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: 2.0330
Rewards/chosen: -2.7266
Rewards/rejected: -2.6680
Rewards/accuracies: 0.5160
Rewards/margins: -0.0586
Logps/rejected: -266.8027
Logps/chosen: -272.6577
Logits/rejected: -0.7176
Logits/chosen: -0.7199
Nll Loss: 0.9493

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: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 8
total_train_batch_size: 128
total_eval_batch_size: 16
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	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen	Nll Loss
17.0014	0.4188	200	2.0831	-2.7051	-2.5590	0.5020	-0.1461	-255.9008	-270.5104	-0.6742	-0.6767	0.9401
16.5359	0.8377	400	2.0330	-2.7266	-2.6680	0.5160	-0.0586	-266.8027	-272.6577	-0.7176	-0.7199	0.9493

Framework versions

Transformers 4.51.0
Pytorch 2.3.1+cu121
Datasets 2.21.0
Tokenizers 0.21.4

jackf857/llama-3-8b-base-cpo-ultrafeedback-4xH200-batch-128-rerun

llama-3-8b-base-cpo-ultrafeedback-4xh200-batch-128

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

jackf857/
llama-3-8b-base-cpo-ultrafeedback-4xH200-batch-128-rerun