milhidaka

FROM python:3.9-buster

ENV PYTHONUNBUFFERED=1

RUN mkdir app

WORKDIR /app

RUN pip install --upgrade diffusers transformers scipy jupyter matplotlib

milhidaka

const express = require('express');
const app = express();
const PORT = 8080;

app.use(express.static('public', {
  setHeaders: function (res, path) {
    // no allow cache
    res.set("Cache-Control", "no-cache");
  
    // CORS

milhidaka

#include <stdio.h>
#include <stdint.h>
#include <assert.h>

#define DATA_SIZE 2052

float decode(uint16_t float16_value)
{
  // MSB -> LSB
  // float16=1bit: sign, 5bit: exponent, 10bit: fraction

milhidaka

# runs model trained by imdb_lstm.py and gets input-output pair

import numpy as np
import keras
from keras.preprocessing import sequence
from keras.datasets import imdb


max_features = 20000
maxlen = 80  # cut texts after this number of words (among top max_features most common words)

milhidaka

# reproduce lstm prediction by basic numpy operations
# model trained on imdb_lstm.py
# based on https://github.com/fchollet/keras/blob/master/keras/layers/recurrent.py#L1130
import numpy as np
from scipy.special import expit  # logistic function
import h5py

"""
{'class_name': 'Sequential',
 'config': [{'class_name': 'Embedding',

milhidaka

'''Trains a LSTM on the IMDB sentiment classification task.
The dataset is actually too small for LSTM to be of any advantage
compared to simpler, much faster methods such as TF-IDF + LogReg.
Notes:

- RNNs are tricky. Choice of batch size is important,
choice of loss and optimizer is critical, etc.
Some configurations won't converge.

- LSTM loss decrease patterns during training can be quite different

milhidaka

#include <iostream>
#include <Eigen/Dense>
using namespace std;
using namespace Eigen;

// matrix multiplication using provided memory area
// compile with -std=c++11
int main()
{
  const int size = 2;