Model save
Browse files- README.md +76 -0
- all_results.json +8 -0
- train_results.json +8 -0
- trainer_state.json +356 -0
README.md
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 9 |
+
model-index:
|
| 10 |
+
- name: nash_dpo_rank4_iter_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 |
+
# nash_dpo_rank4_iter_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 the None dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5471
|
| 22 |
+
- Rewards/chosen: -0.3098
|
| 23 |
+
- Rewards/rejected: -0.8270
|
| 24 |
+
- Rewards/accuracies: 0.7140
|
| 25 |
+
- Rewards/margins: 0.5173
|
| 26 |
+
- Logps/rejected: -380.2009
|
| 27 |
+
- Logps/chosen: -346.7321
|
| 28 |
+
- Logits/rejected: -0.5385
|
| 29 |
+
- Logits/chosen: -0.9596
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-06
|
| 49 |
+
- train_batch_size: 2
|
| 50 |
+
- eval_batch_size: 2
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 4
|
| 54 |
+
- gradient_accumulation_steps: 16
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 8
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.5347 | 0.49 | 100 | 0.5570 | -0.3268 | -0.8259 | 0.7120 | 0.4991 | -380.0872 | -348.4370 | -0.5705 | -0.9798 |
|
| 67 |
+
| 0.5154 | 0.98 | 200 | 0.5471 | -0.3098 | -0.8270 | 0.7140 | 0.5173 | -380.2009 | -346.7321 | -0.5385 | -0.9596 |
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
### Framework versions
|
| 71 |
+
|
| 72 |
+
- PEFT 0.7.1
|
| 73 |
+
- Transformers 4.36.2
|
| 74 |
+
- Pytorch 2.1.2+cu121
|
| 75 |
+
- Datasets 2.14.6
|
| 76 |
+
- Tokenizers 0.15.2
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5623768936185276,
|
| 4 |
+
"train_runtime": 9619.6269,
|
| 5 |
+
"train_samples": 26135,
|
| 6 |
+
"train_samples_per_second": 2.717,
|
| 7 |
+
"train_steps_per_second": 0.021
|
| 8 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5623768936185276,
|
| 4 |
+
"train_runtime": 9619.6269,
|
| 5 |
+
"train_samples": 26135,
|
| 6 |
+
"train_samples_per_second": 2.717,
|
| 7 |
+
"train_steps_per_second": 0.021
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,356 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9990817263544536,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 204,
|
| 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 |
+
"learning_rate": 2.3809523809523811e-07,
|
| 14 |
+
"logits/chosen": -2.5948691368103027,
|
| 15 |
+
"logits/rejected": -2.452101707458496,
|
| 16 |
+
"logps/chosen": -288.6771240234375,
|
| 17 |
+
"logps/rejected": -270.8803405761719,
|
| 18 |
+
"loss": 0.6931,
|
| 19 |
+
"rewards/accuracies": 0.0,
|
| 20 |
+
"rewards/chosen": 0.0,
|
| 21 |
+
"rewards/margins": 0.0,
|
| 22 |
+
"rewards/rejected": 0.0,
|
| 23 |
+
"step": 1
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.05,
|
| 27 |
+
"learning_rate": 2.380952380952381e-06,
|
| 28 |
+
"logits/chosen": -2.4060661792755127,
|
| 29 |
+
"logits/rejected": -2.3895983695983887,
|
| 30 |
+
"logps/chosen": -248.78944396972656,
|
| 31 |
+
"logps/rejected": -272.0387878417969,
|
| 32 |
+
"loss": 0.6902,
|
| 33 |
+
"rewards/accuracies": 0.5138888955116272,
|
| 34 |
+
"rewards/chosen": -0.012765132822096348,
|
| 35 |
+
"rewards/margins": 0.0059960586950182915,
|
| 36 |
+
"rewards/rejected": -0.01876119151711464,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.1,
|
| 41 |
+
"learning_rate": 4.761904761904762e-06,
|
| 42 |
+
"logits/chosen": -2.352856159210205,
|
| 43 |
+
"logits/rejected": -2.2108333110809326,
|
| 44 |
+
"logps/chosen": -330.4653625488281,
|
| 45 |
+
"logps/rejected": -309.5657043457031,
|
| 46 |
+
"loss": 0.6543,
|
| 47 |
+
"rewards/accuracies": 0.659375011920929,
|
| 48 |
+
"rewards/chosen": -0.05031045526266098,
|
| 49 |
+
"rewards/margins": 0.09622685611248016,
|
| 50 |
+
"rewards/rejected": -0.14653730392456055,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.15,
|
| 55 |
+
"learning_rate": 4.970219740227693e-06,
|
| 56 |
+
"logits/chosen": -1.890295386314392,
|
| 57 |
+
"logits/rejected": -1.7602672576904297,
|
| 58 |
+
"logps/chosen": -304.2027282714844,
|
| 59 |
+
"logps/rejected": -310.0560607910156,
|
| 60 |
+
"loss": 0.6122,
|
| 61 |
+
"rewards/accuracies": 0.6656249761581421,
|
| 62 |
+
"rewards/chosen": -0.05764678865671158,
|
| 63 |
+
"rewards/margins": 0.20967629551887512,
|
| 64 |
+
"rewards/rejected": -0.2673230767250061,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.2,
|
| 69 |
+
"learning_rate": 4.868186180746792e-06,
|
| 70 |
+
"logits/chosen": -1.6625322103500366,
|
| 71 |
+
"logits/rejected": -1.408319115638733,
|
| 72 |
+
"logps/chosen": -332.21380615234375,
|
| 73 |
+
"logps/rejected": -359.38226318359375,
|
| 74 |
+
"loss": 0.5921,
|
| 75 |
+
"rewards/accuracies": 0.706250011920929,
|
| 76 |
+
"rewards/chosen": -0.0873696357011795,
|
| 77 |
+
"rewards/margins": 0.3436659276485443,
|
| 78 |
+
"rewards/rejected": -0.431035578250885,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.24,
|
| 83 |
+
"learning_rate": 4.696530612642871e-06,
|
| 84 |
+
"logits/chosen": -1.2149860858917236,
|
| 85 |
+
"logits/rejected": -0.9494598507881165,
|
| 86 |
+
"logps/chosen": -344.01007080078125,
|
| 87 |
+
"logps/rejected": -387.1136779785156,
|
| 88 |
+
"loss": 0.5646,
|
| 89 |
+
"rewards/accuracies": 0.746874988079071,
|
| 90 |
+
"rewards/chosen": -0.3123381733894348,
|
| 91 |
+
"rewards/margins": 0.41091781854629517,
|
| 92 |
+
"rewards/rejected": -0.72325599193573,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.29,
|
| 97 |
+
"learning_rate": 4.460299516441777e-06,
|
| 98 |
+
"logits/chosen": -1.3099194765090942,
|
| 99 |
+
"logits/rejected": -1.0444400310516357,
|
| 100 |
+
"logps/chosen": -308.3590393066406,
|
| 101 |
+
"logps/rejected": -356.48309326171875,
|
| 102 |
+
"loss": 0.5712,
|
| 103 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 104 |
+
"rewards/chosen": -0.07990916073322296,
|
| 105 |
+
"rewards/margins": 0.42140746116638184,
|
| 106 |
+
"rewards/rejected": -0.5013166666030884,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.34,
|
| 111 |
+
"learning_rate": 4.1664378205239085e-06,
|
| 112 |
+
"logits/chosen": -1.1891834735870361,
|
| 113 |
+
"logits/rejected": -0.8295331001281738,
|
| 114 |
+
"logps/chosen": -328.58770751953125,
|
| 115 |
+
"logps/rejected": -367.3239440917969,
|
| 116 |
+
"loss": 0.565,
|
| 117 |
+
"rewards/accuracies": 0.659375011920929,
|
| 118 |
+
"rewards/chosen": -0.2833566665649414,
|
| 119 |
+
"rewards/margins": 0.43687087297439575,
|
| 120 |
+
"rewards/rejected": -0.7202275395393372,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.39,
|
| 125 |
+
"learning_rate": 3.8235847280454626e-06,
|
| 126 |
+
"logits/chosen": -1.118951439857483,
|
| 127 |
+
"logits/rejected": -0.7426460385322571,
|
| 128 |
+
"logps/chosen": -330.37640380859375,
|
| 129 |
+
"logps/rejected": -368.56915283203125,
|
| 130 |
+
"loss": 0.5259,
|
| 131 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 132 |
+
"rewards/chosen": -0.24417515099048615,
|
| 133 |
+
"rewards/margins": 0.4794433116912842,
|
| 134 |
+
"rewards/rejected": -0.7236183881759644,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.44,
|
| 139 |
+
"learning_rate": 3.441819734087963e-06,
|
| 140 |
+
"logits/chosen": -1.1813547611236572,
|
| 141 |
+
"logits/rejected": -0.7541359663009644,
|
| 142 |
+
"logps/chosen": -337.6168518066406,
|
| 143 |
+
"logps/rejected": -367.4094543457031,
|
| 144 |
+
"loss": 0.5487,
|
| 145 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 146 |
+
"rewards/chosen": -0.11910835653543472,
|
| 147 |
+
"rewards/margins": 0.5179867744445801,
|
| 148 |
+
"rewards/rejected": -0.6370951533317566,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.49,
|
| 153 |
+
"learning_rate": 3.0323662998460396e-06,
|
| 154 |
+
"logits/chosen": -1.077327013015747,
|
| 155 |
+
"logits/rejected": -0.598007082939148,
|
| 156 |
+
"logps/chosen": -346.1620788574219,
|
| 157 |
+
"logps/rejected": -371.0578918457031,
|
| 158 |
+
"loss": 0.5347,
|
| 159 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 160 |
+
"rewards/chosen": -0.2570808529853821,
|
| 161 |
+
"rewards/margins": 0.4765000343322754,
|
| 162 |
+
"rewards/rejected": -0.7335808873176575,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.49,
|
| 167 |
+
"eval_logits/chosen": -0.9797883033752441,
|
| 168 |
+
"eval_logits/rejected": -0.5705389380455017,
|
| 169 |
+
"eval_logps/chosen": -348.43701171875,
|
| 170 |
+
"eval_logps/rejected": -380.08721923828125,
|
| 171 |
+
"eval_loss": 0.5570356845855713,
|
| 172 |
+
"eval_rewards/accuracies": 0.7120000123977661,
|
| 173 |
+
"eval_rewards/chosen": -0.3268292248249054,
|
| 174 |
+
"eval_rewards/margins": 0.49906718730926514,
|
| 175 |
+
"eval_rewards/rejected": -0.8258963227272034,
|
| 176 |
+
"eval_runtime": 384.1665,
|
| 177 |
+
"eval_samples_per_second": 5.206,
|
| 178 |
+
"eval_steps_per_second": 0.651,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.54,
|
| 183 |
+
"learning_rate": 2.6072618954988867e-06,
|
| 184 |
+
"logits/chosen": -1.0047966241836548,
|
| 185 |
+
"logits/rejected": -0.5113744735717773,
|
| 186 |
+
"logps/chosen": -352.7798767089844,
|
| 187 |
+
"logps/rejected": -395.8839416503906,
|
| 188 |
+
"loss": 0.5652,
|
| 189 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 190 |
+
"rewards/chosen": -0.37514516711235046,
|
| 191 |
+
"rewards/margins": 0.49685588479042053,
|
| 192 |
+
"rewards/rejected": -0.872001051902771,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.59,
|
| 197 |
+
"learning_rate": 2.1790041121336223e-06,
|
| 198 |
+
"logits/chosen": -1.1386340856552124,
|
| 199 |
+
"logits/rejected": -0.762003481388092,
|
| 200 |
+
"logps/chosen": -347.37567138671875,
|
| 201 |
+
"logps/rejected": -381.8182373046875,
|
| 202 |
+
"loss": 0.5298,
|
| 203 |
+
"rewards/accuracies": 0.721875011920929,
|
| 204 |
+
"rewards/chosen": -0.32593947649002075,
|
| 205 |
+
"rewards/margins": 0.47125840187072754,
|
| 206 |
+
"rewards/rejected": -0.7971979379653931,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.64,
|
| 211 |
+
"learning_rate": 1.760183246631777e-06,
|
| 212 |
+
"logits/chosen": -0.9937411546707153,
|
| 213 |
+
"logits/rejected": -0.5911135673522949,
|
| 214 |
+
"logps/chosen": -341.10308837890625,
|
| 215 |
+
"logps/rejected": -390.31622314453125,
|
| 216 |
+
"loss": 0.5382,
|
| 217 |
+
"rewards/accuracies": 0.731249988079071,
|
| 218 |
+
"rewards/chosen": -0.22077293694019318,
|
| 219 |
+
"rewards/margins": 0.6057422757148743,
|
| 220 |
+
"rewards/rejected": -0.8265151977539062,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.69,
|
| 225 |
+
"learning_rate": 1.3631121611097364e-06,
|
| 226 |
+
"logits/chosen": -0.8524423837661743,
|
| 227 |
+
"logits/rejected": -0.5514906048774719,
|
| 228 |
+
"logps/chosen": -347.8735046386719,
|
| 229 |
+
"logps/rejected": -380.4819641113281,
|
| 230 |
+
"loss": 0.5419,
|
| 231 |
+
"rewards/accuracies": 0.703125,
|
| 232 |
+
"rewards/chosen": -0.3039913773536682,
|
| 233 |
+
"rewards/margins": 0.49449166655540466,
|
| 234 |
+
"rewards/rejected": -0.7984830141067505,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.73,
|
| 239 |
+
"learning_rate": 9.994642986290797e-07,
|
| 240 |
+
"logits/chosen": -0.9539240002632141,
|
| 241 |
+
"logits/rejected": -0.48703765869140625,
|
| 242 |
+
"logps/chosen": -338.36041259765625,
|
| 243 |
+
"logps/rejected": -382.77618408203125,
|
| 244 |
+
"loss": 0.5449,
|
| 245 |
+
"rewards/accuracies": 0.737500011920929,
|
| 246 |
+
"rewards/chosen": -0.3446277379989624,
|
| 247 |
+
"rewards/margins": 0.566214382648468,
|
| 248 |
+
"rewards/rejected": -0.9108421206474304,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.78,
|
| 253 |
+
"learning_rate": 6.799304971075383e-07,
|
| 254 |
+
"logits/chosen": -0.9565147161483765,
|
| 255 |
+
"logits/rejected": -0.4371515214443207,
|
| 256 |
+
"logps/chosen": -359.60650634765625,
|
| 257 |
+
"logps/rejected": -392.94366455078125,
|
| 258 |
+
"loss": 0.533,
|
| 259 |
+
"rewards/accuracies": 0.746874988079071,
|
| 260 |
+
"rewards/chosen": -0.3234427571296692,
|
| 261 |
+
"rewards/margins": 0.5231602787971497,
|
| 262 |
+
"rewards/rejected": -0.8466030359268188,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.83,
|
| 267 |
+
"learning_rate": 4.1390469071538183e-07,
|
| 268 |
+
"logits/chosen": -0.8408013582229614,
|
| 269 |
+
"logits/rejected": -0.5169156789779663,
|
| 270 |
+
"logps/chosen": -344.161376953125,
|
| 271 |
+
"logps/rejected": -401.06927490234375,
|
| 272 |
+
"loss": 0.5348,
|
| 273 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 274 |
+
"rewards/chosen": -0.2773435711860657,
|
| 275 |
+
"rewards/margins": 0.7136318683624268,
|
| 276 |
+
"rewards/rejected": -0.9909754991531372,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.88,
|
| 281 |
+
"learning_rate": 2.092077387824884e-07,
|
| 282 |
+
"logits/chosen": -1.0086092948913574,
|
| 283 |
+
"logits/rejected": -0.6254789233207703,
|
| 284 |
+
"logps/chosen": -348.04437255859375,
|
| 285 |
+
"logps/rejected": -398.12921142578125,
|
| 286 |
+
"loss": 0.5478,
|
| 287 |
+
"rewards/accuracies": 0.721875011920929,
|
| 288 |
+
"rewards/chosen": -0.273074209690094,
|
| 289 |
+
"rewards/margins": 0.5786523818969727,
|
| 290 |
+
"rewards/rejected": -0.8517265319824219,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.93,
|
| 295 |
+
"learning_rate": 7.185750133542168e-08,
|
| 296 |
+
"logits/chosen": -0.9045581817626953,
|
| 297 |
+
"logits/rejected": -0.6502401828765869,
|
| 298 |
+
"logps/chosen": -323.7873840332031,
|
| 299 |
+
"logps/rejected": -377.43109130859375,
|
| 300 |
+
"loss": 0.5416,
|
| 301 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 302 |
+
"rewards/chosen": -0.3004080057144165,
|
| 303 |
+
"rewards/margins": 0.5847684741020203,
|
| 304 |
+
"rewards/rejected": -0.8851765394210815,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.98,
|
| 309 |
+
"learning_rate": 5.891920784984184e-09,
|
| 310 |
+
"logits/chosen": -1.004695177078247,
|
| 311 |
+
"logits/rejected": -0.5996636748313904,
|
| 312 |
+
"logps/chosen": -338.21856689453125,
|
| 313 |
+
"logps/rejected": -390.23956298828125,
|
| 314 |
+
"loss": 0.5154,
|
| 315 |
+
"rewards/accuracies": 0.778124988079071,
|
| 316 |
+
"rewards/chosen": -0.263592392206192,
|
| 317 |
+
"rewards/margins": 0.6431443691253662,
|
| 318 |
+
"rewards/rejected": -0.9067367315292358,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.98,
|
| 323 |
+
"eval_logits/chosen": -0.9595882296562195,
|
| 324 |
+
"eval_logits/rejected": -0.5385043621063232,
|
| 325 |
+
"eval_logps/chosen": -346.7320556640625,
|
| 326 |
+
"eval_logps/rejected": -380.200927734375,
|
| 327 |
+
"eval_loss": 0.5470749735832214,
|
| 328 |
+
"eval_rewards/accuracies": 0.7139999866485596,
|
| 329 |
+
"eval_rewards/chosen": -0.3097793161869049,
|
| 330 |
+
"eval_rewards/margins": 0.5172543525695801,
|
| 331 |
+
"eval_rewards/rejected": -0.8270336985588074,
|
| 332 |
+
"eval_runtime": 384.197,
|
| 333 |
+
"eval_samples_per_second": 5.206,
|
| 334 |
+
"eval_steps_per_second": 0.651,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 1.0,
|
| 339 |
+
"step": 204,
|
| 340 |
+
"total_flos": 0.0,
|
| 341 |
+
"train_loss": 0.5623768936185276,
|
| 342 |
+
"train_runtime": 9619.6269,
|
| 343 |
+
"train_samples_per_second": 2.717,
|
| 344 |
+
"train_steps_per_second": 0.021
|
| 345 |
+
}
|
| 346 |
+
],
|
| 347 |
+
"logging_steps": 10,
|
| 348 |
+
"max_steps": 204,
|
| 349 |
+
"num_input_tokens_seen": 0,
|
| 350 |
+
"num_train_epochs": 1,
|
| 351 |
+
"save_steps": 100,
|
| 352 |
+
"total_flos": 0.0,
|
| 353 |
+
"train_batch_size": 2,
|
| 354 |
+
"trial_name": null,
|
| 355 |
+
"trial_params": null
|
| 356 |
+
}
|