Reference Links:
https://github.com/bermanmaxim/jaccardSegment/blob/master/ckpt_to_dd.py
A. Tensorflow model turn pytorch model
import tensorflow as tf import deepdish as dd import argparse import os import numpy as np def tr(v): # tensorflow weights to pytorch weights if .I == 4: return np.ascontiguousarray(v.transpose(3,2,0,1)) elif .I == 2: return np.ascontiguousarray(v.transpose()) return v def read_ckpt(ckpt): # https://github.com/tensorflow/tensorflow/issues/1823 reader = tf.train.NewCheckpointReader(ckpt) weights = {n: reader.get_tensor(n) for (n, _) in reader.get_variable_to_shape_map().items()} pyweights = {k: tr(v) for (k, v) in weights.items()} return pyweights if __name__ == '__main__': parser = argparse.ArgumentParser(description="Converts ckpt weights to deepdish hdf5") parser.add_argument("infile", type=str, help="Path to the ckpt.") # ***model.ckpt-22177*** parser.add_argument("outfile", type=str, nargs='?', default='', help="Output file (inferred if missing).") args = parser.parse_args() if args.outfile == '': args.outfile = os.path.splitext(args.infile)[0] + '.h5' outdir = os.path.dirname(args.outfile) if not os.path.exists(outdir): os.makedirs (OutDir) weights = read_ckpt(args.infile) dd.io.save(args.outfile, weights)
Model.h5 model will be 1. Run the code, as follows:
Note: python consistent and version tensorflow used pytorch
2. Use: load change in the model pytorch:
It is assumed that the network parameter stored consistent naming
net = ... import torch import deepdish as dd net = resnet50(..) model_dict = net.state_dict() # First parameter value converted to tensor form numpy pretrained_dict = = dd.io.load('./model.h5') new_pre_dict = {} for k,v in pretrained_dict.items(): new_pre_dict[k] = torch.Tensor(v) # Update model_dict.update(new_pre_dict) #load net.load_state_dict(model_dict)
Two. Pytorch turn tensorflow (Continued ..)
Original: https: //blog.csdn.net/weixin_42699651/article/details/88932670