prune with a list of compression ratio

prune with a list of compression ratio

search pruning strategy with reinforcement learning

Insert Quantization nodes to the eval graph

Insert quantization nodes to the evaluation graph.

Insert Quantization nodes to the training graph

Insert quantization nodes to the training graph.

Record transitions of the current roll-out. Args: * states_n_actions: list of state vector and action pairs * reward: reward of the curre

Restore a pre-trained model with the variable scope renamed.

Restore a model from the latest checkpoint files. Args: * is_train: whether to restore a model for training

Retrain the network with layerwise regression & network fine-tuning.

TODO: Check the name of bucket_type, the value of bucket_size

TODO: Check the name of bucket_type, the value of bucket_size

save a in progress training model with a max evaluation result

linear scale function Args: * w: A Tensor (weights or activation output), the shape is [bucket_size, bucekt_num] if use_buckets else

Setup the action noise. Args: * a_min: minimal value in the action vector * a_max: maximal value in the action vector Returns: *

Setup actor network's initialization & update operations. Returns: * actor_loss: actor network's loss

Setup critic network's initialization & update operations. Returns: * critic_loss: critic network's loss

Setup the parameter noise. Args: * normalized_states: normalized state vectors * a_dims: number of action vector's dimensions * a_min

Setup target network's initialization & update operations.

Sample inputs & outputs of sub-regions from full feature maps. Args: Returns:

Solve the sparsity-constrained regression problem. Args: * inputs_np_list: list of input feature maps (one per input channel, N x k^2) *

Create bucket Args: * w: A Tensor (weights) Returns: * A Tensor, with shape [bucket_size, multiple] * An integer: the number of

Train pruned model

Uniform quantization function Args: * x: A Tensor (weights or activation output) * mbits: A scalar Tensor, tf.int64, spicifying number of

Calculate extra storage for the bucket scalling factors Args: * bucket_num: a Tensor, the number of buckets, and 2*bucket_num scalling facto

Resize images preserving the original aspect ratio. Args: image: A 3-D image `Tensor`. resize_min: A python integer or scalar `Tensor` indi

Wrapper for inserting a list of bytes features into Example proto.

Performs central crops of the given image list. Args: image: a 3-D image tensor crop_height: the height of the image following the crop.

Compute bounding-box regression targets for an image.

Compute bounding-box regression targets for an image. that is : [label, tx, ty, tw, th]

Build an Example proto for an example. Args: filename: string, path to an image file, e.g., '/path/to/example.JPG' image_buffer: string, JP

Crops the given image to a random part of the image, and randomly flips. We use the fused decode_and_crop op, which performs better than the two op

Find the bounding boxes for a given image file. Args: directory: string; the path of all datas. cur_record: list of strings; the first of w

Pads the input along the spatial dimensions independently of input size. Pads the input such that if it was used in a convolution with 'VALID' padd

Bounding-box regression targets (bbox_target_data) are stored in a compact form N x (class, tx, ty, tw, th) This function expands those targe

Subtracts the given means from each image channel. For example: means = [123.68, 116.779, 103.939] image = _mean_image_subtraction(image, m

Returns a variable scope that the model should be created under. If self.dtype is a castable type, model variable will be created in fp32 the

Process a single image file. Args: filename: string, path to an image file e.g., '/path/to/example.JPG'. coder: instance of ImageCoder to p

Process and save list of images as TFRecord of Example protos. Args: name: string, unique identifier specifying the data set directory: str

Define kernel size which is automatically reduced for small input. If the shape of the input images is unknown at graph construction time this fu

Generate a random sample of RoIs comprising foreground and background examples. all_rois shape is [-1, 4] gt_boxes shape is [-1, 5]. that

Computes new shape with the smallest side equal to `smallest_side`. Computes new shape with the smallest side equal to `smallest_side` while pres

Apply fake pruning to the convolutional kernel. Args: * kernel: original convolutional kernel Returns: * kernel: randomly pruned convolution

Apply fake pruning to the convolutional kernel. Args: * kernel: original convolutional kernel Returns: * kernel: randomly pruned convolution

Computes func(x, sel), with sel sampled from [0...num_cases-1]. Args: x: input Tensor. func: Python function to apply. num_cases: Pytho

Parameters ---------- boxes: (N, 4) ndarray of float query_boxes: (K, 4) ndarray of float Returns ------- overlaps: (N, K) ndarray of o

Parameters ---------- boxes: (N, 4) ndarray of float query_boxes: (K, 4) ndarray of float Returns ------- overlaps: (N, K) ndarray of o

Creates one layer of blocks for the ResNet model. Args: inputs: A tensor of size [batch, channels, height_in, width_in] or [batch, height

Build the environment and an RL agent to solve it. Args: * sess: TensorFlow session Returns: * env: environment * agent: RL agent

Build the environment and an RL agent to solve it. Args: * sess: TensorFlow session Returns: * env: environment * agent: RL agent

Calculate the loss function's value. Args: * objects: one tensor with all the annotations packed together * outputs: a dictionary of outp

Calculate the loss function's value. Args: * objects: one tensor with all the annotations packed together * outputs: a dictionary of output ten

Calculate the state vector for the chosen maskable variable. Args: * idx: index to the chosen maskable variable Returns: * state: st

return the rl state

return the rl state

Args: * action: An ndarray with shape (1,1), the output of actor network * idx: A scalar, the id of layer * num_weights: A list of sc

Generatee the action that satisfies the resource constraint Args: * action: an ndarray with shape (1,1), the output of actor network; *

compute which channels to be pruned by lasso

Convert the *.pb model to a *.tflite model. Args: * net: network configurations * pb_path: path to the *.pb file * tflite_path: path to the *

Convert *.pb model to a *.tflite model. Args: * file_path_pb: file path to the *.pb model * file_path_tflite: file path to the *.tflite model

Convert action to pruning ratio for the chosen maskable variable. Args: * idx: index to the chosen maskable variable * action: action's v

Distort the color of a Tensor image. Each color distortion is non-commutative and thus ordering of the color ops matters. Ideally we would random

overlap_matrix: num_gt * num_anchors

use draw lines to draw rectangle. since the draw_rectangle func can not modify the width of rectangle :param draw_obj: :param box: [x1, y

ratio = h /w :param anchors: :param anchor_ratios: :return:

Evaluate the RL agent through multiple roll-outs. Args: * sess: TensorFlow session * env: environment * agent: RL agent

Evaluate the RL agent through multiple roll-outs. Args: * sess: TensorFlow session * env: environment * agent: RL agent

Export *.pb & *.tflite models from checkpoint files. Args: * net: network configurations * file_path_meta: file path to the *.meta data * fil

Export *.pb & *.tflite models from checkpoint files. Args: * net: network configurations * meta_path: path to the *.meta file * pb_path: path

Export a *.tflite model from checkpoint files. Args: * input_coll: input collection's name * output_coll: output collection's name Returns: