Bug and Issue list

[colinyoyo26]

[Bug] low accuracy of pretrained model, problem may be in imgnet.load_datas

import tensorflow as tf
from tensorflow.keras.applications import NASNetMobile, EfficientNetB0, MobileNet, ResNet50
from tensorflow.keras.metrics import sparse_categorical_accuracy
from importlib import reload
from utils import imgnet
reload(imgnet)
model = NASNetMobile()
model.compile(metrics=sparse_categorical_accuracy)

xs, ys = imgnet.load_datas(100)
model.evaluate(xs, ys)

4/4 [==============================] - 5s 57ms/step - loss: 0.0000e+00 - sparse_categorical_accuracy: 0.0000e+00
[0.0, 0.0]

[colinyoyo26]

[Issue] Rewite nvlog package in pandas

[colinyoyo26]

[Issue] Figure out how to deal with xla jit

[colinyoyo26]

[Issue] Support pruned version of nasnet

[colinyoyo26]

[Issue] Figure out how nvprof replay the same kernel launching