pse-ecn / pose-sensitive-embedding

Pose Sensitive Embedding for Person Re-Identification (PSE)
112 stars 36 forks source link

How to solve the error "NaN loss during training"? thanks #32

Closed datar001 closed 4 years ago

datar001 commented 4 years ago

environment: tensorflow -gpu1.10 python3.65 no matlab(only train model) I comment line 8、11 and 35 in trainer_preid.py, and set do_evaluation=False.

from evaluation.evaluation_helper import get_evaluation_summary_writer

#from predictor_preid import run_evaluation_conserving_best
#evaluation_summary_writer = get_evaluation_summary_writer(do_evaluation, output_directory)

When i execute the following code: python trainer_preid.py --output=outputs/ --data=data/market/ --dataset-name=market1501 --batch-size=16 --num-epochs=100 --network-name=resnet_v1_50 --initial-checkpoint=checkpoint/resnet_v1_50.ckpt --checkpoint-exclude-scopes=resnet_v1_50/logits --trainable-scopes=resnet_v1_50/logits --no-evaluation The error occured: tensorflow.python.training.basic_session_run_hooks.NanLossDuringTrainingError: NaN loss during training.