griver

If you are like me you find yourself cloning a repo, making some proposed changes and then deciding to later contributing back using the GitHub Flow convention. Below is a set of instructions I've developed for myself on how to deal with this scenario and an explanation of why it matters based on jagregory's gist.

To follow GitHub flow you should really have created a fork initially as a public representation of the forked repository and the clone that instead. My understanding is that the typical setup would have your local repository pointing to your fork as origin and the original forked repository as upstream so that you can use these keywords in other git commands.

1. Clone some repo (you've probably already done this step).

   git clone [email protected]

griver

@classmethod
def update_from_checkpoint(Cls, save_folder, network, optimizer=None,
                           use_best=False, use_cpu=False, ignore_layers=tuple()):
    """
    Update network and optimizer(if specified) from the checkpoints in the save folder.
    If use_best is True then the data is loaded from the Cls.CHECKPOINT_BEST file
    otherwise the data is loaded from the Cls.CHECKPOINT_LAST file
    Returns the number of global steps past for the loaded checkpoint.
    Arguments:
        save_folder (str): a path to a folder containing summaries and weights

griver

tmux and screen commands

---

Action	tmux	screen
start a new session	tmux tmux new tmux new-session	screen
start a new session with a name	tmux new -s name	screen -S name
re-attach a detached session	tmux attach tmux attach-session	screen -r
re-attach a detached session with a name	tmux attach -t name tmux a -t name	screen -r name
re-attach an attached session (detaching it from elsewhere)	tmux attach -dtmux attach-session -d	screen -dr

griver

import gym
import tensorflow as tf
import numpy as np
import itertools
import tensorflow.contrib.layers as layers
from tqdm import trange
from gym.spaces import Discrete, Box

def get_traj(agent, env, max_episode_steps, render, deterministic_acts=False):
  '''