Downloads · 30 days
4
13% of all-time downloads
model_hh_usp1_dpo1
guoyu-zhang/model_hh_usp1_dpo1
model_hh_usp1_dpo1 is a machine learning model from guoyu-zhang. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft.
should probably proofread and complete it, then remove this comment. --
All-time downloads
30
Public
Repo size
12.6 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors25.2 MB · 93%
At a glance
- Library
- peft
- Access
- Public
- Created
- Apr 16, 2024
- Updated
- Apr 16, 2024
- SHA
- eded45aa
Base models
Model card
From the Hugging Face model README
<!-- 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. -->
model_hh_usp1_dpo1
This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.3334
- Rewards/chosen: -12.5084
- Rewards/rejected: -14.9821
- Rewards/accuracies: 0.6800
- Rewards/margins: 2.4737
- Logps/rejected: -265.9678
- Logps/chosen: -237.7022
- Logits/rejected: -1.1144
- Logits/chosen: -1.1457
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0688 | 2.67 | 100 | 0.9409 | -3.4509 | -4.7255 | 0.6900 | 1.2746 | -163.4019 | -147.1273 | -1.4436 | -1.4535 |
| 0.0013 | 5.33 | 200 | 1.2767 | -10.7944 | -12.7570 | 0.6700 | 1.9626 | -243.7165 | -220.5616 | -1.3151 | -1.3441 |
| 0.0001 | 8.0 | 300 | 1.3152 | -12.2279 | -14.6398 | 0.6800 | 2.4118 | -262.5441 | -234.8975 | -1.1347 | -1.1653 |
| 0.0001 | 10.67 | 400 | 1.3231 | -12.3403 | -14.7735 | 0.6800 | 2.4333 | -263.8819 | -236.0207 | -1.1265 | -1.1574 |
| 0.0001 | 13.33 | 500 | 1.3286 | -12.4240 | -14.8779 | 0.6800 | 2.4538 | -264.9250 | -236.8581 | -1.1200 | -1.1512 |
| 0.0001 | 16.0 | 600 | 1.3344 | -12.4832 | -14.9513 | 0.6800 | 2.4681 | -265.6594 | -237.4504 | -1.1165 | -1.1478 |
| 0.0001 | 18.67 | 700 | 1.3322 | -12.5009 | -14.9737 | 0.6800 | 2.4728 | -265.8835 | -237.6271 | -1.1150 | -1.1460 |
| 0.0001 | 21.33 | 800 | 1.3324 | -12.5055 | -14.9833 | 0.6800 | 2.4778 | -265.9792 | -237.6730 | -1.1147 | -1.1457 |
| 0.0001 | 24.0 | 900 | 1.3390 | -12.5121 | -14.9772 | 0.6800 | 2.4650 | -265.9181 | -237.7395 | -1.1143 | -1.1457 |
| 0.0 | 26.67 | 1000 | 1.3334 | -12.5084 | -14.9821 | 0.6800 | 2.4737 | -265.9678 | -237.7022 | -1.1144 | -1.1457 |
Framework versions
- PEFT 0.10.0
- Transformers 4.39.3
- Pytorch 2.2.2+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Library
- peft
- Created
- Apr 16, 2024
- Updated
- Apr 16, 2024