Llama-3.1-8B-Magpie-SFT-500K-UltraDPO-04
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: 0.4482
- Rewards/chosen: -3.4339
- Rewards/rejected: -4.6039
- Rewards/accuracies: 0.7947
- Rewards/margins: 1.1700
- Logps/rejected: -730.4814
- Logps/chosen: -616.4595
- Logits/rejected: -0.7595
- Logits/chosen: -0.7422
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: 4e-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 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.6171 | 0.2138 | 100 | 0.5675 | -1.5275 | -1.9904 | 0.7073 | 0.4630 | -469.1371 | -425.8197 | -0.6912 | -0.6727 |
0.5646 | 0.4275 | 200 | 0.4914 | -2.5135 | -3.3305 | 0.7663 | 0.8170 | -603.1467 | -524.4237 | -0.7024 | -0.6822 |
0.4275 | 0.6413 | 300 | 0.4622 | -2.9305 | -3.9095 | 0.7907 | 0.9790 | -661.0425 | -566.1201 | -0.7343 | -0.7142 |
0.5239 | 0.8550 | 400 | 0.4492 | -3.2972 | -4.4189 | 0.7927 | 1.1217 | -711.9878 | -602.7961 | -0.7358 | -0.7173 |
Framework versions
- Transformers 4.43.2
- Pytorch 2.3.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1
- Downloads last month
- 0
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.