Skip to content

Instantly share code, notes, and snippets.

View tbenst's full-sized avatar

Tyler Benster tbenst

38 followers · 44 following
View GitHub Profile
@tbenst
tbenst / 2020-03-05 letter to nvidia
Created March 5, 2020 16:50
To whom it may concern,
I wanted to follow up on my previous emails.
My understanding after consulting with others is that what we do in the nixpkgs derivation for CUDA does not preclude binary caching & redistribution, as we only modify the library metadata such as the dynamic section (i.e. DT_RUNPATH, setting interpreter, setting RPATH). As I understand, object code refers to machine code that is executed by the processor, and thus my understanding is that we leave the object code untouched. I previously shared these post-distribution patches on 2/3/2020 for your review.
Furthermore, my understanding under Section 2.3 is that we are ok to redistribute the SDK in full as long as this redistribution only happens under Linux.
Thus, we plan on proceeding with setting up a binary cache for distributing CUDA and packages requiring CUDA using nixpkgs.
@jonathanBieler
jonathanBieler / TypedFunction.jl
Last active March 10, 2021 15:24
TypedFunction
struct TypedFunction{T <: Function,I,O}
f::T
end
(f::TypedFunction)(args...) = f.f(args...)
function tupleize(args) # I don't know how to go from (Int,Int) to Tuple{Int,Int}
t = Tuple{Any}
t.parameters = Core.svec(args...)
@karpathy
karpathy / pg-pong.py
Created May 30, 2016 22:50
Training a Neural Network ATARI Pong agent with Policy Gradients from raw pixels
""" Trains an agent with (stochastic) Policy Gradients on Pong. Uses OpenAI Gym. """
import numpy as np
import cPickle as pickle
import gym
# hyperparameters
H = 200 # number of hidden layer neurons
batch_size = 10 # every how many episodes to do a param update?
learning_rate = 1e-4
gamma = 0.99 # discount factor for reward
@dantonnoriega
dantonnoriega / python_blocks_for_repl.py
Last active September 18, 2021 03:19
Using SublimeREPL, this allows one to transfer AND evaluate blocks of python code. The code automatically detect python blocks, executes them, and skips white space, comment blocks (""") and comment lines (# ...).
"""
DESCRIPTION:
Using SublimeREPL, this plugin allows one to easily transfer AND
evaluate blocks of python code. The code automatically detect python
blocks and executes only code lines, omitting empty space and comments.
One can skips space, comment blocks and comment lines by executing on
empty lines, comments etc.
REQUIRES:
working with only 2 groups in the window. the main group (group 0)
@MohamedAlaa
MohamedAlaa / tmux-cheatsheet.markdown
Last active December 10, 2025 13:44
tmux shortcuts & cheatsheet

tmux shortcuts & cheatsheet

start new:

tmux

start new with session name:

tmux new -s myname