Model save
Browse files- README.md +77 -0
- adapter_model.safetensors +1 -1
- all_results.json +8 -0
- train_results.json +8 -0
- trainer_state.json +526 -0
README.md
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-7b-dpo-qlora
|
| 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: 0.5735
|
| 22 |
+
- Rewards/chosen: -0.6770
|
| 23 |
+
- Rewards/rejected: -1.1070
|
| 24 |
+
- Rewards/accuracies: 0.6940
|
| 25 |
+
- Rewards/margins: 0.4300
|
| 26 |
+
- Logps/rejected: -351.8942
|
| 27 |
+
- Logps/chosen: -331.1508
|
| 28 |
+
- Logits/rejected: -1.4599
|
| 29 |
+
- Logits/chosen: -1.7015
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-06
|
| 49 |
+
- train_batch_size: 2
|
| 50 |
+
- eval_batch_size: 2
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 4
|
| 54 |
+
- gradient_accumulation_steps: 4
|
| 55 |
+
- total_train_batch_size: 32
|
| 56 |
+
- total_eval_batch_size: 8
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.6269 | 0.32 | 100 | 0.6269 | -0.2377 | -0.4431 | 0.6820 | 0.2054 | -285.4985 | -287.2169 | -2.2566 | -2.3666 |
|
| 67 |
+
| 0.6332 | 0.64 | 200 | 0.5821 | -0.5909 | -0.9588 | 0.7060 | 0.3679 | -337.0687 | -322.5442 | -1.6871 | -1.8938 |
|
| 68 |
+
| 0.5648 | 0.96 | 300 | 0.5735 | -0.6770 | -1.1070 | 0.6940 | 0.4300 | -351.8942 | -331.1508 | -1.4599 | -1.7015 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- PEFT 0.7.1
|
| 74 |
+
- Transformers 4.36.2
|
| 75 |
+
- Pytorch 2.2.2+cu121
|
| 76 |
+
- Datasets 2.14.6
|
| 77 |
+
- Tokenizers 0.15.2
|
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 sha256:28defe27fdc102149779b9769d25766751de171b48982e943278effb731ebb99
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6116275028922619,
|
| 4 |
+
"train_runtime": 6907.8509,
|
| 5 |
+
"train_samples": 10000,
|
| 6 |
+
"train_samples_per_second": 1.448,
|
| 7 |
+
"train_steps_per_second": 0.045
|
| 8 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6116275028922619,
|
| 4 |
+
"train_runtime": 6907.8509,
|
| 5 |
+
"train_samples": 10000,
|
| 6 |
+
"train_samples_per_second": 1.448,
|
| 7 |
+
"train_steps_per_second": 0.045
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,526 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9984,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 312,
|
| 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": 1.5625e-07,
|
| 14 |
+
"logits/chosen": -2.7731900215148926,
|
| 15 |
+
"logits/rejected": -2.6362287998199463,
|
| 16 |
+
"logps/chosen": -356.1260070800781,
|
| 17 |
+
"logps/rejected": -311.3892822265625,
|
| 18 |
+
"loss": 0.6931,
|
| 19 |
+
"rewards/accuracies": 0.0,
|
| 20 |
+
"rewards/chosen": 0.0,
|
| 21 |
+
"rewards/margins": 0.0,
|
| 22 |
+
"rewards/rejected": 0.0,
|
| 23 |
+
"step": 1
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.03,
|
| 27 |
+
"learning_rate": 1.5625e-06,
|
| 28 |
+
"logits/chosen": -2.3915464878082275,
|
| 29 |
+
"logits/rejected": -2.3424172401428223,
|
| 30 |
+
"logps/chosen": -243.08827209472656,
|
| 31 |
+
"logps/rejected": -240.88124084472656,
|
| 32 |
+
"loss": 0.6927,
|
| 33 |
+
"rewards/accuracies": 0.5,
|
| 34 |
+
"rewards/chosen": 0.004824994597584009,
|
| 35 |
+
"rewards/margins": 0.001562346238642931,
|
| 36 |
+
"rewards/rejected": 0.003262649057433009,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.06,
|
| 41 |
+
"learning_rate": 3.125e-06,
|
| 42 |
+
"logits/chosen": -2.6855998039245605,
|
| 43 |
+
"logits/rejected": -2.503112316131592,
|
| 44 |
+
"logps/chosen": -276.1568908691406,
|
| 45 |
+
"logps/rejected": -245.57150268554688,
|
| 46 |
+
"loss": 0.687,
|
| 47 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 48 |
+
"rewards/chosen": 0.04376252368092537,
|
| 49 |
+
"rewards/margins": 0.011996113695204258,
|
| 50 |
+
"rewards/rejected": 0.03176640719175339,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.1,
|
| 55 |
+
"learning_rate": 4.6875000000000004e-06,
|
| 56 |
+
"logits/chosen": -2.5015687942504883,
|
| 57 |
+
"logits/rejected": -2.448686122894287,
|
| 58 |
+
"logps/chosen": -244.99642944335938,
|
| 59 |
+
"logps/rejected": -249.10916137695312,
|
| 60 |
+
"loss": 0.6832,
|
| 61 |
+
"rewards/accuracies": 0.625,
|
| 62 |
+
"rewards/chosen": 0.05081823468208313,
|
| 63 |
+
"rewards/margins": 0.018069546669721603,
|
| 64 |
+
"rewards/rejected": 0.032748688012361526,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.13,
|
| 69 |
+
"learning_rate": 4.989935734988098e-06,
|
| 70 |
+
"logits/chosen": -2.5158021450042725,
|
| 71 |
+
"logits/rejected": -2.337573289871216,
|
| 72 |
+
"logps/chosen": -272.7821350097656,
|
| 73 |
+
"logps/rejected": -227.36007690429688,
|
| 74 |
+
"loss": 0.6668,
|
| 75 |
+
"rewards/accuracies": 0.675000011920929,
|
| 76 |
+
"rewards/chosen": 0.0395994558930397,
|
| 77 |
+
"rewards/margins": 0.04740050435066223,
|
| 78 |
+
"rewards/rejected": -0.00780104985460639,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.16,
|
| 83 |
+
"learning_rate": 4.949188496058089e-06,
|
| 84 |
+
"logits/chosen": -2.430145502090454,
|
| 85 |
+
"logits/rejected": -2.4263150691986084,
|
| 86 |
+
"logps/chosen": -249.273681640625,
|
| 87 |
+
"logps/rejected": -266.5956726074219,
|
| 88 |
+
"loss": 0.6528,
|
| 89 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 90 |
+
"rewards/chosen": 0.02035255916416645,
|
| 91 |
+
"rewards/margins": 0.059511054307222366,
|
| 92 |
+
"rewards/rejected": -0.03915848955512047,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.19,
|
| 97 |
+
"learning_rate": 4.8776412907378845e-06,
|
| 98 |
+
"logits/chosen": -2.4893181324005127,
|
| 99 |
+
"logits/rejected": -2.418604612350464,
|
| 100 |
+
"logps/chosen": -301.8047790527344,
|
| 101 |
+
"logps/rejected": -252.42892456054688,
|
| 102 |
+
"loss": 0.6591,
|
| 103 |
+
"rewards/accuracies": 0.612500011920929,
|
| 104 |
+
"rewards/chosen": -0.06781601160764694,
|
| 105 |
+
"rewards/margins": 0.07723621279001236,
|
| 106 |
+
"rewards/rejected": -0.1450522094964981,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.22,
|
| 111 |
+
"learning_rate": 4.7761938666470405e-06,
|
| 112 |
+
"logits/chosen": -2.4578957557678223,
|
| 113 |
+
"logits/rejected": -2.4078097343444824,
|
| 114 |
+
"logps/chosen": -259.1146545410156,
|
| 115 |
+
"logps/rejected": -255.2762908935547,
|
| 116 |
+
"loss": 0.6412,
|
| 117 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 118 |
+
"rewards/chosen": -0.09445185959339142,
|
| 119 |
+
"rewards/margins": 0.15250881016254425,
|
| 120 |
+
"rewards/rejected": -0.24696068465709686,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.26,
|
| 125 |
+
"learning_rate": 4.646121984004666e-06,
|
| 126 |
+
"logits/chosen": -2.5219717025756836,
|
| 127 |
+
"logits/rejected": -2.3697924613952637,
|
| 128 |
+
"logps/chosen": -289.8721618652344,
|
| 129 |
+
"logps/rejected": -306.9769287109375,
|
| 130 |
+
"loss": 0.6357,
|
| 131 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 132 |
+
"rewards/chosen": -0.21299012005329132,
|
| 133 |
+
"rewards/margins": 0.16702046990394592,
|
| 134 |
+
"rewards/rejected": -0.38001060485839844,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.29,
|
| 139 |
+
"learning_rate": 4.4890613722044526e-06,
|
| 140 |
+
"logits/chosen": -2.427962064743042,
|
| 141 |
+
"logits/rejected": -2.326305866241455,
|
| 142 |
+
"logps/chosen": -265.3756408691406,
|
| 143 |
+
"logps/rejected": -262.7252197265625,
|
| 144 |
+
"loss": 0.6272,
|
| 145 |
+
"rewards/accuracies": 0.737500011920929,
|
| 146 |
+
"rewards/chosen": -0.16604574024677277,
|
| 147 |
+
"rewards/margins": 0.19639183580875397,
|
| 148 |
+
"rewards/rejected": -0.36243754625320435,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.32,
|
| 153 |
+
"learning_rate": 4.3069871595684795e-06,
|
| 154 |
+
"logits/chosen": -2.213723659515381,
|
| 155 |
+
"logits/rejected": -2.217102527618408,
|
| 156 |
+
"logps/chosen": -245.6179962158203,
|
| 157 |
+
"logps/rejected": -293.12518310546875,
|
| 158 |
+
"loss": 0.6269,
|
| 159 |
+
"rewards/accuracies": 0.612500011920929,
|
| 160 |
+
"rewards/chosen": -0.38440248370170593,
|
| 161 |
+
"rewards/margins": 0.126637801527977,
|
| 162 |
+
"rewards/rejected": -0.5110402703285217,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.32,
|
| 167 |
+
"eval_logits/chosen": -2.3665878772735596,
|
| 168 |
+
"eval_logits/rejected": -2.256598949432373,
|
| 169 |
+
"eval_logps/chosen": -287.2168884277344,
|
| 170 |
+
"eval_logps/rejected": -285.49847412109375,
|
| 171 |
+
"eval_loss": 0.6268974542617798,
|
| 172 |
+
"eval_rewards/accuracies": 0.6819999814033508,
|
| 173 |
+
"eval_rewards/chosen": -0.23765824735164642,
|
| 174 |
+
"eval_rewards/margins": 0.2054254114627838,
|
| 175 |
+
"eval_rewards/rejected": -0.44308364391326904,
|
| 176 |
+
"eval_runtime": 543.2957,
|
| 177 |
+
"eval_samples_per_second": 3.681,
|
| 178 |
+
"eval_steps_per_second": 0.46,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.35,
|
| 183 |
+
"learning_rate": 4.102189034962561e-06,
|
| 184 |
+
"logits/chosen": -2.338050603866577,
|
| 185 |
+
"logits/rejected": -2.2199347019195557,
|
| 186 |
+
"logps/chosen": -304.7019958496094,
|
| 187 |
+
"logps/rejected": -283.5575256347656,
|
| 188 |
+
"loss": 0.6232,
|
| 189 |
+
"rewards/accuracies": 0.6875,
|
| 190 |
+
"rewards/chosen": -0.18706437945365906,
|
| 191 |
+
"rewards/margins": 0.24619019031524658,
|
| 192 |
+
"rewards/rejected": -0.43325456976890564,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.38,
|
| 197 |
+
"learning_rate": 3.8772424536302565e-06,
|
| 198 |
+
"logits/chosen": -2.199939250946045,
|
| 199 |
+
"logits/rejected": -2.1462173461914062,
|
| 200 |
+
"logps/chosen": -280.5738525390625,
|
| 201 |
+
"logps/rejected": -272.75537109375,
|
| 202 |
+
"loss": 0.6256,
|
| 203 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 204 |
+
"rewards/chosen": -0.23016035556793213,
|
| 205 |
+
"rewards/margins": 0.2529276907444,
|
| 206 |
+
"rewards/rejected": -0.48308807611465454,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.42,
|
| 211 |
+
"learning_rate": 3.634976249348867e-06,
|
| 212 |
+
"logits/chosen": -2.4285922050476074,
|
| 213 |
+
"logits/rejected": -2.252119541168213,
|
| 214 |
+
"logps/chosen": -337.8984375,
|
| 215 |
+
"logps/rejected": -329.4248962402344,
|
| 216 |
+
"loss": 0.6299,
|
| 217 |
+
"rewards/accuracies": 0.737500011920929,
|
| 218 |
+
"rewards/chosen": -0.27586501836776733,
|
| 219 |
+
"rewards/margins": 0.285078763961792,
|
| 220 |
+
"rewards/rejected": -0.5609437823295593,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.45,
|
| 225 |
+
"learning_rate": 3.3784370602033572e-06,
|
| 226 |
+
"logits/chosen": -2.072373628616333,
|
| 227 |
+
"logits/rejected": -1.9053455591201782,
|
| 228 |
+
"logps/chosen": -251.76571655273438,
|
| 229 |
+
"logps/rejected": -285.0694885253906,
|
| 230 |
+
"loss": 0.6067,
|
| 231 |
+
"rewards/accuracies": 0.6875,
|
| 232 |
+
"rewards/chosen": -0.5010747909545898,
|
| 233 |
+
"rewards/margins": 0.26867786049842834,
|
| 234 |
+
"rewards/rejected": -0.7697526216506958,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.48,
|
| 239 |
+
"learning_rate": 3.1108510153447352e-06,
|
| 240 |
+
"logits/chosen": -2.21221661567688,
|
| 241 |
+
"logits/rejected": -2.136280059814453,
|
| 242 |
+
"logps/chosen": -338.2016296386719,
|
| 243 |
+
"logps/rejected": -331.0526428222656,
|
| 244 |
+
"loss": 0.608,
|
| 245 |
+
"rewards/accuracies": 0.625,
|
| 246 |
+
"rewards/chosen": -0.639680027961731,
|
| 247 |
+
"rewards/margins": 0.2550516426563263,
|
| 248 |
+
"rewards/rejected": -0.8947317004203796,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.51,
|
| 253 |
+
"learning_rate": 2.835583164544139e-06,
|
| 254 |
+
"logits/chosen": -2.2209646701812744,
|
| 255 |
+
"logits/rejected": -2.022948980331421,
|
| 256 |
+
"logps/chosen": -377.3534851074219,
|
| 257 |
+
"logps/rejected": -344.77252197265625,
|
| 258 |
+
"loss": 0.5937,
|
| 259 |
+
"rewards/accuracies": 0.737500011920929,
|
| 260 |
+
"rewards/chosen": -0.7194479703903198,
|
| 261 |
+
"rewards/margins": 0.39620086550712585,
|
| 262 |
+
"rewards/rejected": -1.115648865699768,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.54,
|
| 267 |
+
"learning_rate": 2.556095160739513e-06,
|
| 268 |
+
"logits/chosen": -2.1350314617156982,
|
| 269 |
+
"logits/rejected": -1.85476553440094,
|
| 270 |
+
"logps/chosen": -351.29638671875,
|
| 271 |
+
"logps/rejected": -354.8650817871094,
|
| 272 |
+
"loss": 0.6069,
|
| 273 |
+
"rewards/accuracies": 0.612500011920929,
|
| 274 |
+
"rewards/chosen": -0.7903974652290344,
|
| 275 |
+
"rewards/margins": 0.24958536028862,
|
| 276 |
+
"rewards/rejected": -1.039982795715332,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.58,
|
| 281 |
+
"learning_rate": 2.2759017277414165e-06,
|
| 282 |
+
"logits/chosen": -2.0943400859832764,
|
| 283 |
+
"logits/rejected": -1.8893616199493408,
|
| 284 |
+
"logps/chosen": -322.147216796875,
|
| 285 |
+
"logps/rejected": -327.81304931640625,
|
| 286 |
+
"loss": 0.6252,
|
| 287 |
+
"rewards/accuracies": 0.625,
|
| 288 |
+
"rewards/chosen": -0.6768954992294312,
|
| 289 |
+
"rewards/margins": 0.20395174622535706,
|
| 290 |
+
"rewards/rejected": -0.8808472752571106,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.61,
|
| 295 |
+
"learning_rate": 1.9985264605418185e-06,
|
| 296 |
+
"logits/chosen": -1.9419981241226196,
|
| 297 |
+
"logits/rejected": -1.7324016094207764,
|
| 298 |
+
"logps/chosen": -328.23760986328125,
|
| 299 |
+
"logps/rejected": -314.13922119140625,
|
| 300 |
+
"loss": 0.584,
|
| 301 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 302 |
+
"rewards/chosen": -0.4452829360961914,
|
| 303 |
+
"rewards/margins": 0.4189114570617676,
|
| 304 |
+
"rewards/rejected": -0.864194393157959,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.64,
|
| 309 |
+
"learning_rate": 1.7274575140626318e-06,
|
| 310 |
+
"logits/chosen": -2.144902229309082,
|
| 311 |
+
"logits/rejected": -1.7156444787979126,
|
| 312 |
+
"logps/chosen": -362.327880859375,
|
| 313 |
+
"logps/rejected": -322.9747619628906,
|
| 314 |
+
"loss": 0.6332,
|
| 315 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 316 |
+
"rewards/chosen": -0.4486660957336426,
|
| 317 |
+
"rewards/margins": 0.3436250388622284,
|
| 318 |
+
"rewards/rejected": -0.7922911047935486,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.64,
|
| 323 |
+
"eval_logits/chosen": -1.893760323524475,
|
| 324 |
+
"eval_logits/rejected": -1.6871448755264282,
|
| 325 |
+
"eval_logps/chosen": -322.544189453125,
|
| 326 |
+
"eval_logps/rejected": -337.0687255859375,
|
| 327 |
+
"eval_loss": 0.5820500254631042,
|
| 328 |
+
"eval_rewards/accuracies": 0.7059999704360962,
|
| 329 |
+
"eval_rewards/chosen": -0.5909315943717957,
|
| 330 |
+
"eval_rewards/margins": 0.3678547739982605,
|
| 331 |
+
"eval_rewards/rejected": -0.9587863683700562,
|
| 332 |
+
"eval_runtime": 543.1459,
|
| 333 |
+
"eval_samples_per_second": 3.682,
|
| 334 |
+
"eval_steps_per_second": 0.46,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.67,
|
| 339 |
+
"learning_rate": 1.466103737583699e-06,
|
| 340 |
+
"logits/chosen": -1.8559290170669556,
|
| 341 |
+
"logits/rejected": -1.7014697790145874,
|
| 342 |
+
"logps/chosen": -324.19256591796875,
|
| 343 |
+
"logps/rejected": -352.70697021484375,
|
| 344 |
+
"loss": 0.552,
|
| 345 |
+
"rewards/accuracies": 0.762499988079071,
|
| 346 |
+
"rewards/chosen": -0.6294658780097961,
|
| 347 |
+
"rewards/margins": 0.45733365416526794,
|
| 348 |
+
"rewards/rejected": -1.0867995023727417,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.7,
|
| 353 |
+
"learning_rate": 1.217751806485235e-06,
|
| 354 |
+
"logits/chosen": -1.8568174839019775,
|
| 355 |
+
"logits/rejected": -1.6362855434417725,
|
| 356 |
+
"logps/chosen": -356.0939636230469,
|
| 357 |
+
"logps/rejected": -389.1434326171875,
|
| 358 |
+
"loss": 0.5765,
|
| 359 |
+
"rewards/accuracies": 0.75,
|
| 360 |
+
"rewards/chosen": -0.6206706762313843,
|
| 361 |
+
"rewards/margins": 0.5877247452735901,
|
| 362 |
+
"rewards/rejected": -1.2083956003189087,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.74,
|
| 367 |
+
"learning_rate": 9.855248903979505e-07,
|
| 368 |
+
"logits/chosen": -1.9677798748016357,
|
| 369 |
+
"logits/rejected": -1.8001766204833984,
|
| 370 |
+
"logps/chosen": -333.63409423828125,
|
| 371 |
+
"logps/rejected": -372.82232666015625,
|
| 372 |
+
"loss": 0.5961,
|
| 373 |
+
"rewards/accuracies": 0.75,
|
| 374 |
+
"rewards/chosen": -0.6273213028907776,
|
| 375 |
+
"rewards/margins": 0.4957484304904938,
|
| 376 |
+
"rewards/rejected": -1.1230696439743042,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.77,
|
| 381 |
+
"learning_rate": 7.723433775328385e-07,
|
| 382 |
+
"logits/chosen": -1.6269299983978271,
|
| 383 |
+
"logits/rejected": -1.5314247608184814,
|
| 384 |
+
"logps/chosen": -343.7135314941406,
|
| 385 |
+
"logps/rejected": -360.920166015625,
|
| 386 |
+
"loss": 0.5733,
|
| 387 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 388 |
+
"rewards/chosen": -0.6133186221122742,
|
| 389 |
+
"rewards/margins": 0.3855026662349701,
|
| 390 |
+
"rewards/rejected": -0.9988213777542114,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.8,
|
| 395 |
+
"learning_rate": 5.808881491049723e-07,
|
| 396 |
+
"logits/chosen": -1.8092960119247437,
|
| 397 |
+
"logits/rejected": -1.4363586902618408,
|
| 398 |
+
"logps/chosen": -302.46234130859375,
|
| 399 |
+
"logps/rejected": -305.09393310546875,
|
| 400 |
+
"loss": 0.5822,
|
| 401 |
+
"rewards/accuracies": 0.6875,
|
| 402 |
+
"rewards/chosen": -0.7107473611831665,
|
| 403 |
+
"rewards/margins": 0.3086285889148712,
|
| 404 |
+
"rewards/rejected": -1.0193760395050049,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.83,
|
| 409 |
+
"learning_rate": 4.1356686569674344e-07,
|
| 410 |
+
"logits/chosen": -2.0522053241729736,
|
| 411 |
+
"logits/rejected": -1.6467043161392212,
|
| 412 |
+
"logps/chosen": -370.14019775390625,
|
| 413 |
+
"logps/rejected": -350.79364013671875,
|
| 414 |
+
"loss": 0.5594,
|
| 415 |
+
"rewards/accuracies": 0.699999988079071,
|
| 416 |
+
"rewards/chosen": -0.6428462266921997,
|
| 417 |
+
"rewards/margins": 0.41562938690185547,
|
| 418 |
+
"rewards/rejected": -1.0584756135940552,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.86,
|
| 423 |
+
"learning_rate": 2.7248368952908055e-07,
|
| 424 |
+
"logits/chosen": -1.736702561378479,
|
| 425 |
+
"logits/rejected": -1.5174537897109985,
|
| 426 |
+
"logps/chosen": -293.9969177246094,
|
| 427 |
+
"logps/rejected": -317.34844970703125,
|
| 428 |
+
"loss": 0.5803,
|
| 429 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 430 |
+
"rewards/chosen": -0.5701481103897095,
|
| 431 |
+
"rewards/margins": 0.500605583190918,
|
| 432 |
+
"rewards/rejected": -1.0707536935806274,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.9,
|
| 437 |
+
"learning_rate": 1.59412823400657e-07,
|
| 438 |
+
"logits/chosen": -1.6159837245941162,
|
| 439 |
+
"logits/rejected": -1.2289941310882568,
|
| 440 |
+
"logps/chosen": -342.9421081542969,
|
| 441 |
+
"logps/rejected": -372.6164855957031,
|
| 442 |
+
"loss": 0.5578,
|
| 443 |
+
"rewards/accuracies": 0.737500011920929,
|
| 444 |
+
"rewards/chosen": -0.7729519605636597,
|
| 445 |
+
"rewards/margins": 0.49912238121032715,
|
| 446 |
+
"rewards/rejected": -1.2720743417739868,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.93,
|
| 451 |
+
"learning_rate": 7.577619905828281e-08,
|
| 452 |
+
"logits/chosen": -1.502423644065857,
|
| 453 |
+
"logits/rejected": -1.456081509590149,
|
| 454 |
+
"logps/chosen": -329.8805236816406,
|
| 455 |
+
"logps/rejected": -356.34417724609375,
|
| 456 |
+
"loss": 0.5827,
|
| 457 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 458 |
+
"rewards/chosen": -0.678324818611145,
|
| 459 |
+
"rewards/margins": 0.42007485032081604,
|
| 460 |
+
"rewards/rejected": -1.0983997583389282,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.96,
|
| 465 |
+
"learning_rate": 2.262559558016325e-08,
|
| 466 |
+
"logits/chosen": -1.6769917011260986,
|
| 467 |
+
"logits/rejected": -1.406165361404419,
|
| 468 |
+
"logps/chosen": -328.3318786621094,
|
| 469 |
+
"logps/rejected": -353.73968505859375,
|
| 470 |
+
"loss": 0.5648,
|
| 471 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 472 |
+
"rewards/chosen": -0.6493935585021973,
|
| 473 |
+
"rewards/margins": 0.43068727850914,
|
| 474 |
+
"rewards/rejected": -1.0800807476043701,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.96,
|
| 479 |
+
"eval_logits/chosen": -1.7015434503555298,
|
| 480 |
+
"eval_logits/rejected": -1.4598934650421143,
|
| 481 |
+
"eval_logps/chosen": -331.1508483886719,
|
| 482 |
+
"eval_logps/rejected": -351.8941955566406,
|
| 483 |
+
"eval_loss": 0.5735270977020264,
|
| 484 |
+
"eval_rewards/accuracies": 0.6940000057220459,
|
| 485 |
+
"eval_rewards/chosen": -0.6769981980323792,
|
| 486 |
+
"eval_rewards/margins": 0.4300425946712494,
|
| 487 |
+
"eval_rewards/rejected": -1.1070406436920166,
|
| 488 |
+
"eval_runtime": 542.9185,
|
| 489 |
+
"eval_samples_per_second": 3.684,
|
| 490 |
+
"eval_steps_per_second": 0.46,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.99,
|
| 495 |
+
"learning_rate": 6.294126437336734e-10,
|
| 496 |
+
"logits/chosen": -1.7649977207183838,
|
| 497 |
+
"logits/rejected": -1.521240234375,
|
| 498 |
+
"logps/chosen": -326.1722717285156,
|
| 499 |
+
"logps/rejected": -356.2889709472656,
|
| 500 |
+
"loss": 0.5603,
|
| 501 |
+
"rewards/accuracies": 0.737500011920929,
|
| 502 |
+
"rewards/chosen": -0.6321894526481628,
|
| 503 |
+
"rewards/margins": 0.4693359434604645,
|
| 504 |
+
"rewards/rejected": -1.1015253067016602,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 1.0,
|
| 509 |
+
"step": 312,
|
| 510 |
+
"total_flos": 0.0,
|
| 511 |
+
"train_loss": 0.6116275028922619,
|
| 512 |
+
"train_runtime": 6907.8509,
|
| 513 |
+
"train_samples_per_second": 1.448,
|
| 514 |
+
"train_steps_per_second": 0.045
|
| 515 |
+
}
|
| 516 |
+
],
|
| 517 |
+
"logging_steps": 10,
|
| 518 |
+
"max_steps": 312,
|
| 519 |
+
"num_input_tokens_seen": 0,
|
| 520 |
+
"num_train_epochs": 1,
|
| 521 |
+
"save_steps": 100,
|
| 522 |
+
"total_flos": 0.0,
|
| 523 |
+
"train_batch_size": 2,
|
| 524 |
+
"trial_name": null,
|
| 525 |
+
"trial_params": null
|
| 526 |
+
}
|