Hugging Face's logo

chchen
/
Mistral-Nemo-12B-Instruct-SAA

PEFT
Safetensors
llama-factory
lora
trl
dpo
Generated from Trainer
Model card Files Community
Mistral-Nemo-12B-Instruct-SAA / README.md
chchen's picture
chchen
End of training
fdf211e verified
preview code
|
raw
history blame contribute delete
1.64 kB
metadata
base_model: mistralai/Mistral-Nemo-Instruct-2407
library_name: peft
license: apache-2.0
tags:
  - llama-factory
  - lora
  - trl
  - dpo
  - generated_from_trainer
model-index:
  - name: Mistral-Nemo-12B-Instruct-SAA
    results: []

Mistral-Nemo-12B-Instruct-SAA

This model is a fine-tuned version of mistralai/Mistral-Nemo-Instruct-2407 on the bct_non_cot_dpo_1000 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1609
  • Rewards/chosen: -0.0115
  • Rewards/rejected: -0.0450
  • Rewards/accuracies: 0.7900
  • Rewards/margins: 0.0335
  • Logps/rejected: -0.4501
  • Logps/chosen: -0.1152
  • Logits/rejected: -2.2660
  • Logits/chosen: -2.2422
  • Sft Loss: 0.0143
  • Odds Ratio Loss: 1.4658

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
  • gradient_accumulation_steps: 8
  • total_train_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: 3.0

Training results

Framework versions

  • PEFT 0.12.0
  • Transformers 4.45.2
  • Pytorch 2.3.0
  • Datasets 2.19.0
  • Tokenizers 0.20.0