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-0.0-0.2-polynomial-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-0.0-0.2-polynomial-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: 0.6467
|
| 22 |
+
- Rewards/chosen: -0.1914
|
| 23 |
+
- Rewards/rejected: -0.6247
|
| 24 |
+
- Rewards/accuracies: 0.3313
|
| 25 |
+
- Rewards/margins: 0.4333
|
| 26 |
+
- Logps/rejected: -83.7654
|
| 27 |
+
- Logps/chosen: -75.2992
|
| 28 |
+
- Logits/rejected: -2.5078
|
| 29 |
+
- Logits/chosen: -2.5251
|
| 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: 2.1445950529764786e-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: polynomial
|
| 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 |
+
| 0.5502 | 1.0417 | 100 | 0.6403 | 0.2781 | 0.0281 | 0.3254 | 0.2500 | -81.0099 | -73.3173 | -2.5124 | -2.5290 |
|
| 67 |
+
| 0.377 | 2.0833 | 200 | 0.6467 | -0.1914 | -0.6247 | 0.3313 | 0.4333 | -83.7654 | -75.2992 | -2.5078 | -2.5251 |
|
| 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": 0.4920133708251847,
|
| 5 |
+
"train_runtime": 3526.2726,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.201,
|
| 8 |
+
"train_steps_per_second": 0.082
|
| 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": 0.4920133708251847,
|
| 5 |
+
"train_runtime": 3526.2726,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.201,
|
| 8 |
+
"train_steps_per_second": 0.082
|
| 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": 155.97537788651405,
|
| 14 |
+
"learning_rate": 7.395155355091305e-09,
|
| 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": 140.96156802230377,
|
| 29 |
+
"learning_rate": 7.395155355091307e-08,
|
| 30 |
+
"logits/chosen": -2.5568971633911133,
|
| 31 |
+
"logits/rejected": -2.538863182067871,
|
| 32 |
+
"logps/chosen": -87.95579528808594,
|
| 33 |
+
"logps/rejected": -81.03325653076172,
|
| 34 |
+
"loss": 0.6946,
|
| 35 |
+
"rewards/accuracies": 0.1875,
|
| 36 |
+
"rewards/chosen": -0.009626655839383602,
|
| 37 |
+
"rewards/margins": -0.01086943969130516,
|
| 38 |
+
"rewards/rejected": 0.001242783502675593,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.20833333333333334,
|
| 43 |
+
"grad_norm": 125.98597666420928,
|
| 44 |
+
"learning_rate": 1.4790310710182613e-07,
|
| 45 |
+
"logits/chosen": -2.615262269973755,
|
| 46 |
+
"logits/rejected": -2.566425323486328,
|
| 47 |
+
"logps/chosen": -103.3228759765625,
|
| 48 |
+
"logps/rejected": -89.88175964355469,
|
| 49 |
+
"loss": 0.6912,
|
| 50 |
+
"rewards/accuracies": 0.3062500059604645,
|
| 51 |
+
"rewards/chosen": 0.006651306990534067,
|
| 52 |
+
"rewards/margins": 0.006033898331224918,
|
| 53 |
+
"rewards/rejected": 0.0006174088339321315,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.3125,
|
| 58 |
+
"grad_norm": 163.17755441844716,
|
| 59 |
+
"learning_rate": 2.1401757670576506e-07,
|
| 60 |
+
"logits/chosen": -2.5203137397766113,
|
| 61 |
+
"logits/rejected": -2.5346357822418213,
|
| 62 |
+
"logps/chosen": -67.3207778930664,
|
| 63 |
+
"logps/rejected": -74.93171691894531,
|
| 64 |
+
"loss": 0.684,
|
| 65 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 66 |
+
"rewards/chosen": 0.027593865990638733,
|
| 67 |
+
"rewards/margins": 0.028976192697882652,
|
| 68 |
+
"rewards/rejected": -0.0013823278713971376,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.4166666666666667,
|
| 73 |
+
"grad_norm": 101.98216597077327,
|
| 74 |
+
"learning_rate": 2.0959829078693693e-07,
|
| 75 |
+
"logits/chosen": -2.5797293186187744,
|
| 76 |
+
"logits/rejected": -2.567654848098755,
|
| 77 |
+
"logps/chosen": -71.80469512939453,
|
| 78 |
+
"logps/rejected": -70.60755920410156,
|
| 79 |
+
"loss": 0.6752,
|
| 80 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 81 |
+
"rewards/chosen": 0.08097527176141739,
|
| 82 |
+
"rewards/margins": 0.05593211576342583,
|
| 83 |
+
"rewards/rejected": 0.02504315972328186,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.5208333333333334,
|
| 88 |
+
"grad_norm": 116.53954973852585,
|
| 89 |
+
"learning_rate": 2.0517900486810885e-07,
|
| 90 |
+
"logits/chosen": -2.491152048110962,
|
| 91 |
+
"logits/rejected": -2.503134250640869,
|
| 92 |
+
"logps/chosen": -50.373291015625,
|
| 93 |
+
"logps/rejected": -58.62458419799805,
|
| 94 |
+
"loss": 0.6651,
|
| 95 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 96 |
+
"rewards/chosen": 0.1250298023223877,
|
| 97 |
+
"rewards/margins": 0.05763009935617447,
|
| 98 |
+
"rewards/rejected": 0.06739969551563263,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.625,
|
| 103 |
+
"grad_norm": 118.26991196128117,
|
| 104 |
+
"learning_rate": 2.0075971894928074e-07,
|
| 105 |
+
"logits/chosen": -2.5747580528259277,
|
| 106 |
+
"logits/rejected": -2.556131601333618,
|
| 107 |
+
"logps/chosen": -77.9587173461914,
|
| 108 |
+
"logps/rejected": -78.81073760986328,
|
| 109 |
+
"loss": 0.6597,
|
| 110 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 111 |
+
"rewards/chosen": 0.26304519176483154,
|
| 112 |
+
"rewards/margins": 0.13826420903205872,
|
| 113 |
+
"rewards/rejected": 0.12478096783161163,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.7291666666666666,
|
| 118 |
+
"grad_norm": 109.77312584060142,
|
| 119 |
+
"learning_rate": 1.9634043303045264e-07,
|
| 120 |
+
"logits/chosen": -2.552567720413208,
|
| 121 |
+
"logits/rejected": -2.5432896614074707,
|
| 122 |
+
"logps/chosen": -94.4707260131836,
|
| 123 |
+
"logps/rejected": -84.51776885986328,
|
| 124 |
+
"loss": 0.6456,
|
| 125 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 126 |
+
"rewards/chosen": 0.4439457952976227,
|
| 127 |
+
"rewards/margins": 0.22308318316936493,
|
| 128 |
+
"rewards/rejected": 0.22086259722709656,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.8333333333333334,
|
| 133 |
+
"grad_norm": 178.1381395653738,
|
| 134 |
+
"learning_rate": 1.9192114711162454e-07,
|
| 135 |
+
"logits/chosen": -2.552009105682373,
|
| 136 |
+
"logits/rejected": -2.497727632522583,
|
| 137 |
+
"logps/chosen": -83.81471252441406,
|
| 138 |
+
"logps/rejected": -78.81379699707031,
|
| 139 |
+
"loss": 0.6258,
|
| 140 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 141 |
+
"rewards/chosen": 0.4336014688014984,
|
| 142 |
+
"rewards/margins": 0.29889458417892456,
|
| 143 |
+
"rewards/rejected": 0.13470691442489624,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.9375,
|
| 148 |
+
"grad_norm": 158.54231900826267,
|
| 149 |
+
"learning_rate": 1.8750186119279643e-07,
|
| 150 |
+
"logits/chosen": -2.4831385612487793,
|
| 151 |
+
"logits/rejected": -2.477928400039673,
|
| 152 |
+
"logps/chosen": -52.954994201660156,
|
| 153 |
+
"logps/rejected": -63.29545974731445,
|
| 154 |
+
"loss": 0.645,
|
| 155 |
+
"rewards/accuracies": 0.2562499940395355,
|
| 156 |
+
"rewards/chosen": 0.2427663803100586,
|
| 157 |
+
"rewards/margins": 0.20185653865337372,
|
| 158 |
+
"rewards/rejected": 0.04090985283255577,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 1.0416666666666667,
|
| 163 |
+
"grad_norm": 59.1130579303193,
|
| 164 |
+
"learning_rate": 1.8308257527396833e-07,
|
| 165 |
+
"logits/chosen": -2.492121458053589,
|
| 166 |
+
"logits/rejected": -2.470026731491089,
|
| 167 |
+
"logps/chosen": -71.71488952636719,
|
| 168 |
+
"logps/rejected": -72.61717224121094,
|
| 169 |
+
"loss": 0.5502,
|
| 170 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 171 |
+
"rewards/chosen": 0.42804664373397827,
|
| 172 |
+
"rewards/margins": 0.5683026909828186,
|
| 173 |
+
"rewards/rejected": -0.14025607705116272,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 1.0416666666666667,
|
| 178 |
+
"eval_logits/chosen": -2.5289955139160156,
|
| 179 |
+
"eval_logits/rejected": -2.5124404430389404,
|
| 180 |
+
"eval_logps/chosen": -73.31734466552734,
|
| 181 |
+
"eval_logps/rejected": -81.00989532470703,
|
| 182 |
+
"eval_loss": 0.6403160691261292,
|
| 183 |
+
"eval_rewards/accuracies": 0.3253968358039856,
|
| 184 |
+
"eval_rewards/chosen": 0.2780722677707672,
|
| 185 |
+
"eval_rewards/margins": 0.25001367926597595,
|
| 186 |
+
"eval_rewards/rejected": 0.028058595955371857,
|
| 187 |
+
"eval_runtime": 142.317,
|
| 188 |
+
"eval_samples_per_second": 14.053,
|
| 189 |
+
"eval_steps_per_second": 0.443,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 1.1458333333333333,
|
| 194 |
+
"grad_norm": 47.71650195966285,
|
| 195 |
+
"learning_rate": 1.7866328935514022e-07,
|
| 196 |
+
"logits/chosen": -2.499016284942627,
|
| 197 |
+
"logits/rejected": -2.532212495803833,
|
| 198 |
+
"logps/chosen": -62.74967575073242,
|
| 199 |
+
"logps/rejected": -83.47831726074219,
|
| 200 |
+
"loss": 0.4501,
|
| 201 |
+
"rewards/accuracies": 0.45625001192092896,
|
| 202 |
+
"rewards/chosen": 0.6203856468200684,
|
| 203 |
+
"rewards/margins": 1.092772126197815,
|
| 204 |
+
"rewards/rejected": -0.4723864197731018,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 1.25,
|
| 209 |
+
"grad_norm": 34.72113709789245,
|
| 210 |
+
"learning_rate": 1.7424400343631214e-07,
|
| 211 |
+
"logits/chosen": -2.566894054412842,
|
| 212 |
+
"logits/rejected": -2.53535795211792,
|
| 213 |
+
"logps/chosen": -105.34260559082031,
|
| 214 |
+
"logps/rejected": -97.03143310546875,
|
| 215 |
+
"loss": 0.4582,
|
| 216 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 217 |
+
"rewards/chosen": 0.630843997001648,
|
| 218 |
+
"rewards/margins": 1.5061585903167725,
|
| 219 |
+
"rewards/rejected": -0.8753145337104797,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 1.3541666666666667,
|
| 224 |
+
"grad_norm": 44.58377424440547,
|
| 225 |
+
"learning_rate": 1.69824717517484e-07,
|
| 226 |
+
"logits/chosen": -2.4931442737579346,
|
| 227 |
+
"logits/rejected": -2.4973056316375732,
|
| 228 |
+
"logps/chosen": -85.21261596679688,
|
| 229 |
+
"logps/rejected": -93.46207427978516,
|
| 230 |
+
"loss": 0.4423,
|
| 231 |
+
"rewards/accuracies": 0.4375,
|
| 232 |
+
"rewards/chosen": 0.3205886781215668,
|
| 233 |
+
"rewards/margins": 1.2978527545928955,
|
| 234 |
+
"rewards/rejected": -0.9772640466690063,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.4583333333333333,
|
| 239 |
+
"grad_norm": 73.130106273938,
|
| 240 |
+
"learning_rate": 1.6540543159865593e-07,
|
| 241 |
+
"logits/chosen": -2.4969403743743896,
|
| 242 |
+
"logits/rejected": -2.513062000274658,
|
| 243 |
+
"logps/chosen": -47.37117385864258,
|
| 244 |
+
"logps/rejected": -67.8313980102539,
|
| 245 |
+
"loss": 0.4504,
|
| 246 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 247 |
+
"rewards/chosen": 0.4540034234523773,
|
| 248 |
+
"rewards/margins": 1.0959008932113647,
|
| 249 |
+
"rewards/rejected": -0.641897439956665,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 1.5625,
|
| 254 |
+
"grad_norm": 40.370233820639314,
|
| 255 |
+
"learning_rate": 1.609861456798278e-07,
|
| 256 |
+
"logits/chosen": -2.5404396057128906,
|
| 257 |
+
"logits/rejected": -2.5168280601501465,
|
| 258 |
+
"logps/chosen": -69.16145324707031,
|
| 259 |
+
"logps/rejected": -76.45948791503906,
|
| 260 |
+
"loss": 0.4389,
|
| 261 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 262 |
+
"rewards/chosen": 0.7074292302131653,
|
| 263 |
+
"rewards/margins": 1.3904778957366943,
|
| 264 |
+
"rewards/rejected": -0.6830486059188843,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 1.6666666666666665,
|
| 269 |
+
"grad_norm": 35.06923806180356,
|
| 270 |
+
"learning_rate": 1.5656685976099972e-07,
|
| 271 |
+
"logits/chosen": -2.502528190612793,
|
| 272 |
+
"logits/rejected": -2.5127906799316406,
|
| 273 |
+
"logps/chosen": -77.15453338623047,
|
| 274 |
+
"logps/rejected": -91.57726287841797,
|
| 275 |
+
"loss": 0.4299,
|
| 276 |
+
"rewards/accuracies": 0.53125,
|
| 277 |
+
"rewards/chosen": 0.8931992650032043,
|
| 278 |
+
"rewards/margins": 1.7554128170013428,
|
| 279 |
+
"rewards/rejected": -0.8622136116027832,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 1.7708333333333335,
|
| 284 |
+
"grad_norm": 55.605591268891615,
|
| 285 |
+
"learning_rate": 1.521475738421716e-07,
|
| 286 |
+
"logits/chosen": -2.4732327461242676,
|
| 287 |
+
"logits/rejected": -2.4644830226898193,
|
| 288 |
+
"logps/chosen": -71.61512756347656,
|
| 289 |
+
"logps/rejected": -86.00133514404297,
|
| 290 |
+
"loss": 0.4399,
|
| 291 |
+
"rewards/accuracies": 0.518750011920929,
|
| 292 |
+
"rewards/chosen": 0.8539541363716125,
|
| 293 |
+
"rewards/margins": 1.5681732892990112,
|
| 294 |
+
"rewards/rejected": -0.7142192125320435,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 1.875,
|
| 299 |
+
"grad_norm": 53.2168247664102,
|
| 300 |
+
"learning_rate": 1.4772828792334352e-07,
|
| 301 |
+
"logits/chosen": -2.4987406730651855,
|
| 302 |
+
"logits/rejected": -2.4796464443206787,
|
| 303 |
+
"logps/chosen": -68.8732681274414,
|
| 304 |
+
"logps/rejected": -77.2318344116211,
|
| 305 |
+
"loss": 0.4149,
|
| 306 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 307 |
+
"rewards/chosen": 0.7064772248268127,
|
| 308 |
+
"rewards/margins": 1.5638738870620728,
|
| 309 |
+
"rewards/rejected": -0.8573967814445496,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 1.9791666666666665,
|
| 314 |
+
"grad_norm": 45.792720550959615,
|
| 315 |
+
"learning_rate": 1.433090020045154e-07,
|
| 316 |
+
"logits/chosen": -2.5336966514587402,
|
| 317 |
+
"logits/rejected": -2.5320465564727783,
|
| 318 |
+
"logps/chosen": -85.67411041259766,
|
| 319 |
+
"logps/rejected": -96.0043716430664,
|
| 320 |
+
"loss": 0.4271,
|
| 321 |
+
"rewards/accuracies": 0.518750011920929,
|
| 322 |
+
"rewards/chosen": 0.6603367328643799,
|
| 323 |
+
"rewards/margins": 1.9354482889175415,
|
| 324 |
+
"rewards/rejected": -1.275111436843872,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 2.0833333333333335,
|
| 329 |
+
"grad_norm": 5.83003196296594,
|
| 330 |
+
"learning_rate": 1.388897160856873e-07,
|
| 331 |
+
"logits/chosen": -2.482923746109009,
|
| 332 |
+
"logits/rejected": -2.505110263824463,
|
| 333 |
+
"logps/chosen": -86.26466369628906,
|
| 334 |
+
"logps/rejected": -109.96630859375,
|
| 335 |
+
"loss": 0.377,
|
| 336 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 337 |
+
"rewards/chosen": 0.6489728093147278,
|
| 338 |
+
"rewards/margins": 2.6458792686462402,
|
| 339 |
+
"rewards/rejected": -1.9969062805175781,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 2.0833333333333335,
|
| 344 |
+
"eval_logits/chosen": -2.5251481533050537,
|
| 345 |
+
"eval_logits/rejected": -2.5078439712524414,
|
| 346 |
+
"eval_logps/chosen": -75.29920959472656,
|
| 347 |
+
"eval_logps/rejected": -83.76541137695312,
|
| 348 |
+
"eval_loss": 0.6467146873474121,
|
| 349 |
+
"eval_rewards/accuracies": 0.3313491940498352,
|
| 350 |
+
"eval_rewards/chosen": -0.19139644503593445,
|
| 351 |
+
"eval_rewards/margins": 0.43327686190605164,
|
| 352 |
+
"eval_rewards/rejected": -0.6246733069419861,
|
| 353 |
+
"eval_runtime": 144.2917,
|
| 354 |
+
"eval_samples_per_second": 13.861,
|
| 355 |
+
"eval_steps_per_second": 0.437,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.1875,
|
| 360 |
+
"grad_norm": 8.844561075287848,
|
| 361 |
+
"learning_rate": 1.344704301668592e-07,
|
| 362 |
+
"logits/chosen": -2.526526927947998,
|
| 363 |
+
"logits/rejected": -2.5184459686279297,
|
| 364 |
+
"logps/chosen": -60.16546630859375,
|
| 365 |
+
"logps/rejected": -74.23641967773438,
|
| 366 |
+
"loss": 0.3939,
|
| 367 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 368 |
+
"rewards/chosen": 0.8121669888496399,
|
| 369 |
+
"rewards/margins": 2.1258318424224854,
|
| 370 |
+
"rewards/rejected": -1.3136647939682007,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 2.2916666666666665,
|
| 375 |
+
"grad_norm": 5.363217741474801,
|
| 376 |
+
"learning_rate": 1.300511442480311e-07,
|
| 377 |
+
"logits/chosen": -2.5266761779785156,
|
| 378 |
+
"logits/rejected": -2.477085590362549,
|
| 379 |
+
"logps/chosen": -98.9388198852539,
|
| 380 |
+
"logps/rejected": -102.80915832519531,
|
| 381 |
+
"loss": 0.3463,
|
| 382 |
+
"rewards/accuracies": 0.574999988079071,
|
| 383 |
+
"rewards/chosen": 0.736892819404602,
|
| 384 |
+
"rewards/margins": 3.152064800262451,
|
| 385 |
+
"rewards/rejected": -2.4151718616485596,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 2.3958333333333335,
|
| 390 |
+
"grad_norm": 3.636762082821054,
|
| 391 |
+
"learning_rate": 1.25631858329203e-07,
|
| 392 |
+
"logits/chosen": -2.5067200660705566,
|
| 393 |
+
"logits/rejected": -2.4729416370391846,
|
| 394 |
+
"logps/chosen": -57.91691970825195,
|
| 395 |
+
"logps/rejected": -63.94114303588867,
|
| 396 |
+
"loss": 0.3891,
|
| 397 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 398 |
+
"rewards/chosen": 0.45365357398986816,
|
| 399 |
+
"rewards/margins": 2.347585678100586,
|
| 400 |
+
"rewards/rejected": -1.8939321041107178,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 2.5,
|
| 405 |
+
"grad_norm": 10.817240165720927,
|
| 406 |
+
"learning_rate": 1.212125724103749e-07,
|
| 407 |
+
"logits/chosen": -2.4893410205841064,
|
| 408 |
+
"logits/rejected": -2.438758373260498,
|
| 409 |
+
"logps/chosen": -82.23040771484375,
|
| 410 |
+
"logps/rejected": -85.9568862915039,
|
| 411 |
+
"loss": 0.3836,
|
| 412 |
+
"rewards/accuracies": 0.5,
|
| 413 |
+
"rewards/chosen": 1.3419866561889648,
|
| 414 |
+
"rewards/margins": 2.889305591583252,
|
| 415 |
+
"rewards/rejected": -1.547318696975708,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 2.6041666666666665,
|
| 420 |
+
"grad_norm": 13.517868605826433,
|
| 421 |
+
"learning_rate": 1.167932864915468e-07,
|
| 422 |
+
"logits/chosen": -2.4413411617279053,
|
| 423 |
+
"logits/rejected": -2.429288387298584,
|
| 424 |
+
"logps/chosen": -70.37754821777344,
|
| 425 |
+
"logps/rejected": -79.62659454345703,
|
| 426 |
+
"loss": 0.3565,
|
| 427 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 428 |
+
"rewards/chosen": 0.542807400226593,
|
| 429 |
+
"rewards/margins": 2.719435930252075,
|
| 430 |
+
"rewards/rejected": -2.176628589630127,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 2.7083333333333335,
|
| 435 |
+
"grad_norm": 4.587653594042731,
|
| 436 |
+
"learning_rate": 1.1237400057271868e-07,
|
| 437 |
+
"logits/chosen": -2.440176010131836,
|
| 438 |
+
"logits/rejected": -2.458115577697754,
|
| 439 |
+
"logps/chosen": -57.05979537963867,
|
| 440 |
+
"logps/rejected": -83.13993835449219,
|
| 441 |
+
"loss": 0.3849,
|
| 442 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 443 |
+
"rewards/chosen": 0.25371766090393066,
|
| 444 |
+
"rewards/margins": 2.664337635040283,
|
| 445 |
+
"rewards/rejected": -2.4106199741363525,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 2.8125,
|
| 450 |
+
"grad_norm": 17.548402893009843,
|
| 451 |
+
"learning_rate": 1.0795471465389056e-07,
|
| 452 |
+
"logits/chosen": -2.4485597610473633,
|
| 453 |
+
"logits/rejected": -2.4837899208068848,
|
| 454 |
+
"logps/chosen": -74.96304321289062,
|
| 455 |
+
"logps/rejected": -89.0799331665039,
|
| 456 |
+
"loss": 0.3625,
|
| 457 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 458 |
+
"rewards/chosen": 0.405341237783432,
|
| 459 |
+
"rewards/margins": 2.9616785049438477,
|
| 460 |
+
"rewards/rejected": -2.556337356567383,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 2.9166666666666665,
|
| 465 |
+
"grad_norm": 9.09052295820177,
|
| 466 |
+
"learning_rate": 1.0353542873506248e-07,
|
| 467 |
+
"logits/chosen": -2.5102744102478027,
|
| 468 |
+
"logits/rejected": -2.474376678466797,
|
| 469 |
+
"logps/chosen": -81.9388656616211,
|
| 470 |
+
"logps/rejected": -92.2305679321289,
|
| 471 |
+
"loss": 0.3734,
|
| 472 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 473 |
+
"rewards/chosen": 0.8098124265670776,
|
| 474 |
+
"rewards/margins": 3.1447136402130127,
|
| 475 |
+
"rewards/rejected": -2.3349013328552246,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 3.0,
|
| 480 |
+
"step": 288,
|
| 481 |
+
"total_flos": 0.0,
|
| 482 |
+
"train_loss": 0.4920133708251847,
|
| 483 |
+
"train_runtime": 3526.2726,
|
| 484 |
+
"train_samples_per_second": 5.201,
|
| 485 |
+
"train_steps_per_second": 0.082
|
| 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 |
+
}
|