metadata
library_name: transformers
license: other
base_model: /home/bl3615/data/Goedel-Prover-SFT
tags:
- llama-factory
- full
- generated_from_trainer
model-index:
- name: dpo_dpo_lean_0_b0.03_f0_lr5e-6_e2
results: []
dpo_dpo_lean_0_b0.03_f0_lr5e-6_e2
This model is a fine-tuned version of /home/bl3615/data/Goedel-Prover-SFT on the dpo_lean_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5357
- Rewards/chosen: -4.1002
- Rewards/rejected: -5.8239
- Rewards/accuracies: 0.7566
- Rewards/margins: 1.7237
- Logps/chosen: -209.7086
- Logps/rejected: -266.5144
- Logits/chosen: -13.4579
- Logits/rejected: -12.9861
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: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- total_eval_batch_size: 8
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 2.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/chosen | Logps/rejected | Logits/chosen | Logits/rejected |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5867 | 0.5329 | 500 | 0.5393 | -1.7260 | -2.3840 | 0.7138 | 0.6580 | -130.5684 | -151.8518 | -4.1731 | -3.9948 |
| 0.1708 | 1.0650 | 1000 | 0.5134 | -4.1008 | -5.5590 | 0.7533 | 1.4582 | -209.7289 | -257.6841 | -11.3604 | -10.9174 |
| 0.1112 | 1.5979 | 1500 | 0.5428 | -4.2436 | -5.9501 | 0.7599 | 1.7065 | -214.4901 | -270.7217 | -14.0546 | -13.5623 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.6.0+cu124
- Datasets 3.2.0
- Tokenizers 0.21.0