Closed vladmandic closed 3 years ago
Quite a few new models from TFHub fail on coversion:
vlado@Spectre:~/dev/models $ tfjs_graph_converter imagenet-efficientnet-b5 imagenet-efficientnet-b5.pb TensorFlow.js Graph Model Converter Graph model: imagenet-efficientnet-b5 Output: imagenet-efficientnet-b5.pb Target format: tf_frozen_model Converting.... Error: Node module_apply_default/efficientnet-b5/model/blocks_0/depthwise_conv2d/depthwise: unsupported op FusedDepthwiseConv2dNative
Model in question is published at https://tfhub.dev/google/efficientnet/b5/classification/1
Same error with few other models published recently by Google, such as all AIY models and all EfficientNet models on TFHub.
Nice! Thanks :)
Quite a few new models from TFHub fail on coversion:
Model in question is published at https://tfhub.dev/google/efficientnet/b5/classification/1
Same error with few other models published recently by Google, such as all AIY models and all EfficientNet models on TFHub.