This model is a fine-tuned version of slm-research-vn/Qwen2-7B-Instruct-SPPO-Function-call-v2.1 on the slm-research-vn/dpo-format-function-calling-v4, the slm-research-vn/dpo-format-glaive-code-assistant-v3-with-mistral-large-slm-iter4 and the argilla/dpo-mix-7k datasets. It achieves the following results on the evaluation set:

Loss: 0.3152
Rewards/chosen: 1.9961
Rewards/rejected: 0.2161
Rewards/accuracies: 0.8815
Rewards/margins: 1.7800
Logps/rejected: -267.1725
Logps/chosen: -202.5304
Logits/rejected: -0.6205
Logits/chosen: -0.6185

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: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 4
total_train_batch_size: 32
total_eval_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6534	0.1020	100	0.6139	0.2871	0.0961	0.7572	0.1911	-269.5727	-236.7095	-0.6762	-0.6844
0.4902	0.2041	200	0.4530	1.4421	0.5735	0.8064	0.8685	-260.0234	-213.6108	-0.6513	-0.6502
0.391	0.3061	300	0.3935	1.9109	0.6931	0.8382	1.2178	-257.6317	-204.2344	-0.6321	-0.6298
0.3497	0.4082	400	0.3633	1.9715	0.5740	0.8468	1.3975	-260.0141	-203.0221	-0.6323	-0.6313
0.3378	0.5102	500	0.3421	2.0346	0.4602	0.8699	1.5744	-262.2907	-201.7610	-0.6197	-0.6103
0.2904	0.6122	600	0.3287	1.9449	0.3083	0.8757	1.6366	-265.3278	-203.5543	-0.6221	-0.6159
0.3053	0.7143	700	0.3207	1.9933	0.2606	0.8902	1.7327	-266.2818	-202.5857	-0.6162	-0.6111
0.2655	0.8163	800	0.3158	1.9845	0.2262	0.8815	1.7583	-266.9698	-202.7614	-0.6127	-0.6026
0.2943	0.9184	900	0.3144	1.9968	0.2178	0.8844	1.7789	-267.1377	-202.5171	-0.6136	-0.6052

Framework versions

PEFT 0.12.0
Transformers 4.44.0
Pytorch 2.3.1+cu121
Datasets 2.20.0
Tokenizers 0.19.1

khongtrunght/Qwen2-7B-Instruct-SPPO-Function-call-v2.4

Qwen2-7B-Instruct-SPPO-Function-call-v2.4

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

khongtrunght/
Qwen2-7B-Instruct-SPPO-Function-call-v2.4