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: 0.6808
Logps: -90.9674
Logits: -1.6164
Objective: 0.6836
Dpo Loss: 0.6836
Regularize: 0.6836
Ranking Simple: 0.5331
Ranking Idealized: 0.6030
Ranking Idealized Expo: 0.5223
Wo Beta: 7.8643

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-06
train_batch_size: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 3
gradient_accumulation_steps: 12
total_train_batch_size: 144
total_eval_batch_size: 12
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	Wo Beta
0.689	0.1417	50	0.6875	-90.0815	-1.4869	0.6892	0.6892	0.6892	0.5259	0.6030	0.5223	7.8857
0.6673	0.2834	100	0.6808	-90.9674	-1.6164	0.6836	0.6836	0.6836	0.5331	0.6030	0.5223	7.8643
0.6376	0.4251	150	0.6785	-94.6386	-1.6873	0.6833	0.6833	0.6833	0.5342	0.6030	0.5223	8.1745
0.5955	0.5668	200	0.6808	-100.2786	-1.8583	0.6818	0.6818	0.6818	0.5342	0.6030	0.5223	7.9037
0.5623	0.7085	250	0.6757	-97.3034	-1.9407	0.6757	0.6757	0.6757	0.5362	0.6030	0.5223	7.9161
0.5255	0.8503	300	0.7037	-102.4820	-2.0313	0.7119	0.7119	0.7119	0.5352	0.6030	0.5223	8.7956
0.4939	0.9920	350	0.6897	-102.1435	-1.9358	0.6916	0.6916	0.6916	0.5419	0.6030	0.5223	8.3961

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-DPO-ES2-0.1

qwen2.5-0.5b-expo-DPO-ES2-0.1

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
qwen2.5-0.5b-expo-DPO-ES2-0.1