File size: 3,827 Bytes
1e1c504 a7c9c69 1e1c504 a7c9c69 1e1c504 a7c9c69 1e1c504 c5be93f 1e1c504 c5be93f 1e1c504 c5be93f 1e1c504 |
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 81 82 83 84 85 86 87 |
---
license: mit
library_name: peft
tags:
- alignment-handbook
- generated_from_trainer
- trl
- dpo
- generated_from_trainer
datasets:
- HuggingFaceH4/ultrafeedback_binarized
base_model: microsoft/phi-2
model-index:
- name: phi-2-dpo-ultrachat-lora
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. -->
# phi-2-dpo-ultrachat-lora
This model is a fine-tuned version of [lole25/phi-2-sft-ultrachat-lora](https://huggingface.co/lole25/phi-2-sft-ultrachat-lora) on the HuggingFaceH4/ultrafeedback_binarized dataset.
It achieves the following results on the evaluation set:
- Loss: 0.6872
- Rewards/chosen: -0.0312
- Rewards/rejected: -0.0436
- Rewards/accuracies: 0.3340
- Rewards/margins: 0.0124
- Logps/rejected: -98.5542
- Logps/chosen: -94.8435
- Logits/rejected: 0.7532
- Logits/chosen: 0.7326
## 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-06
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 4
- 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: 2
### Training results
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
| 0.693 | 0.21 | 100 | 0.7998 | 0.8176 | -91.7748 | -94.2804 | 0.6931 | 0.2680 | -0.0005 | 0.0004 | -0.0008 |
| 0.6922 | 0.42 | 200 | 0.7941 | 0.8121 | -91.9068 | -94.5141 | 0.6924 | 0.3020 | -0.0018 | 0.0014 | -0.0032 |
| 0.6917 | 0.63 | 300 | 0.7870 | 0.8057 | -92.2189 | -94.9659 | 0.6917 | 0.3100 | -0.0049 | 0.0028 | -0.0077 |
| 0.6905 | 0.84 | 400 | 0.7827 | 0.8012 | -92.4247 | -95.2509 | 0.6913 | 0.3280 | -0.0070 | 0.0036 | -0.0105 |
| 0.6898 | 1.05 | 500 | 0.6900 | -0.0142 | -0.0205 | 0.3360 | 0.0064 | -96.2490 | -93.1429 | 0.7903 | 0.7711 |
| 0.6882 | 1.26 | 600 | 0.6887 | -0.0217 | -0.0306 | 0.3340 | 0.0089 | -97.2594 | -93.8981 | 0.7722 | 0.7527 |
| 0.6858 | 1.47 | 700 | 0.6879 | -0.0274 | -0.0383 | 0.3280 | 0.0108 | -98.0249 | -94.4717 | 0.7600 | 0.7395 |
| 0.6857 | 1.67 | 800 | 0.6874 | -0.0303 | -0.0423 | 0.3340 | 0.0120 | -98.4270 | -94.7618 | 0.7548 | 0.7341 |
| 0.6866 | 1.88 | 900 | 0.6872 | -0.0313 | -0.0437 | 0.3420 | 0.0124 | -98.5655 | -94.8550 | 0.7528 | 0.7321 |
### Framework versions
- PEFT 0.7.1
- Transformers 4.36.2
- Pytorch 2.1.2+cu118
- Datasets 2.14.6
- Tokenizers 0.15.2 |