JossWhittle

def find_spans(sample, common_words):
    
    spans = []
    for word in common_words:
        count = 0
        s = re.finditer(f'(^|\s){re.escape(word)}[^a-zA-Z]', sample, flags=re.IGNORECASE)
        for idx in s:
            start, end = idx.span()
            count += 1
            spans.append([start, end])

JossWhittle

Global Step: 0 Script: data/scripts/Friends S07E18.txt

TITLE: The One With Joey's Award

SCENE: [Scene: Central Perk, Phoebe, Monica, and Rachel are hanging out.]

Rachel: So Pheebs, you know I was just at this batchelorette party and they had a cake that in the shape of a man's....thing-a-ding-ding, and I was wondering, what you'd thought of that: offensive or amusing?

Phoebe: I am not sure. Monica what do you think. >

JossWhittle

import transformers
from transformers import GPT2Tokenizer
from transformers import TFGPT2LMHeadModel

tokenizer =     GPT2Tokenizer.from_pretrained('gpt2-large')
model     = TFGPT2LMHeadModel.from_pretrained('gpt2-large')

print(sample_sequence('What does the fox say?', tokenizer, model, max_length=32, top_k=100, temperature=1.0, repetition_penalty=2.0))

JossWhittle

#include <iostream>

int main(int argc, char* argv[]) {
	
	int i = 2;

	((i == 0) ? ([]() {
		std::cout << "Branch 1" << std::endl;
	})() : ([]() {
		std::cout << "Branch 2" << std::endl;

JossWhittle

// Compile: nvcc hello.cu -o hello
// Run:     ./hello

#include <iostream> 
#include <assert.h>

// The GPU kernel
// Computes c[i] = a[i] + b[i] for all i
__global__ 
void hello(const int num_elements, int *in_a, int *in_b, int *out_c) {

JossWhittle

JossWhittle

JossWhittle

import glob, os, re

output_file_name = 'output.csv'
print('Writing output to %s...' % (output_file_name))

with open(output_file_name, 'w') as output_file:
    
    file_list = glob.glob('*.txt')
    print('Found %d files to process...' % (len(file_list)))
    

JossWhittle

# Perform 2d convolution using runtime-initialization constant evaluation
#
def dense_with_scaling(inputs, fan_out, activation=None, name='dense_with_scaling'):
    with tf.variable_scope(name, reuse=tf.AUTO_REUSE):
        # Features from previous tensor
        fan_in      = int(inputs.get_shape()[-1])
        # Compute He initialization constant
        C           = np.sqrt(1.3 * 2.0 / fan_in)
        W           = tf.get_variable('W', shape=(fan_in, fan_out), 
                                      initializer=tf.initializers.truncated_normal()) #stddev=C

JossWhittle