This model is a fine-tuned version of hZzy/qwen2.5-0.5b-sft-25-1 on the hZzy/train_pairwise_strong_new dataset. It achieves the following results on the evaluation set:

Loss: 0.7822
Objective: 0.7915
Ranking Simple: 0.5149
Reward Accuracy: 0.5679
Logp Accuracy: 0.5149
Log Diff Policy: 2.9432
Chosen Logps: -121.7912
Rejected Logps: -124.7345
Chosen Rewards: -2.6453
Rejected Rewards: -2.9455
Logits: -2.0615

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: 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: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Objective	Ranking Simple	Reward Accuracy	Logp Accuracy	Log Diff Policy	Chosen Logps	Rejected Logps	Chosen Rewards	Rejected Rewards	Logits
0.6382	0.3212	50	0.6749	0.6697	0.4959	0.5897	0.4959	0.9686	-98.4585	-99.4270	-0.3120	-0.4148	-1.3144
0.5716	0.6424	100	0.6872	0.6975	0.5095	0.5910	0.5095	1.8554	-112.9981	-114.8535	-1.7660	-1.9574	-1.5117
0.511	0.9636	150	0.6720	0.6765	0.5082	0.5965	0.5082	2.0840	-107.7762	-109.8602	-1.2438	-1.4581	-1.6335
0.4011	1.2848	200	0.7015	0.7154	0.5109	0.5734	0.5109	2.3167	-114.3844	-116.7011	-1.9046	-2.1422	-1.8366
0.3821	1.6060	250	0.7388	0.7540	0.5122	0.5870	0.5122	2.4988	-116.4216	-118.9203	-2.1083	-2.3641	-1.9767
0.3495	1.9272	300	0.7455	0.7536	0.5041	0.5883	0.5041	2.8123	-116.9906	-119.8029	-2.1652	-2.4523	-1.9868
0.3128	2.2484	350	0.7826	0.7957	0.5109	0.5625	0.5109	2.8056	-122.6670	-125.4726	-2.7329	-3.0193	-2.0636
0.2916	2.5696	400	0.7804	0.7867	0.5149	0.5720	0.5149	2.9527	-121.3108	-124.2635	-2.5973	-2.8984	-2.0567
0.2713	2.8908	450	0.7821	0.7915	0.5149	0.5679	0.5149	2.9428	-121.7890	-124.7318	-2.6451	-2.9452	-2.0615

Framework versions

Transformers 4.42.0
Pytorch 2.3.0+cu121
Datasets 3.2.0
Tokenizers 0.19.1

hZzy/qwen2.5-0.5b-expo-DPO-25-3

qwen2.5-0.5b-expo-DPO-25-3

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
qwen2.5-0.5b-expo-DPO-25-3