This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.6450
Rewards/chosen: 0.6594
Rewards/rejected: 0.1472
Rewards/accuracies: 0.3532
Rewards/margins: 0.5122
Logps/rejected: -80.8012
Logps/chosen: -73.0258
Logits/rejected: -2.5362
Logits/chosen: -2.5527

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: 4e-07
train_batch_size: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 2
total_train_batch_size: 64
total_eval_batch_size: 32
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
0.6677	0.3484	100	0.6483	0.8134	0.3763	0.3512	0.4372	-80.2922	-72.6836	-2.5394	-2.5556
0.6658	0.6969	200	0.6494	0.7541	0.2411	0.3552	0.5130	-80.5926	-72.8155	-2.5245	-2.5411

Framework versions

Transformers 4.44.2
Pytorch 2.4.0
Datasets 2.21.0
Tokenizers 0.19.1

taicheng/zephyr-7b-align-scan-4e-07-0.45-cosine-1.0

zephyr-7b-align-scan-4e-07-0.45-cosine-1.0

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

taicheng/
zephyr-7b-align-scan-4e-07-0.45-cosine-1.0