tomsherborne

fairseq-train qa_en_small-bin \
--log-interval=10 \
--log-format=json \
--tensorboard-logdir=/users/tom/ed/sp/pretrain/tests/fairseq/bart_en_small/logs \
--seed=1 \
--cpu \
--min-loss-scale=0.0001 \
--model-parallel-size=1 \
--criterion=cross_entropy \

eddex

How to install CUDA 10.1 on Ubuntu 18.04

A clean installation of Ubuntu 18.04.02 LTS was used.

This gist is an extension to the official docs, adding missing parts and instructions.

2 pre-install actions

follow the pre-installation actions on:

thomwolf

from torch.optim import Optimizer

class AdamW(Optimizer):
    """
       Implements Adam algorithm with weight decay fix in PyTorch
       Paper: Fixing Weight Decay Regularization in Adam by Ilya Loshchilov, Frank Hutter
       https://arxiv.org/abs/1711.05101
    """
    def __init__(self, params, lr, b1=0.9, b2=0.999, e=1e-8, l2=0,
                 vector_l2=False, max_grad_norm=-1, **kwargs):

yuzeh

# MIT License
#
# Copyright (c) 2018 Yuze Huang (hi@yuzeh.com)
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:

ahmedhosny

import nrrd # pip install pynrrd
import nibabel as nib # pip install nibabel
import numpy as np

# load nrrd 
_nrrd = nrrd.read('/path/to/nrrd.nrrd')
data = _nrrd[0]
header = _nrrd[1]
print data.shape, header

jihunchoi

def _sequence_mask(sequence_length, max_len=None):
    if max_len is None:
        max_len = sequence_length.data.max()
    batch_size = sequence_length.size(0)
    seq_range = torch.range(0, max_len - 1).long()
    seq_range_expand = seq_range.unsqueeze(0).expand(batch_size, max_len)
    seq_range_expand = Variable(seq_range_expand)
    if sequence_length.is_cuda:
        seq_range_expand = seq_range_expand.cuda()
    seq_length_expand = (sequence_length.unsqueeze(1)

yossorion

What I Wish I'd Known About Equity Before Joining A Unicorn

Disclaimer: This piece is written anonymously. The names of a few particular companies are mentioned, but as common examples only.

This is a short write-up on things that I wish I'd known and considered before joining a private company (aka startup, aka unicorn in some cases). I'm not trying to make the case that you should never join a private company, but the power imbalance between founder and employee is extreme, and that potential candidates would

sunskyhsh

def qlearning(env, policy, num_iter1, alpha, gamma):
    actions = policy.actions
    for i in xrange(len(policy.theta)):
        policy.theta[i] = 0.1

    for iter1 in xrange(num_iter1):
        s_f       = env.reset()
        a         = policy.epsilon_greedy(s_f)
        count     = 0
        t         = False

brotherofken

# In[]
import gym
import numpy as np
import theano
import theano.tensor as T
import lasagne
import sklearn.preprocessing
np.set_printoptions(precision=2)

pylover

#! /usr/bin/env python3
"""Fixing bluetooth stereo headphone/headset problem in debian distros.

Workaround for bug: https://bugs.launchpad.net/ubuntu/+source/indicator-sound/+bug/1577197
Run it with python3.5 or higher after pairing/connecting the bluetooth stereo headphone.

This will be only fixes the bluez5 problem mentioned above .

Licence: Freeware