metadata
license: llama3
base_model: tsavage68/Summary_L3_1000steps_1e7rate_SFT2
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: Summary_L3_1000steps_1e5rate_01beta_CSFTDPO
results: []
Summary_L3_1000steps_1e5rate_01beta_CSFTDPO
This model is a fine-tuned version of tsavage68/Summary_L3_1000steps_1e7rate_SFT2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5961
- Rewards/chosen: -0.8715
- Rewards/rejected: -3.9531
- Rewards/accuracies: 0.1400
- Rewards/margins: 3.0816
- Logps/rejected: -54.7948
- Logps/chosen: -18.0977
- Logits/rejected: -1.3576
- Logits/chosen: -1.3527
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5546 | 0.2004 | 50 | 0.5961 | -0.8720 | -3.9451 | 0.1400 | 3.0730 | -54.7146 | -18.1031 | -1.3571 | -1.3522 |
| 0.6585 | 0.4008 | 100 | 0.5961 | -0.8712 | -3.9495 | 0.1400 | 3.0783 | -54.7588 | -18.0949 | -1.3575 | -1.3526 |
| 0.6238 | 0.6012 | 150 | 0.5961 | -0.8681 | -3.9389 | 0.1400 | 3.0707 | -54.6525 | -18.0641 | -1.3563 | -1.3514 |
| 0.6065 | 0.8016 | 200 | 0.5961 | -0.8725 | -3.9499 | 0.1400 | 3.0774 | -54.7626 | -18.1074 | -1.3568 | -1.3519 |
| 0.6238 | 1.0020 | 250 | 0.5961 | -0.8717 | -3.9513 | 0.1400 | 3.0796 | -54.7771 | -18.1000 | -1.3576 | -1.3527 |
| 0.6238 | 1.2024 | 300 | 0.5961 | -0.8725 | -3.9481 | 0.1400 | 3.0756 | -54.7450 | -18.1078 | -1.3571 | -1.3522 |
| 0.6238 | 1.4028 | 350 | 0.5961 | -0.8727 | -3.9498 | 0.1400 | 3.0771 | -54.7614 | -18.1094 | -1.3572 | -1.3523 |
| 0.5718 | 1.6032 | 400 | 0.5961 | -0.8724 | -3.9505 | 0.1400 | 3.0781 | -54.7691 | -18.1072 | -1.3573 | -1.3524 |
| 0.5892 | 1.8036 | 450 | 0.5961 | -0.8726 | -3.9502 | 0.1400 | 3.0776 | -54.7655 | -18.1083 | -1.3573 | -1.3523 |
| 0.5718 | 2.0040 | 500 | 0.5961 | -0.8717 | -3.9446 | 0.1400 | 3.0728 | -54.7095 | -18.1001 | -1.3575 | -1.3526 |
| 0.5718 | 2.2044 | 550 | 0.5961 | -0.8733 | -3.9538 | 0.1400 | 3.0805 | -54.8019 | -18.1157 | -1.3569 | -1.3521 |
| 0.5545 | 2.4048 | 600 | 0.5961 | -0.8691 | -3.9509 | 0.1400 | 3.0818 | -54.7729 | -18.0740 | -1.3573 | -1.3524 |
| 0.5199 | 2.6052 | 650 | 0.5961 | -0.8731 | -3.9531 | 0.1400 | 3.0800 | -54.7946 | -18.1135 | -1.3573 | -1.3524 |
| 0.6238 | 2.8056 | 700 | 0.5961 | -0.8719 | -3.9544 | 0.1400 | 3.0826 | -54.8080 | -18.1013 | -1.3581 | -1.3532 |
| 0.6065 | 3.0060 | 750 | 0.5961 | -0.8719 | -3.9517 | 0.1400 | 3.0798 | -54.7812 | -18.1017 | -1.3575 | -1.3526 |
| 0.6412 | 3.2064 | 800 | 0.5961 | -0.8706 | -3.9530 | 0.1400 | 3.0824 | -54.7941 | -18.0886 | -1.3574 | -1.3525 |
| 0.6585 | 3.4068 | 850 | 0.5961 | -0.8715 | -3.9512 | 0.1400 | 3.0798 | -54.7760 | -18.0975 | -1.3577 | -1.3529 |
| 0.6238 | 3.6072 | 900 | 0.5961 | -0.8715 | -3.9512 | 0.1400 | 3.0798 | -54.7760 | -18.0975 | -1.3577 | -1.3529 |
| 0.5372 | 3.8076 | 950 | 0.5961 | -0.8715 | -3.9531 | 0.1400 | 3.0816 | -54.7948 | -18.0977 | -1.3576 | -1.3527 |
| 0.6238 | 4.0080 | 1000 | 0.5961 | -0.8715 | -3.9531 | 0.1400 | 3.0816 | -54.7948 | -18.0977 | -1.3576 | -1.3527 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.0.0+cu117
- Datasets 2.20.0
- Tokenizers 0.19.1