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: 0.3951
Logps: -195.4572
Logits: -3.2699
Objective: 0.3956
Dpo Loss: 0.6771
Regularize: 0.3956
Ranking Simple: 0.5661
Ranking Idealized: 0.9194
Ranking Idealized Expo: 0.5310

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: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 6
gradient_accumulation_steps: 12
total_train_batch_size: 288
total_eval_batch_size: 24
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Logps	Logits	Objective	Dpo Loss	Regularize	Ranking Simple	Ranking Idealized	Ranking Idealized Expo
0.4052	0.2834	50	0.4107	-129.0883	-1.8292	0.4120	0.6914	0.4120	0.5372	0.9194	0.5310
0.3407	0.5668	100	0.4017	-173.3319	-2.5066	0.4063	0.6839	0.4063	0.5548	0.9194	0.5310
0.2596	0.8503	150	0.4017	-188.6395	-2.4464	0.4052	0.6806	0.4052	0.5424	0.9194	0.5310
0.1965	1.1337	200	0.4002	-193.1247	-2.5977	0.4041	0.6801	0.4041	0.5589	0.9194	0.5310
0.1784	1.4171	250	0.3990	-189.4701	-2.7528	0.4023	0.6802	0.4023	0.5620	0.9194	0.5310
0.1717	1.7005	300	0.4021	-195.7304	-2.8777	0.4042	0.6799	0.4042	0.5455	0.9194	0.5310
0.1527	1.9839	350	0.3960	-211.6068	-3.1101	0.3970	0.6760	0.3970	0.5558	0.9194	0.5310
0.1267	2.2674	400	0.3981	-201.0368	-3.2515	0.3998	0.6776	0.3998	0.5620	0.9194	0.5310
0.1121	2.5508	450	0.3957	-192.7809	-2.9523	0.3976	0.6782	0.3976	0.5620	0.9194	0.5310
0.1063	2.8342	500	0.3941	-195.7920	-3.2835	0.3949	0.6760	0.3949	0.5671	0.9194	0.5310
0.0891	3.1176	550	0.3956	-196.1659	-3.1953	0.3960	0.6777	0.3960	0.5610	0.9194	0.5310
0.0749	3.4010	600	0.3962	-194.1237	-3.1966	0.3973	0.6781	0.3973	0.5744	0.9194	0.5310
0.062	3.6845	650	0.3956	-195.3244	-3.2412	0.3967	0.6778	0.3967	0.5702	0.9194	0.5310
0.0583	3.9679	700	0.3956	-196.4469	-3.2432	0.3961	0.6772	0.3961	0.5640	0.9194	0.5310
0.0451	4.2513	750	0.3952	-195.4398	-3.2666	0.3955	0.6771	0.3955	0.5671	0.9194	0.5310
0.0438	4.5347	800	0.3952	-195.2319	-3.2693	0.3956	0.6771	0.3956	0.5661	0.9194	0.5310
0.0408	4.8181	850	0.3951	-195.5095	-3.2704	0.3956	0.6771	0.3956	0.5661	0.9194	0.5310

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-EXPERIMENT-0.005-5e6

qwen2.5-0.5b-expo-L2EXPO-EXPERIMENT-0.005-5e6

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
qwen2.5-0.5b-expo-L2EXPO-EXPERIMENT-0.005-5e6