This model is a fine-tuned version of JunxiongWang/mamba_0_5_dpo_ep3 on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.7141
Rewards/chosen: -5.3346
Rewards/rejected: -8.3118
Rewards/accuracies: 0.7891
Rewards/margins: 2.9772
Logps/rejected: -337.4994
Logps/chosen: -304.9619
Logits/rejected: -2.7812
Logits/chosen: -2.8272

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: 8
seed: 42
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 32
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: 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
0.1171	1.0466	2000	0.5329	-1.4521	-2.9272	0.7734	1.4750	-283.6535	-266.1376	-2.8897	-2.9362
0.0086	2.0931	4000	0.7141	-5.3346	-8.3118	0.7891	2.9772	-337.4994	-304.9619	-2.7812	-2.8272

Framework versions

Transformers 4.41.2
Pytorch 2.1.0+cu118
Datasets 2.20.0
Tokenizers 0.19.1

MambaInLlama

@article{junxiongdaniele2024mambainllama,
  title   = {The Mamba in the Llama: Distilling and Accelerating Hybrid Models},
  author  = {Junxiong Wang and Daniele Paliotta and Avner May and Alexander M. Rush and Tri Dao},
  journal = {arXiv preprint arXiv:2408.15237},
  year    = {2024}
}