval_logps/chosen": -666.1113891601562,
|
| 273 |
+
"eval_logps/rejected": -5907.1767578125,
|
| 274 |
+
"eval_loss": 0.0005207537906244397,
|
| 275 |
+
"eval_rewards/accuracies": 0.9979838728904724,
|
| 276 |
+
"eval_rewards/chosen": -2.9912962913513184,
|
| 277 |
+
"eval_rewards/margins": 49.67258834838867,
|
| 278 |
+
"eval_rewards/rejected": -52.66389083862305,
|
| 279 |
+
"eval_runtime": 192.4622,
|
| 280 |
+
"eval_samples_per_second": 20.29,
|
| 281 |
+
"eval_steps_per_second": 0.322,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 1.5165876777251186,
|
| 286 |
+
"grad_norm": 0.010481668812329812,
|
| 287 |
+
"learning_rate": 2.8732590479375165e-07,
|
| 288 |
+
"logits/chosen": 0.743170440196991,
|
| 289 |
+
"logits/rejected": 3.9760921001434326,
|
| 290 |
+
"logps/chosen": -743.1651611328125,
|
| 291 |
+
"logps/rejected": -4806.0869140625,
|
| 292 |
+
"loss": 0.0027,
|
| 293 |
+
"rewards/accuracies": 1.0,
|
| 294 |
+
"rewards/chosen": -3.295560121536255,
|
| 295 |
+
"rewards/margins": 39.42099380493164,
|
| 296 |
+
"rewards/rejected": -42.716552734375,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 1.6113744075829384,
|
| 301 |
+
"grad_norm": 0.09321288882655707,
|
| 302 |
+
"learning_rate": 2.597109611334169e-07,
|
| 303 |
+
"logits/chosen": 1.2140353918075562,
|
| 304 |
+
"logits/rejected": 4.648254871368408,
|
| 305 |
+
"logps/chosen": -728.6572265625,
|
| 306 |
+
"logps/rejected": -4531.54150390625,
|
| 307 |
+
"loss": 0.0008,
|
| 308 |
+
"rewards/accuracies": 1.0,
|
| 309 |
+
"rewards/chosen": -3.5468170642852783,
|
| 310 |
+
"rewards/margins": 37.09358596801758,
|
| 311 |
+
"rewards/rejected": -40.640403747558594,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 1.7061611374407581,
|
| 316 |
+
"grad_norm": 0.10729010910304806,
|
| 317 |
+
"learning_rate": 2.3197646927086694e-07,
|
| 318 |
+
"logits/chosen": 1.2630093097686768,
|
| 319 |
+
"logits/rejected": 4.423764228820801,
|
| 320 |
+
"logps/chosen": -708.4100341796875,
|
| 321 |
+
"logps/rejected": -4733.248046875,
|
| 322 |
+
"loss": 0.0002,
|
| 323 |
+
"rewards/accuracies": 1.0,
|
| 324 |
+
"rewards/chosen": -3.5483551025390625,
|
| 325 |
+
"rewards/margins": 38.949485778808594,
|
| 326 |
+
"rewards/rejected": -42.497840881347656,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 1.8009478672985781,
|
| 331 |
+
"grad_norm": 0.24468759681680258,
|
| 332 |
+
"learning_rate": 2.0446385870993467e-07,
|
| 333 |
+
"logits/chosen": 1.4046242237091064,
|
| 334 |
+
"logits/rejected": 4.30928897857666,
|
| 335 |
+
"logps/chosen": -701.5442504882812,
|
| 336 |
+
"logps/rejected": -5078.294921875,
|
| 337 |
+
"loss": 0.0023,
|
| 338 |
+
"rewards/accuracies": 1.0,
|
| 339 |
+
"rewards/chosen": -3.5222976207733154,
|
| 340 |
+
"rewards/margins": 41.7343635559082,
|
| 341 |
+
"rewards/rejected": -45.25666427612305,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 1.8957345971563981,
|
| 346 |
+
"grad_norm": 0.3903600557186432,
|
| 347 |
+
"learning_rate": 1.775118274523545e-07,
|
| 348 |
+
"logits/chosen": 1.221166729927063,
|
| 349 |
+
"logits/rejected": 4.167235851287842,
|
| 350 |
+
"logps/chosen": -732.6544189453125,
|
| 351 |
+
"logps/rejected": -4501.24951171875,
|
| 352 |
+
"loss": 0.0015,
|
| 353 |
+
"rewards/accuracies": 0.9937499761581421,
|
| 354 |
+
"rewards/chosen": -3.7073051929473877,
|
| 355 |
+
"rewards/margins": 36.32033157348633,
|
| 356 |
+
"rewards/rejected": -40.02763366699219,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 1.8957345971563981,
|
| 361 |
+
"eval_logits/chosen": 1.0487316846847534,
|
| 362 |
+
"eval_logits/rejected": 4.071253299713135,
|
| 363 |
+
"eval_logps/chosen": -726.3069458007812,
|
| 364 |
+
"eval_logps/rejected": -6069.36572265625,
|
| 365 |
+
"eval_loss": 0.0006494168192148209,
|
| 366 |
+
"eval_rewards/accuracies": 0.9979838728904724,
|
| 367 |
+
"eval_rewards/chosen": -3.593252658843994,
|
| 368 |
+
"eval_rewards/margins": 50.6925163269043,
|
| 369 |
+
"eval_rewards/rejected": -54.285770416259766,
|
| 370 |
+
"eval_runtime": 194.2327,
|
| 371 |
+
"eval_samples_per_second": 20.105,
|
| 372 |
+
"eval_steps_per_second": 0.319,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 1.9905213270142181,
|
| 377 |
+
"grad_norm": 0.0026076475546211236,
|
| 378 |
+
"learning_rate": 1.514521724066537e-07,
|
| 379 |
+
"logits/chosen": 1.5519434213638306,
|
| 380 |
+
"logits/rejected": 4.584723472595215,
|
| 381 |
+
"logps/chosen": -740.3926391601562,
|
| 382 |
+
"logps/rejected": -4949.19970703125,
|
| 383 |
+
"loss": 0.0011,
|
| 384 |
+
"rewards/accuracies": 1.0,
|
| 385 |
+
"rewards/chosen": -3.849459409713745,
|
| 386 |
+
"rewards/margins": 40.864036560058594,
|
| 387 |
+
"rewards/rejected": -44.71349334716797,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 2.085308056872038,
|
| 392 |
+
"grad_norm": 0.06001104822177154,
|
| 393 |
+
"learning_rate": 1.266057047539568e-07,
|
| 394 |
+
"logits/chosen": 1.105302095413208,
|
| 395 |
+
"logits/rejected": 3.880047559738159,
|
| 396 |
+
"logps/chosen": -792.5735473632812,
|
| 397 |
+
"logps/rejected": -5075.94384765625,
|
| 398 |
+
"loss": 0.0002,
|
| 399 |
+
"rewards/accuracies": 1.0,
|
| 400 |
+
"rewards/chosen": -4.032340049743652,
|
| 401 |
+
"rewards/margins": 41.21745681762695,
|
| 402 |
+
"rewards/rejected": -45.24979782104492,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 2.1800947867298577,
|
| 407 |
+
"grad_norm": 0.015291332826346503,
|
| 408 |
+
"learning_rate": 1.032783005551884e-07,
|
| 409 |
+
"logits/chosen": 1.3159544467926025,
|
| 410 |
+
"logits/rejected": 4.874353885650635,
|
| 411 |
+
"logps/chosen": -772.8294677734375,
|
| 412 |
+
"logps/rejected": -4771.970703125,
|
| 413 |
+
"loss": 0.0,
|
| 414 |
+
"rewards/accuracies": 1.0,
|
| 415 |
+
"rewards/chosen": -4.134366035461426,
|
| 416 |
+
"rewards/margins": 38.77143096923828,
|
| 417 |
+
"rewards/rejected": -42.90580368041992,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 2.2748815165876777,
|
| 422 |
+
"grad_norm": 0.019790975881220592,
|
| 423 |
+
"learning_rate": 8.175713521924976e-08,
|
| 424 |
+
"logits/chosen": 1.721247673034668,
|
| 425 |
+
"logits/rejected": 4.219584941864014,
|
| 426 |
+
"logps/chosen": -779.8192138671875,
|
| 427 |
+
"logps/rejected": -5218.1943359375,
|
| 428 |
+
"loss": 0.0,
|
| 429 |
+
"rewards/accuracies": 1.0,
|
| 430 |
+
"rewards/chosen": -4.025120735168457,
|
| 431 |
+
"rewards/margins": 42.607852935791016,
|
| 432 |
+
"rewards/rejected": -46.632972717285156,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 2.3696682464454977,
|
| 437 |
+
"grad_norm": 0.029539030378801425,
|
| 438 |
+
"learning_rate": 6.230714818829733e-08,
|
| 439 |
+
"logits/chosen": 1.274677038192749,
|
| 440 |
+
"logits/rejected": 4.052763938903809,
|
| 441 |
+
"logps/chosen": -783.1781005859375,
|
| 442 |
+
"logps/rejected": -5148.7236328125,
|
| 443 |
+
"loss": 0.0,
|
| 444 |
+
"rewards/accuracies": 1.0,
|
| 445 |
+
"rewards/chosen": -4.11904764175415,
|
| 446 |
+
"rewards/margins": 41.87529754638672,
|
| 447 |
+
"rewards/rejected": -45.99434280395508,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 2.3696682464454977,
|
| 452 |
+
"eval_logits/chosen": 1.259900450706482,
|
| 453 |
+
"eval_logits/rejected": 4.24884557723999,
|
| 454 |
+
"eval_logps/chosen": -760.6976928710938,
|
| 455 |
+
"eval_logps/rejected": -6404.8037109375,
|
| 456 |
+
"eval_loss": 0.0009418130503036082,
|
| 457 |
+
"eval_rewards/accuracies": 0.9979838728904724,
|
| 458 |
+
"eval_rewards/chosen": -3.937159299850464,
|
| 459 |
+
"eval_rewards/margins": 53.70299530029297,
|
| 460 |
+
"eval_rewards/rejected": -57.640159606933594,
|
| 461 |
+
"eval_runtime": 193.6285,
|
| 462 |
+
"eval_samples_per_second": 20.167,
|
| 463 |
+
"eval_steps_per_second": 0.32,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 2.4644549763033177,
|
| 468 |
+
"grad_norm": 0.002101891811001152,
|
| 469 |
+
"learning_rate": 4.516778136213037e-08,
|
| 470 |
+
"logits/chosen": 1.4295393228530884,
|
| 471 |
+
"logits/rejected": 4.259391784667969,
|
| 472 |
+
"logps/chosen": -791.4283447265625,
|
| 473 |
+
"logps/rejected": -5040.88037109375,
|
| 474 |
+
"loss": 0.0,
|
| 475 |
+
"rewards/accuracies": 1.0,
|
| 476 |
+
"rewards/chosen": -4.1635847091674805,
|
| 477 |
+
"rewards/margins": 40.99254608154297,
|
| 478 |
+
"rewards/rejected": -45.15613555908203,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 2.5592417061611377,
|
| 483 |
+
"grad_norm": 0.007793292774055604,
|
| 484 |
+
"learning_rate": 3.055003141378948e-08,
|
| 485 |
+
"logits/chosen": 1.430039644241333,
|
| 486 |
+
"logits/rejected": 4.239804267883301,
|
| 487 |
+
"logps/chosen": -771.2034301757812,
|
| 488 |
+
"logps/rejected": -5207.9248046875,
|
| 489 |
+
"loss": 0.0001,
|
| 490 |
+
"rewards/accuracies": 1.0,
|
| 491 |
+
"rewards/chosen": -4.072695255279541,
|
| 492 |
+
"rewards/margins": 42.79798126220703,
|
| 493 |
+
"rewards/rejected": -46.87067794799805,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 2.654028436018957,
|
| 498 |
+
"grad_norm": 0.005849603176213166,
|
| 499 |
+
"learning_rate": 1.8633852284264508e-08,
|
| 500 |
+
"logits/chosen": 1.6255613565444946,
|
| 501 |
+
"logits/rejected": 4.967244625091553,
|
| 502 |
+
"logps/chosen": -777.3895874023438,
|
| 503 |
+
"logps/rejected": -4971.68994140625,
|
| 504 |
+
"loss": 0.0001,
|
| 505 |
+
"rewards/accuracies": 1.0,
|
| 506 |
+
"rewards/chosen": -3.9707741737365723,
|
| 507 |
+
"rewards/margins": 40.53221893310547,
|
| 508 |
+
"rewards/rejected": -44.502994537353516,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 2.748815165876777,
|
| 513 |
+
"grad_norm": 0.0343313276372006,
|
| 514 |
+
"learning_rate": 9.56593983327919e-09,
|
| 515 |
+
"logits/chosen": 1.3324038982391357,
|
| 516 |
+
"logits/rejected": 4.286895275115967,
|
| 517 |
+
"logps/chosen": -791.8960571289062,
|
| 518 |
+
"logps/rejected": -5403.97119140625,
|
| 519 |
+
"loss": 0.0,
|
| 520 |
+
"rewards/accuracies": 1.0,
|
| 521 |
+
"rewards/chosen": -4.294528007507324,
|
| 522 |
+
"rewards/margins": 44.62022399902344,
|
| 523 |
+
"rewards/rejected": -48.91474914550781,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 2.843601895734597,
|
| 528 |
+
"grad_norm": 0.052240981983852935,
|
| 529 |
+
"learning_rate": 3.4579259185321398e-09,
|
| 530 |
+
"logits/chosen": 1.404039740562439,
|
| 531 |
+
"logits/rejected": 3.9035141468048096,
|
| 532 |
+
"logps/chosen": -799.1031494140625,
|
| 533 |
+
"logps/rejected": -5045.74658203125,
|
| 534 |
+
"loss": 0.0001,
|
| 535 |
+
"rewards/accuracies": 1.0,
|
| 536 |
+
"rewards/chosen": -4.179960250854492,
|
| 537 |
+
"rewards/margins": 40.62794876098633,
|
| 538 |
+
"rewards/rejected": -44.80791473388672,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 2.843601895734597,
|
| 543 |
+
"eval_logits/chosen": 1.2520259618759155,
|
| 544 |
+
"eval_logits/rejected": 4.261307716369629,
|
| 545 |
+
"eval_logps/chosen": -763.88330078125,
|
| 546 |
+
"eval_logps/rejected": -6457.25146484375,
|
| 547 |
+
"eval_loss": 0.0009674043976701796,
|
| 548 |
+
"eval_rewards/accuracies": 0.9979838728904724,
|
| 549 |
+
"eval_rewards/chosen": -3.969015598297119,
|
| 550 |
+
"eval_rewards/margins": 54.195613861083984,
|
| 551 |
+
"eval_rewards/rejected": -58.16463851928711,
|
| 552 |
+
"eval_runtime": 192.9853,
|
| 553 |
+
"eval_samples_per_second": 20.235,
|
| 554 |
+
"eval_steps_per_second": 0.321,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 2.938388625592417,
|
| 559 |
+
"grad_norm": 0.08742324749616634,
|
| 560 |
+
"learning_rate": 3.850041354441502e-10,
|
| 561 |
+
"logits/chosen": 1.2420690059661865,
|
| 562 |
+
"logits/rejected": 3.870642900466919,
|
| 563 |
+
"logps/chosen": -808.8611450195312,
|
| 564 |
+
"logps/rejected": -5520.935546875,
|
| 565 |
+
"loss": 0.0001,
|
| 566 |
+
"rewards/accuracies": 1.0,
|
| 567 |
+
"rewards/chosen": -4.237000465393066,
|
| 568 |
+
"rewards/margins": 45.446434020996094,
|
| 569 |
+
"rewards/rejected": -49.68342971801758,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 2.985781990521327,
|
| 574 |
+
"step": 315,
|
| 575 |
+
"total_flos": 0.0,
|
| 576 |
+
"train_loss": 0.04501082229597535,
|
| 577 |
+
"train_runtime": 9068.4154,
|
| 578 |
+
"train_samples_per_second": 4.466,
|
| 579 |
+
"train_steps_per_second": 0.035
|
| 580 |
+
}
|
| 581 |
+
],
|
| 582 |
+
"logging_steps": 10,
|
| 583 |
+
"max_steps": 315,
|
| 584 |
+
"num_input_tokens_seen": 0,
|
| 585 |
+
"num_train_epochs": 3,
|
| 586 |
+
"save_steps": 100,
|
| 587 |
+
"stateful_callbacks": {
|
| 588 |
+
"TrainerControl": {
|
| 589 |
+
"args": {
|
| 590 |
+
"should_epoch_stop": false,
|
| 591 |
+
"should_evaluate": false,
|
| 592 |
+
"should_log": false,
|
| 593 |
+
"should_save": true,
|
| 594 |
+
"should_training_stop": true
|
| 595 |
+
},
|
| 596 |
+
"attributes": {}
|
| 597 |
+
}
|
| 598 |
+
},
|
| 599 |
+
"total_flos": 0.0,
|
| 600 |
+
"train_batch_size": 8,
|
| 601 |
+
"trial_name": null,
|
| 602 |
+
"trial_params": null
|
| 603 |
+
}
|