This model is a fine-tuned version of W-61/llama-3-8b-base-sft-ultrachat-8xh200 on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.5358
Margin Dpo/margin Mean: 72.1584
Margin Dpo/margin Std: 96.1592
Logps/chosen: -403.2485
Logps/rejected: -474.8625
Logps/ref Chosen: -280.7076
Logps/ref Rejected: -280.1632
Logits/chosen: -0.8223
Logits/rejected: -0.8121

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-07
train_batch_size: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 2
total_train_batch_size: 128
total_eval_batch_size: 64
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Margin Dpo/margin Mean	Margin Dpo/margin Std	Logps/chosen	Logps/rejected	Logps/ref Chosen	Logps/ref Rejected	Logits/chosen	Logits/rejected
1.0912	0.4188	200	0.5626	60.1408	83.6690	-374.4195	-434.0159	-280.7076	-280.1632	-0.8522	-0.8391
1.1088	0.8377	400	0.5358	72.1584	96.1592	-403.2485	-474.8625	-280.7076	-280.1632	-0.8223	-0.8121

Framework versions

Transformers 4.51.0
Pytorch 2.3.1+cu121
Datasets 2.21.0
Tokenizers 0.21.4

seanyhan/llama-3-8b-base-margin-dpo-ultrafeedback-8xh200

llama-3-8b-base-margin-dpo-ultrafeedback-8xh200-20260410-155037

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

seanyhan/
llama-3-8b-base-margin-dpo-ultrafeedback-8xh200