About performance on market1501 for global learning and mutual learning

[vincentman]

I try to train with global learning and mutual learning on market1501, and then mAP for model 1 is 68.82%, mAP for model 2 is 68.56%. These performances are worse than yours(your mAP is around 75%).

Snapshot from AlignedReID-Scores.xlsx:

What are your parameters for training? I post my training script as following:
python3 script/experiment/train_ml.py
-d "((0,), (1,))"
-r 1
--num_models 2
--dataset market1501
--ids_per_batch 32
--ims_per_id 4
--normalize_feature false
-gm 0.3
-glw 1
-llw 0
-idlw 0
-pmlw 0
-gdmlw 1
-ldmlw 0
--base_lr 2e-4
--lr_decay_type exp
--exp_decay_at_epoch 151
--total_epochs 300