Commit
•
8703e79
1
Parent(s):
494ea79
Model save
Browse files- README.md +74 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +434 -0
README.md
ADDED
|
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
base_model: allenai/tulu-2-7b
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: ultrafeedback-binarized-tulu-2-7b-dpo-full
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# ultrafeedback-binarized-tulu-2-7b-dpo-full
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [allenai/tulu-2-7b](https://huggingface.co/allenai/tulu-2-7b) on an unknown dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.6640
|
| 21 |
+
- Rewards/chosen: 0.0423
|
| 22 |
+
- Rewards/rejected: -0.0311
|
| 23 |
+
- Rewards/accuracies: 0.6706
|
| 24 |
+
- Rewards/margins: 0.0734
|
| 25 |
+
- Logps/rejected: -317.2082
|
| 26 |
+
- Logps/chosen: -335.1042
|
| 27 |
+
- Logits/rejected: -1.2523
|
| 28 |
+
- Logits/chosen: -1.1794
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 4
|
| 50 |
+
- seed: 42
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 4
|
| 54 |
+
- total_train_batch_size: 256
|
| 55 |
+
- total_eval_batch_size: 32
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.6762 | 0.4184 | 100 | 0.6753 | 0.0546 | 0.0122 | 0.6627 | 0.0424 | -312.8761 | -333.8717 | -1.2638 | -1.1861 |
|
| 66 |
+
| 0.6604 | 0.8368 | 200 | 0.6640 | 0.0423 | -0.0311 | 0.6706 | 0.0734 | -317.2082 | -335.1042 | -1.2523 | -1.1794 |
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
### Framework versions
|
| 70 |
+
|
| 71 |
+
- Transformers 4.44.1
|
| 72 |
+
- Pytorch 2.1.2+cu121
|
| 73 |
+
- Datasets 2.21.0
|
| 74 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6732059223382543,
|
| 5 |
+
"train_runtime": 6951.1203,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 8.795,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.1"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6732059223382543,
|
| 5 |
+
"train_runtime": 6951.1203,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 8.795,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,434 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 239,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0041841004184100415,
|
| 13 |
+
"grad_norm": 2.335511797805867,
|
| 14 |
+
"learning_rate": 2.083333333333333e-08,
|
| 15 |
+
"logits/chosen": -1.168128490447998,
|
| 16 |
+
"logits/rejected": -1.1001684665679932,
|
| 17 |
+
"logps/chosen": -345.7468566894531,
|
| 18 |
+
"logps/rejected": -388.03436279296875,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.04184100418410042,
|
| 28 |
+
"grad_norm": 3.0538125395530042,
|
| 29 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 30 |
+
"logits/chosen": -1.1523456573486328,
|
| 31 |
+
"logits/rejected": -1.216726541519165,
|
| 32 |
+
"logps/chosen": -327.39593505859375,
|
| 33 |
+
"logps/rejected": -330.5606689453125,
|
| 34 |
+
"loss": 0.6932,
|
| 35 |
+
"rewards/accuracies": 0.4583333432674408,
|
| 36 |
+
"rewards/chosen": 0.00016736745601519942,
|
| 37 |
+
"rewards/margins": 0.00036640118923969567,
|
| 38 |
+
"rewards/rejected": -0.00019903367501683533,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.08368200836820083,
|
| 43 |
+
"grad_norm": 2.051052061032609,
|
| 44 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 45 |
+
"logits/chosen": -1.1201739311218262,
|
| 46 |
+
"logits/rejected": -1.1966865062713623,
|
| 47 |
+
"logps/chosen": -337.3428649902344,
|
| 48 |
+
"logps/rejected": -310.3174743652344,
|
| 49 |
+
"loss": 0.6931,
|
| 50 |
+
"rewards/accuracies": 0.4906249940395355,
|
| 51 |
+
"rewards/chosen": 0.0008397102355957031,
|
| 52 |
+
"rewards/margins": -0.00019551152945496142,
|
| 53 |
+
"rewards/rejected": 0.001035221852362156,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.12552301255230125,
|
| 58 |
+
"grad_norm": 2.051405044620473,
|
| 59 |
+
"learning_rate": 4.990398100856366e-07,
|
| 60 |
+
"logits/chosen": -1.1592100858688354,
|
| 61 |
+
"logits/rejected": -1.1792274713516235,
|
| 62 |
+
"logps/chosen": -327.45831298828125,
|
| 63 |
+
"logps/rejected": -347.738037109375,
|
| 64 |
+
"loss": 0.6922,
|
| 65 |
+
"rewards/accuracies": 0.546875,
|
| 66 |
+
"rewards/chosen": 0.004632706753909588,
|
| 67 |
+
"rewards/margins": 0.001139953499659896,
|
| 68 |
+
"rewards/rejected": 0.0034927534870803356,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.16736401673640167,
|
| 73 |
+
"grad_norm": 2.280425541026252,
|
| 74 |
+
"learning_rate": 4.931986719649298e-07,
|
| 75 |
+
"logits/chosen": -1.2329559326171875,
|
| 76 |
+
"logits/rejected": -1.273472547531128,
|
| 77 |
+
"logps/chosen": -352.55316162109375,
|
| 78 |
+
"logps/rejected": -314.9006652832031,
|
| 79 |
+
"loss": 0.6902,
|
| 80 |
+
"rewards/accuracies": 0.6343749761581421,
|
| 81 |
+
"rewards/chosen": 0.013729018159210682,
|
| 82 |
+
"rewards/margins": 0.00675433874130249,
|
| 83 |
+
"rewards/rejected": 0.006974679883569479,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.20920502092050208,
|
| 88 |
+
"grad_norm": 1.953444430114674,
|
| 89 |
+
"learning_rate": 4.821741763807186e-07,
|
| 90 |
+
"logits/chosen": -1.2172247171401978,
|
| 91 |
+
"logits/rejected": -1.2943127155303955,
|
| 92 |
+
"logps/chosen": -316.37371826171875,
|
| 93 |
+
"logps/rejected": -312.96661376953125,
|
| 94 |
+
"loss": 0.6874,
|
| 95 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 96 |
+
"rewards/chosen": 0.020605359226465225,
|
| 97 |
+
"rewards/margins": 0.01330882590264082,
|
| 98 |
+
"rewards/rejected": 0.007296533323824406,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.2510460251046025,
|
| 103 |
+
"grad_norm": 2.228539317195908,
|
| 104 |
+
"learning_rate": 4.662012913161997e-07,
|
| 105 |
+
"logits/chosen": -1.2346419095993042,
|
| 106 |
+
"logits/rejected": -1.3438832759857178,
|
| 107 |
+
"logps/chosen": -348.8750915527344,
|
| 108 |
+
"logps/rejected": -317.3927307128906,
|
| 109 |
+
"loss": 0.6851,
|
| 110 |
+
"rewards/accuracies": 0.621874988079071,
|
| 111 |
+
"rewards/chosen": 0.03193451464176178,
|
| 112 |
+
"rewards/margins": 0.016508478671312332,
|
| 113 |
+
"rewards/rejected": 0.01542603224515915,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.2928870292887029,
|
| 118 |
+
"grad_norm": 2.1140708636197845,
|
| 119 |
+
"learning_rate": 4.456204510851956e-07,
|
| 120 |
+
"logits/chosen": -1.225719928741455,
|
| 121 |
+
"logits/rejected": -1.2811986207962036,
|
| 122 |
+
"logps/chosen": -344.2880554199219,
|
| 123 |
+
"logps/rejected": -338.8671569824219,
|
| 124 |
+
"loss": 0.6818,
|
| 125 |
+
"rewards/accuracies": 0.640625,
|
| 126 |
+
"rewards/chosen": 0.04076681658625603,
|
| 127 |
+
"rewards/margins": 0.02389165200293064,
|
| 128 |
+
"rewards/rejected": 0.016875162720680237,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.33472803347280333,
|
| 133 |
+
"grad_norm": 1.8046487610538895,
|
| 134 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 135 |
+
"logits/chosen": -1.2622222900390625,
|
| 136 |
+
"logits/rejected": -1.3684368133544922,
|
| 137 |
+
"logps/chosen": -328.14263916015625,
|
| 138 |
+
"logps/rejected": -317.73193359375,
|
| 139 |
+
"loss": 0.6804,
|
| 140 |
+
"rewards/accuracies": 0.6031249761581421,
|
| 141 |
+
"rewards/chosen": 0.03804489225149155,
|
| 142 |
+
"rewards/margins": 0.022086812183260918,
|
| 143 |
+
"rewards/rejected": 0.01595807820558548,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.37656903765690375,
|
| 148 |
+
"grad_norm": 2.1792714189731988,
|
| 149 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 150 |
+
"logits/chosen": -1.2576462030410767,
|
| 151 |
+
"logits/rejected": -1.3489463329315186,
|
| 152 |
+
"logps/chosen": -303.93096923828125,
|
| 153 |
+
"logps/rejected": -290.1899719238281,
|
| 154 |
+
"loss": 0.6759,
|
| 155 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 156 |
+
"rewards/chosen": 0.050195373594760895,
|
| 157 |
+
"rewards/margins": 0.038325682282447815,
|
| 158 |
+
"rewards/rejected": 0.011869693174958229,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.41841004184100417,
|
| 163 |
+
"grad_norm": 2.022836388367397,
|
| 164 |
+
"learning_rate": 3.610497133404795e-07,
|
| 165 |
+
"logits/chosen": -1.284611701965332,
|
| 166 |
+
"logits/rejected": -1.3313934803009033,
|
| 167 |
+
"logps/chosen": -319.56182861328125,
|
| 168 |
+
"logps/rejected": -312.11492919921875,
|
| 169 |
+
"loss": 0.6762,
|
| 170 |
+
"rewards/accuracies": 0.706250011920929,
|
| 171 |
+
"rewards/chosen": 0.05392398685216904,
|
| 172 |
+
"rewards/margins": 0.04721994698047638,
|
| 173 |
+
"rewards/rejected": 0.006704038940370083,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.41841004184100417,
|
| 178 |
+
"eval_logits/chosen": -1.1861072778701782,
|
| 179 |
+
"eval_logits/rejected": -1.2637689113616943,
|
| 180 |
+
"eval_logps/chosen": -333.8717041015625,
|
| 181 |
+
"eval_logps/rejected": -312.8760681152344,
|
| 182 |
+
"eval_loss": 0.6753271818161011,
|
| 183 |
+
"eval_rewards/accuracies": 0.6626983880996704,
|
| 184 |
+
"eval_rewards/chosen": 0.054593343287706375,
|
| 185 |
+
"eval_rewards/margins": 0.042364537715911865,
|
| 186 |
+
"eval_rewards/rejected": 0.01222880370914936,
|
| 187 |
+
"eval_runtime": 91.2196,
|
| 188 |
+
"eval_samples_per_second": 21.925,
|
| 189 |
+
"eval_steps_per_second": 0.691,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.4602510460251046,
|
| 194 |
+
"grad_norm": 1.8857411098055814,
|
| 195 |
+
"learning_rate": 3.272542485937368e-07,
|
| 196 |
+
"logits/chosen": -1.261444330215454,
|
| 197 |
+
"logits/rejected": -1.3659017086029053,
|
| 198 |
+
"logps/chosen": -330.69903564453125,
|
| 199 |
+
"logps/rejected": -304.2332458496094,
|
| 200 |
+
"loss": 0.6739,
|
| 201 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 202 |
+
"rewards/chosen": 0.0468447282910347,
|
| 203 |
+
"rewards/margins": 0.04751918837428093,
|
| 204 |
+
"rewards/rejected": -0.0006744601996615529,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.502092050209205,
|
| 209 |
+
"grad_norm": 2.320788210793853,
|
| 210 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 211 |
+
"logits/chosen": -1.2688654661178589,
|
| 212 |
+
"logits/rejected": -1.3098180294036865,
|
| 213 |
+
"logps/chosen": -307.9932556152344,
|
| 214 |
+
"logps/rejected": -306.85675048828125,
|
| 215 |
+
"loss": 0.6702,
|
| 216 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 217 |
+
"rewards/chosen": 0.05479610711336136,
|
| 218 |
+
"rewards/margins": 0.05313174054026604,
|
| 219 |
+
"rewards/rejected": 0.001664361683651805,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.5439330543933054,
|
| 224 |
+
"grad_norm": 1.8823812053819848,
|
| 225 |
+
"learning_rate": 2.55479083351317e-07,
|
| 226 |
+
"logits/chosen": -1.2392737865447998,
|
| 227 |
+
"logits/rejected": -1.3330614566802979,
|
| 228 |
+
"logps/chosen": -348.48931884765625,
|
| 229 |
+
"logps/rejected": -315.95013427734375,
|
| 230 |
+
"loss": 0.6691,
|
| 231 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 232 |
+
"rewards/chosen": 0.06406990438699722,
|
| 233 |
+
"rewards/margins": 0.06992693990468979,
|
| 234 |
+
"rewards/rejected": -0.005857035517692566,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.5857740585774058,
|
| 239 |
+
"grad_norm": 5.44373598859388,
|
| 240 |
+
"learning_rate": 2.19029145890313e-07,
|
| 241 |
+
"logits/chosen": -1.2415839433670044,
|
| 242 |
+
"logits/rejected": -1.3235524892807007,
|
| 243 |
+
"logps/chosen": -326.3031005859375,
|
| 244 |
+
"logps/rejected": -311.26214599609375,
|
| 245 |
+
"loss": 0.6678,
|
| 246 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 247 |
+
"rewards/chosen": 0.054818809032440186,
|
| 248 |
+
"rewards/margins": 0.05220777913928032,
|
| 249 |
+
"rewards/rejected": 0.002611032919958234,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.6276150627615062,
|
| 254 |
+
"grad_norm": 1.939512273689077,
|
| 255 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 256 |
+
"logits/chosen": -1.227124571800232,
|
| 257 |
+
"logits/rejected": -1.3289756774902344,
|
| 258 |
+
"logps/chosen": -326.6729736328125,
|
| 259 |
+
"logps/rejected": -308.0332336425781,
|
| 260 |
+
"loss": 0.666,
|
| 261 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 262 |
+
"rewards/chosen": 0.035384777933359146,
|
| 263 |
+
"rewards/margins": 0.051828302443027496,
|
| 264 |
+
"rewards/rejected": -0.01644352450966835,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.6694560669456067,
|
| 269 |
+
"grad_norm": 1.9253405708401141,
|
| 270 |
+
"learning_rate": 1.488723393865766e-07,
|
| 271 |
+
"logits/chosen": -1.264524221420288,
|
| 272 |
+
"logits/rejected": -1.307389736175537,
|
| 273 |
+
"logps/chosen": -344.7317199707031,
|
| 274 |
+
"logps/rejected": -312.7427978515625,
|
| 275 |
+
"loss": 0.6612,
|
| 276 |
+
"rewards/accuracies": 0.684374988079071,
|
| 277 |
+
"rewards/chosen": 0.04708511754870415,
|
| 278 |
+
"rewards/margins": 0.07312874495983124,
|
| 279 |
+
"rewards/rejected": -0.02604362927377224,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.7112970711297071,
|
| 284 |
+
"grad_norm": 1.9507035640244035,
|
| 285 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 286 |
+
"logits/chosen": -1.2524497509002686,
|
| 287 |
+
"logits/rejected": -1.2847630977630615,
|
| 288 |
+
"logps/chosen": -332.83990478515625,
|
| 289 |
+
"logps/rejected": -329.17608642578125,
|
| 290 |
+
"loss": 0.6645,
|
| 291 |
+
"rewards/accuracies": 0.643750011920929,
|
| 292 |
+
"rewards/chosen": 0.03406905010342598,
|
| 293 |
+
"rewards/margins": 0.05726125091314316,
|
| 294 |
+
"rewards/rejected": -0.023192204535007477,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.7531380753138075,
|
| 299 |
+
"grad_norm": 1.8449874305732954,
|
| 300 |
+
"learning_rate": 8.729103716819111e-08,
|
| 301 |
+
"logits/chosen": -1.2439481019973755,
|
| 302 |
+
"logits/rejected": -1.2792072296142578,
|
| 303 |
+
"logps/chosen": -332.6515197753906,
|
| 304 |
+
"logps/rejected": -310.3517150878906,
|
| 305 |
+
"loss": 0.6615,
|
| 306 |
+
"rewards/accuracies": 0.6781250238418579,
|
| 307 |
+
"rewards/chosen": 0.04834641143679619,
|
| 308 |
+
"rewards/margins": 0.07643269002437592,
|
| 309 |
+
"rewards/rejected": -0.028086284175515175,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.7949790794979079,
|
| 314 |
+
"grad_norm": 2.2113847121012986,
|
| 315 |
+
"learning_rate": 6.138919252022435e-08,
|
| 316 |
+
"logits/chosen": -1.270702600479126,
|
| 317 |
+
"logits/rejected": -1.3235771656036377,
|
| 318 |
+
"logps/chosen": -313.1623229980469,
|
| 319 |
+
"logps/rejected": -310.3594665527344,
|
| 320 |
+
"loss": 0.6592,
|
| 321 |
+
"rewards/accuracies": 0.65625,
|
| 322 |
+
"rewards/chosen": 0.020944807678461075,
|
| 323 |
+
"rewards/margins": 0.0681905671954155,
|
| 324 |
+
"rewards/rejected": -0.04724575951695442,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.8368200836820083,
|
| 329 |
+
"grad_norm": 1.9219438856219404,
|
| 330 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 331 |
+
"logits/chosen": -1.2548792362213135,
|
| 332 |
+
"logits/rejected": -1.2600829601287842,
|
| 333 |
+
"logps/chosen": -339.12457275390625,
|
| 334 |
+
"logps/rejected": -337.1640625,
|
| 335 |
+
"loss": 0.6604,
|
| 336 |
+
"rewards/accuracies": 0.671875,
|
| 337 |
+
"rewards/chosen": 0.040566906332969666,
|
| 338 |
+
"rewards/margins": 0.07623802870512009,
|
| 339 |
+
"rewards/rejected": -0.03567112237215042,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.8368200836820083,
|
| 344 |
+
"eval_logits/chosen": -1.1793808937072754,
|
| 345 |
+
"eval_logits/rejected": -1.2523409128189087,
|
| 346 |
+
"eval_logps/chosen": -335.1041564941406,
|
| 347 |
+
"eval_logps/rejected": -317.20819091796875,
|
| 348 |
+
"eval_loss": 0.6639689803123474,
|
| 349 |
+
"eval_rewards/accuracies": 0.670634925365448,
|
| 350 |
+
"eval_rewards/chosen": 0.04226858913898468,
|
| 351 |
+
"eval_rewards/margins": 0.07336094230413437,
|
| 352 |
+
"eval_rewards/rejected": -0.031092360615730286,
|
| 353 |
+
"eval_runtime": 91.7242,
|
| 354 |
+
"eval_samples_per_second": 21.804,
|
| 355 |
+
"eval_steps_per_second": 0.687,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.8786610878661087,
|
| 360 |
+
"grad_norm": 2.0461724170667646,
|
| 361 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 362 |
+
"logits/chosen": -1.1933133602142334,
|
| 363 |
+
"logits/rejected": -1.3109954595565796,
|
| 364 |
+
"logps/chosen": -347.81976318359375,
|
| 365 |
+
"logps/rejected": -335.7594909667969,
|
| 366 |
+
"loss": 0.6613,
|
| 367 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 368 |
+
"rewards/chosen": 0.03740222379565239,
|
| 369 |
+
"rewards/margins": 0.05939198657870293,
|
| 370 |
+
"rewards/rejected": -0.021989766508340836,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.9205020920502092,
|
| 375 |
+
"grad_norm": 3.392645562933325,
|
| 376 |
+
"learning_rate": 9.57301420397924e-09,
|
| 377 |
+
"logits/chosen": -1.1992428302764893,
|
| 378 |
+
"logits/rejected": -1.243272066116333,
|
| 379 |
+
"logps/chosen": -338.14166259765625,
|
| 380 |
+
"logps/rejected": -323.56817626953125,
|
| 381 |
+
"loss": 0.6585,
|
| 382 |
+
"rewards/accuracies": 0.637499988079071,
|
| 383 |
+
"rewards/chosen": 0.04126737266778946,
|
| 384 |
+
"rewards/margins": 0.07338305562734604,
|
| 385 |
+
"rewards/rejected": -0.032115690410137177,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.9623430962343096,
|
| 390 |
+
"grad_norm": 1.9082006258080877,
|
| 391 |
+
"learning_rate": 2.158697848236607e-09,
|
| 392 |
+
"logits/chosen": -1.2281768321990967,
|
| 393 |
+
"logits/rejected": -1.2996705770492554,
|
| 394 |
+
"logps/chosen": -334.5081787109375,
|
| 395 |
+
"logps/rejected": -336.0547790527344,
|
| 396 |
+
"loss": 0.6611,
|
| 397 |
+
"rewards/accuracies": 0.606249988079071,
|
| 398 |
+
"rewards/chosen": 0.029226357117295265,
|
| 399 |
+
"rewards/margins": 0.054667878895998,
|
| 400 |
+
"rewards/rejected": -0.025441521778702736,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 1.0,
|
| 405 |
+
"step": 239,
|
| 406 |
+
"total_flos": 0.0,
|
| 407 |
+
"train_loss": 0.6732059223382543,
|
| 408 |
+
"train_runtime": 6951.1203,
|
| 409 |
+
"train_samples_per_second": 8.795,
|
| 410 |
+
"train_steps_per_second": 0.034
|
| 411 |
+
}
|
| 412 |
+
],
|
| 413 |
+
"logging_steps": 10,
|
| 414 |
+
"max_steps": 239,
|
| 415 |
+
"num_input_tokens_seen": 0,
|
| 416 |
+
"num_train_epochs": 1,
|
| 417 |
+
"save_steps": 100,
|
| 418 |
+
"stateful_callbacks": {
|
| 419 |
+
"TrainerControl": {
|
| 420 |
+
"args": {
|
| 421 |
+
"should_epoch_stop": false,
|
| 422 |
+
"should_evaluate": false,
|
| 423 |
+
"should_log": false,
|
| 424 |
+
"should_save": true,
|
| 425 |
+
"should_training_stop": true
|
| 426 |
+
},
|
| 427 |
+
"attributes": {}
|
| 428 |
+
}
|
| 429 |
+
},
|
| 430 |
+
"total_flos": 0.0,
|
| 431 |
+
"train_batch_size": 8,
|
| 432 |
+
"trial_name": null,
|
| 433 |
+
"trial_params": null
|
| 434 |
+
}
|