wphicks

lsawaniewski

# Htop
sudo apt install -y htop

# Byobu
sudo apt install -y byobu

# Gdebi
sudo apt install -y gdebi-core

# GIT

jmp84

import argparse
import logging

import torch
from fairseq.checkpoint_utils import load_model_ensemble_and_task
from fairseq.sequence_generator import SequenceGenerator


def get_args():
    parser = argparse.ArgumentParser(

akurniawan

import itertools
import torch
from torchtext.experimental.datasets.translation import DATASETS, TranslationDataset
from torchtext.vocab import build_vocab_from_iterator
from torchtext.experimental.functional import (
    vocab_func,
    totensor,
    sequential_transforms,
)
from torchtext.data.utils import get_tokenizer

mingfeima

General guidelines for CPU performance on PyTorch

This file serves a BKM to get better performance on CPU for PyTorch, mostly focusing on inference or deployment. Chinese version available here.

1. Use channels last memory format

Right now, on PyTorch CPU path, you may choose to use 3 types of memory formats.

• torch.contiguous_format: default memory format, also referred as NHCW.
• torch.channels_last: also referred as NHWC.
• torch._mkldnn: mkldnn blocked format.

mingfeima

benchmark

Based on huggingface repo for performance evaluation, actual benchmark run script placed at repo. How to reproduce performance:

1. prepare dataset according to link.
2. update GLUE_DIR to actual dataset path in run_inference.sh.
3. change env settings, the default setting is using 20 cores;

MKL v.s. MKLDNN

Inference performance result on Xeon 6148 (2x20 cores), single socket and single thread.

akiross

vinhkhuc

"""
Vanilla Char-RNN using TensorFlow by Vinh Khuc (@knvinh).
Adapted from Karpathy's min-char-rnn.py
https://gist.github.com/karpathy/d4dee566867f8291f086
Requires tensorflow>=1.0
BSD License
"""
import random
import numpy as np
import tensorflow as tf

Reddine

sudo apt-get install build-essential
sudo apt-get update
# BLAS → LAPACK → ATLAS → numpy → scipy → Theano
# remove numpy and scipy
sudo apt-get remove python-numpy
sudo apt-get remove python-scipy
# Instalation commands
sudo apt-get install gfortran
sudo apt-get install libopenblas-dev
sudo apt-get install liblapack-dev

afolarin

#Container Resource Allocation Options in docker-run

now see: https://docs.docker.com/engine/reference/run/#runtime-constraints-on-resources

You have various options for controlling resources (cpu, memory, disk) in docker. These are principally via the docker-run command options.

##Dynamic CPU Allocation -c, --cpu-shares=0
CPU shares (relative weight, specify some numeric value which is used to allocate relative cpu share)