hi ,I want to load a trained yolov3 tensorflow model use Tensorrt.
First, I converted a pb model to uff format.
convert-to-uff frozen_inference_graph.pb -p config.py
the config.py is:
import graphsurgeon as gs
import tensorflow as tf
Input = gs.create_node(“input”, op=“Placeholder”,dtype=tf.float32,shape=[1, 3, 416, 416])
upsample_layer = gs.create_plugin_node(name=“My_upsamply_layer”, op=“upsamply_layer_TRT”, dtype=tf.float32, scale=2)
namespace_plugin_map = {
“up_sampling2d_2/ResizeNearestNeighbor”: upsample_layer,
“up_sampling2d_1/ResizeNearestNeighbor”: upsample_layer,
“input_1”: Input
}
def preprocess(dynamic_graph):
dynamic_graph.collapse_namespaces(namespace_plugin_map)
when I use Tensorrt load the uff model is failde:
UffParser: Graph error: Cycle graph detected