This model is a fine-tuned version of four-two-labs/tinyllama-moe-base on the see code dataset. It achieves the following results on the evaluation set:

Loss: 1.1757
Rewards/chosen: -0.1708
Rewards/rejected: -0.1682
Rewards/accuracies: 0.5003
Rewards/margins: -0.0027
Logps/rejected: -1.6818
Logps/chosen: -1.7085
Logits/rejected: -2.6114
Logits/chosen: -2.6139
Nll Loss: 1.0859
Log Odds Ratio: -0.8989
Log Odds Chosen: -0.1073

Model description

More information needed

Training and evaluation data

python

from datasets import load_dataset
from datasets import interleave_datasets


def format_chat_template(row):
    for key in ['prompt', 'chosen', 'rejected']: 
        row[key] = tokenizer.apply_chat_template(row[key], tokenize=False)
        
    return row


dataset = (
    interleave_datasets([
        (
            interleave_datasets(
                load_dataset(
                    'four-two-labs/orpo-dpo-mix-40k-multilang-fixed',                 
                    token=hf_token,
                )
                .values()
            )
            .select_columns(['prompt', 'chosen', 'rejected'])
        ),
        (
            load_dataset(
                'four-two-labs/translations-5M-DPO',
                split='train', 
                token=hf_token,                
            )
            .shuffle(42)
            .select(range(250_000))
            .select_columns(['prompt', 'chosen', 'rejected'])
        ),
        (
            load_dataset(
                'four-two-labs/ultrafeedback_binarized-fixed', 
                split='train_prefs', 
                token=hf_token,
            )
            .select_columns(['prompt', 'chosen', 'rejected'])
        ),        
    ])
    .shuffle(seed=42)
    #.select(range(1000))
    .map(format_chat_template, num_proc=32)    
    .train_test_split(test_size=0.01)
)

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-06
train_batch_size: 12
eval_batch_size: 12
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 48
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 10
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
1.1835	0.6001	2270	1.1764	-0.1711	-0.1684	0.4976	-0.0027	-1.6838	-1.7106	-2.6416	-2.6430	1.0866	-0.8991	-0.1073
1.1494	1.2002	4540	1.1757	-0.1709	-0.1682	0.5003	-0.0026	-1.6820	-1.7085	-2.5529	-2.5573	1.0860	-0.8988	-0.1070
1.233	1.8003	6810	1.1757	-0.1709	-0.1682	0.4993	-0.0027	-1.6819	-1.7086	-2.5859	-2.5892	1.0859	-0.8989	-0.1073
1.2344	2.4004	9080	1.1757	-0.1708	-0.1682	0.5003	-0.0027	-1.6818	-1.7085	-2.6114	-2.6139	1.0859	-0.8989	-0.1073

Framework versions

Transformers 4.40.0
Pytorch 2.1.2+cu121
Datasets 2.19.0
Tokenizers 0.19.1

four-two-labs/tinyllama-moe-base-mix-orpo

tinyllama-moe-base-mix-orpo