Llama-3.1-8B-Magpie-SFT-500K-UltraDPO-02
This model is a fine-tuned version of Magpie-Align/Llama-3.1-8B-Magpie-SFT-500K on the princeton-nlp/llama3-ultrafeedback-armorm dataset. It achieves the following results on the evaluation set:
- Loss: 0.5227
- Rewards/chosen: -2.2584
- Rewards/rejected: -2.9580
- Rewards/accuracies: 0.7317
- Rewards/margins: 0.6997
- Logps/rejected: -565.8957
- Logps/chosen: -498.9081
- Logits/rejected: -0.7276
- Logits/chosen: -0.7074
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: 2e-07
- train_batch_size: 2
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 16
- total_train_batch_size: 128
- 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 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.6681 | 0.2138 | 100 | 0.6601 | -0.1862 | -0.2613 | 0.6667 | 0.0751 | -296.2193 | -291.6893 | -0.7011 | -0.6901 |
0.6268 | 0.4275 | 200 | 0.5620 | -1.5976 | -2.0872 | 0.7114 | 0.4897 | -478.8154 | -432.8277 | -0.6499 | -0.6307 |
0.5079 | 0.6413 | 300 | 0.5348 | -1.9620 | -2.5797 | 0.7154 | 0.6177 | -528.0609 | -469.2682 | -0.7145 | -0.6949 |
0.6177 | 0.8550 | 400 | 0.5233 | -2.2484 | -2.9461 | 0.7236 | 0.6977 | -564.7011 | -497.9148 | -0.7064 | -0.6860 |
Framework versions
- Transformers 4.43.2
- Pytorch 2.3.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1
- Downloads last month
- 7
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.