This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the HuggingFaceH4/ultrafeedback_binarized and the HuggingFaceH4/deita-10k-v0-sft datasets. It achieves the following results on the evaluation set:

Loss: 0.7869

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: 2e-05
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: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.896	0.2013	48	0.8305
0.782	0.4025	96	0.8111
0.8159	0.6038	144	0.7999
0.8269	0.8050	192	0.7923
0.6927	1.0063	240	0.7913
0.7249	1.2075	288	0.7919
0.7109	1.4088	336	0.7894
0.7045	1.6101	384	0.7878
0.7154	1.8113	432	0.7873

Framework versions

Transformers 4.41.2
Pytorch 2.3.1
Datasets 2.19.2
Tokenizers 0.19.1

skymizer/Llama2-7b-chat-sft-no-template

Llama2-7b-chat-sft-no-template

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

skymizer/
Llama2-7b-chat-sft-no-template