I can not set batch_size more than 4,my GPU is GTX980 Ti, I do not why, so when i train the model, the loss just decrease to 10.

georgesung / ssd_tensorflow_traffic_sign_detection

Implementation of Single Shot MultiBox Detector in TensorFlow, to detect and classify traffic signs

MIT License

530 stars 222 forks source link

I can not set batch_size more than 4,my GPU is GTX980 Ti, I do not why, so when i train the model, the loss just decrease to 10. #7

Closed Suodislie closed 7 years ago

georgesung commented 7 years ago

The GTX 980 Ti has 6GB VRAM which is less than the 8GB VRAM of the reference GTX 1080 I used, so that may explain why you had to use a smaller batch size. As to the loss, maybe resume learning with a lower learning rate.