lemonhat/Qwen2.5-7B-Instruct-agenttuning_v2_15k_tag4

agenttuning_v2_15k_tag4

This model is a fine-tuned version of Qwen/Qwen2.5-7B-Instruct on the agenttuning_v2_15k_tag4 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3589

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: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 4
• total_eval_batch_size: 4
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
0.7145	0.0286	100	0.4777
0.5212	0.0572	200	0.4385
0.4558	0.0858	300	0.4370
0.4211	0.1144	400	0.4323
0.2821	0.1430	500	0.4279
0.4548	0.1716	600	0.4251
0.3571	0.2002	700	0.4241
0.2706	0.2288	800	0.4255
0.3376	0.2574	900	0.4176
0.4035	0.2860	1000	0.4220
0.4495	0.3146	1100	0.4090
0.3477	0.3432	1200	0.4125
0.3128	0.3717	1300	0.4060
0.3558	0.4003	1400	0.4107
0.3695	0.4289	1500	0.4061
0.343	0.4575	1600	0.4066
0.5475	0.4861	1700	0.4010
0.2354	0.5147	1800	0.3888
0.5055	0.5433	1900	0.3893
0.3117	0.5719	2000	0.3871
0.5256	0.6005	2100	0.3764
0.4246	0.6291	2200	0.3756
0.2611	0.6577	2300	0.3704
0.3354	0.6863	2400	0.3718
0.3213	0.7149	2500	0.3674
0.4258	0.7435	2600	0.3647
0.3216	0.7721	2700	0.3601
0.2982	0.8007	2800	0.3620
0.3676	0.8293	2900	0.3625
0.3482	0.8579	3000	0.3620
0.4485	0.8865	3100	0.3605
0.3042	0.9151	3200	0.3600
0.3637	0.9437	3300	0.3600
0.2896	0.9723	3400	0.3593

Framework versions

• Transformers 4.46.1
• Pytorch 2.8.0+cu128
• Datasets 3.1.0
• Tokenizers 0.20.3