tensorflow 将ckpt中的参数存储为 npy 格式。

最新推荐文章于 2022-12-27 00:10:00 发布

小田同学tdx

最新推荐文章于 2022-12-27 00:10:00 发布

阅读量2k

点赞数

分类专栏： Tensorflow CNN 文章标签： tensorflow

本文链接：https://blog.csdn.net/tiandd12/article/details/80998149

版权

Tensorflow 同时被 2 个专栏收录

14 篇文章 0 订阅

订阅专栏

CNN

3 篇文章 0 订阅

订阅专栏

reader=pywrap_tensorflow.NewCheckpointReader(FILE_PATH)
    var_to_shape_map=reader.get_variable_to_shape_map()
    for key in var_to_shape_map:
        print(key)


    layers = ['conv1_1', 'conv1_2', 'conv2_1', 'conv2_2','conv3_1', 'conv3_2','conv3_3', 'conv4_1', 'conv4_2','conv4_3','conv5_1','conv5_2', 'conv5_3', 'fc6', 'fc7', 'cls_score','bbox_pred','rpn_cls_score','rpn_bbox_pred']
    data = {
        'conv1_1': [],
        'conv1_2': [],
        'conv2_1': [],
        'conv2_2': [],
        'conv3_1': [],
        'conv3_2': [],
        'conv3_3': [],
        'conv4_1': [],
        'conv4_2': [],
        'conv4_3': [],
        'conv5_1': [],
        'conv5_2': [],
        'conv5_3': [],
        'fc6': [],
        'fc7': [],
        'cls_score_na': [],
        'bbox_pred_na':[],
        'rpn_cls_score_na': [],
        'rpn_bbox_pred_na':[]
    }

    for op_name in layers:

        biases_variable = reader.get_tensor(op_name+'/biases')
        weights_variable = reader.get_tensor(op_name+'/weights')
        tmp={'biases':biases_variable,'weights':weights_variable}

        data[op_name] = tmp
    np.save(OUTPUT_FILE, data)