You can deploy this model and start earning money today!

17Lab/
qwen14b-dpo-reversed-s42

Name: qwen14b-dpo-reversed-s42
Availability: InStock

Views

No views yet

qwen14b_dpo_reversed_from_sft_s42_s42

This model is a fine-tuned version of Qwen/Qwen2.5-14B on the assimilation_dpo_v1_reversed dataset. It achieves the following results on the evaluation set:

Loss: 0.0947
Rewards/chosen: -0.1340
Rewards/rejected: -4.2220
Rewards/accuracies: 0.9881
Rewards/margins: 4.0879
Logps/chosen: -19.5065
Logps/rejected: -45.6895
Logits/chosen: -0.8035
Logits/rejected: -0.7973

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-06
train_batch_size: 2
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 4
total_train_batch_size: 16
total_eval_batch_size: 2
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_steps: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/chosen	Logps/rejected	Logits/chosen	Logits/rejected
0.1263	0.5038	50	0.1257	0.0158	-3.4143	0.9762	3.4300	-18.0083	-37.6123	-0.7283	-0.7223
0.1230	1.0	100	0.0947	-0.1340	-4.2220	0.9881	4.0879	-19.5065	-45.6895	-0.8035	-0.7973

Framework versions

PEFT 0.18.1
Transformers 5.6.0
Pytorch 2.7.0+cu128
Datasets 4.0.0
Tokenizers 0.22.2

17Lab/qwen14b-dpo-reversed-s42

qwen14b_dpo_reversed_from_sft_s42_s42

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

17Lab/
qwen14b-dpo-reversed-s42