rohit-gupta

def shrink_loss(pseudo_label, logits_u_s, conf_thresh):
    removed_class_idx = []
    loss_u_shrink_batch = 0
    
    B, C = pseudo_label.shape
    
    max_probs = pseudo_label.max(dim=-1)[0]
    mask = pseudo_label.ge(conf_thresh).float()
    
    sorted_prob_w, sorted_idx = pseudo_label.topk(C, dim=-1, sorted=True)

rohit-gupta

ps -AF | grep ssh
pkill -t pts/9

rohit-gupta

// ==UserScript==
// @name         New Userscript
// @namespace    http://tampermonkey.net/
// @version      0.1
// @description  try to take over the world!
// @author       You
// @match        <$URL$>
// @icon         <$ICON$>
// @grant        none
// ==/UserScript==

rohit-gupta

.open yfcc100m_dataset.sql
.databases
.tables
.schema yfcc100m_dataset
SELECT * from yfcc100m_dataset WHERE photoid=2932067831 LIMIT 1;
.mode csv
.headers on
.out yfcc100m_dump.csv
SELECT photoid, title, description, downloadurl FROM yfcc100m_dataset;

rohit-gupta

def var_covar_loss(Z, alpha=1.0, beta=0.01):
  
  eps = 1e-5
  K = z.shape[1]
  
  # covariance matrix
  C = torch.cov(Z.t())
  
  # Push sqrt of diagonal terms to 1 (std dev = 1.0)

rohit-gupta

#!/bin/bash

# TODO might be more missing packages
# TODO Shard train set


# Install dependencies
cd /tmp
wget https://repo.anaconda.com/archive/Anaconda3-2020.02-Linux-x86_64.sh
bash Anaconda3-2020.02-Linux-x86_64.sh -b -p $HOME/anaconda3

rohit-gupta

from torch.autograd import Function


class GradientReversal(Function):
    @staticmethod
    def forward(ctx, x):
        ctx.save_for_backward(x)
        output = x
        return output

rohit-gupta

# Pascal Titan X, Titan Xp, 1080Ti, 1080, 1070Ti, 1070 ... = 6.1
# Volta Titan V = 7.0
# Turing Titan RTX, 2080Ti ... = 7.5
export TORCH_CUDA_ARCH_LIST="6.1;7.0;7.5" # Pascal, Volta and Turing on CRCV Cluster
git clone https://github.com/NVIDIA/apex
cd apex
pip install --upgrade --force-reinstall -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

rohit-gupta

rohit-gupta

import tensorflow as tf
import keras.backend as K


gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.25)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))

K.set_session(sess)

## Use normally after this