David Rose cipher982

⌨️

Getting back in the rhythm

I love building cool stuff with Python. ML engineer with a day job in generative AI and recommender systems.

cipher982 / onnx_inference_yolov5.py

Created November 25, 2021 00:37

minimum reproducible Python example for YOLOv5 ONNX inference

	# Setup imports
	import os
	from pathlib import Path

	import cv2
	from matplotlib import pyplot as plt
	import numpy as np
	import onnx
	import onnxruntime
	import torch

cipher982 / list_set_lookup_speed.py

Created December 12, 2019 19:56

List versus set lookup speed comparison

	import random
	import time

	# Create lists of 100,000 random numbers each
	random_list_1 = [random.randint(1,1e6) for i in range(int(1e5))]
	random_list_2 = [random.randint(1,1e6) for i in range(int(1e5))]

	start = time.time()
	matching_numbers = 0
	for number in random_list_1:

cipher982 / learn_soft_update_dqn.py

Created January 3, 2019 20:30

learn and soft update steps for the DQN

	def learn(self, experiences, gamma):
	"""
	Update value parameters using given batch of experience tuples

	Params
	======
	experiences (Tuple[torch.Tensor]): tuple of (s, a, r, s', done) tuples
	gamma (float): discount factor
	"""
	states, actions, rewards, next_states, dones = experiences

cipher982 / sample_replay_buffer.py

Created January 3, 2019 20:22

replay buffer for pytorch DQN

	def sample(self):
	"""Randomly sample a batch of experiences from memory"""
	experiences = random.sample(self.memory, k=self.batch_size)

	states = torch.from_numpy(np.vstack([e.state for e in experiences if e is not None])).float().to(device)
	actions = torch.from_numpy(np.vstack([e.action for e in experiences if e is not None])).long().to(device)
	rewards = torch.from_numpy(np.vstack([e.reward for e in experiences if e is not None])).float().to(device)
	next_states = torch.from_numpy(np.vstack([e.next_state for e in experiences if e is not None])).float().to(device)
	dones = torch.from_numpy(np.vstack([e.done for e in experiences if e is not None]).astype(np.uint8)).float().to(device)

cipher982 / QNetwork.py

Created January 3, 2019 20:07

QNetwork

	class QNetwork(nn.Module):
	"""Actor (Policy) Model"""

	def __init__(self, state_size, action_size, seed, fc1_units=64, fc2_units=64):
	"""
	Initialize parameters and build model

	Params
	======
	state_size (int): Dimension of each state

cipher982 / push x y with spline.cpp

Last active February 16, 2018 02:41

	// define the (x,y) points we will use for the planner
	vector<double> next_x_vals;
	vector<double> next_y_vals;
	// start with previous path points from above
	for (int i = 0; i < prev_size; i++)
	{
	next_x_vals.push_back(previous_path_x[i]);
	next_y_vals.push_back(previous_path_y[i]);
	}
	// calculate how to break up spline points to travel at desired reference velocity

cipher982 / spline.cpp

Created February 16, 2018 02:39

cipher982 / vehicle-decision-state-machine.cpp

Created February 16, 2018 00:44

	if (car_ahead)
	{
	cout << "Car Ahead!!!\n\n\n\n\n\n"
	<< endl;
	if (!car_left && lane > 0) // no left-car, yes left-lane
	{
	lane--; // Change lane left
	}
	else if (!car_right && lane != 2) // no right-car, yes right-lane
	{

cipher982 / lane_state_ideas.cpp

Last active February 16, 2018 00:40

	// estimate longitudinal position of a car (s)
	check_car_s += ((double)prev_size * 0.02 * check_speed);

	if (car_lane == lane) // if the other car is in 'lane' (our lane)
	{
	// is it within 30m ahead of us?
	car_ahead \|= check_car_s > car_s && check_car_s - car_s < 30;
	}
	else if (car_lane - lane == -1) // if left of us
	{

cipher982 / lane_detection.cpp

Created February 16, 2018 00:29

	if (d > 0 && d < 4) // d represents location in meters from left-most edge
	{
	car_lane = 0; // 0-4 meters = lane 0, (left-lane)
	}
	else if (d > 4 && d < 8)
	{
	car_lane = 1; // middle-lane
	}
	else if (d > 8 && d < 12)
	{