This model is a fine-tuned version of hZzy/qwen2.5-0.5b-sft-news-IFT on the hZzy/train_pairwise dataset. It achieves the following results on the evaluation set:

Loss: 0.4106
Logps: -103.3770
Logits: -1.4557
Objective: 0.4142
Dpo Loss: 0.6907
Regularize: 0.4142
Ranking Simple: 0.5345
Ranking Idealized: 0.8584
Ranking Idealized Expo: 0.5207

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: 1e-07
train_batch_size: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 12
total_train_batch_size: 192
total_eval_batch_size: 16
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	Logps	Logits	Objective	Dpo Loss	Regularize	Ranking Simple	Ranking Idealized	Ranking Idealized Expo
0.4151	0.1889	50	0.4153	-97.3580	-1.3067	0.4159	0.6927	0.4159	0.5180	0.8584	0.5207
0.3989	0.3778	100	0.4140	-97.6116	-1.3266	0.4149	0.6921	0.4149	0.5207	0.8584	0.5207
0.4205	0.5668	150	0.4131	-97.5505	-1.3609	0.4142	0.6916	0.4142	0.5283	0.8584	0.5207
0.4006	0.7557	200	0.4119	-99.2361	-1.3908	0.4139	0.6912	0.4139	0.5297	0.8584	0.5207
0.4158	0.9446	250	0.4114	-100.9730	-1.4099	0.4140	0.6912	0.4140	0.5318	0.8584	0.5207
0.41	1.1335	300	0.4107	-101.8556	-1.4325	0.4136	0.6908	0.4136	0.5338	0.8584	0.5207
0.4037	1.3224	350	0.4106	-102.7009	-1.4417	0.4139	0.6908	0.4139	0.5331	0.8584	0.5207
0.4169	1.5113	400	0.4107	-103.3040	-1.4516	0.4141	0.6908	0.4141	0.5345	0.8584	0.5207
0.4049	1.7003	450	0.4106	-103.2897	-1.4550	0.4141	0.6907	0.4141	0.5338	0.8584	0.5207
0.3916	1.8892	500	0.4106	-103.3759	-1.4556	0.4142	0.6907	0.4142	0.5345	0.8584	0.5207

Framework versions

Transformers 4.42.0
Pytorch 2.3.0+cu121
Datasets 2.19.1
Tokenizers 0.19.1

hZzy/qwen2.5-0.5b-expo-L2EXPO-0.01

qwen2.5-0.5b-expo-L2EXPO-0.01

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
qwen2.5-0.5b-expo-L2EXPO-0.01