metadata
base_model: >-
/data/junxiong/Llama-Mamba-3.2-3B-teacher-Llama-3.1-70B-Instruct-kl1.0-ce0.0-update/
tags:
- alignment-handbook
- generated_from_trainer
datasets:
- HuggingFaceH4/ultrafeedback_binarized
- HuggingFaceH4/orca_dpo_pairs
- JunxiongWang/llama3-ultrafeedback-armorm
model-index:
- name: >-
Llama-Mamba-3.2-3B-teacher-Llama-3.1-70B-Instruct-kl1.0-ce0.0-update-dpo-short
results: []
Llama-Mamba-3.2-3B-teacher-Llama-3.1-70B-Instruct-kl1.0-ce0.0-update-dpo-short
This model is a fine-tuned version of /data/junxiong/Llama-Mamba-3.2-3B-teacher-Llama-3.1-70B-Instruct-kl1.0-ce0.0-update/ on the HuggingFaceH4/ultrafeedback_binarized, the HuggingFaceH4/orca_dpo_pairs and the JunxiongWang/llama3-ultrafeedback-armorm datasets. It achieves the following results on the evaluation set:
- Loss: 0.4802
- Rewards/chosen: -2.0035
- Rewards/rejected: -4.1751
- Rewards/accuracies: 0.7929
- Rewards/margins: 2.1716
- Logps/rejected: -691.1746
- Logps/chosen: -472.6584
- Logits/rejected: -1.5357
- Logits/chosen: -1.5952
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-07
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- total_train_batch_size: 32
- total_eval_batch_size: 64
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5034 | 0.4798 | 2000 | 0.4988 | -1.5060 | -3.1448 | 0.7982 | 1.6388 | -588.1365 | -422.9025 | -1.5466 | -1.5856 |
| 0.4894 | 0.9597 | 4000 | 0.4802 | -2.0035 | -4.1751 | 0.7929 | 2.1716 | -691.1746 | -472.6584 | -1.5357 | -1.5952 |
Framework versions
- Transformers 4.43.1
- Pytorch 2.1.1+cu118
- Datasets 2.20.0
- Tokenizers 0.19.1