Model save
Browse files- README.md +75 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +509 -0
README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-align-scan-0.0-0.3-linear-3
|
| 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-align-scan-0.0-0.3-linear-3
|
| 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 an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.8316
|
| 22 |
+
- Rewards/chosen: -3.2765
|
| 23 |
+
- Rewards/rejected: -4.2890
|
| 24 |
+
- Rewards/accuracies: 0.3413
|
| 25 |
+
- Rewards/margins: 1.0125
|
| 26 |
+
- Logps/rejected: -94.3157
|
| 27 |
+
- Logps/chosen: -84.5654
|
| 28 |
+
- Logits/rejected: -2.4456
|
| 29 |
+
- Logits/chosen: -2.4656
|
| 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: 8.165546278140361e-07
|
| 49 |
+
- train_batch_size: 8
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 4
|
| 54 |
+
- gradient_accumulation_steps: 2
|
| 55 |
+
- total_train_batch_size: 64
|
| 56 |
+
- total_eval_batch_size: 32
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: linear
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 3
|
| 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 |
+
| 0.5554 | 1.0417 | 100 | 0.7309 | 0.3343 | -0.2184 | 0.3393 | 0.5527 | -81.8000 | -73.4634 | -2.4356 | -2.4525 |
|
| 67 |
+
| 0.3557 | 2.0833 | 200 | 0.8316 | -3.2765 | -4.2890 | 0.3413 | 1.0125 | -94.3157 | -84.5654 | -2.4456 | -2.4656 |
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
### Framework versions
|
| 71 |
+
|
| 72 |
+
- Transformers 4.44.2
|
| 73 |
+
- Pytorch 2.4.0
|
| 74 |
+
- Datasets 2.21.0
|
| 75 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.4743386713994874,
|
| 5 |
+
"train_runtime": 3246.9146,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.648,
|
| 8 |
+
"train_steps_per_second": 0.089
|
| 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": "4.44.2"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.4743386713994874,
|
| 5 |
+
"train_runtime": 3246.9146,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.648,
|
| 8 |
+
"train_steps_per_second": 0.089
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,509 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 288,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.010416666666666666,
|
| 13 |
+
"grad_norm": 214.15386464224227,
|
| 14 |
+
"learning_rate": 2.8157056131518488e-08,
|
| 15 |
+
"logits/chosen": -2.590585231781006,
|
| 16 |
+
"logits/rejected": -2.5664222240448,
|
| 17 |
+
"logps/chosen": -80.29847717285156,
|
| 18 |
+
"logps/rejected": -53.10200881958008,
|
| 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": 0.10416666666666667,
|
| 28 |
+
"grad_norm": 194.60148666589583,
|
| 29 |
+
"learning_rate": 2.815705613151849e-07,
|
| 30 |
+
"logits/chosen": -2.556077241897583,
|
| 31 |
+
"logits/rejected": -2.538231611251831,
|
| 32 |
+
"logps/chosen": -87.93022918701172,
|
| 33 |
+
"logps/rejected": -81.02066040039062,
|
| 34 |
+
"loss": 0.6969,
|
| 35 |
+
"rewards/accuracies": 0.1805555522441864,
|
| 36 |
+
"rewards/chosen": -0.004902619402855635,
|
| 37 |
+
"rewards/margins": -0.010706111788749695,
|
| 38 |
+
"rewards/rejected": 0.00580349238589406,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.20833333333333334,
|
| 43 |
+
"grad_norm": 140.77132706513083,
|
| 44 |
+
"learning_rate": 5.631411226303698e-07,
|
| 45 |
+
"logits/chosen": -2.611011505126953,
|
| 46 |
+
"logits/rejected": -2.561887264251709,
|
| 47 |
+
"logps/chosen": -102.95599365234375,
|
| 48 |
+
"logps/rejected": -89.57844543457031,
|
| 49 |
+
"loss": 0.67,
|
| 50 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 51 |
+
"rewards/chosen": 0.12845759093761444,
|
| 52 |
+
"rewards/margins": 0.028963133692741394,
|
| 53 |
+
"rewards/rejected": 0.09949447214603424,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.3125,
|
| 58 |
+
"grad_norm": 216.34715329753536,
|
| 59 |
+
"learning_rate": 8.134019072433256e-07,
|
| 60 |
+
"logits/chosen": -2.5004661083221436,
|
| 61 |
+
"logits/rejected": -2.5154635906219482,
|
| 62 |
+
"logps/chosen": -66.19593048095703,
|
| 63 |
+
"logps/rejected": -74.33724212646484,
|
| 64 |
+
"loss": 0.6736,
|
| 65 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 66 |
+
"rewards/chosen": 0.40372389554977417,
|
| 67 |
+
"rewards/margins": 0.21228119730949402,
|
| 68 |
+
"rewards/rejected": 0.19144263863563538,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.4166666666666667,
|
| 73 |
+
"grad_norm": 168.00522909028686,
|
| 74 |
+
"learning_rate": 7.818747015362199e-07,
|
| 75 |
+
"logits/chosen": -2.5674819946289062,
|
| 76 |
+
"logits/rejected": -2.5553228855133057,
|
| 77 |
+
"logps/chosen": -70.16609191894531,
|
| 78 |
+
"logps/rejected": -69.43184661865234,
|
| 79 |
+
"loss": 0.6795,
|
| 80 |
+
"rewards/accuracies": 0.26875001192092896,
|
| 81 |
+
"rewards/chosen": 0.6441055536270142,
|
| 82 |
+
"rewards/margins": 0.22733807563781738,
|
| 83 |
+
"rewards/rejected": 0.41676750779151917,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.5208333333333334,
|
| 88 |
+
"grad_norm": 180.7175984450008,
|
| 89 |
+
"learning_rate": 7.503474958291143e-07,
|
| 90 |
+
"logits/chosen": -2.4915037155151367,
|
| 91 |
+
"logits/rejected": -2.5040123462677,
|
| 92 |
+
"logps/chosen": -47.24058532714844,
|
| 93 |
+
"logps/rejected": -55.82140350341797,
|
| 94 |
+
"loss": 0.6926,
|
| 95 |
+
"rewards/accuracies": 0.23125000298023224,
|
| 96 |
+
"rewards/chosen": 1.1905267238616943,
|
| 97 |
+
"rewards/margins": 0.18629805743694305,
|
| 98 |
+
"rewards/rejected": 1.0042288303375244,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.625,
|
| 103 |
+
"grad_norm": 145.01566527061468,
|
| 104 |
+
"learning_rate": 7.188202901220087e-07,
|
| 105 |
+
"logits/chosen": -2.6002345085144043,
|
| 106 |
+
"logits/rejected": -2.5831518173217773,
|
| 107 |
+
"logps/chosen": -73.40350341796875,
|
| 108 |
+
"logps/rejected": -74.82201385498047,
|
| 109 |
+
"loss": 0.6936,
|
| 110 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 111 |
+
"rewards/chosen": 1.8426650762557983,
|
| 112 |
+
"rewards/margins": 0.37407809495925903,
|
| 113 |
+
"rewards/rejected": 1.4685871601104736,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.7291666666666666,
|
| 118 |
+
"grad_norm": 179.82436828387043,
|
| 119 |
+
"learning_rate": 6.87293084414903e-07,
|
| 120 |
+
"logits/chosen": -2.5965964794158936,
|
| 121 |
+
"logits/rejected": -2.5863566398620605,
|
| 122 |
+
"logps/chosen": -90.09097290039062,
|
| 123 |
+
"logps/rejected": -80.5506591796875,
|
| 124 |
+
"loss": 0.6864,
|
| 125 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 126 |
+
"rewards/chosen": 2.033968448638916,
|
| 127 |
+
"rewards/margins": 0.44049301743507385,
|
| 128 |
+
"rewards/rejected": 1.5934756994247437,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.8333333333333334,
|
| 133 |
+
"grad_norm": 271.72190744530945,
|
| 134 |
+
"learning_rate": 6.557658787077973e-07,
|
| 135 |
+
"logits/chosen": -2.5700366497039795,
|
| 136 |
+
"logits/rejected": -2.522472620010376,
|
| 137 |
+
"logps/chosen": -79.72940826416016,
|
| 138 |
+
"logps/rejected": -75.4540023803711,
|
| 139 |
+
"loss": 0.6721,
|
| 140 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 141 |
+
"rewards/chosen": 1.9240013360977173,
|
| 142 |
+
"rewards/margins": 0.6463297605514526,
|
| 143 |
+
"rewards/rejected": 1.2776715755462646,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.9375,
|
| 148 |
+
"grad_norm": 209.9071476755963,
|
| 149 |
+
"learning_rate": 6.242386730006917e-07,
|
| 150 |
+
"logits/chosen": -2.461777687072754,
|
| 151 |
+
"logits/rejected": -2.4585537910461426,
|
| 152 |
+
"logps/chosen": -51.07688522338867,
|
| 153 |
+
"logps/rejected": -61.461647033691406,
|
| 154 |
+
"loss": 0.7091,
|
| 155 |
+
"rewards/accuracies": 0.23749999701976776,
|
| 156 |
+
"rewards/chosen": 0.9441378712654114,
|
| 157 |
+
"rewards/margins": 0.2915518581867218,
|
| 158 |
+
"rewards/rejected": 0.6525859832763672,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 1.0416666666666667,
|
| 163 |
+
"grad_norm": 12.21362061398562,
|
| 164 |
+
"learning_rate": 5.927114672935861e-07,
|
| 165 |
+
"logits/chosen": -2.4158740043640137,
|
| 166 |
+
"logits/rejected": -2.395545244216919,
|
| 167 |
+
"logps/chosen": -66.05335235595703,
|
| 168 |
+
"logps/rejected": -75.04288482666016,
|
| 169 |
+
"loss": 0.5554,
|
| 170 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 171 |
+
"rewards/chosen": 2.4290213584899902,
|
| 172 |
+
"rewards/margins": 3.4105141162872314,
|
| 173 |
+
"rewards/rejected": -0.9814925193786621,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 1.0416666666666667,
|
| 178 |
+
"eval_logits/chosen": -2.452505588531494,
|
| 179 |
+
"eval_logits/rejected": -2.4356331825256348,
|
| 180 |
+
"eval_logps/chosen": -73.46339416503906,
|
| 181 |
+
"eval_logps/rejected": -81.79998779296875,
|
| 182 |
+
"eval_loss": 0.7309352159500122,
|
| 183 |
+
"eval_rewards/accuracies": 0.3392857015132904,
|
| 184 |
+
"eval_rewards/chosen": 0.33428940176963806,
|
| 185 |
+
"eval_rewards/margins": 0.5527295470237732,
|
| 186 |
+
"eval_rewards/rejected": -0.21844016015529633,
|
| 187 |
+
"eval_runtime": 113.7572,
|
| 188 |
+
"eval_samples_per_second": 17.581,
|
| 189 |
+
"eval_steps_per_second": 0.554,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 1.1458333333333333,
|
| 194 |
+
"grad_norm": 9.404688750875733,
|
| 195 |
+
"learning_rate": 5.611842615864804e-07,
|
| 196 |
+
"logits/chosen": -2.4337010383605957,
|
| 197 |
+
"logits/rejected": -2.464585781097412,
|
| 198 |
+
"logps/chosen": -55.87408447265625,
|
| 199 |
+
"logps/rejected": -93.3639907836914,
|
| 200 |
+
"loss": 0.3638,
|
| 201 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 202 |
+
"rewards/chosen": 3.0879528522491455,
|
| 203 |
+
"rewards/margins": 6.951683044433594,
|
| 204 |
+
"rewards/rejected": -3.8637306690216064,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 1.25,
|
| 209 |
+
"grad_norm": 11.93242109022535,
|
| 210 |
+
"learning_rate": 5.296570558793748e-07,
|
| 211 |
+
"logits/chosen": -2.466701030731201,
|
| 212 |
+
"logits/rejected": -2.429455280303955,
|
| 213 |
+
"logps/chosen": -96.85403442382812,
|
| 214 |
+
"logps/rejected": -117.70040130615234,
|
| 215 |
+
"loss": 0.3857,
|
| 216 |
+
"rewards/accuracies": 0.581250011920929,
|
| 217 |
+
"rewards/chosen": 3.6269073486328125,
|
| 218 |
+
"rewards/margins": 11.550947189331055,
|
| 219 |
+
"rewards/rejected": -7.924038887023926,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 1.3541666666666667,
|
| 224 |
+
"grad_norm": 24.581368760955932,
|
| 225 |
+
"learning_rate": 4.981298501722692e-07,
|
| 226 |
+
"logits/chosen": -2.3588976860046387,
|
| 227 |
+
"logits/rejected": -2.3589892387390137,
|
| 228 |
+
"logps/chosen": -80.81659698486328,
|
| 229 |
+
"logps/rejected": -112.35628509521484,
|
| 230 |
+
"loss": 0.3718,
|
| 231 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 232 |
+
"rewards/chosen": 1.8699009418487549,
|
| 233 |
+
"rewards/margins": 9.356704711914062,
|
| 234 |
+
"rewards/rejected": -7.486802577972412,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.4583333333333333,
|
| 239 |
+
"grad_norm": 22.835446345258582,
|
| 240 |
+
"learning_rate": 4.6660264446516346e-07,
|
| 241 |
+
"logits/chosen": -2.3836002349853516,
|
| 242 |
+
"logits/rejected": -2.4029014110565186,
|
| 243 |
+
"logps/chosen": -48.08272171020508,
|
| 244 |
+
"logps/rejected": -81.94709777832031,
|
| 245 |
+
"loss": 0.3721,
|
| 246 |
+
"rewards/accuracies": 0.4437499940395355,
|
| 247 |
+
"rewards/chosen": 0.3919129967689514,
|
| 248 |
+
"rewards/margins": 5.86413049697876,
|
| 249 |
+
"rewards/rejected": -5.472217559814453,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 1.5625,
|
| 254 |
+
"grad_norm": 1.3144367942764572,
|
| 255 |
+
"learning_rate": 4.350754387580579e-07,
|
| 256 |
+
"logits/chosen": -2.4465391635894775,
|
| 257 |
+
"logits/rejected": -2.4150843620300293,
|
| 258 |
+
"logps/chosen": -68.7873306274414,
|
| 259 |
+
"logps/rejected": -94.09663391113281,
|
| 260 |
+
"loss": 0.3885,
|
| 261 |
+
"rewards/accuracies": 0.4625000059604645,
|
| 262 |
+
"rewards/chosen": 1.0929625034332275,
|
| 263 |
+
"rewards/margins": 7.766973972320557,
|
| 264 |
+
"rewards/rejected": -6.67401123046875,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 1.6666666666666665,
|
| 269 |
+
"grad_norm": 5.382498184271839,
|
| 270 |
+
"learning_rate": 4.0354823305095224e-07,
|
| 271 |
+
"logits/chosen": -2.4038569927215576,
|
| 272 |
+
"logits/rejected": -2.408721923828125,
|
| 273 |
+
"logps/chosen": -78.36112976074219,
|
| 274 |
+
"logps/rejected": -118.96110534667969,
|
| 275 |
+
"loss": 0.369,
|
| 276 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 277 |
+
"rewards/chosen": 0.8339115977287292,
|
| 278 |
+
"rewards/margins": 10.923874855041504,
|
| 279 |
+
"rewards/rejected": -10.089962005615234,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 1.7708333333333335,
|
| 284 |
+
"grad_norm": 29.49369822653688,
|
| 285 |
+
"learning_rate": 3.7202102734384655e-07,
|
| 286 |
+
"logits/chosen": -2.345175266265869,
|
| 287 |
+
"logits/rejected": -2.3313803672790527,
|
| 288 |
+
"logps/chosen": -74.82723999023438,
|
| 289 |
+
"logps/rejected": -113.85337829589844,
|
| 290 |
+
"loss": 0.3923,
|
| 291 |
+
"rewards/accuracies": 0.53125,
|
| 292 |
+
"rewards/chosen": 0.12776944041252136,
|
| 293 |
+
"rewards/margins": 10.1668119430542,
|
| 294 |
+
"rewards/rejected": -10.039042472839355,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 1.875,
|
| 299 |
+
"grad_norm": 19.55786574804487,
|
| 300 |
+
"learning_rate": 3.404938216367409e-07,
|
| 301 |
+
"logits/chosen": -2.3961374759674072,
|
| 302 |
+
"logits/rejected": -2.376530647277832,
|
| 303 |
+
"logps/chosen": -72.12110137939453,
|
| 304 |
+
"logps/rejected": -100.91358947753906,
|
| 305 |
+
"loss": 0.3623,
|
| 306 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 307 |
+
"rewards/chosen": -0.0863330215215683,
|
| 308 |
+
"rewards/margins": 8.792970657348633,
|
| 309 |
+
"rewards/rejected": -8.879304885864258,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 1.9791666666666665,
|
| 314 |
+
"grad_norm": 21.215069783043003,
|
| 315 |
+
"learning_rate": 3.0896661592963533e-07,
|
| 316 |
+
"logits/chosen": -2.4835853576660156,
|
| 317 |
+
"logits/rejected": -2.4758810997009277,
|
| 318 |
+
"logps/chosen": -83.96082305908203,
|
| 319 |
+
"logps/rejected": -122.45509338378906,
|
| 320 |
+
"loss": 0.3768,
|
| 321 |
+
"rewards/accuracies": 0.512499988079071,
|
| 322 |
+
"rewards/chosen": 1.4638442993164062,
|
| 323 |
+
"rewards/margins": 11.81721305847168,
|
| 324 |
+
"rewards/rejected": -10.353367805480957,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 2.0833333333333335,
|
| 329 |
+
"grad_norm": 0.14888221388054043,
|
| 330 |
+
"learning_rate": 2.7743941022252964e-07,
|
| 331 |
+
"logits/chosen": -2.4281277656555176,
|
| 332 |
+
"logits/rejected": -2.450599431991577,
|
| 333 |
+
"logps/chosen": -85.7192153930664,
|
| 334 |
+
"logps/rejected": -136.96456909179688,
|
| 335 |
+
"loss": 0.3557,
|
| 336 |
+
"rewards/accuracies": 0.543749988079071,
|
| 337 |
+
"rewards/chosen": 1.0684223175048828,
|
| 338 |
+
"rewards/margins": 12.590888023376465,
|
| 339 |
+
"rewards/rejected": -11.522466659545898,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 2.0833333333333335,
|
| 344 |
+
"eval_logits/chosen": -2.465620279312134,
|
| 345 |
+
"eval_logits/rejected": -2.4456026554107666,
|
| 346 |
+
"eval_logps/chosen": -84.56539154052734,
|
| 347 |
+
"eval_logps/rejected": -94.31573486328125,
|
| 348 |
+
"eval_loss": 0.8315721750259399,
|
| 349 |
+
"eval_rewards/accuracies": 0.341269850730896,
|
| 350 |
+
"eval_rewards/chosen": -3.2764618396759033,
|
| 351 |
+
"eval_rewards/margins": 1.0125247240066528,
|
| 352 |
+
"eval_rewards/rejected": -4.288985729217529,
|
| 353 |
+
"eval_runtime": 113.6105,
|
| 354 |
+
"eval_samples_per_second": 17.604,
|
| 355 |
+
"eval_steps_per_second": 0.555,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.1875,
|
| 360 |
+
"grad_norm": 0.11711860425549066,
|
| 361 |
+
"learning_rate": 2.45912204515424e-07,
|
| 362 |
+
"logits/chosen": -2.4621033668518066,
|
| 363 |
+
"logits/rejected": -2.448930501937866,
|
| 364 |
+
"logps/chosen": -62.363304138183594,
|
| 365 |
+
"logps/rejected": -93.92267608642578,
|
| 366 |
+
"loss": 0.38,
|
| 367 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 368 |
+
"rewards/chosen": 0.40027302503585815,
|
| 369 |
+
"rewards/margins": 8.606541633605957,
|
| 370 |
+
"rewards/rejected": -8.206268310546875,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 2.2916666666666665,
|
| 375 |
+
"grad_norm": 2.554878231095572,
|
| 376 |
+
"learning_rate": 2.1438499880831834e-07,
|
| 377 |
+
"logits/chosen": -2.4758799076080322,
|
| 378 |
+
"logits/rejected": -2.428165912628174,
|
| 379 |
+
"logps/chosen": -96.42266845703125,
|
| 380 |
+
"logps/rejected": -130.16854858398438,
|
| 381 |
+
"loss": 0.3276,
|
| 382 |
+
"rewards/accuracies": 0.574999988079071,
|
| 383 |
+
"rewards/chosen": 1.8300745487213135,
|
| 384 |
+
"rewards/margins": 14.0442533493042,
|
| 385 |
+
"rewards/rejected": -12.214178085327148,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 2.3958333333333335,
|
| 390 |
+
"grad_norm": 0.19011116947531465,
|
| 391 |
+
"learning_rate": 1.8285779310121273e-07,
|
| 392 |
+
"logits/chosen": -2.446868419647217,
|
| 393 |
+
"logits/rejected": -2.4058339595794678,
|
| 394 |
+
"logps/chosen": -58.83857345581055,
|
| 395 |
+
"logps/rejected": -79.16459655761719,
|
| 396 |
+
"loss": 0.3785,
|
| 397 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 398 |
+
"rewards/chosen": 0.32310181856155396,
|
| 399 |
+
"rewards/margins": 7.874607086181641,
|
| 400 |
+
"rewards/rejected": -7.551505088806152,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 2.5,
|
| 405 |
+
"grad_norm": 71.38953195661851,
|
| 406 |
+
"learning_rate": 1.5133058739410707e-07,
|
| 407 |
+
"logits/chosen": -2.4237632751464844,
|
| 408 |
+
"logits/rejected": -2.3682377338409424,
|
| 409 |
+
"logps/chosen": -86.28116607666016,
|
| 410 |
+
"logps/rejected": -109.75992584228516,
|
| 411 |
+
"loss": 0.3763,
|
| 412 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 413 |
+
"rewards/chosen": 0.5250676870346069,
|
| 414 |
+
"rewards/margins": 10.391054153442383,
|
| 415 |
+
"rewards/rejected": -9.865986824035645,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 2.6041666666666665,
|
| 420 |
+
"grad_norm": 0.21856365716272738,
|
| 421 |
+
"learning_rate": 1.1980338168700146e-07,
|
| 422 |
+
"logits/chosen": -2.3821611404418945,
|
| 423 |
+
"logits/rejected": -2.366961717605591,
|
| 424 |
+
"logps/chosen": -71.19404602050781,
|
| 425 |
+
"logps/rejected": -100.2869644165039,
|
| 426 |
+
"loss": 0.3438,
|
| 427 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 428 |
+
"rewards/chosen": 0.47970885038375854,
|
| 429 |
+
"rewards/margins": 10.1876220703125,
|
| 430 |
+
"rewards/rejected": -9.70791244506836,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 2.7083333333333335,
|
| 435 |
+
"grad_norm": 0.2713321343286941,
|
| 436 |
+
"learning_rate": 8.82761759798958e-08,
|
| 437 |
+
"logits/chosen": -2.395521402359009,
|
| 438 |
+
"logits/rejected": -2.403886318206787,
|
| 439 |
+
"logps/chosen": -56.680870056152344,
|
| 440 |
+
"logps/rejected": -104.11289978027344,
|
| 441 |
+
"loss": 0.3752,
|
| 442 |
+
"rewards/accuracies": 0.45625001192092896,
|
| 443 |
+
"rewards/chosen": 0.4715859293937683,
|
| 444 |
+
"rewards/margins": 10.60243034362793,
|
| 445 |
+
"rewards/rejected": -10.130845069885254,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 2.8125,
|
| 450 |
+
"grad_norm": 2.290935978349522,
|
| 451 |
+
"learning_rate": 5.6748970272790155e-08,
|
| 452 |
+
"logits/chosen": -2.407195568084717,
|
| 453 |
+
"logits/rejected": -2.4316117763519287,
|
| 454 |
+
"logps/chosen": -77.73648071289062,
|
| 455 |
+
"logps/rejected": -109.90128326416016,
|
| 456 |
+
"loss": 0.3519,
|
| 457 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 458 |
+
"rewards/chosen": -0.3454963266849518,
|
| 459 |
+
"rewards/margins": 9.936100006103516,
|
| 460 |
+
"rewards/rejected": -10.281596183776855,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 2.9166666666666665,
|
| 465 |
+
"grad_norm": 0.34209721575617014,
|
| 466 |
+
"learning_rate": 2.5221764565684515e-08,
|
| 467 |
+
"logits/chosen": -2.470424175262451,
|
| 468 |
+
"logits/rejected": -2.4299635887145996,
|
| 469 |
+
"logps/chosen": -82.2805404663086,
|
| 470 |
+
"logps/rejected": -117.57357025146484,
|
| 471 |
+
"loss": 0.366,
|
| 472 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 473 |
+
"rewards/chosen": 1.000727891921997,
|
| 474 |
+
"rewards/margins": 12.448894500732422,
|
| 475 |
+
"rewards/rejected": -11.44816780090332,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 3.0,
|
| 480 |
+
"step": 288,
|
| 481 |
+
"total_flos": 0.0,
|
| 482 |
+
"train_loss": 0.4743386713994874,
|
| 483 |
+
"train_runtime": 3246.9146,
|
| 484 |
+
"train_samples_per_second": 5.648,
|
| 485 |
+
"train_steps_per_second": 0.089
|
| 486 |
+
}
|
| 487 |
+
],
|
| 488 |
+
"logging_steps": 10,
|
| 489 |
+
"max_steps": 288,
|
| 490 |
+
"num_input_tokens_seen": 0,
|
| 491 |
+
"num_train_epochs": 3,
|
| 492 |
+
"save_steps": 100,
|
| 493 |
+
"stateful_callbacks": {
|
| 494 |
+
"TrainerControl": {
|
| 495 |
+
"args": {
|
| 496 |
+
"should_epoch_stop": false,
|
| 497 |
+
"should_evaluate": false,
|
| 498 |
+
"should_log": false,
|
| 499 |
+
"should_save": true,
|
| 500 |
+
"should_training_stop": true
|
| 501 |
+
},
|
| 502 |
+
"attributes": {}
|
| 503 |
+
}
|
| 504 |
+
},
|
| 505 |
+
"total_flos": 0.0,
|
| 506 |
+
"train_batch_size": 8,
|
| 507 |
+
"trial_name": null,
|
| 508 |
+
"trial_params": null
|
| 509 |
+
}
|