metadata

license: apache-2.0
base_model: mistralai/Mistral-7B-v0.1
tags:
  - trl
  - orpo
  - generated_from_trainer
model-index:
  - name: mistral-orpo-mix-beta
    results: []

mistral-orpo-mix-beta

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.2152
Rewards/chosen: -0.1113
Rewards/rejected: -0.1461
Rewards/accuracies: 0.6562
Rewards/margins: 0.0348
Logps/rejected: -1.4607
Logps/chosen: -1.1130
Logits/rejected: -2.4201
Logits/chosen: -2.4178
Nll Loss: 1.1711
Log Odds Ratio: -0.6108
Log Odds Chosen: 0.5395

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: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
total_train_batch_size: 32
total_eval_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: inverse_sqrt
lr_scheduler_warmup_ratio: 0.1
lr_scheduler_warmup_steps: 100
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen	Nll Loss	Log Odds Ratio	Log Odds Chosen
0.9961	1.0	211	0.9199	-0.0831	-0.1055	0.6302	0.0224	-1.0547	-0.8311	-2.3542	-2.3590	0.8774	-0.6294	0.3896
0.4999	2.0	422	0.9813	-0.0894	-0.1184	0.6615	0.0290	-1.1836	-0.8937	-2.5184	-2.5177	0.9404	-0.6012	0.4994
0.1899	3.0	633	1.2152	-0.1113	-0.1461	0.6562	0.0348	-1.4607	-1.1130	-2.4201	-2.4178	1.1711	-0.6108	0.5395

Framework versions

Transformers 4.39.0
Pytorch 2.1.1+cu121
Datasets 2.16.1
Tokenizers 0.15.2