Model save
Browse files- README.md +25 -26
- adapter_config.json +3 -3
- adapter_model.safetensors +1 -1
- all_results.json +16 -16
- eval_results.json +12 -12
- train_results.json +4 -4
- trainer_state.json +581 -1331
- training_args.bin +1 -1
README.md
CHANGED
|
@@ -1,8 +1,11 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
-
|
| 4 |
tags:
|
|
|
|
|
|
|
| 5 |
- generated_from_trainer
|
|
|
|
| 6 |
model-index:
|
| 7 |
- name: zephyr-7b-dpo-qlora
|
| 8 |
results: []
|
|
@@ -15,15 +18,15 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 15 |
|
| 16 |
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
|
| 17 |
It achieves the following results on the evaluation set:
|
| 18 |
-
- Loss:
|
| 19 |
-
- Rewards/chosen:
|
| 20 |
-
- Rewards/rejected: -
|
| 21 |
-
- Rewards/accuracies: 0.
|
| 22 |
-
- Rewards/margins:
|
| 23 |
-
- Logps/rejected: -
|
| 24 |
-
- Logps/chosen: -
|
| 25 |
-
-
|
| 26 |
-
-
|
| 27 |
|
| 28 |
## Model description
|
| 29 |
|
|
@@ -43,13 +46,13 @@ More information needed
|
|
| 43 |
|
| 44 |
The following hyperparameters were used during training:
|
| 45 |
- learning_rate: 5e-06
|
| 46 |
-
- train_batch_size:
|
| 47 |
- eval_batch_size: 8
|
| 48 |
- seed: 42
|
| 49 |
- distributed_type: multi-GPU
|
| 50 |
- num_devices: 4
|
| 51 |
-
- gradient_accumulation_steps:
|
| 52 |
-
- total_train_batch_size:
|
| 53 |
- total_eval_batch_size: 32
|
| 54 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 55 |
- lr_scheduler_type: cosine
|
|
@@ -58,22 +61,18 @@ The following hyperparameters were used during training:
|
|
| 58 |
|
| 59 |
### Training results
|
| 60 |
|
| 61 |
-
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen |
|
| 62 |
-
|
| 63 |
-
|
|
| 64 |
-
|
|
| 65 |
-
|
|
| 66 |
-
|
|
| 67 |
-
| 0.0182 | 0.52 | 500 | 0.0148 | -331.5956 | -289.6645 | 0.4464 | -41.9311 | -29219.6113 | -33432.5625 | 256.0 | 0.0 |
|
| 68 |
-
| 0.013 | 0.63 | 600 | 0.0143 | -356.6841 | -312.4188 | 0.4544 | -44.2654 | -31495.0312 | -35941.4141 | 256.0 | 0.0 |
|
| 69 |
-
| 0.0165 | 0.73 | 700 | 0.0143 | -353.6940 | -310.5345 | 0.4504 | -43.1595 | -31306.6094 | -35642.4023 | 256.0 | 0.0 |
|
| 70 |
-
| 0.0145 | 0.84 | 800 | 0.0135 | -374.0797 | -328.2772 | 0.4544 | -45.8026 | -33080.8789 | -37680.9766 | 256.0 | 0.0 |
|
| 71 |
-
| 0.0195 | 0.94 | 900 | 0.0137 | -376.5184 | -330.4032 | 0.4544 | -46.1152 | -33293.4727 | -37924.8398 | 256.0 | 0.0 |
|
| 72 |
|
| 73 |
|
| 74 |
### Framework versions
|
| 75 |
|
| 76 |
-
-
|
|
|
|
| 77 |
- Pytorch 2.1.1+cu121
|
| 78 |
- Datasets 2.14.6
|
| 79 |
-
- Tokenizers 0.
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
library_name: peft
|
| 4 |
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
- generated_from_trainer
|
| 8 |
+
base_model: mistralai/Mistral-7B-v0.1
|
| 9 |
model-index:
|
| 10 |
- name: zephyr-7b-dpo-qlora
|
| 11 |
results: []
|
|
|
|
| 18 |
|
| 19 |
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 2121.0452
|
| 22 |
+
- Rewards/chosen: 0.0578
|
| 23 |
+
- Rewards/rejected: -0.0912
|
| 24 |
+
- Rewards/accuracies: 0.7599
|
| 25 |
+
- Rewards/margins: 0.1490
|
| 26 |
+
- Logps/rejected: -253.8891
|
| 27 |
+
- Logps/chosen: -259.2458
|
| 28 |
+
- Logits/rejected: -2.2028
|
| 29 |
+
- Logits/chosen: -2.2552
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 46 |
|
| 47 |
The following hyperparameters were used during training:
|
| 48 |
- learning_rate: 5e-06
|
| 49 |
+
- train_batch_size: 2
|
| 50 |
- eval_batch_size: 8
|
| 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: 32
|
| 57 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
- lr_scheduler_type: cosine
|
|
|
|
| 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 |
+
| 2149.4746 | 0.21 | 100 | 2190.7666 | 0.0445 | -0.0848 | 0.7460 | 0.1293 | -253.2523 | -260.5782 | -2.1770 | -2.2229 |
|
| 67 |
+
| 2105.1256 | 0.42 | 200 | 2151.1555 | 0.0543 | -0.0961 | 0.7599 | 0.1504 | -254.3840 | -259.5941 | -2.2074 | -2.2603 |
|
| 68 |
+
| 2135.4973 | 0.63 | 300 | 2129.0896 | 0.0626 | -0.0799 | 0.7560 | 0.1425 | -252.7585 | -258.7624 | -2.2232 | -2.2765 |
|
| 69 |
+
| 2099.8018 | 0.84 | 400 | 2121.6672 | 0.0538 | -0.0959 | 0.7540 | 0.1497 | -254.3591 | -259.6440 | -2.2016 | -2.2541 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
|
| 71 |
|
| 72 |
### Framework versions
|
| 73 |
|
| 74 |
+
- PEFT 0.7.1
|
| 75 |
+
- Transformers 4.36.2
|
| 76 |
- Pytorch 2.1.1+cu121
|
| 77 |
- Datasets 2.14.6
|
| 78 |
+
- Tokenizers 0.15.2
|
adapter_config.json
CHANGED
|
@@ -19,13 +19,13 @@
|
|
| 19 |
"rank_pattern": {},
|
| 20 |
"revision": null,
|
| 21 |
"target_modules": [
|
|
|
|
|
|
|
| 22 |
"q_proj",
|
| 23 |
"gate_proj",
|
| 24 |
-
"up_proj",
|
| 25 |
-
"k_proj",
|
| 26 |
"down_proj",
|
| 27 |
"o_proj",
|
| 28 |
-
"
|
| 29 |
],
|
| 30 |
"task_type": "CAUSAL_LM"
|
| 31 |
}
|
|
|
|
| 19 |
"rank_pattern": {},
|
| 20 |
"revision": null,
|
| 21 |
"target_modules": [
|
| 22 |
+
"up_proj",
|
| 23 |
+
"v_proj",
|
| 24 |
"q_proj",
|
| 25 |
"gate_proj",
|
|
|
|
|
|
|
| 26 |
"down_proj",
|
| 27 |
"o_proj",
|
| 28 |
+
"k_proj"
|
| 29 |
],
|
| 30 |
"task_type": "CAUSAL_LM"
|
| 31 |
}
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c6f7d1a95415a6c03799926b1b3b0647c3602207bcc4fb5c48fa957c5b2fea04
|
| 3 |
size 671150064
|
all_results.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"
|
| 6 |
-
"
|
| 7 |
-
"
|
| 8 |
-
"eval_rewards/
|
| 9 |
-
"eval_rewards/
|
| 10 |
-
"eval_rewards/
|
| 11 |
-
"
|
|
|
|
| 12 |
"eval_samples": 2000,
|
| 13 |
-
"eval_samples_per_second":
|
| 14 |
-
"
|
| 15 |
-
"
|
| 16 |
-
"
|
| 17 |
-
"train_runtime": 42177.7253,
|
| 18 |
"train_samples": 61135,
|
| 19 |
-
"train_samples_per_second": 1.
|
| 20 |
-
"train_steps_per_second": 0.
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -2.255185127258301,
|
| 4 |
+
"eval_logits/rejected": -2.2027812004089355,
|
| 5 |
+
"eval_logps/chosen": -259.245849609375,
|
| 6 |
+
"eval_logps/rejected": -253.8891143798828,
|
| 7 |
+
"eval_loss": 2121.045166015625,
|
| 8 |
+
"eval_rewards/accuracies": 0.7599206566810608,
|
| 9 |
+
"eval_rewards/chosen": 0.057787273079156876,
|
| 10 |
+
"eval_rewards/margins": 0.14896924793720245,
|
| 11 |
+
"eval_rewards/rejected": -0.09118196368217468,
|
| 12 |
+
"eval_runtime": 547.839,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 3.651,
|
| 15 |
+
"eval_steps_per_second": 0.115,
|
| 16 |
+
"train_loss": 2164.5614415454666,
|
| 17 |
+
"train_runtime": 32346.8016,
|
|
|
|
| 18 |
"train_samples": 61135,
|
| 19 |
+
"train_samples_per_second": 1.89,
|
| 20 |
+
"train_steps_per_second": 0.015
|
| 21 |
}
|
eval_results.json
CHANGED
|
@@ -1,16 +1,16 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"
|
| 6 |
-
"
|
| 7 |
-
"
|
| 8 |
-
"eval_rewards/
|
| 9 |
-
"eval_rewards/
|
| 10 |
-
"eval_rewards/
|
| 11 |
-
"
|
|
|
|
| 12 |
"eval_samples": 2000,
|
| 13 |
-
"eval_samples_per_second":
|
| 14 |
-
"
|
| 15 |
-
"eval_steps_per_second": 0.065
|
| 16 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -2.255185127258301,
|
| 4 |
+
"eval_logits/rejected": -2.2027812004089355,
|
| 5 |
+
"eval_logps/chosen": -259.245849609375,
|
| 6 |
+
"eval_logps/rejected": -253.8891143798828,
|
| 7 |
+
"eval_loss": 2121.045166015625,
|
| 8 |
+
"eval_rewards/accuracies": 0.7599206566810608,
|
| 9 |
+
"eval_rewards/chosen": 0.057787273079156876,
|
| 10 |
+
"eval_rewards/margins": 0.14896924793720245,
|
| 11 |
+
"eval_rewards/rejected": -0.09118196368217468,
|
| 12 |
+
"eval_runtime": 547.839,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 3.651,
|
| 15 |
+
"eval_steps_per_second": 0.115
|
|
|
|
| 16 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 61135,
|
| 6 |
-
"train_samples_per_second": 1.
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 2164.5614415454666,
|
| 4 |
+
"train_runtime": 32346.8016,
|
| 5 |
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 1.89,
|
| 7 |
+
"train_steps_per_second": 0.015
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,1516 +1,766 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 0.
|
| 5 |
"eval_steps": 100,
|
| 6 |
-
"global_step":
|
| 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":
|
| 14 |
-
"
|
| 15 |
-
"
|
| 16 |
-
"
|
| 17 |
-
"
|
|
|
|
| 18 |
"rewards/accuracies": 0.0,
|
| 19 |
"rewards/chosen": 0.0,
|
| 20 |
"rewards/margins": 0.0,
|
| 21 |
"rewards/rejected": 0.0,
|
| 22 |
-
"selected": 0.0,
|
| 23 |
"step": 1
|
| 24 |
},
|
| 25 |
-
{
|
| 26 |
-
"epoch": 0.01,
|
| 27 |
-
"learning_rate": 5.208333333333334e-07,
|
| 28 |
-
"logps/chosen": -313.4251708984375,
|
| 29 |
-
"logps/rejected": -277.2637023925781,
|
| 30 |
-
"loss": 0.693,
|
| 31 |
-
"neglected": 90.0,
|
| 32 |
-
"rewards/accuracies": 0.4861111044883728,
|
| 33 |
-
"rewards/chosen": 8.512949716532603e-05,
|
| 34 |
-
"rewards/margins": 0.0002041187253780663,
|
| 35 |
-
"rewards/rejected": -0.00011898923548869789,
|
| 36 |
-
"selected": 0.0,
|
| 37 |
-
"step": 10
|
| 38 |
-
},
|
| 39 |
{
|
| 40 |
"epoch": 0.02,
|
| 41 |
"learning_rate": 1.0416666666666667e-06,
|
| 42 |
-
"
|
| 43 |
-
"
|
| 44 |
-
"
|
| 45 |
-
"
|
| 46 |
-
"
|
| 47 |
-
"rewards/
|
| 48 |
-
"rewards/
|
| 49 |
-
"rewards/
|
| 50 |
-
"
|
| 51 |
-
"step":
|
| 52 |
-
},
|
| 53 |
-
{
|
| 54 |
-
"epoch": 0.03,
|
| 55 |
-
"learning_rate": 1.5625e-06,
|
| 56 |
-
"logps/chosen": -270.5987548828125,
|
| 57 |
-
"logps/rejected": -244.8210906982422,
|
| 58 |
-
"loss": 0.693,
|
| 59 |
-
"neglected": 402.0,
|
| 60 |
-
"rewards/accuracies": 0.5,
|
| 61 |
-
"rewards/chosen": -2.5821285817073658e-05,
|
| 62 |
-
"rewards/margins": 0.00013345989282242954,
|
| 63 |
-
"rewards/rejected": -0.00015928114589769393,
|
| 64 |
-
"selected": 0.0,
|
| 65 |
-
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
"epoch": 0.04,
|
| 69 |
"learning_rate": 2.0833333333333334e-06,
|
| 70 |
-
"
|
| 71 |
-
"
|
| 72 |
-
"
|
| 73 |
-
"
|
| 74 |
-
"
|
| 75 |
-
"rewards/
|
| 76 |
-
"rewards/
|
| 77 |
-
"rewards/
|
| 78 |
-
"
|
| 79 |
-
"step":
|
| 80 |
-
},
|
| 81 |
-
{
|
| 82 |
-
"epoch": 0.05,
|
| 83 |
-
"learning_rate": 2.604166666666667e-06,
|
| 84 |
-
"logps/chosen": -255.3240509033203,
|
| 85 |
-
"logps/rejected": -249.5677490234375,
|
| 86 |
-
"loss": 0.6925,
|
| 87 |
-
"neglected": 722.0,
|
| 88 |
-
"rewards/accuracies": 0.581250011920929,
|
| 89 |
-
"rewards/chosen": 0.0003948220401071012,
|
| 90 |
-
"rewards/margins": 0.001298406976275146,
|
| 91 |
-
"rewards/rejected": -0.0009035851107910275,
|
| 92 |
-
"selected": 0.0,
|
| 93 |
-
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"epoch": 0.06,
|
| 97 |
"learning_rate": 3.125e-06,
|
| 98 |
-
"
|
| 99 |
-
"
|
| 100 |
-
"
|
| 101 |
-
"
|
| 102 |
-
"
|
| 103 |
-
"rewards/
|
| 104 |
-
"rewards/
|
| 105 |
-
"rewards/
|
| 106 |
-
"
|
| 107 |
-
"step":
|
| 108 |
-
},
|
| 109 |
-
{
|
| 110 |
-
"epoch": 0.07,
|
| 111 |
-
"learning_rate": 3.6458333333333333e-06,
|
| 112 |
-
"logps/chosen": -285.159423828125,
|
| 113 |
-
"logps/rejected": -263.88519287109375,
|
| 114 |
-
"loss": 0.6909,
|
| 115 |
-
"neglected": 1042.0,
|
| 116 |
-
"rewards/accuracies": 0.7124999761581421,
|
| 117 |
-
"rewards/chosen": 0.0027831769548356533,
|
| 118 |
-
"rewards/margins": 0.004454310052096844,
|
| 119 |
-
"rewards/rejected": -0.0016711335629224777,
|
| 120 |
-
"selected": 0.0,
|
| 121 |
-
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
"epoch": 0.08,
|
| 125 |
"learning_rate": 4.166666666666667e-06,
|
| 126 |
-
"
|
| 127 |
-
"
|
| 128 |
-
"
|
| 129 |
-
"
|
| 130 |
-
"
|
| 131 |
-
"rewards/
|
| 132 |
-
"rewards/
|
| 133 |
-
"rewards/
|
| 134 |
-
"
|
| 135 |
-
"step":
|
| 136 |
-
},
|
| 137 |
-
{
|
| 138 |
-
"epoch": 0.09,
|
| 139 |
-
"learning_rate": 4.6875000000000004e-06,
|
| 140 |
-
"logps/chosen": -280.253662109375,
|
| 141 |
-
"logps/rejected": -257.877197265625,
|
| 142 |
-
"loss": 0.6827,
|
| 143 |
-
"neglected": 1362.0,
|
| 144 |
-
"rewards/accuracies": 0.71875,
|
| 145 |
-
"rewards/chosen": 0.007530213333666325,
|
| 146 |
-
"rewards/margins": 0.01851402223110199,
|
| 147 |
-
"rewards/rejected": -0.010983810760080814,
|
| 148 |
-
"selected": 0.0,
|
| 149 |
-
"step": 90
|
| 150 |
-
},
|
| 151 |
-
{
|
| 152 |
-
"epoch": 0.1,
|
| 153 |
-
"learning_rate": 4.9997324926814375e-06,
|
| 154 |
-
"logps/chosen": -262.7320251464844,
|
| 155 |
-
"logps/rejected": -284.05206298828125,
|
| 156 |
-
"loss": 0.6727,
|
| 157 |
-
"neglected": 1522.0,
|
| 158 |
-
"rewards/accuracies": 0.699999988079071,
|
| 159 |
-
"rewards/chosen": 0.008031011559069157,
|
| 160 |
-
"rewards/margins": 0.026352444663643837,
|
| 161 |
-
"rewards/rejected": -0.018321430310606956,
|
| 162 |
-
"selected": 0.0,
|
| 163 |
-
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
"epoch": 0.1,
|
| 167 |
-
"
|
| 168 |
-
"
|
| 169 |
-
"
|
| 170 |
-
"
|
| 171 |
-
"
|
| 172 |
-
"
|
| 173 |
-
"
|
| 174 |
-
"
|
| 175 |
-
"
|
| 176 |
-
"
|
| 177 |
-
"
|
| 178 |
-
"eval_steps_per_second": 0.094,
|
| 179 |
-
"step": 100
|
| 180 |
-
},
|
| 181 |
-
{
|
| 182 |
-
"epoch": 0.12,
|
| 183 |
-
"learning_rate": 4.996723692767927e-06,
|
| 184 |
-
"logps/chosen": -256.07342529296875,
|
| 185 |
-
"logps/rejected": -245.7953643798828,
|
| 186 |
-
"loss": 0.6539,
|
| 187 |
-
"neglected": 586.0,
|
| 188 |
-
"rewards/accuracies": 0.6937500238418579,
|
| 189 |
-
"rewards/chosen": -0.0058533623814582825,
|
| 190 |
-
"rewards/margins": 0.04092119634151459,
|
| 191 |
-
"rewards/rejected": -0.04677455872297287,
|
| 192 |
-
"selected": 0.0,
|
| 193 |
-
"step": 110
|
| 194 |
},
|
| 195 |
{
|
| 196 |
"epoch": 0.13,
|
| 197 |
-
"learning_rate": 4.
|
| 198 |
-
"
|
| 199 |
-
"
|
| 200 |
-
"
|
| 201 |
-
"
|
| 202 |
-
"
|
| 203 |
-
"rewards/
|
| 204 |
-
"rewards/
|
| 205 |
-
"rewards/
|
| 206 |
-
"
|
| 207 |
-
"step":
|
| 208 |
-
},
|
| 209 |
-
{
|
| 210 |
-
"epoch": 0.14,
|
| 211 |
-
"learning_rate": 4.980697142834315e-06,
|
| 212 |
-
"logps/chosen": -312.620849609375,
|
| 213 |
-
"logps/rejected": -286.03900146484375,
|
| 214 |
-
"loss": 0.588,
|
| 215 |
-
"neglected": 906.0,
|
| 216 |
-
"rewards/accuracies": 0.6499999761581421,
|
| 217 |
-
"rewards/chosen": -0.1690162718296051,
|
| 218 |
-
"rewards/margins": 0.05675806850194931,
|
| 219 |
-
"rewards/rejected": -0.2257743626832962,
|
| 220 |
-
"selected": 0.0,
|
| 221 |
-
"step": 130
|
| 222 |
},
|
| 223 |
{
|
| 224 |
"epoch": 0.15,
|
| 225 |
-
"learning_rate": 4.
|
| 226 |
-
"
|
| 227 |
-
"
|
| 228 |
-
"
|
| 229 |
-
"
|
| 230 |
-
"
|
| 231 |
-
"rewards/
|
| 232 |
-
"rewards/
|
| 233 |
-
"rewards/
|
| 234 |
-
"
|
| 235 |
-
"step":
|
| 236 |
-
},
|
| 237 |
-
{
|
| 238 |
-
"epoch": 0.16,
|
| 239 |
-
"learning_rate": 4.951404179843963e-06,
|
| 240 |
-
"logps/chosen": -450.36004638671875,
|
| 241 |
-
"logps/rejected": -442.3370056152344,
|
| 242 |
-
"loss": 0.3643,
|
| 243 |
-
"neglected": 1226.0,
|
| 244 |
-
"rewards/accuracies": 0.606249988079071,
|
| 245 |
-
"rewards/chosen": -1.6727848052978516,
|
| 246 |
-
"rewards/margins": 0.14853505790233612,
|
| 247 |
-
"rewards/rejected": -1.821319818496704,
|
| 248 |
-
"selected": 0.0,
|
| 249 |
-
"step": 150
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.17,
|
| 253 |
-
"learning_rate": 4.
|
| 254 |
-
"
|
| 255 |
-
"
|
| 256 |
-
"
|
| 257 |
-
"
|
| 258 |
-
"
|
| 259 |
-
"rewards/
|
| 260 |
-
"rewards/
|
| 261 |
-
"rewards/
|
| 262 |
-
"
|
| 263 |
-
"step":
|
| 264 |
-
},
|
| 265 |
-
{
|
| 266 |
-
"epoch": 0.18,
|
| 267 |
-
"learning_rate": 4.909001458367867e-06,
|
| 268 |
-
"logps/chosen": -2515.188232421875,
|
| 269 |
-
"logps/rejected": -2068.07666015625,
|
| 270 |
-
"loss": 0.0935,
|
| 271 |
-
"neglected": 1546.0,
|
| 272 |
-
"rewards/accuracies": 0.5,
|
| 273 |
-
"rewards/chosen": -22.28314781188965,
|
| 274 |
-
"rewards/margins": -4.024808406829834,
|
| 275 |
-
"rewards/rejected": -18.25834083557129,
|
| 276 |
-
"selected": 0.0,
|
| 277 |
-
"step": 170
|
| 278 |
},
|
| 279 |
{
|
| 280 |
"epoch": 0.19,
|
| 281 |
-
"learning_rate": 4.
|
| 282 |
-
"
|
| 283 |
-
"
|
| 284 |
-
"
|
| 285 |
-
"
|
| 286 |
-
"
|
| 287 |
-
"rewards/
|
| 288 |
-
"rewards/
|
| 289 |
-
"rewards/
|
| 290 |
-
"
|
| 291 |
-
"step":
|
| 292 |
-
},
|
| 293 |
-
{
|
| 294 |
-
"epoch": 0.2,
|
| 295 |
-
"learning_rate": 4.853715742087947e-06,
|
| 296 |
-
"logps/chosen": -5398.57421875,
|
| 297 |
-
"logps/rejected": -4498.2265625,
|
| 298 |
-
"loss": 0.0402,
|
| 299 |
-
"neglected": 1866.0,
|
| 300 |
-
"rewards/accuracies": 0.4375,
|
| 301 |
-
"rewards/chosen": -51.07048416137695,
|
| 302 |
-
"rewards/margins": -8.596471786499023,
|
| 303 |
-
"rewards/rejected": -42.47401428222656,
|
| 304 |
-
"selected": 0.0,
|
| 305 |
-
"step": 190
|
| 306 |
},
|
| 307 |
{
|
| 308 |
"epoch": 0.21,
|
| 309 |
-
"learning_rate": 4.
|
| 310 |
-
"
|
| 311 |
-
"
|
| 312 |
-
"
|
| 313 |
-
"
|
| 314 |
-
"
|
| 315 |
-
"rewards/
|
| 316 |
-
"rewards/
|
| 317 |
-
"rewards/
|
| 318 |
-
"
|
| 319 |
-
"step":
|
| 320 |
},
|
| 321 |
{
|
| 322 |
"epoch": 0.21,
|
| 323 |
-
"
|
| 324 |
-
"
|
| 325 |
-
"
|
| 326 |
-
"
|
| 327 |
-
"
|
| 328 |
-
"eval_rewards/
|
| 329 |
-
"eval_rewards/
|
| 330 |
-
"eval_rewards/
|
| 331 |
-
"
|
| 332 |
-
"
|
| 333 |
-
"
|
| 334 |
-
"eval_steps_per_second": 0.
|
| 335 |
-
"step":
|
| 336 |
-
},
|
| 337 |
-
{
|
| 338 |
-
"epoch": 0.22,
|
| 339 |
-
"learning_rate": 4.7858426910973435e-06,
|
| 340 |
-
"logps/chosen": -12588.5732421875,
|
| 341 |
-
"logps/rejected": -11012.65625,
|
| 342 |
-
"loss": 0.0212,
|
| 343 |
-
"neglected": 586.0,
|
| 344 |
-
"rewards/accuracies": 0.44999998807907104,
|
| 345 |
-
"rewards/chosen": -122.9972152709961,
|
| 346 |
-
"rewards/margins": -15.338798522949219,
|
| 347 |
-
"rewards/rejected": -107.6584243774414,
|
| 348 |
-
"selected": 0.0,
|
| 349 |
-
"step": 210
|
| 350 |
},
|
| 351 |
{
|
| 352 |
"epoch": 0.23,
|
| 353 |
-
"learning_rate": 4.
|
| 354 |
-
"
|
| 355 |
-
"
|
| 356 |
-
"
|
| 357 |
-
"
|
| 358 |
-
"
|
| 359 |
-
"rewards/
|
| 360 |
-
"rewards/
|
| 361 |
-
"rewards/
|
| 362 |
-
"
|
| 363 |
-
"step":
|
| 364 |
-
},
|
| 365 |
-
{
|
| 366 |
-
"epoch": 0.24,
|
| 367 |
-
"learning_rate": 4.705745280752586e-06,
|
| 368 |
-
"logps/chosen": -9174.8818359375,
|
| 369 |
-
"logps/rejected": -8592.8603515625,
|
| 370 |
-
"loss": 0.0293,
|
| 371 |
-
"neglected": 906.0,
|
| 372 |
-
"rewards/accuracies": 0.46875,
|
| 373 |
-
"rewards/chosen": -88.80033874511719,
|
| 374 |
-
"rewards/margins": -5.414186000823975,
|
| 375 |
-
"rewards/rejected": -83.38614654541016,
|
| 376 |
-
"selected": 0.0,
|
| 377 |
-
"step": 230
|
| 378 |
},
|
| 379 |
{
|
| 380 |
"epoch": 0.25,
|
| 381 |
-
"learning_rate": 4.
|
| 382 |
-
"
|
| 383 |
-
"
|
| 384 |
-
"
|
| 385 |
-
"
|
| 386 |
-
"
|
| 387 |
-
"rewards/
|
| 388 |
-
"rewards/
|
| 389 |
-
"rewards/
|
| 390 |
-
"
|
| 391 |
-
"step":
|
| 392 |
-
},
|
| 393 |
-
{
|
| 394 |
-
"epoch": 0.26,
|
| 395 |
-
"learning_rate": 4.613851860533367e-06,
|
| 396 |
-
"logps/chosen": -12696.328125,
|
| 397 |
-
"logps/rejected": -12253.8046875,
|
| 398 |
-
"loss": 0.0216,
|
| 399 |
-
"neglected": 1226.0,
|
| 400 |
-
"rewards/accuracies": 0.5062500238418579,
|
| 401 |
-
"rewards/chosen": -124.18550872802734,
|
| 402 |
-
"rewards/margins": -4.2357869148254395,
|
| 403 |
-
"rewards/rejected": -119.94972229003906,
|
| 404 |
-
"selected": 0.0,
|
| 405 |
-
"step": 250
|
| 406 |
},
|
| 407 |
{
|
| 408 |
"epoch": 0.27,
|
| 409 |
-
"learning_rate": 4.
|
| 410 |
-
"
|
| 411 |
-
"
|
| 412 |
-
"
|
| 413 |
-
"
|
| 414 |
-
"
|
| 415 |
-
"rewards/
|
| 416 |
-
"rewards/
|
| 417 |
-
"rewards/
|
| 418 |
-
"
|
| 419 |
-
"step":
|
| 420 |
-
},
|
| 421 |
-
{
|
| 422 |
-
"epoch": 0.28,
|
| 423 |
-
"learning_rate": 4.510653863290871e-06,
|
| 424 |
-
"logps/chosen": -21082.57421875,
|
| 425 |
-
"logps/rejected": -18679.11328125,
|
| 426 |
-
"loss": 0.0214,
|
| 427 |
-
"neglected": 1546.0,
|
| 428 |
-
"rewards/accuracies": 0.48124998807907104,
|
| 429 |
-
"rewards/chosen": -208.0191192626953,
|
| 430 |
-
"rewards/margins": -23.887981414794922,
|
| 431 |
-
"rewards/rejected": -184.1311492919922,
|
| 432 |
-
"selected": 0.0,
|
| 433 |
-
"step": 270
|
| 434 |
},
|
| 435 |
{
|
| 436 |
"epoch": 0.29,
|
| 437 |
-
"learning_rate": 4.
|
| 438 |
-
"
|
| 439 |
-
"
|
| 440 |
-
"
|
| 441 |
-
"
|
| 442 |
-
"
|
| 443 |
-
"rewards/
|
| 444 |
-
"rewards/
|
| 445 |
-
"rewards/
|
| 446 |
-
"
|
| 447 |
-
"step":
|
| 448 |
-
},
|
| 449 |
-
{
|
| 450 |
-
"epoch": 0.3,
|
| 451 |
-
"learning_rate": 4.396703177135262e-06,
|
| 452 |
-
"logps/chosen": -25520.37890625,
|
| 453 |
-
"logps/rejected": -20765.65234375,
|
| 454 |
-
"loss": 0.0181,
|
| 455 |
-
"neglected": 1866.0,
|
| 456 |
-
"rewards/accuracies": 0.375,
|
| 457 |
-
"rewards/chosen": -252.4154815673828,
|
| 458 |
-
"rewards/margins": -46.95801544189453,
|
| 459 |
-
"rewards/rejected": -205.4574432373047,
|
| 460 |
-
"selected": 0.0,
|
| 461 |
-
"step": 290
|
| 462 |
-
},
|
| 463 |
-
{
|
| 464 |
-
"epoch": 0.31,
|
| 465 |
-
"learning_rate": 4.335883851539693e-06,
|
| 466 |
-
"logps/chosen": -26487.619140625,
|
| 467 |
-
"logps/rejected": -21971.16796875,
|
| 468 |
-
"loss": 0.0208,
|
| 469 |
-
"neglected": 2026.0,
|
| 470 |
-
"rewards/accuracies": 0.4124999940395355,
|
| 471 |
-
"rewards/chosen": -262.32843017578125,
|
| 472 |
-
"rewards/margins": -44.91776657104492,
|
| 473 |
-
"rewards/rejected": -217.41067504882812,
|
| 474 |
-
"selected": 0.0,
|
| 475 |
-
"step": 300
|
| 476 |
},
|
| 477 |
{
|
| 478 |
"epoch": 0.31,
|
| 479 |
-
"
|
| 480 |
-
"
|
| 481 |
-
"
|
| 482 |
-
"
|
| 483 |
-
"
|
| 484 |
-
"
|
| 485 |
-
"
|
| 486 |
-
"
|
| 487 |
-
"
|
| 488 |
-
"
|
| 489 |
-
"
|
| 490 |
-
"eval_steps_per_second": 0.125,
|
| 491 |
-
"step": 300
|
| 492 |
-
},
|
| 493 |
-
{
|
| 494 |
-
"epoch": 0.32,
|
| 495 |
-
"learning_rate": 4.2726091940171055e-06,
|
| 496 |
-
"logps/chosen": -23575.23046875,
|
| 497 |
-
"logps/rejected": -23384.05078125,
|
| 498 |
-
"loss": 0.0186,
|
| 499 |
-
"neglected": 586.0,
|
| 500 |
-
"rewards/accuracies": 0.5062500238418579,
|
| 501 |
-
"rewards/chosen": -233.19082641601562,
|
| 502 |
-
"rewards/margins": -2.1331238746643066,
|
| 503 |
-
"rewards/rejected": -231.0576934814453,
|
| 504 |
-
"selected": 0.0,
|
| 505 |
-
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.33,
|
| 509 |
-
"learning_rate": 4.
|
| 510 |
-
"
|
| 511 |
-
"
|
| 512 |
-
"
|
| 513 |
-
"
|
| 514 |
-
"
|
| 515 |
-
"rewards/
|
| 516 |
-
"rewards/
|
| 517 |
-
"rewards/
|
| 518 |
-
"
|
| 519 |
-
"step":
|
| 520 |
-
},
|
| 521 |
-
{
|
| 522 |
-
"epoch": 0.35,
|
| 523 |
-
"learning_rate": 4.139035550786495e-06,
|
| 524 |
-
"logps/chosen": -23703.537109375,
|
| 525 |
-
"logps/rejected": -19463.728515625,
|
| 526 |
-
"loss": 0.022,
|
| 527 |
-
"neglected": 906.0,
|
| 528 |
-
"rewards/accuracies": 0.4749999940395355,
|
| 529 |
-
"rewards/chosen": -234.44473266601562,
|
| 530 |
-
"rewards/margins": -42.07261657714844,
|
| 531 |
-
"rewards/rejected": -192.3721160888672,
|
| 532 |
-
"selected": 0.0,
|
| 533 |
-
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.36,
|
| 537 |
-
"learning_rate": 4.
|
| 538 |
-
"
|
| 539 |
-
"
|
| 540 |
-
"
|
| 541 |
-
"
|
| 542 |
-
"
|
| 543 |
-
"rewards/
|
| 544 |
-
"rewards/
|
| 545 |
-
"rewards/
|
| 546 |
-
"
|
| 547 |
-
"step":
|
| 548 |
-
},
|
| 549 |
-
{
|
| 550 |
-
"epoch": 0.37,
|
| 551 |
-
"learning_rate": 3.996696580158211e-06,
|
| 552 |
-
"logps/chosen": -30706.958984375,
|
| 553 |
-
"logps/rejected": -28025.75390625,
|
| 554 |
-
"loss": 0.0174,
|
| 555 |
-
"neglected": 1226.0,
|
| 556 |
-
"rewards/accuracies": 0.4124999940395355,
|
| 557 |
-
"rewards/chosen": -304.2236328125,
|
| 558 |
-
"rewards/margins": -26.65300941467285,
|
| 559 |
-
"rewards/rejected": -277.57061767578125,
|
| 560 |
-
"selected": 0.0,
|
| 561 |
-
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.38,
|
| 565 |
-
"learning_rate": 3.
|
| 566 |
-
"
|
| 567 |
-
"
|
| 568 |
-
"
|
| 569 |
-
"
|
| 570 |
-
"
|
| 571 |
-
"rewards/
|
| 572 |
-
"rewards/
|
| 573 |
-
"rewards/
|
| 574 |
-
"
|
| 575 |
-
"step":
|
| 576 |
-
},
|
| 577 |
-
{
|
| 578 |
-
"epoch": 0.39,
|
| 579 |
-
"learning_rate": 3.846353490562664e-06,
|
| 580 |
-
"logps/chosen": -31143.537109375,
|
| 581 |
-
"logps/rejected": -27110.509765625,
|
| 582 |
-
"loss": 0.016,
|
| 583 |
-
"neglected": 1546.0,
|
| 584 |
-
"rewards/accuracies": 0.4625000059604645,
|
| 585 |
-
"rewards/chosen": -308.7967529296875,
|
| 586 |
-
"rewards/margins": -40.30244827270508,
|
| 587 |
-
"rewards/rejected": -268.49432373046875,
|
| 588 |
-
"selected": 0.0,
|
| 589 |
-
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.4,
|
| 593 |
-
"learning_rate": 3.
|
| 594 |
-
"
|
| 595 |
-
"
|
| 596 |
-
"
|
| 597 |
-
"
|
| 598 |
-
"
|
| 599 |
-
"rewards/
|
| 600 |
-
"rewards/
|
| 601 |
-
"rewards/
|
| 602 |
-
"
|
| 603 |
-
"step":
|
| 604 |
-
},
|
| 605 |
-
{
|
| 606 |
-
"epoch": 0.41,
|
| 607 |
-
"learning_rate": 3.6888102953122307e-06,
|
| 608 |
-
"logps/chosen": -30651.95703125,
|
| 609 |
-
"logps/rejected": -26855.337890625,
|
| 610 |
-
"loss": 0.0167,
|
| 611 |
-
"neglected": 1866.0,
|
| 612 |
-
"rewards/accuracies": 0.4312500059604645,
|
| 613 |
-
"rewards/chosen": -303.84075927734375,
|
| 614 |
-
"rewards/margins": -37.63256072998047,
|
| 615 |
-
"rewards/rejected": -266.208251953125,
|
| 616 |
-
"selected": 0.0,
|
| 617 |
-
"step": 390
|
| 618 |
},
|
| 619 |
{
|
| 620 |
"epoch": 0.42,
|
| 621 |
-
"learning_rate": 3.
|
| 622 |
-
"
|
| 623 |
-
"
|
| 624 |
-
"
|
| 625 |
-
"
|
| 626 |
-
"
|
| 627 |
-
"rewards/
|
| 628 |
-
"rewards/
|
| 629 |
-
"rewards/
|
| 630 |
-
"
|
| 631 |
-
"step":
|
| 632 |
},
|
| 633 |
{
|
| 634 |
"epoch": 0.42,
|
| 635 |
-
"
|
| 636 |
-
"
|
| 637 |
-
"
|
| 638 |
-
"
|
| 639 |
-
"
|
| 640 |
-
"eval_rewards/
|
| 641 |
-
"eval_rewards/
|
| 642 |
-
"eval_rewards/
|
| 643 |
-
"
|
| 644 |
-
"
|
| 645 |
-
"
|
| 646 |
-
"eval_steps_per_second": 0.
|
| 647 |
-
"step":
|
| 648 |
-
},
|
| 649 |
-
{
|
| 650 |
-
"epoch": 0.43,
|
| 651 |
-
"learning_rate": 3.5249095128531863e-06,
|
| 652 |
-
"logps/chosen": -35328.84375,
|
| 653 |
-
"logps/rejected": -30211.0,
|
| 654 |
-
"loss": 0.0123,
|
| 655 |
-
"neglected": 586.0,
|
| 656 |
-
"rewards/accuracies": 0.4437499940395355,
|
| 657 |
-
"rewards/chosen": -350.4105529785156,
|
| 658 |
-
"rewards/margins": -50.80678176879883,
|
| 659 |
-
"rewards/rejected": -299.60382080078125,
|
| 660 |
-
"selected": 0.0,
|
| 661 |
-
"step": 410
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.44,
|
| 665 |
-
"learning_rate": 3.
|
| 666 |
-
"
|
| 667 |
-
"
|
| 668 |
-
"
|
| 669 |
-
"
|
| 670 |
-
"
|
| 671 |
-
"rewards/
|
| 672 |
-
"rewards/
|
| 673 |
-
"rewards/
|
| 674 |
-
"
|
| 675 |
-
"step":
|
| 676 |
-
},
|
| 677 |
-
{
|
| 678 |
-
"epoch": 0.45,
|
| 679 |
-
"learning_rate": 3.355527661097728e-06,
|
| 680 |
-
"logps/chosen": -32917.82421875,
|
| 681 |
-
"logps/rejected": -31499.80078125,
|
| 682 |
-
"loss": 0.0179,
|
| 683 |
-
"neglected": 906.0,
|
| 684 |
-
"rewards/accuracies": 0.4749999940395355,
|
| 685 |
-
"rewards/chosen": -326.5057678222656,
|
| 686 |
-
"rewards/margins": -14.0189790725708,
|
| 687 |
-
"rewards/rejected": -312.4867858886719,
|
| 688 |
-
"selected": 0.0,
|
| 689 |
-
"step": 430
|
| 690 |
},
|
| 691 |
{
|
| 692 |
"epoch": 0.46,
|
| 693 |
-
"learning_rate": 3.
|
| 694 |
-
"
|
| 695 |
-
"
|
| 696 |
-
"
|
| 697 |
-
"
|
| 698 |
-
"
|
| 699 |
-
"rewards/
|
| 700 |
-
"rewards/
|
| 701 |
-
"rewards/
|
| 702 |
-
"
|
| 703 |
-
"step":
|
| 704 |
-
},
|
| 705 |
-
{
|
| 706 |
-
"epoch": 0.47,
|
| 707 |
-
"learning_rate": 3.181570569931697e-06,
|
| 708 |
-
"logps/chosen": -28051.92578125,
|
| 709 |
-
"logps/rejected": -26678.359375,
|
| 710 |
-
"loss": 0.025,
|
| 711 |
-
"neglected": 1226.0,
|
| 712 |
-
"rewards/accuracies": 0.41874998807907104,
|
| 713 |
-
"rewards/chosen": -278.347900390625,
|
| 714 |
-
"rewards/margins": -13.726778984069824,
|
| 715 |
-
"rewards/rejected": -264.62115478515625,
|
| 716 |
-
"selected": 0.0,
|
| 717 |
-
"step": 450
|
| 718 |
},
|
| 719 |
{
|
| 720 |
"epoch": 0.48,
|
| 721 |
-
"learning_rate": 3.
|
| 722 |
-
"
|
| 723 |
-
"
|
| 724 |
-
"
|
| 725 |
-
"
|
| 726 |
-
"
|
| 727 |
-
"rewards/
|
| 728 |
-
"rewards/
|
| 729 |
-
"rewards/
|
| 730 |
-
"
|
| 731 |
-
"step":
|
| 732 |
-
},
|
| 733 |
-
{
|
| 734 |
-
"epoch": 0.49,
|
| 735 |
-
"learning_rate": 3.0039685369660785e-06,
|
| 736 |
-
"logps/chosen": -34295.48046875,
|
| 737 |
-
"logps/rejected": -30349.291015625,
|
| 738 |
-
"loss": 0.0128,
|
| 739 |
-
"neglected": 1546.0,
|
| 740 |
-
"rewards/accuracies": 0.48124998807907104,
|
| 741 |
-
"rewards/chosen": -340.40753173828125,
|
| 742 |
-
"rewards/margins": -39.258506774902344,
|
| 743 |
-
"rewards/rejected": -301.1490173339844,
|
| 744 |
-
"selected": 0.0,
|
| 745 |
-
"step": 470
|
| 746 |
},
|
| 747 |
{
|
| 748 |
"epoch": 0.5,
|
| 749 |
-
"learning_rate": 2.
|
| 750 |
-
"
|
| 751 |
-
"
|
| 752 |
-
"
|
| 753 |
-
"
|
| 754 |
-
"
|
| 755 |
-
"rewards/
|
| 756 |
-
"rewards/
|
| 757 |
-
"rewards/
|
| 758 |
-
"
|
| 759 |
-
"step":
|
| 760 |
-
},
|
| 761 |
-
{
|
| 762 |
-
"epoch": 0.51,
|
| 763 |
-
"learning_rate": 2.8236713524386085e-06,
|
| 764 |
-
"logps/chosen": -32674.708984375,
|
| 765 |
-
"logps/rejected": -27154.64453125,
|
| 766 |
-
"loss": 0.0173,
|
| 767 |
-
"neglected": 1866.0,
|
| 768 |
-
"rewards/accuracies": 0.42500001192092896,
|
| 769 |
-
"rewards/chosen": -324.16522216796875,
|
| 770 |
-
"rewards/margins": -54.92888641357422,
|
| 771 |
-
"rewards/rejected": -269.236328125,
|
| 772 |
-
"selected": 0.0,
|
| 773 |
-
"step": 490
|
| 774 |
-
},
|
| 775 |
-
{
|
| 776 |
-
"epoch": 0.52,
|
| 777 |
-
"learning_rate": 2.7328129695107205e-06,
|
| 778 |
-
"logps/chosen": -33942.0390625,
|
| 779 |
-
"logps/rejected": -26798.068359375,
|
| 780 |
-
"loss": 0.0182,
|
| 781 |
-
"neglected": 2026.0,
|
| 782 |
-
"rewards/accuracies": 0.39375001192092896,
|
| 783 |
-
"rewards/chosen": -336.67034912109375,
|
| 784 |
-
"rewards/margins": -71.16191864013672,
|
| 785 |
-
"rewards/rejected": -265.5083923339844,
|
| 786 |
-
"selected": 0.0,
|
| 787 |
-
"step": 500
|
| 788 |
},
|
| 789 |
{
|
| 790 |
"epoch": 0.52,
|
| 791 |
-
"
|
| 792 |
-
"
|
| 793 |
-
"
|
| 794 |
-
"
|
| 795 |
-
"
|
| 796 |
-
"
|
| 797 |
-
"
|
| 798 |
-
"
|
| 799 |
-
"
|
| 800 |
-
"
|
| 801 |
-
"
|
| 802 |
-
"eval_steps_per_second": 0.125,
|
| 803 |
-
"step": 500
|
| 804 |
-
},
|
| 805 |
-
{
|
| 806 |
-
"epoch": 0.53,
|
| 807 |
-
"learning_rate": 2.641643219871597e-06,
|
| 808 |
-
"logps/chosen": -31474.45703125,
|
| 809 |
-
"logps/rejected": -25789.703125,
|
| 810 |
-
"loss": 0.0149,
|
| 811 |
-
"neglected": 586.0,
|
| 812 |
-
"rewards/accuracies": 0.46875,
|
| 813 |
-
"rewards/chosen": -312.119384765625,
|
| 814 |
-
"rewards/margins": -56.47794723510742,
|
| 815 |
-
"rewards/rejected": -255.6414337158203,
|
| 816 |
-
"selected": 0.0,
|
| 817 |
-
"step": 510
|
| 818 |
},
|
| 819 |
{
|
| 820 |
"epoch": 0.54,
|
| 821 |
-
"learning_rate": 2.
|
| 822 |
-
"
|
| 823 |
-
"
|
| 824 |
-
"
|
| 825 |
-
"
|
| 826 |
-
"
|
| 827 |
-
"rewards/
|
| 828 |
-
"rewards/
|
| 829 |
-
"rewards/
|
| 830 |
-
"
|
| 831 |
-
"step":
|
| 832 |
-
},
|
| 833 |
-
{
|
| 834 |
-
"epoch": 0.55,
|
| 835 |
-
"learning_rate": 2.4588575996495797e-06,
|
| 836 |
-
"logps/chosen": -34159.5703125,
|
| 837 |
-
"logps/rejected": -30448.79296875,
|
| 838 |
-
"loss": 0.0138,
|
| 839 |
-
"neglected": 906.0,
|
| 840 |
-
"rewards/accuracies": 0.40625,
|
| 841 |
-
"rewards/chosen": -338.7040710449219,
|
| 842 |
-
"rewards/margins": -37.0126953125,
|
| 843 |
-
"rewards/rejected": -301.69134521484375,
|
| 844 |
-
"selected": 0.0,
|
| 845 |
-
"step": 530
|
| 846 |
},
|
| 847 |
{
|
| 848 |
"epoch": 0.57,
|
| 849 |
-
"learning_rate": 2.
|
| 850 |
-
"
|
| 851 |
-
"
|
| 852 |
-
"
|
| 853 |
-
"
|
| 854 |
-
"
|
| 855 |
-
"rewards/
|
| 856 |
-
"rewards/
|
| 857 |
-
"rewards/
|
| 858 |
-
"
|
| 859 |
-
"step":
|
| 860 |
-
},
|
| 861 |
-
{
|
| 862 |
-
"epoch": 0.58,
|
| 863 |
-
"learning_rate": 2.276292003092593e-06,
|
| 864 |
-
"logps/chosen": -33080.16796875,
|
| 865 |
-
"logps/rejected": -29123.927734375,
|
| 866 |
-
"loss": 0.0145,
|
| 867 |
-
"neglected": 1226.0,
|
| 868 |
-
"rewards/accuracies": 0.4312500059604645,
|
| 869 |
-
"rewards/chosen": -328.01007080078125,
|
| 870 |
-
"rewards/margins": -39.278717041015625,
|
| 871 |
-
"rewards/rejected": -288.7312927246094,
|
| 872 |
-
"selected": 0.0,
|
| 873 |
-
"step": 550
|
| 874 |
},
|
| 875 |
{
|
| 876 |
"epoch": 0.59,
|
| 877 |
-
"learning_rate": 2.
|
| 878 |
-
"
|
| 879 |
-
"
|
| 880 |
-
"
|
| 881 |
-
"
|
| 882 |
-
"
|
| 883 |
-
"rewards/
|
| 884 |
-
"rewards/
|
| 885 |
-
"rewards/
|
| 886 |
-
"
|
| 887 |
-
"step":
|
| 888 |
-
},
|
| 889 |
-
{
|
| 890 |
-
"epoch": 0.6,
|
| 891 |
-
"learning_rate": 2.0949227648656194e-06,
|
| 892 |
-
"logps/chosen": -34481.3203125,
|
| 893 |
-
"logps/rejected": -30493.02734375,
|
| 894 |
-
"loss": 0.0172,
|
| 895 |
-
"neglected": 1546.0,
|
| 896 |
-
"rewards/accuracies": 0.4312500059604645,
|
| 897 |
-
"rewards/chosen": -342.2419738769531,
|
| 898 |
-
"rewards/margins": -39.841697692871094,
|
| 899 |
-
"rewards/rejected": -302.4002685546875,
|
| 900 |
-
"selected": 0.0,
|
| 901 |
-
"step": 570
|
| 902 |
},
|
| 903 |
{
|
| 904 |
"epoch": 0.61,
|
| 905 |
-
"learning_rate":
|
| 906 |
-
"
|
| 907 |
-
"
|
| 908 |
-
"
|
| 909 |
-
"
|
| 910 |
-
"
|
| 911 |
-
"rewards/
|
| 912 |
-
"rewards/
|
| 913 |
-
"rewards/
|
| 914 |
-
"
|
| 915 |
-
"step":
|
| 916 |
-
},
|
| 917 |
-
{
|
| 918 |
-
"epoch": 0.62,
|
| 919 |
-
"learning_rate": 1.915719821680624e-06,
|
| 920 |
-
"logps/chosen": -32696.44921875,
|
| 921 |
-
"logps/rejected": -30430.255859375,
|
| 922 |
-
"loss": 0.0167,
|
| 923 |
-
"neglected": 1866.0,
|
| 924 |
-
"rewards/accuracies": 0.45625001192092896,
|
| 925 |
-
"rewards/chosen": -324.4429626464844,
|
| 926 |
-
"rewards/margins": -22.577922821044922,
|
| 927 |
-
"rewards/rejected": -301.8650207519531,
|
| 928 |
-
"selected": 0.0,
|
| 929 |
-
"step": 590
|
| 930 |
},
|
| 931 |
{
|
| 932 |
"epoch": 0.63,
|
| 933 |
-
"learning_rate": 1.
|
| 934 |
-
"
|
| 935 |
-
"
|
| 936 |
-
"
|
| 937 |
-
"
|
| 938 |
-
"
|
| 939 |
-
"rewards/
|
| 940 |
-
"rewards/
|
| 941 |
-
"rewards/
|
| 942 |
-
"
|
| 943 |
-
"step":
|
| 944 |
},
|
| 945 |
{
|
| 946 |
"epoch": 0.63,
|
| 947 |
-
"
|
| 948 |
-
"
|
| 949 |
-
"
|
| 950 |
-
"
|
| 951 |
-
"
|
| 952 |
-
"eval_rewards/
|
| 953 |
-
"eval_rewards/
|
| 954 |
-
"eval_rewards/
|
| 955 |
-
"
|
| 956 |
-
"
|
| 957 |
-
"
|
| 958 |
-
"eval_steps_per_second": 0.
|
| 959 |
-
"step":
|
| 960 |
-
},
|
| 961 |
-
{
|
| 962 |
-
"epoch": 0.64,
|
| 963 |
-
"learning_rate": 1.739641525213929e-06,
|
| 964 |
-
"logps/chosen": -36224.6953125,
|
| 965 |
-
"logps/rejected": -28970.59765625,
|
| 966 |
-
"loss": 0.0178,
|
| 967 |
-
"neglected": 586.0,
|
| 968 |
-
"rewards/accuracies": 0.4124999940395355,
|
| 969 |
-
"rewards/chosen": -359.3634948730469,
|
| 970 |
-
"rewards/margins": -71.96965789794922,
|
| 971 |
-
"rewards/rejected": -287.39385986328125,
|
| 972 |
-
"selected": 0.0,
|
| 973 |
-
"step": 610
|
| 974 |
},
|
| 975 |
{
|
| 976 |
"epoch": 0.65,
|
| 977 |
-
"learning_rate": 1.
|
| 978 |
-
"
|
| 979 |
-
"
|
| 980 |
-
"
|
| 981 |
-
"
|
| 982 |
-
"
|
| 983 |
-
"rewards/
|
| 984 |
-
"rewards/
|
| 985 |
-
"rewards/
|
| 986 |
-
"
|
| 987 |
-
"step":
|
| 988 |
-
},
|
| 989 |
-
{
|
| 990 |
-
"epoch": 0.66,
|
| 991 |
-
"learning_rate": 1.5676295169786864e-06,
|
| 992 |
-
"logps/chosen": -33129.5703125,
|
| 993 |
-
"logps/rejected": -29490.744140625,
|
| 994 |
-
"loss": 0.0122,
|
| 995 |
-
"neglected": 906.0,
|
| 996 |
-
"rewards/accuracies": 0.4437499940395355,
|
| 997 |
-
"rewards/chosen": -328.56658935546875,
|
| 998 |
-
"rewards/margins": -36.11338806152344,
|
| 999 |
-
"rewards/rejected": -292.45318603515625,
|
| 1000 |
-
"selected": 0.0,
|
| 1001 |
-
"step": 630
|
| 1002 |
},
|
| 1003 |
{
|
| 1004 |
"epoch": 0.67,
|
| 1005 |
-
"learning_rate": 1.
|
| 1006 |
-
"
|
| 1007 |
-
"
|
| 1008 |
-
"
|
| 1009 |
-
"
|
| 1010 |
-
"
|
| 1011 |
-
"rewards/
|
| 1012 |
-
"rewards/
|
| 1013 |
-
"rewards/
|
| 1014 |
-
"
|
| 1015 |
-
"step":
|
| 1016 |
-
},
|
| 1017 |
-
{
|
| 1018 |
-
"epoch": 0.68,
|
| 1019 |
-
"learning_rate": 1.4006036925609245e-06,
|
| 1020 |
-
"logps/chosen": -37346.015625,
|
| 1021 |
-
"logps/rejected": -32676.51171875,
|
| 1022 |
-
"loss": 0.0115,
|
| 1023 |
-
"neglected": 1226.0,
|
| 1024 |
-
"rewards/accuracies": 0.4124999940395355,
|
| 1025 |
-
"rewards/chosen": -370.58929443359375,
|
| 1026 |
-
"rewards/margins": -46.472129821777344,
|
| 1027 |
-
"rewards/rejected": -324.11724853515625,
|
| 1028 |
-
"selected": 0.0,
|
| 1029 |
-
"step": 650
|
| 1030 |
},
|
| 1031 |
{
|
| 1032 |
"epoch": 0.69,
|
| 1033 |
-
"learning_rate": 1.
|
| 1034 |
-
"
|
| 1035 |
-
"
|
| 1036 |
-
"
|
| 1037 |
-
"
|
| 1038 |
-
"
|
| 1039 |
-
"rewards/
|
| 1040 |
-
"rewards/
|
| 1041 |
-
"rewards/
|
| 1042 |
-
"
|
| 1043 |
-
"step":
|
| 1044 |
-
},
|
| 1045 |
-
{
|
| 1046 |
-
"epoch": 0.7,
|
| 1047 |
-
"learning_rate": 1.2394572821496953e-06,
|
| 1048 |
-
"logps/chosen": -34126.65234375,
|
| 1049 |
-
"logps/rejected": -28228.837890625,
|
| 1050 |
-
"loss": 0.0213,
|
| 1051 |
-
"neglected": 1546.0,
|
| 1052 |
-
"rewards/accuracies": 0.4375,
|
| 1053 |
-
"rewards/chosen": -338.5702209472656,
|
| 1054 |
-
"rewards/margins": -58.62493896484375,
|
| 1055 |
-
"rewards/rejected": -279.9453125,
|
| 1056 |
-
"selected": 0.0,
|
| 1057 |
-
"step": 670
|
| 1058 |
},
|
| 1059 |
{
|
| 1060 |
"epoch": 0.71,
|
| 1061 |
-
"learning_rate": 1.
|
| 1062 |
-
"
|
| 1063 |
-
"
|
| 1064 |
-
"
|
| 1065 |
-
"
|
| 1066 |
-
"
|
| 1067 |
-
"rewards/
|
| 1068 |
-
"rewards/
|
| 1069 |
-
"rewards/
|
| 1070 |
-
"
|
| 1071 |
-
"step":
|
| 1072 |
-
},
|
| 1073 |
-
{
|
| 1074 |
-
"epoch": 0.72,
|
| 1075 |
-
"learning_rate": 1.0850520736699362e-06,
|
| 1076 |
-
"logps/chosen": -30460.240234375,
|
| 1077 |
-
"logps/rejected": -24525.88671875,
|
| 1078 |
-
"loss": 0.0166,
|
| 1079 |
-
"neglected": 1866.0,
|
| 1080 |
-
"rewards/accuracies": 0.38749998807907104,
|
| 1081 |
-
"rewards/chosen": -302.1998291015625,
|
| 1082 |
-
"rewards/margins": -58.95698165893555,
|
| 1083 |
-
"rewards/rejected": -243.2428436279297,
|
| 1084 |
-
"selected": 0.0,
|
| 1085 |
-
"step": 690
|
| 1086 |
-
},
|
| 1087 |
-
{
|
| 1088 |
-
"epoch": 0.73,
|
| 1089 |
-
"learning_rate": 1.0106369933615043e-06,
|
| 1090 |
-
"logps/chosen": -34674.8515625,
|
| 1091 |
-
"logps/rejected": -27218.890625,
|
| 1092 |
-
"loss": 0.0165,
|
| 1093 |
-
"neglected": 2026.0,
|
| 1094 |
-
"rewards/accuracies": 0.40625,
|
| 1095 |
-
"rewards/chosen": -344.1193542480469,
|
| 1096 |
-
"rewards/margins": -74.25706481933594,
|
| 1097 |
-
"rewards/rejected": -269.8622741699219,
|
| 1098 |
-
"selected": 0.0,
|
| 1099 |
-
"step": 700
|
| 1100 |
},
|
| 1101 |
{
|
| 1102 |
"epoch": 0.73,
|
| 1103 |
-
"
|
| 1104 |
-
"
|
| 1105 |
-
"
|
| 1106 |
-
"
|
| 1107 |
-
"
|
| 1108 |
-
"
|
| 1109 |
-
"
|
| 1110 |
-
"
|
| 1111 |
-
"
|
| 1112 |
-
"
|
| 1113 |
-
"
|
| 1114 |
-
"eval_steps_per_second": 0.125,
|
| 1115 |
-
"step": 700
|
| 1116 |
-
},
|
| 1117 |
-
{
|
| 1118 |
-
"epoch": 0.74,
|
| 1119 |
-
"learning_rate": 9.382138040640714e-07,
|
| 1120 |
-
"logps/chosen": -35340.9296875,
|
| 1121 |
-
"logps/rejected": -29694.728515625,
|
| 1122 |
-
"loss": 0.0173,
|
| 1123 |
-
"neglected": 586.0,
|
| 1124 |
-
"rewards/accuracies": 0.41874998807907104,
|
| 1125 |
-
"rewards/chosen": -350.610595703125,
|
| 1126 |
-
"rewards/margins": -55.991127014160156,
|
| 1127 |
-
"rewards/rejected": -294.61944580078125,
|
| 1128 |
-
"selected": 0.0,
|
| 1129 |
-
"step": 710
|
| 1130 |
},
|
| 1131 |
{
|
| 1132 |
"epoch": 0.75,
|
| 1133 |
-
"learning_rate": 8.
|
| 1134 |
-
"
|
| 1135 |
-
"
|
| 1136 |
-
"
|
| 1137 |
-
"
|
| 1138 |
-
"
|
| 1139 |
-
"rewards/
|
| 1140 |
-
"rewards/
|
| 1141 |
-
"rewards/
|
| 1142 |
-
"
|
| 1143 |
-
"step":
|
| 1144 |
-
},
|
| 1145 |
-
{
|
| 1146 |
-
"epoch": 0.76,
|
| 1147 |
-
"learning_rate": 7.997277433690984e-07,
|
| 1148 |
-
"logps/chosen": -33704.45703125,
|
| 1149 |
-
"logps/rejected": -27931.458984375,
|
| 1150 |
-
"loss": 0.0181,
|
| 1151 |
-
"neglected": 906.0,
|
| 1152 |
-
"rewards/accuracies": 0.36250001192092896,
|
| 1153 |
-
"rewards/chosen": -334.1654052734375,
|
| 1154 |
-
"rewards/margins": -57.19194793701172,
|
| 1155 |
-
"rewards/rejected": -276.97344970703125,
|
| 1156 |
-
"selected": 0.0,
|
| 1157 |
-
"step": 730
|
| 1158 |
},
|
| 1159 |
{
|
| 1160 |
"epoch": 0.77,
|
| 1161 |
-
"learning_rate": 7.
|
| 1162 |
-
"
|
| 1163 |
-
"
|
| 1164 |
-
"
|
| 1165 |
-
"
|
| 1166 |
-
"
|
| 1167 |
-
"rewards/
|
| 1168 |
-
"rewards/
|
| 1169 |
-
"rewards/
|
| 1170 |
-
"
|
| 1171 |
-
"step":
|
| 1172 |
-
},
|
| 1173 |
-
{
|
| 1174 |
-
"epoch": 0.79,
|
| 1175 |
-
"learning_rate": 6.70334495204884e-07,
|
| 1176 |
-
"logps/chosen": -35638.7421875,
|
| 1177 |
-
"logps/rejected": -31301.58984375,
|
| 1178 |
-
"loss": 0.0156,
|
| 1179 |
-
"neglected": 1226.0,
|
| 1180 |
-
"rewards/accuracies": 0.39375001192092896,
|
| 1181 |
-
"rewards/chosen": -353.8833312988281,
|
| 1182 |
-
"rewards/margins": -43.371212005615234,
|
| 1183 |
-
"rewards/rejected": -310.51214599609375,
|
| 1184 |
-
"selected": 0.0,
|
| 1185 |
-
"step": 750
|
| 1186 |
},
|
| 1187 |
{
|
| 1188 |
"epoch": 0.8,
|
| 1189 |
-
"learning_rate": 6.
|
| 1190 |
-
"
|
| 1191 |
-
"
|
| 1192 |
-
"
|
| 1193 |
-
"
|
| 1194 |
-
"
|
| 1195 |
-
"rewards/
|
| 1196 |
-
"rewards/
|
| 1197 |
-
"rewards/
|
| 1198 |
-
"
|
| 1199 |
-
"step":
|
| 1200 |
-
},
|
| 1201 |
-
{
|
| 1202 |
-
"epoch": 0.81,
|
| 1203 |
-
"learning_rate": 5.507260361320738e-07,
|
| 1204 |
-
"logps/chosen": -36556.7890625,
|
| 1205 |
-
"logps/rejected": -33627.23828125,
|
| 1206 |
-
"loss": 0.0106,
|
| 1207 |
-
"neglected": 1546.0,
|
| 1208 |
-
"rewards/accuracies": 0.4312500059604645,
|
| 1209 |
-
"rewards/chosen": -362.6523742675781,
|
| 1210 |
-
"rewards/margins": -29.13754653930664,
|
| 1211 |
-
"rewards/rejected": -333.51483154296875,
|
| 1212 |
-
"selected": 0.0,
|
| 1213 |
-
"step": 770
|
| 1214 |
},
|
| 1215 |
{
|
| 1216 |
"epoch": 0.82,
|
| 1217 |
-
"learning_rate": 4.
|
| 1218 |
-
"
|
| 1219 |
-
"
|
| 1220 |
-
"
|
| 1221 |
-
"
|
| 1222 |
-
"
|
| 1223 |
-
"rewards/
|
| 1224 |
-
"rewards/
|
| 1225 |
-
"rewards/
|
| 1226 |
-
"
|
| 1227 |
-
"step":
|
| 1228 |
-
},
|
| 1229 |
-
{
|
| 1230 |
-
"epoch": 0.83,
|
| 1231 |
-
"learning_rate": 4.4154201506053985e-07,
|
| 1232 |
-
"logps/chosen": -35076.3515625,
|
| 1233 |
-
"logps/rejected": -33320.109375,
|
| 1234 |
-
"loss": 0.0146,
|
| 1235 |
-
"neglected": 1866.0,
|
| 1236 |
-
"rewards/accuracies": 0.45625001192092896,
|
| 1237 |
-
"rewards/chosen": -348.1772155761719,
|
| 1238 |
-
"rewards/margins": -17.483577728271484,
|
| 1239 |
-
"rewards/rejected": -330.693603515625,
|
| 1240 |
-
"selected": 0.0,
|
| 1241 |
-
"step": 790
|
| 1242 |
},
|
| 1243 |
{
|
| 1244 |
"epoch": 0.84,
|
| 1245 |
-
"learning_rate": 3.
|
| 1246 |
-
"
|
| 1247 |
-
"
|
| 1248 |
-
"
|
| 1249 |
-
"
|
| 1250 |
-
"
|
| 1251 |
-
"rewards/
|
| 1252 |
-
"rewards/
|
| 1253 |
-
"rewards/
|
| 1254 |
-
"
|
| 1255 |
-
"step":
|
| 1256 |
},
|
| 1257 |
{
|
| 1258 |
"epoch": 0.84,
|
| 1259 |
-
"
|
| 1260 |
-
"
|
| 1261 |
-
"
|
| 1262 |
-
"
|
| 1263 |
-
"
|
| 1264 |
-
"eval_rewards/
|
| 1265 |
-
"eval_rewards/
|
| 1266 |
-
"eval_rewards/
|
| 1267 |
-
"
|
| 1268 |
-
"
|
| 1269 |
-
"
|
| 1270 |
-
"eval_steps_per_second": 0.
|
| 1271 |
-
"step":
|
| 1272 |
-
},
|
| 1273 |
-
{
|
| 1274 |
-
"epoch": 0.85,
|
| 1275 |
-
"learning_rate": 3.4336633249862084e-07,
|
| 1276 |
-
"logps/chosen": -34459.21484375,
|
| 1277 |
-
"logps/rejected": -27397.837890625,
|
| 1278 |
-
"loss": 0.0064,
|
| 1279 |
-
"neglected": 586.0,
|
| 1280 |
-
"rewards/accuracies": 0.375,
|
| 1281 |
-
"rewards/chosen": -341.8682556152344,
|
| 1282 |
-
"rewards/margins": -70.08294677734375,
|
| 1283 |
-
"rewards/rejected": -271.78533935546875,
|
| 1284 |
-
"selected": 0.0,
|
| 1285 |
-
"step": 810
|
| 1286 |
},
|
| 1287 |
{
|
| 1288 |
"epoch": 0.86,
|
| 1289 |
-
"learning_rate": 2.
|
| 1290 |
-
"
|
| 1291 |
-
"
|
| 1292 |
-
"
|
| 1293 |
-
"
|
| 1294 |
-
"
|
| 1295 |
-
"rewards/
|
| 1296 |
-
"rewards/
|
| 1297 |
-
"rewards/
|
| 1298 |
-
"
|
| 1299 |
-
"step":
|
| 1300 |
-
},
|
| 1301 |
-
{
|
| 1302 |
-
"epoch": 0.87,
|
| 1303 |
-
"learning_rate": 2.5672401793681854e-07,
|
| 1304 |
-
"logps/chosen": -35177.1796875,
|
| 1305 |
-
"logps/rejected": -35886.7734375,
|
| 1306 |
-
"loss": 0.0127,
|
| 1307 |
-
"neglected": 906.0,
|
| 1308 |
-
"rewards/accuracies": 0.550000011920929,
|
| 1309 |
-
"rewards/chosen": -349.24237060546875,
|
| 1310 |
-
"rewards/margins": 6.9106035232543945,
|
| 1311 |
-
"rewards/rejected": -356.1529235839844,
|
| 1312 |
-
"selected": 0.0,
|
| 1313 |
-
"step": 830
|
| 1314 |
},
|
| 1315 |
{
|
| 1316 |
"epoch": 0.88,
|
| 1317 |
-
"learning_rate": 2.
|
| 1318 |
-
"
|
| 1319 |
-
"
|
| 1320 |
-
"
|
| 1321 |
-
"
|
| 1322 |
-
"
|
| 1323 |
-
"rewards/
|
| 1324 |
-
"rewards/
|
| 1325 |
-
"rewards/
|
| 1326 |
-
"
|
| 1327 |
-
"step":
|
| 1328 |
-
},
|
| 1329 |
-
{
|
| 1330 |
-
"epoch": 0.89,
|
| 1331 |
-
"learning_rate": 1.820784220652766e-07,
|
| 1332 |
-
"logps/chosen": -36996.6171875,
|
| 1333 |
-
"logps/rejected": -31543.29296875,
|
| 1334 |
-
"loss": 0.0169,
|
| 1335 |
-
"neglected": 1226.0,
|
| 1336 |
-
"rewards/accuracies": 0.40625,
|
| 1337 |
-
"rewards/chosen": -367.1126708984375,
|
| 1338 |
-
"rewards/margins": -54.04352951049805,
|
| 1339 |
-
"rewards/rejected": -313.0691833496094,
|
| 1340 |
-
"selected": 0.0,
|
| 1341 |
-
"step": 850
|
| 1342 |
},
|
| 1343 |
{
|
| 1344 |
"epoch": 0.9,
|
| 1345 |
-
"learning_rate": 1.
|
| 1346 |
-
"
|
| 1347 |
-
"
|
| 1348 |
-
"
|
| 1349 |
-
"
|
| 1350 |
-
"
|
| 1351 |
-
"rewards/
|
| 1352 |
-
"rewards/
|
| 1353 |
-
"rewards/
|
| 1354 |
-
"
|
| 1355 |
-
"step":
|
| 1356 |
-
},
|
| 1357 |
-
{
|
| 1358 |
-
"epoch": 0.91,
|
| 1359 |
-
"learning_rate": 1.1982873884064466e-07,
|
| 1360 |
-
"logps/chosen": -31830.9375,
|
| 1361 |
-
"logps/rejected": -30581.083984375,
|
| 1362 |
-
"loss": 0.0158,
|
| 1363 |
-
"neglected": 1546.0,
|
| 1364 |
-
"rewards/accuracies": 0.543749988079071,
|
| 1365 |
-
"rewards/chosen": -315.9645690917969,
|
| 1366 |
-
"rewards/margins": -12.488165855407715,
|
| 1367 |
-
"rewards/rejected": -303.4764404296875,
|
| 1368 |
-
"selected": 0.0,
|
| 1369 |
-
"step": 870
|
| 1370 |
},
|
| 1371 |
{
|
| 1372 |
"epoch": 0.92,
|
| 1373 |
-
"learning_rate": 9.
|
| 1374 |
-
"
|
| 1375 |
-
"
|
| 1376 |
-
"
|
| 1377 |
-
"
|
| 1378 |
-
"
|
| 1379 |
-
"rewards/
|
| 1380 |
-
"rewards/
|
| 1381 |
-
"rewards/
|
| 1382 |
-
"
|
| 1383 |
-
"step":
|
| 1384 |
-
},
|
| 1385 |
-
{
|
| 1386 |
-
"epoch": 0.93,
|
| 1387 |
-
"learning_rate": 7.030787065396866e-08,
|
| 1388 |
-
"logps/chosen": -39424.8828125,
|
| 1389 |
-
"logps/rejected": -34691.7265625,
|
| 1390 |
-
"loss": 0.0135,
|
| 1391 |
-
"neglected": 1866.0,
|
| 1392 |
-
"rewards/accuracies": 0.4312500059604645,
|
| 1393 |
-
"rewards/chosen": -391.5462341308594,
|
| 1394 |
-
"rewards/margins": -47.24760437011719,
|
| 1395 |
-
"rewards/rejected": -344.29864501953125,
|
| 1396 |
-
"selected": 0.0,
|
| 1397 |
-
"step": 890
|
| 1398 |
-
},
|
| 1399 |
-
{
|
| 1400 |
-
"epoch": 0.94,
|
| 1401 |
-
"learning_rate": 5.0406202043228604e-08,
|
| 1402 |
-
"logps/chosen": -36448.4765625,
|
| 1403 |
-
"logps/rejected": -30919.54296875,
|
| 1404 |
-
"loss": 0.0195,
|
| 1405 |
-
"neglected": 2026.0,
|
| 1406 |
-
"rewards/accuracies": 0.44999998807907104,
|
| 1407 |
-
"rewards/chosen": -361.8363037109375,
|
| 1408 |
-
"rewards/margins": -55.14497756958008,
|
| 1409 |
-
"rewards/rejected": -306.69134521484375,
|
| 1410 |
-
"selected": 0.0,
|
| 1411 |
-
"step": 900
|
| 1412 |
},
|
| 1413 |
{
|
| 1414 |
"epoch": 0.94,
|
| 1415 |
-
"
|
| 1416 |
-
"
|
| 1417 |
-
"
|
| 1418 |
-
"
|
| 1419 |
-
"
|
| 1420 |
-
"
|
| 1421 |
-
"
|
| 1422 |
-
"
|
| 1423 |
-
"
|
| 1424 |
-
"
|
| 1425 |
-
"
|
| 1426 |
-
"eval_steps_per_second": 0.065,
|
| 1427 |
-
"step": 900
|
| 1428 |
-
},
|
| 1429 |
-
{
|
| 1430 |
-
"epoch": 0.95,
|
| 1431 |
-
"learning_rate": 3.378064801637687e-08,
|
| 1432 |
-
"logps/chosen": -37312.3515625,
|
| 1433 |
-
"logps/rejected": -31717.978515625,
|
| 1434 |
-
"loss": 0.0123,
|
| 1435 |
-
"neglected": 586.0,
|
| 1436 |
-
"rewards/accuracies": 0.4375,
|
| 1437 |
-
"rewards/chosen": -370.2989196777344,
|
| 1438 |
-
"rewards/margins": -55.649620056152344,
|
| 1439 |
-
"rewards/rejected": -314.6493225097656,
|
| 1440 |
-
"selected": 0.0,
|
| 1441 |
-
"step": 910
|
| 1442 |
},
|
| 1443 |
{
|
| 1444 |
"epoch": 0.96,
|
| 1445 |
-
"learning_rate":
|
| 1446 |
-
"
|
| 1447 |
-
"
|
| 1448 |
-
"
|
| 1449 |
-
"
|
| 1450 |
-
"
|
| 1451 |
-
"rewards/
|
| 1452 |
-
"rewards/
|
| 1453 |
-
"rewards/
|
| 1454 |
-
"
|
| 1455 |
-
"step":
|
| 1456 |
-
},
|
| 1457 |
-
{
|
| 1458 |
-
"epoch": 0.97,
|
| 1459 |
-
"learning_rate": 1.0442413283435759e-08,
|
| 1460 |
-
"logps/chosen": -37030.9140625,
|
| 1461 |
-
"logps/rejected": -29036.62109375,
|
| 1462 |
-
"loss": 0.0174,
|
| 1463 |
-
"neglected": 906.0,
|
| 1464 |
-
"rewards/accuracies": 0.45625001192092896,
|
| 1465 |
-
"rewards/chosen": -367.7157897949219,
|
| 1466 |
-
"rewards/margins": -79.69428253173828,
|
| 1467 |
-
"rewards/rejected": -288.02154541015625,
|
| 1468 |
-
"selected": 0.0,
|
| 1469 |
-
"step": 930
|
| 1470 |
},
|
| 1471 |
{
|
| 1472 |
"epoch": 0.98,
|
| 1473 |
-
"learning_rate": 3.
|
| 1474 |
-
"
|
| 1475 |
-
"
|
| 1476 |
-
"
|
| 1477 |
-
"
|
| 1478 |
-
"
|
| 1479 |
-
"rewards/
|
| 1480 |
-
"rewards/
|
| 1481 |
-
"rewards/
|
| 1482 |
-
"
|
| 1483 |
-
"step":
|
| 1484 |
-
},
|
| 1485 |
-
{
|
| 1486 |
-
"epoch": 0.99,
|
| 1487 |
-
"learning_rate": 4.1797599220405605e-10,
|
| 1488 |
-
"logps/chosen": -34646.1640625,
|
| 1489 |
-
"logps/rejected": -31544.458984375,
|
| 1490 |
-
"loss": 0.0168,
|
| 1491 |
-
"neglected": 1226.0,
|
| 1492 |
-
"rewards/accuracies": 0.48750001192092896,
|
| 1493 |
-
"rewards/chosen": -343.90411376953125,
|
| 1494 |
-
"rewards/margins": -30.896869659423828,
|
| 1495 |
-
"rewards/rejected": -313.00726318359375,
|
| 1496 |
-
"selected": 0.0,
|
| 1497 |
-
"step": 950
|
| 1498 |
},
|
| 1499 |
{
|
| 1500 |
"epoch": 1.0,
|
| 1501 |
-
"step":
|
| 1502 |
"total_flos": 0.0,
|
| 1503 |
-
"train_loss":
|
| 1504 |
-
"train_runtime":
|
| 1505 |
-
"train_samples_per_second": 1.
|
| 1506 |
-
"train_steps_per_second": 0.
|
| 1507 |
}
|
| 1508 |
],
|
| 1509 |
"logging_steps": 10,
|
| 1510 |
-
"max_steps":
|
|
|
|
| 1511 |
"num_train_epochs": 1,
|
| 1512 |
"save_steps": 100,
|
| 1513 |
"total_flos": 0.0,
|
|
|
|
| 1514 |
"trial_name": null,
|
| 1515 |
"trial_params": null
|
| 1516 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.998691442030882,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 477,
|
| 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": 1.0416666666666667e-07,
|
| 14 |
+
"logits/chosen": -2.856400966644287,
|
| 15 |
+
"logits/rejected": -2.6539194583892822,
|
| 16 |
+
"logps/chosen": -302.289794921875,
|
| 17 |
+
"logps/rejected": -253.04373168945312,
|
| 18 |
+
"loss": 2500.0,
|
| 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": 0.02,
|
| 27 |
"learning_rate": 1.0416666666666667e-06,
|
| 28 |
+
"logits/chosen": -2.5851330757141113,
|
| 29 |
+
"logits/rejected": -2.6188478469848633,
|
| 30 |
+
"logps/chosen": -265.6952209472656,
|
| 31 |
+
"logps/rejected": -261.4213562011719,
|
| 32 |
+
"loss": 2495.385,
|
| 33 |
+
"rewards/accuracies": 0.4375,
|
| 34 |
+
"rewards/chosen": 0.005977082531899214,
|
| 35 |
+
"rewards/margins": 0.0005994850071147084,
|
| 36 |
+
"rewards/rejected": 0.005377596709877253,
|
| 37 |
+
"step": 10
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
},
|
| 39 |
{
|
| 40 |
"epoch": 0.04,
|
| 41 |
"learning_rate": 2.0833333333333334e-06,
|
| 42 |
+
"logits/chosen": -2.6101512908935547,
|
| 43 |
+
"logits/rejected": -2.5939109325408936,
|
| 44 |
+
"logps/chosen": -255.68185424804688,
|
| 45 |
+
"logps/rejected": -248.1254119873047,
|
| 46 |
+
"loss": 2457.86,
|
| 47 |
+
"rewards/accuracies": 0.628125011920929,
|
| 48 |
+
"rewards/chosen": 0.013690793886780739,
|
| 49 |
+
"rewards/margins": 0.00916606467217207,
|
| 50 |
+
"rewards/rejected": 0.004524729214608669,
|
| 51 |
+
"step": 20
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
},
|
| 53 |
{
|
| 54 |
"epoch": 0.06,
|
| 55 |
"learning_rate": 3.125e-06,
|
| 56 |
+
"logits/chosen": -2.604323148727417,
|
| 57 |
+
"logits/rejected": -2.598053455352783,
|
| 58 |
+
"logps/chosen": -254.423095703125,
|
| 59 |
+
"logps/rejected": -226.73153686523438,
|
| 60 |
+
"loss": 2402.3988,
|
| 61 |
+
"rewards/accuracies": 0.703125,
|
| 62 |
+
"rewards/chosen": 0.01266755722463131,
|
| 63 |
+
"rewards/margins": 0.024019470438361168,
|
| 64 |
+
"rewards/rejected": -0.01135191135108471,
|
| 65 |
+
"step": 30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
},
|
| 67 |
{
|
| 68 |
"epoch": 0.08,
|
| 69 |
"learning_rate": 4.166666666666667e-06,
|
| 70 |
+
"logits/chosen": -2.6043972969055176,
|
| 71 |
+
"logits/rejected": -2.582412004470825,
|
| 72 |
+
"logps/chosen": -279.12042236328125,
|
| 73 |
+
"logps/rejected": -241.2065887451172,
|
| 74 |
+
"loss": 2290.4264,
|
| 75 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 76 |
+
"rewards/chosen": 0.024520257487893105,
|
| 77 |
+
"rewards/margins": 0.0557018406689167,
|
| 78 |
+
"rewards/rejected": -0.031181585043668747,
|
| 79 |
+
"step": 40
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
},
|
| 81 |
{
|
| 82 |
"epoch": 0.1,
|
| 83 |
+
"learning_rate": 4.999731868769027e-06,
|
| 84 |
+
"logits/chosen": -2.531161308288574,
|
| 85 |
+
"logits/rejected": -2.5264387130737305,
|
| 86 |
+
"logps/chosen": -252.51846313476562,
|
| 87 |
+
"logps/rejected": -247.7227325439453,
|
| 88 |
+
"loss": 2291.9322,
|
| 89 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 90 |
+
"rewards/chosen": 0.029673133045434952,
|
| 91 |
+
"rewards/margins": 0.08245684206485748,
|
| 92 |
+
"rewards/rejected": -0.05278371647000313,
|
| 93 |
+
"step": 50
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"epoch": 0.13,
|
| 97 |
+
"learning_rate": 4.9903533134293035e-06,
|
| 98 |
+
"logits/chosen": -2.545037031173706,
|
| 99 |
+
"logits/rejected": -2.5416412353515625,
|
| 100 |
+
"logps/chosen": -260.83905029296875,
|
| 101 |
+
"logps/rejected": -239.8417205810547,
|
| 102 |
+
"loss": 2269.9371,
|
| 103 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 104 |
+
"rewards/chosen": 0.03231300041079521,
|
| 105 |
+
"rewards/margins": 0.09112317860126495,
|
| 106 |
+
"rewards/rejected": -0.05881017446517944,
|
| 107 |
+
"step": 60
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 108 |
},
|
| 109 |
{
|
| 110 |
"epoch": 0.15,
|
| 111 |
+
"learning_rate": 4.967625656594782e-06,
|
| 112 |
+
"logits/chosen": -2.5832419395446777,
|
| 113 |
+
"logits/rejected": -2.564356565475464,
|
| 114 |
+
"logps/chosen": -275.95452880859375,
|
| 115 |
+
"logps/rejected": -264.7611083984375,
|
| 116 |
+
"loss": 2236.1113,
|
| 117 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 118 |
+
"rewards/chosen": 0.036882974207401276,
|
| 119 |
+
"rewards/margins": 0.08578891307115555,
|
| 120 |
+
"rewards/rejected": -0.048905935138463974,
|
| 121 |
+
"step": 70
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 122 |
},
|
| 123 |
{
|
| 124 |
"epoch": 0.17,
|
| 125 |
+
"learning_rate": 4.93167072587771e-06,
|
| 126 |
+
"logits/chosen": -2.552919864654541,
|
| 127 |
+
"logits/rejected": -2.524970293045044,
|
| 128 |
+
"logps/chosen": -257.78448486328125,
|
| 129 |
+
"logps/rejected": -262.3812561035156,
|
| 130 |
+
"loss": 2220.0893,
|
| 131 |
+
"rewards/accuracies": 0.737500011920929,
|
| 132 |
+
"rewards/chosen": 0.037375591695308685,
|
| 133 |
+
"rewards/margins": 0.11339374631643295,
|
| 134 |
+
"rewards/rejected": -0.07601816952228546,
|
| 135 |
+
"step": 80
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.19,
|
| 139 |
+
"learning_rate": 4.882681251368549e-06,
|
| 140 |
+
"logits/chosen": -2.56257963180542,
|
| 141 |
+
"logits/rejected": -2.5289363861083984,
|
| 142 |
+
"logps/chosen": -239.4860382080078,
|
| 143 |
+
"logps/rejected": -252.36196899414062,
|
| 144 |
+
"loss": 2167.3848,
|
| 145 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 146 |
+
"rewards/chosen": 0.04182355850934982,
|
| 147 |
+
"rewards/margins": 0.10886694490909576,
|
| 148 |
+
"rewards/rejected": -0.06704337894916534,
|
| 149 |
+
"step": 90
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 150 |
},
|
| 151 |
{
|
| 152 |
"epoch": 0.21,
|
| 153 |
+
"learning_rate": 4.8209198325401815e-06,
|
| 154 |
+
"logits/chosen": -2.5551962852478027,
|
| 155 |
+
"logits/rejected": -2.562063455581665,
|
| 156 |
+
"logps/chosen": -266.8739013671875,
|
| 157 |
+
"logps/rejected": -269.649169921875,
|
| 158 |
+
"loss": 2149.4746,
|
| 159 |
+
"rewards/accuracies": 0.762499988079071,
|
| 160 |
+
"rewards/chosen": 0.04759662598371506,
|
| 161 |
+
"rewards/margins": 0.1307816356420517,
|
| 162 |
+
"rewards/rejected": -0.08318501710891724,
|
| 163 |
+
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
"epoch": 0.21,
|
| 167 |
+
"eval_logits/chosen": -2.222931385040283,
|
| 168 |
+
"eval_logits/rejected": -2.1770126819610596,
|
| 169 |
+
"eval_logps/chosen": -260.57818603515625,
|
| 170 |
+
"eval_logps/rejected": -253.25228881835938,
|
| 171 |
+
"eval_loss": 2190.7666015625,
|
| 172 |
+
"eval_rewards/accuracies": 0.7460317611694336,
|
| 173 |
+
"eval_rewards/chosen": 0.044464047998189926,
|
| 174 |
+
"eval_rewards/margins": 0.12927772104740143,
|
| 175 |
+
"eval_rewards/rejected": -0.0848136618733406,
|
| 176 |
+
"eval_runtime": 549.355,
|
| 177 |
+
"eval_samples_per_second": 3.641,
|
| 178 |
+
"eval_steps_per_second": 0.115,
|
| 179 |
+
"step": 100
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 180 |
},
|
| 181 |
{
|
| 182 |
"epoch": 0.23,
|
| 183 |
+
"learning_rate": 4.746717530629565e-06,
|
| 184 |
+
"logits/chosen": -2.5229454040527344,
|
| 185 |
+
"logits/rejected": -2.5105621814727783,
|
| 186 |
+
"logps/chosen": -261.46649169921875,
|
| 187 |
+
"logps/rejected": -256.37835693359375,
|
| 188 |
+
"loss": 2174.1184,
|
| 189 |
+
"rewards/accuracies": 0.746874988079071,
|
| 190 |
+
"rewards/chosen": 0.03517655283212662,
|
| 191 |
+
"rewards/margins": 0.11897265911102295,
|
| 192 |
+
"rewards/rejected": -0.08379611372947693,
|
| 193 |
+
"step": 110
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 194 |
},
|
| 195 |
{
|
| 196 |
"epoch": 0.25,
|
| 197 |
+
"learning_rate": 4.660472094042121e-06,
|
| 198 |
+
"logits/chosen": -2.5114097595214844,
|
| 199 |
+
"logits/rejected": -2.481840133666992,
|
| 200 |
+
"logps/chosen": -246.70370483398438,
|
| 201 |
+
"logps/rejected": -238.27621459960938,
|
| 202 |
+
"loss": 2181.3053,
|
| 203 |
+
"rewards/accuracies": 0.699999988079071,
|
| 204 |
+
"rewards/chosen": 0.044524095952510834,
|
| 205 |
+
"rewards/margins": 0.10293309390544891,
|
| 206 |
+
"rewards/rejected": -0.05840899422764778,
|
| 207 |
+
"step": 120
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
},
|
| 209 |
{
|
| 210 |
"epoch": 0.27,
|
| 211 |
+
"learning_rate": 4.5626458262912745e-06,
|
| 212 |
+
"logits/chosen": -2.4726600646972656,
|
| 213 |
+
"logits/rejected": -2.46514630317688,
|
| 214 |
+
"logps/chosen": -271.7862548828125,
|
| 215 |
+
"logps/rejected": -260.61676025390625,
|
| 216 |
+
"loss": 2175.3252,
|
| 217 |
+
"rewards/accuracies": 0.762499988079071,
|
| 218 |
+
"rewards/chosen": 0.06200919300317764,
|
| 219 |
+
"rewards/margins": 0.12613125145435333,
|
| 220 |
+
"rewards/rejected": -0.06412206590175629,
|
| 221 |
+
"step": 130
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 222 |
},
|
| 223 |
{
|
| 224 |
"epoch": 0.29,
|
| 225 |
+
"learning_rate": 4.453763107901676e-06,
|
| 226 |
+
"logits/chosen": -2.506436586380005,
|
| 227 |
+
"logits/rejected": -2.5005128383636475,
|
| 228 |
+
"logps/chosen": -237.8655242919922,
|
| 229 |
+
"logps/rejected": -249.9298553466797,
|
| 230 |
+
"loss": 2167.2516,
|
| 231 |
+
"rewards/accuracies": 0.734375,
|
| 232 |
+
"rewards/chosen": 0.024008702486753464,
|
| 233 |
+
"rewards/margins": 0.1495535969734192,
|
| 234 |
+
"rewards/rejected": -0.12554487586021423,
|
| 235 |
+
"step": 140
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 236 |
},
|
| 237 |
{
|
| 238 |
"epoch": 0.31,
|
| 239 |
+
"learning_rate": 4.33440758555951e-06,
|
| 240 |
+
"logits/chosen": -2.5227842330932617,
|
| 241 |
+
"logits/rejected": -2.536785364151001,
|
| 242 |
+
"logps/chosen": -260.7518005371094,
|
| 243 |
+
"logps/rejected": -235.9630889892578,
|
| 244 |
+
"loss": 2119.4062,
|
| 245 |
+
"rewards/accuracies": 0.715624988079071,
|
| 246 |
+
"rewards/chosen": 0.04733316972851753,
|
| 247 |
+
"rewards/margins": 0.12345732748508453,
|
| 248 |
+
"rewards/rejected": -0.0761241465806961,
|
| 249 |
+
"step": 150
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.33,
|
| 253 |
+
"learning_rate": 4.205219043576955e-06,
|
| 254 |
+
"logits/chosen": -2.5534234046936035,
|
| 255 |
+
"logits/rejected": -2.4914207458496094,
|
| 256 |
+
"logps/chosen": -254.14065551757812,
|
| 257 |
+
"logps/rejected": -250.95700073242188,
|
| 258 |
+
"loss": 2114.7645,
|
| 259 |
+
"rewards/accuracies": 0.778124988079071,
|
| 260 |
+
"rewards/chosen": 0.06031092256307602,
|
| 261 |
+
"rewards/margins": 0.15202957391738892,
|
| 262 |
+
"rewards/rejected": -0.09171866625547409,
|
| 263 |
+
"step": 160
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 264 |
},
|
| 265 |
{
|
| 266 |
"epoch": 0.36,
|
| 267 |
+
"learning_rate": 4.066889974440757e-06,
|
| 268 |
+
"logits/chosen": -2.5092320442199707,
|
| 269 |
+
"logits/rejected": -2.4965577125549316,
|
| 270 |
+
"logps/chosen": -254.91439819335938,
|
| 271 |
+
"logps/rejected": -242.8040008544922,
|
| 272 |
+
"loss": 2229.8135,
|
| 273 |
+
"rewards/accuracies": 0.675000011920929,
|
| 274 |
+
"rewards/chosen": 0.034448813647031784,
|
| 275 |
+
"rewards/margins": 0.12951095402240753,
|
| 276 |
+
"rewards/rejected": -0.09506212174892426,
|
| 277 |
+
"step": 170
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 278 |
},
|
| 279 |
{
|
| 280 |
"epoch": 0.38,
|
| 281 |
+
"learning_rate": 3.92016186682789e-06,
|
| 282 |
+
"logits/chosen": -2.521221399307251,
|
| 283 |
+
"logits/rejected": -2.533686399459839,
|
| 284 |
+
"logps/chosen": -251.4235382080078,
|
| 285 |
+
"logps/rejected": -259.76220703125,
|
| 286 |
+
"loss": 2175.5213,
|
| 287 |
+
"rewards/accuracies": 0.721875011920929,
|
| 288 |
+
"rewards/chosen": 0.04062749817967415,
|
| 289 |
+
"rewards/margins": 0.120635487139225,
|
| 290 |
+
"rewards/rejected": -0.08000798523426056,
|
| 291 |
+
"step": 180
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 292 |
},
|
| 293 |
{
|
| 294 |
"epoch": 0.4,
|
| 295 |
+
"learning_rate": 3.7658212309857576e-06,
|
| 296 |
+
"logits/chosen": -2.5192363262176514,
|
| 297 |
+
"logits/rejected": -2.4917151927948,
|
| 298 |
+
"logps/chosen": -255.2060089111328,
|
| 299 |
+
"logps/rejected": -250.82022094726562,
|
| 300 |
+
"loss": 2099.443,
|
| 301 |
+
"rewards/accuracies": 0.715624988079071,
|
| 302 |
+
"rewards/chosen": 0.0492943711578846,
|
| 303 |
+
"rewards/margins": 0.14053165912628174,
|
| 304 |
+
"rewards/rejected": -0.09123729914426804,
|
| 305 |
+
"step": 190
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 306 |
},
|
| 307 |
{
|
| 308 |
"epoch": 0.42,
|
| 309 |
+
"learning_rate": 3.604695382782159e-06,
|
| 310 |
+
"logits/chosen": -2.5251801013946533,
|
| 311 |
+
"logits/rejected": -2.5034642219543457,
|
| 312 |
+
"logps/chosen": -269.3675537109375,
|
| 313 |
+
"logps/rejected": -262.86376953125,
|
| 314 |
+
"loss": 2105.1256,
|
| 315 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 316 |
+
"rewards/chosen": 0.0575677752494812,
|
| 317 |
+
"rewards/margins": 0.14340198040008545,
|
| 318 |
+
"rewards/rejected": -0.08583419024944305,
|
| 319 |
+
"step": 200
|
| 320 |
},
|
| 321 |
{
|
| 322 |
"epoch": 0.42,
|
| 323 |
+
"eval_logits/chosen": -2.260270833969116,
|
| 324 |
+
"eval_logits/rejected": -2.2073864936828613,
|
| 325 |
+
"eval_logps/chosen": -259.5941467285156,
|
| 326 |
+
"eval_logps/rejected": -254.3839874267578,
|
| 327 |
+
"eval_loss": 2151.155517578125,
|
| 328 |
+
"eval_rewards/accuracies": 0.7599206566810608,
|
| 329 |
+
"eval_rewards/chosen": 0.05430443957448006,
|
| 330 |
+
"eval_rewards/margins": 0.15043501555919647,
|
| 331 |
+
"eval_rewards/rejected": -0.09613056480884552,
|
| 332 |
+
"eval_runtime": 548.195,
|
| 333 |
+
"eval_samples_per_second": 3.648,
|
| 334 |
+
"eval_steps_per_second": 0.115,
|
| 335 |
+
"step": 200
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 336 |
},
|
| 337 |
{
|
| 338 |
"epoch": 0.44,
|
| 339 |
+
"learning_rate": 3.437648009023905e-06,
|
| 340 |
+
"logits/chosen": -2.533383369445801,
|
| 341 |
+
"logits/rejected": -2.4935860633850098,
|
| 342 |
+
"logps/chosen": -243.6236114501953,
|
| 343 |
+
"logps/rejected": -238.85140991210938,
|
| 344 |
+
"loss": 2145.5416,
|
| 345 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 346 |
+
"rewards/chosen": 0.06410142779350281,
|
| 347 |
+
"rewards/margins": 0.14374245703220367,
|
| 348 |
+
"rewards/rejected": -0.07964102178812027,
|
| 349 |
+
"step": 210
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 350 |
},
|
| 351 |
{
|
| 352 |
"epoch": 0.46,
|
| 353 |
+
"learning_rate": 3.265574537815398e-06,
|
| 354 |
+
"logits/chosen": -2.554565906524658,
|
| 355 |
+
"logits/rejected": -2.56289005279541,
|
| 356 |
+
"logps/chosen": -277.4061584472656,
|
| 357 |
+
"logps/rejected": -253.40048217773438,
|
| 358 |
+
"loss": 2196.8484,
|
| 359 |
+
"rewards/accuracies": 0.675000011920929,
|
| 360 |
+
"rewards/chosen": 0.052330613136291504,
|
| 361 |
+
"rewards/margins": 0.11339585483074188,
|
| 362 |
+
"rewards/rejected": -0.06106524541974068,
|
| 363 |
+
"step": 220
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 364 |
},
|
| 365 |
{
|
| 366 |
"epoch": 0.48,
|
| 367 |
+
"learning_rate": 3.089397338773569e-06,
|
| 368 |
+
"logits/chosen": -2.4857611656188965,
|
| 369 |
+
"logits/rejected": -2.473193407058716,
|
| 370 |
+
"logps/chosen": -247.3427276611328,
|
| 371 |
+
"logps/rejected": -241.8627471923828,
|
| 372 |
+
"loss": 2160.1729,
|
| 373 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 374 |
+
"rewards/chosen": 0.03845102712512016,
|
| 375 |
+
"rewards/margins": 0.11976752430200577,
|
| 376 |
+
"rewards/rejected": -0.0813164934515953,
|
| 377 |
+
"step": 230
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 378 |
},
|
| 379 |
{
|
| 380 |
"epoch": 0.5,
|
| 381 |
+
"learning_rate": 2.9100607788275547e-06,
|
| 382 |
+
"logits/chosen": -2.5121560096740723,
|
| 383 |
+
"logits/rejected": -2.516338586807251,
|
| 384 |
+
"logps/chosen": -257.1769714355469,
|
| 385 |
+
"logps/rejected": -247.3695068359375,
|
| 386 |
+
"loss": 2185.7641,
|
| 387 |
+
"rewards/accuracies": 0.684374988079071,
|
| 388 |
+
"rewards/chosen": 0.0379050187766552,
|
| 389 |
+
"rewards/margins": 0.11140499264001846,
|
| 390 |
+
"rewards/rejected": -0.07349997013807297,
|
| 391 |
+
"step": 240
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 392 |
},
|
| 393 |
{
|
| 394 |
"epoch": 0.52,
|
| 395 |
+
"learning_rate": 2.72852616010567e-06,
|
| 396 |
+
"logits/chosen": -2.5092978477478027,
|
| 397 |
+
"logits/rejected": -2.487090826034546,
|
| 398 |
+
"logps/chosen": -264.5955505371094,
|
| 399 |
+
"logps/rejected": -246.3382110595703,
|
| 400 |
+
"loss": 2136.6197,
|
| 401 |
+
"rewards/accuracies": 0.7406250238418579,
|
| 402 |
+
"rewards/chosen": 0.039962492883205414,
|
| 403 |
+
"rewards/margins": 0.1403963267803192,
|
| 404 |
+
"rewards/rejected": -0.1004338413476944,
|
| 405 |
+
"step": 250
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 406 |
},
|
| 407 |
{
|
| 408 |
"epoch": 0.54,
|
| 409 |
+
"learning_rate": 2.5457665670441937e-06,
|
| 410 |
+
"logits/chosen": -2.5069711208343506,
|
| 411 |
+
"logits/rejected": -2.5030505657196045,
|
| 412 |
+
"logps/chosen": -257.4859619140625,
|
| 413 |
+
"logps/rejected": -231.91958618164062,
|
| 414 |
+
"loss": 2085.2795,
|
| 415 |
+
"rewards/accuracies": 0.762499988079071,
|
| 416 |
+
"rewards/chosen": 0.05723271518945694,
|
| 417 |
+
"rewards/margins": 0.15024301409721375,
|
| 418 |
+
"rewards/rejected": -0.0930103212594986,
|
| 419 |
+
"step": 260
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 420 |
},
|
| 421 |
{
|
| 422 |
"epoch": 0.57,
|
| 423 |
+
"learning_rate": 2.3627616503391813e-06,
|
| 424 |
+
"logits/chosen": -2.525665760040283,
|
| 425 |
+
"logits/rejected": -2.5043163299560547,
|
| 426 |
+
"logps/chosen": -280.7471618652344,
|
| 427 |
+
"logps/rejected": -267.36712646484375,
|
| 428 |
+
"loss": 2089.859,
|
| 429 |
+
"rewards/accuracies": 0.737500011920929,
|
| 430 |
+
"rewards/chosen": 0.05569761246442795,
|
| 431 |
+
"rewards/margins": 0.179846853017807,
|
| 432 |
+
"rewards/rejected": -0.12414924055337906,
|
| 433 |
+
"step": 270
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 434 |
},
|
| 435 |
{
|
| 436 |
"epoch": 0.59,
|
| 437 |
+
"learning_rate": 2.1804923757009885e-06,
|
| 438 |
+
"logits/chosen": -2.500837564468384,
|
| 439 |
+
"logits/rejected": -2.501950740814209,
|
| 440 |
+
"logps/chosen": -270.04193115234375,
|
| 441 |
+
"logps/rejected": -248.61978149414062,
|
| 442 |
+
"loss": 2111.6906,
|
| 443 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 444 |
+
"rewards/chosen": 0.05320361256599426,
|
| 445 |
+
"rewards/margins": 0.1410333216190338,
|
| 446 |
+
"rewards/rejected": -0.08782971650362015,
|
| 447 |
+
"step": 280
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 448 |
},
|
| 449 |
{
|
| 450 |
"epoch": 0.61,
|
| 451 |
+
"learning_rate": 1.9999357655598894e-06,
|
| 452 |
+
"logits/chosen": -2.5122292041778564,
|
| 453 |
+
"logits/rejected": -2.50368070602417,
|
| 454 |
+
"logps/chosen": -258.72686767578125,
|
| 455 |
+
"logps/rejected": -256.91387939453125,
|
| 456 |
+
"loss": 2137.0592,
|
| 457 |
+
"rewards/accuracies": 0.737500011920929,
|
| 458 |
+
"rewards/chosen": 0.053160279989242554,
|
| 459 |
+
"rewards/margins": 0.15454119443893433,
|
| 460 |
+
"rewards/rejected": -0.10138092190027237,
|
| 461 |
+
"step": 290
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 462 |
},
|
| 463 |
{
|
| 464 |
"epoch": 0.63,
|
| 465 |
+
"learning_rate": 1.8220596619089576e-06,
|
| 466 |
+
"logits/chosen": -2.471623659133911,
|
| 467 |
+
"logits/rejected": -2.4690403938293457,
|
| 468 |
+
"logps/chosen": -246.51766967773438,
|
| 469 |
+
"logps/rejected": -251.79257202148438,
|
| 470 |
+
"loss": 2135.4973,
|
| 471 |
+
"rewards/accuracies": 0.71875,
|
| 472 |
+
"rewards/chosen": 0.0453377440571785,
|
| 473 |
+
"rewards/margins": 0.12641170620918274,
|
| 474 |
+
"rewards/rejected": -0.08107397705316544,
|
| 475 |
+
"step": 300
|
| 476 |
},
|
| 477 |
{
|
| 478 |
"epoch": 0.63,
|
| 479 |
+
"eval_logits/chosen": -2.2764506340026855,
|
| 480 |
+
"eval_logits/rejected": -2.2231767177581787,
|
| 481 |
+
"eval_logps/chosen": -258.7624206542969,
|
| 482 |
+
"eval_logps/rejected": -252.75852966308594,
|
| 483 |
+
"eval_loss": 2129.089599609375,
|
| 484 |
+
"eval_rewards/accuracies": 0.7559523582458496,
|
| 485 |
+
"eval_rewards/chosen": 0.06262180209159851,
|
| 486 |
+
"eval_rewards/margins": 0.14249789714813232,
|
| 487 |
+
"eval_rewards/rejected": -0.07987607270479202,
|
| 488 |
+
"eval_runtime": 547.9938,
|
| 489 |
+
"eval_samples_per_second": 3.65,
|
| 490 |
+
"eval_steps_per_second": 0.115,
|
| 491 |
+
"step": 300
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.65,
|
| 495 |
+
"learning_rate": 1.647817538357072e-06,
|
| 496 |
+
"logits/chosen": -2.5041086673736572,
|
| 497 |
+
"logits/rejected": -2.495436191558838,
|
| 498 |
+
"logps/chosen": -264.5109558105469,
|
| 499 |
+
"logps/rejected": -248.3275604248047,
|
| 500 |
+
"loss": 2107.123,
|
| 501 |
+
"rewards/accuracies": 0.731249988079071,
|
| 502 |
+
"rewards/chosen": 0.05480458214879036,
|
| 503 |
+
"rewards/margins": 0.13964474201202393,
|
| 504 |
+
"rewards/rejected": -0.08484016358852386,
|
| 505 |
+
"step": 310
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.67,
|
| 509 |
+
"learning_rate": 1.4781433892011132e-06,
|
| 510 |
+
"logits/chosen": -2.53191876411438,
|
| 511 |
+
"logits/rejected": -2.4989166259765625,
|
| 512 |
+
"logps/chosen": -242.36599731445312,
|
| 513 |
+
"logps/rejected": -243.78067016601562,
|
| 514 |
+
"loss": 2076.0621,
|
| 515 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 516 |
+
"rewards/chosen": 0.05456935614347458,
|
| 517 |
+
"rewards/margins": 0.14978976547718048,
|
| 518 |
+
"rewards/rejected": -0.0952204093337059,
|
| 519 |
+
"step": 320
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.69,
|
| 523 |
+
"learning_rate": 1.3139467229135999e-06,
|
| 524 |
+
"logits/chosen": -2.4768006801605225,
|
| 525 |
+
"logits/rejected": -2.4569873809814453,
|
| 526 |
+
"logps/chosen": -263.0523681640625,
|
| 527 |
+
"logps/rejected": -250.5469207763672,
|
| 528 |
+
"loss": 2112.1141,
|
| 529 |
+
"rewards/accuracies": 0.734375,
|
| 530 |
+
"rewards/chosen": 0.044828929007053375,
|
| 531 |
+
"rewards/margins": 0.13050048053264618,
|
| 532 |
+
"rewards/rejected": -0.0856715738773346,
|
| 533 |
+
"step": 330
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 534 |
},
|
| 535 |
{
|
| 536 |
"epoch": 0.71,
|
| 537 |
+
"learning_rate": 1.1561076868822756e-06,
|
| 538 |
+
"logits/chosen": -2.5158028602600098,
|
| 539 |
+
"logits/rejected": -2.5096983909606934,
|
| 540 |
+
"logps/chosen": -275.6848449707031,
|
| 541 |
+
"logps/rejected": -246.7259979248047,
|
| 542 |
+
"loss": 2151.2445,
|
| 543 |
+
"rewards/accuracies": 0.746874988079071,
|
| 544 |
+
"rewards/chosen": 0.052164845168590546,
|
| 545 |
+
"rewards/margins": 0.15314052999019623,
|
| 546 |
+
"rewards/rejected": -0.10097566992044449,
|
| 547 |
+
"step": 340
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 548 |
},
|
| 549 |
{
|
| 550 |
"epoch": 0.73,
|
| 551 |
+
"learning_rate": 1.0054723495346484e-06,
|
| 552 |
+
"logits/chosen": -2.518799304962158,
|
| 553 |
+
"logits/rejected": -2.4620516300201416,
|
| 554 |
+
"logps/chosen": -249.27401733398438,
|
| 555 |
+
"logps/rejected": -218.7183074951172,
|
| 556 |
+
"loss": 2093.9803,
|
| 557 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 558 |
+
"rewards/chosen": 0.0662151575088501,
|
| 559 |
+
"rewards/margins": 0.14556364715099335,
|
| 560 |
+
"rewards/rejected": -0.07934850454330444,
|
| 561 |
+
"step": 350
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.75,
|
| 565 |
+
"learning_rate": 8.628481651367876e-07,
|
| 566 |
+
"logits/chosen": -2.5340943336486816,
|
| 567 |
+
"logits/rejected": -2.5006654262542725,
|
| 568 |
+
"logps/chosen": -260.32464599609375,
|
| 569 |
+
"logps/rejected": -237.3218536376953,
|
| 570 |
+
"loss": 2094.1246,
|
| 571 |
+
"rewards/accuracies": 0.765625,
|
| 572 |
+
"rewards/chosen": 0.05396001785993576,
|
| 573 |
+
"rewards/margins": 0.15317106246948242,
|
| 574 |
+
"rewards/rejected": -0.09921105206012726,
|
| 575 |
+
"step": 360
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.77,
|
| 579 |
+
"learning_rate": 7.289996455765749e-07,
|
| 580 |
+
"logits/chosen": -2.529265880584717,
|
| 581 |
+
"logits/rejected": -2.515712261199951,
|
| 582 |
+
"logps/chosen": -266.943115234375,
|
| 583 |
+
"logps/rejected": -246.0579376220703,
|
| 584 |
+
"loss": 2115.357,
|
| 585 |
+
"rewards/accuracies": 0.71875,
|
| 586 |
+
"rewards/chosen": 0.052078358829021454,
|
| 587 |
+
"rewards/margins": 0.1462351232767105,
|
| 588 |
+
"rewards/rejected": -0.09415675699710846,
|
| 589 |
+
"step": 370
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 590 |
},
|
| 591 |
{
|
| 592 |
"epoch": 0.8,
|
| 593 |
+
"learning_rate": 6.046442623320145e-07,
|
| 594 |
+
"logits/chosen": -2.4891440868377686,
|
| 595 |
+
"logits/rejected": -2.499753952026367,
|
| 596 |
+
"logps/chosen": -253.51632690429688,
|
| 597 |
+
"logps/rejected": -245.4505615234375,
|
| 598 |
+
"loss": 2082.182,
|
| 599 |
+
"rewards/accuracies": 0.7406250238418579,
|
| 600 |
+
"rewards/chosen": 0.051686953753232956,
|
| 601 |
+
"rewards/margins": 0.1390691101551056,
|
| 602 |
+
"rewards/rejected": -0.08738215267658234,
|
| 603 |
+
"step": 380
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 604 |
},
|
| 605 |
{
|
| 606 |
"epoch": 0.82,
|
| 607 |
+
"learning_rate": 4.904486005914027e-07,
|
| 608 |
+
"logits/chosen": -2.532160997390747,
|
| 609 |
+
"logits/rejected": -2.5001654624938965,
|
| 610 |
+
"logps/chosen": -280.9754333496094,
|
| 611 |
+
"logps/rejected": -279.0588684082031,
|
| 612 |
+
"loss": 2114.3043,
|
| 613 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 614 |
+
"rewards/chosen": 0.0547635443508625,
|
| 615 |
+
"rewards/margins": 0.14076778292655945,
|
| 616 |
+
"rewards/rejected": -0.08600424975156784,
|
| 617 |
+
"step": 390
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 618 |
},
|
| 619 |
{
|
| 620 |
"epoch": 0.84,
|
| 621 |
+
"learning_rate": 3.8702478614051353e-07,
|
| 622 |
+
"logits/chosen": -2.4791765213012695,
|
| 623 |
+
"logits/rejected": -2.4799935817718506,
|
| 624 |
+
"logps/chosen": -246.14102172851562,
|
| 625 |
+
"logps/rejected": -251.533447265625,
|
| 626 |
+
"loss": 2099.8018,
|
| 627 |
+
"rewards/accuracies": 0.703125,
|
| 628 |
+
"rewards/chosen": 0.0392024889588356,
|
| 629 |
+
"rewards/margins": 0.13221651315689087,
|
| 630 |
+
"rewards/rejected": -0.09301402419805527,
|
| 631 |
+
"step": 400
|
| 632 |
},
|
| 633 |
{
|
| 634 |
"epoch": 0.84,
|
| 635 |
+
"eval_logits/chosen": -2.254145860671997,
|
| 636 |
+
"eval_logits/rejected": -2.2016360759735107,
|
| 637 |
+
"eval_logps/chosen": -259.64398193359375,
|
| 638 |
+
"eval_logps/rejected": -254.3590850830078,
|
| 639 |
+
"eval_loss": 2121.667236328125,
|
| 640 |
+
"eval_rewards/accuracies": 0.7539682388305664,
|
| 641 |
+
"eval_rewards/chosen": 0.05380600318312645,
|
| 642 |
+
"eval_rewards/margins": 0.14968746900558472,
|
| 643 |
+
"eval_rewards/rejected": -0.09588146954774857,
|
| 644 |
+
"eval_runtime": 547.9727,
|
| 645 |
+
"eval_samples_per_second": 3.65,
|
| 646 |
+
"eval_steps_per_second": 0.115,
|
| 647 |
+
"step": 400
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 648 |
},
|
| 649 |
{
|
| 650 |
"epoch": 0.86,
|
| 651 |
+
"learning_rate": 2.9492720416985004e-07,
|
| 652 |
+
"logits/chosen": -2.4832329750061035,
|
| 653 |
+
"logits/rejected": -2.463463306427002,
|
| 654 |
+
"logps/chosen": -284.7741394042969,
|
| 655 |
+
"logps/rejected": -252.4269561767578,
|
| 656 |
+
"loss": 2145.448,
|
| 657 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 658 |
+
"rewards/chosen": 0.05263269692659378,
|
| 659 |
+
"rewards/margins": 0.15021036565303802,
|
| 660 |
+
"rewards/rejected": -0.09757767617702484,
|
| 661 |
+
"step": 410
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.88,
|
| 665 |
+
"learning_rate": 2.1464952759020857e-07,
|
| 666 |
+
"logits/chosen": -2.4804348945617676,
|
| 667 |
+
"logits/rejected": -2.457764148712158,
|
| 668 |
+
"logps/chosen": -254.78604125976562,
|
| 669 |
+
"logps/rejected": -278.61346435546875,
|
| 670 |
+
"loss": 2123.6629,
|
| 671 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 672 |
+
"rewards/chosen": 0.033899884670972824,
|
| 673 |
+
"rewards/margins": 0.11116783320903778,
|
| 674 |
+
"rewards/rejected": -0.07726795971393585,
|
| 675 |
+
"step": 420
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 0.9,
|
| 679 |
+
"learning_rate": 1.4662207078575685e-07,
|
| 680 |
+
"logits/chosen": -2.4848549365997314,
|
| 681 |
+
"logits/rejected": -2.485640048980713,
|
| 682 |
+
"logps/chosen": -268.3457336425781,
|
| 683 |
+
"logps/rejected": -268.5885925292969,
|
| 684 |
+
"loss": 2144.4309,
|
| 685 |
+
"rewards/accuracies": 0.721875011920929,
|
| 686 |
+
"rewards/chosen": 0.03841588646173477,
|
| 687 |
+
"rewards/margins": 0.13024446368217468,
|
| 688 |
+
"rewards/rejected": -0.09182857722043991,
|
| 689 |
+
"step": 430
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 690 |
},
|
| 691 |
{
|
| 692 |
"epoch": 0.92,
|
| 693 |
+
"learning_rate": 9.120948298936422e-08,
|
| 694 |
+
"logits/chosen": -2.457054615020752,
|
| 695 |
+
"logits/rejected": -2.4329726696014404,
|
| 696 |
+
"logps/chosen": -231.9584197998047,
|
| 697 |
+
"logps/rejected": -234.6277313232422,
|
| 698 |
+
"loss": 2118.3984,
|
| 699 |
+
"rewards/accuracies": 0.737500011920929,
|
| 700 |
+
"rewards/chosen": 0.038600482046604156,
|
| 701 |
+
"rewards/margins": 0.13669805228710175,
|
| 702 |
+
"rewards/rejected": -0.09809757024049759,
|
| 703 |
+
"step": 440
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.94,
|
| 707 |
+
"learning_rate": 4.870879364444109e-08,
|
| 708 |
+
"logits/chosen": -2.5156655311584473,
|
| 709 |
+
"logits/rejected": -2.563300848007202,
|
| 710 |
+
"logps/chosen": -263.9936218261719,
|
| 711 |
+
"logps/rejected": -265.6227722167969,
|
| 712 |
+
"loss": 2123.5402,
|
| 713 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 714 |
+
"rewards/chosen": 0.04902677983045578,
|
| 715 |
+
"rewards/margins": 0.1260160207748413,
|
| 716 |
+
"rewards/rejected": -0.07698923349380493,
|
| 717 |
+
"step": 450
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 718 |
},
|
| 719 |
{
|
| 720 |
"epoch": 0.96,
|
| 721 |
+
"learning_rate": 1.93478202307823e-08,
|
| 722 |
+
"logits/chosen": -2.470996379852295,
|
| 723 |
+
"logits/rejected": -2.4720451831817627,
|
| 724 |
+
"logps/chosen": -258.21734619140625,
|
| 725 |
+
"logps/rejected": -262.04925537109375,
|
| 726 |
+
"loss": 2078.5094,
|
| 727 |
+
"rewards/accuracies": 0.7281249761581421,
|
| 728 |
+
"rewards/chosen": 0.04391016811132431,
|
| 729 |
+
"rewards/margins": 0.14817874133586884,
|
| 730 |
+
"rewards/rejected": -0.10426857322454453,
|
| 731 |
+
"step": 460
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 732 |
},
|
| 733 |
{
|
| 734 |
"epoch": 0.98,
|
| 735 |
+
"learning_rate": 3.283947088983663e-09,
|
| 736 |
+
"logits/chosen": -2.513140916824341,
|
| 737 |
+
"logits/rejected": -2.535651206970215,
|
| 738 |
+
"logps/chosen": -249.6727752685547,
|
| 739 |
+
"logps/rejected": -248.2782745361328,
|
| 740 |
+
"loss": 2093.2779,
|
| 741 |
+
"rewards/accuracies": 0.75,
|
| 742 |
+
"rewards/chosen": 0.04741714522242546,
|
| 743 |
+
"rewards/margins": 0.143958181142807,
|
| 744 |
+
"rewards/rejected": -0.09654103964567184,
|
| 745 |
+
"step": 470
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 746 |
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
+
"step": 477,
|
| 750 |
"total_flos": 0.0,
|
| 751 |
+
"train_loss": 2164.5614415454666,
|
| 752 |
+
"train_runtime": 32346.8016,
|
| 753 |
+
"train_samples_per_second": 1.89,
|
| 754 |
+
"train_steps_per_second": 0.015
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
+
"max_steps": 477,
|
| 759 |
+
"num_input_tokens_seen": 0,
|
| 760 |
"num_train_epochs": 1,
|
| 761 |
"save_steps": 100,
|
| 762 |
"total_flos": 0.0,
|
| 763 |
+
"train_batch_size": 2,
|
| 764 |
"trial_name": null,
|
| 765 |
"trial_params": null
|
| 766 |
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4920
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89ae9269821a7a76bfccee733cea8c3af1d1b7b751ef31fb40915f4d080f4944
|
| 3 |
size 4920
|