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: 0.8855
Rewards/chosen: -3.8900
Rewards/rejected: -5.1926
Rewards/accuracies: 0.3413
Rewards/margins: 1.3027
Logps/rejected: -97.0942
Logps/chosen: -86.4517
Logits/rejected: -2.4499
Logits/chosen: -2.4705

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: 8.165546278140361e-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: linear
lr_scheduler_warmup_ratio: 0.1
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
0.5554	1.0417	100	0.7309	0.3343	-0.2184	0.3393	0.5527	-81.8000	-73.4634	-2.4356	-2.4525
0.3557	2.0833	200	0.8316	-3.2765	-4.2890	0.3413	1.0125	-94.3157	-84.5654	-2.4456	-2.4656

Framework versions

Transformers 4.44.2
Pytorch 2.4.0
Datasets 2.21.0
Tokenizers 0.19.1

taicheng/zephyr-7b-align-scan-0.0-0.3-linear-3

zephyr-7b-align-scan-0.0-0.3-linear-3

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

taicheng/
zephyr-7b-align-scan-0.0-0.3-linear-3