File size: 3,538 Bytes
f399d89 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 |
---
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: arceeai-cpt-sft-dpo-full
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# arceeai-cpt-sft-dpo-full
This model was trained from scratch on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4290
- Rewards/chosen: -2.1400
- Rewards/rejected: -3.3730
- Rewards/accuracies: 0.7680
- Rewards/margins: 1.2330
- Logps/rejected: -667.6970
- Logps/chosen: -599.8088
- Logits/rejected: -3.8772
- Logits/chosen: -3.7995
## 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: 5e-07
- train_batch_size: 4
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- total_eval_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1
### Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.6051 | 0.1 | 100 | 0.5848 | -0.2214 | -0.5948 | 0.6920 | 0.3733 | -389.8707 | -407.9477 | -3.9800 | -3.8777 |
| 0.5134 | 0.21 | 200 | 0.5025 | -1.5935 | -2.5024 | 0.7160 | 0.9089 | -580.6380 | -545.1561 | -3.9154 | -3.8176 |
| 0.4489 | 0.31 | 300 | 0.4614 | -1.5620 | -2.6097 | 0.7760 | 1.0477 | -591.3610 | -542.0072 | -3.7594 | -3.6703 |
| 0.4359 | 0.42 | 400 | 0.4467 | -2.0879 | -3.2160 | 0.7680 | 1.1281 | -651.9947 | -594.5918 | -3.7022 | -3.6221 |
| 0.4271 | 0.52 | 500 | 0.4441 | -2.0549 | -3.2181 | 0.7840 | 1.1631 | -652.2027 | -591.3006 | -3.8189 | -3.7408 |
| 0.4181 | 0.63 | 600 | 0.4366 | -1.9876 | -3.1678 | 0.7760 | 1.1802 | -647.1777 | -584.5698 | -3.7950 | -3.7170 |
| 0.4 | 0.73 | 700 | 0.4317 | -2.1647 | -3.3521 | 0.7640 | 1.1874 | -665.6046 | -602.2762 | -3.8739 | -3.7970 |
| 0.4123 | 0.84 | 800 | 0.4291 | -2.2039 | -3.4491 | 0.7680 | 1.2453 | -675.3075 | -606.1934 | -3.8606 | -3.7827 |
| 0.4394 | 0.94 | 900 | 0.4292 | -2.1325 | -3.3633 | 0.7680 | 1.2308 | -666.7250 | -599.0574 | -3.8777 | -3.8001 |
### Framework versions
- Transformers 4.36.2
- Pytorch 2.1.2+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0
|