decagon-pytorch/src/icosagon/model.py

from .data import Data
from typing import List
from .trainprep import prepare_training, \
    TrainValTest
import torch
from .convlayer import DecagonLayer
from .input import OneHotInputLayer
from types import FunctionType
from .declayer import DecodeLayer
from .batch import PredictionsBatch


class Model(object):
    def __init__(self, data: Data,
        layer_dimensions: List[int] = [32, 64],
        ratios: TrainValTest = TrainValTest(.8, .1, .1),
        keep_prob: float = 1.,
        rel_activation: Callable[[torch.Tensor], torch.Tensor] = lambda x: x,
        layer_activation: Callable[[torch.Tensor], torch.Tensor] = torch.nn.functional.relu,
        dec_activation: Callable[[torch.Tensor], torch.Tensor] = lambda x: x,
        lr: float = 0.001,
        loss = Callable[[torch.Tensor, torch.Tensor], torch.Tensor] = torch.nn.functional.binary_cross_entropy_with_logits,
        batch_size: int = 100) -> None:

        if not isinstance(data, Data):
            raise TypeError('data must be an instance of Data')

        if not isinstance(layer_sizes, list):
            raise TypeError('layer_dimensions must be a list')

        if not isinstance(ratios, TrainValTest):
            raise TypeError('ratios must be an instance of TrainValTest')

        keep_prob = float(keep_prob)

        if not isinstance(rel_activation, FunctionType):
            raise TypeError('rel_activation must be a function')

        if not isinstance(layer_activation, FunctionType):
            raise TypeError('layer_activation must be a function')

        if not isinstance(dec_activation, FunctionType):
            raise TypeError('dec_activation must be a function')

        lr = float(lr)

        if not isinstance(loss, FunctionType):
            raise TypeError('loss must be a function')

        batch_size = int(batch_size)

        self.data = data
        self.layer_dimensions = layer_dimensions
        self.ratios = ratios
        self.keep_prob = keep_prob
        self.rel_activation = rel_activation
        self.layer_activation = layer_activation
        self.dec_activation = dec_activation
        self.lr = lr
        self.loss = loss
        self.batch_size = batch_size

        self.build()

    def build(self):
        self.prep_d = prepare_training(self.data, self.ratios)

        in_layer = OneHotInputLayer(self.prep_d)
        last_output_dim = in_layer.output_dim
        seq = [ in_layer ]

        for dim in self.layer_dimensions:
            conv_layer = DecagonLayer(input_dim = last_output_dim,
                output_dim = [ dim ] * len(self.prep_d.node_types),
                data = self.prep_d,
                keep_prob = self.keep_prob,
                rel_activation = self.rel_activation,
                layer_activation = self.layer_activation)
            last_output_dim = conv_layer.output_dim
            seq.append(conv_layer)

        dec_layer = DecodeLayer(input_dim = last_output_dim,
            data = self.prep_d,
            keep_prob = self.keep_prob,
            activation = self.dec_activation)
        seq.append(dec_layer)

        seq = torch.nn.Sequential(*seq)
        self.seq = seq

        opt = torch.optim.Adam(seq.parameters(), lr=self.lr)
        self.opt = opt


    def run_epoch(self):
        pred = self.seq(None)
        batch = PredictionsBatch(pred, self.batch_size)
        n = len(list(iter(batch)))
        loss_sum = 0
        for i in range(n - 1):
            self.opt.zero_grad()
            pred = self.seq(None)
            batch = PredictionsBatch(pred, self.batch_size)
            seed = torch.rand(1).item()
            rng_state = torch.get_rng_state()
            torch.manual_seed(seed)
            it = iter(batch)
            torch.set_rng_state(rng_state)
            for k in range(i):
                _ = next(it)
            (input, target) = next(it)
            loss = self.loss(input, target)
            loss.backward()
            self.opt.optimize()
            loss_sum += loss.detach().cpu().item()
        return loss_sum


    def train(self, max_epochs):
        best_loss = None
        best_epoch = None
        for i in range(max_epochs):
            loss = self.run_epoch()
            if best_loss is None or loss < best_loss:
                best_loss = loss
                best_epoch = i
        return loss, best_loss, best_epoch