Merge pull request #25 from MysteryVaibhav/master

neubig · web-flow · commit a9e8be5b101c · 2019-02-06T15:58:21.000-05:00
Adding pytorch code for 05-cnn
diff --git a/05-cnn-pytorch/cnn-activation.py b/05-cnn-pytorch/cnn-activation.py
@@ -0,0 +1,150 @@
+from collections import defaultdict
+import time
+import random
+import torch
+import numpy as np
+
+
+class CNNclass(torch.nn.Module):
+    def __init__(self, nwords, emb_size, num_filters, window_size, ntags):
+        super(CNNclass, self).__init__()
+
+        """ layers """
+        self.embedding = torch.nn.Embedding(nwords, emb_size)
+        # uniform initialization
+        torch.nn.init.uniform_(self.embedding.weight, -0.25, 0.25)
+        # Conv 1d
+        self.conv_1d = torch.nn.Conv1d(in_channels=emb_size, out_channels=num_filters, kernel_size=window_size,
+                                       stride=1, padding=0, dilation=1, groups=1, bias=True)
+        self.relu = torch.nn.ReLU()
+        self.projection_layer = torch.nn.Linear(in_features=num_filters, out_features=ntags, bias=True)
+        # Initializing the projection layer
+        torch.nn.init.xavier_uniform_(self.projection_layer.weight)
+
+    def forward(self, words, return_activations=False):
+        emb = self.embedding(words)                 # nwords x emb_size
+        emb = emb.unsqueeze(0).permute(0, 2, 1)     # 1 x emb_size x nwords
+        h = self.conv_1d(emb)                       # 1 x num_filters x nwords
+        activations = h.squeeze(0).max(dim=1)[1]     # argmax along length of the sentence
+        # Do max pooling
+        h = h.max(dim=2)[0]                         # 1 x num_filters
+        h = self.relu(h)
+        features = h.squeeze(0)
+        out = self.projection_layer(h)              # size(out) = 1 x ntags
+        if return_activations:
+            return out, activations.data.cpu().numpy(), features.data.cpu().numpy()
+        return out
+
+
+np.set_printoptions(linewidth=np.nan, threshold=np.nan)
+
+# Functions to read in the corpus
+w2i = defaultdict(lambda: len(w2i))
+UNK = w2i["<unk>"]
+def read_dataset(filename):
+    with open(filename, "r") as f:
+        for line in f:
+            tag, words = line.lower().strip().split(" ||| ")
+            words = words.split(" ")
+            yield (words, [w2i[x] for x in words], int(tag))
+
+
+# Read in the data
+train = list(read_dataset("../data/classes/train.txt"))
+w2i = defaultdict(lambda: UNK, w2i)
+dev = list(read_dataset("../data/classes/test.txt"))
+nwords = len(w2i)
+ntags = 5
+
+# Define the model
+EMB_SIZE = 10
+WIN_SIZE = 3
+FILTER_SIZE = 8
+
+# initialize the model
+model = CNNclass(nwords, EMB_SIZE, FILTER_SIZE, WIN_SIZE, ntags)
+criterion = torch.nn.CrossEntropyLoss()
+optimizer = torch.optim.Adam(model.parameters())
+
+type = torch.LongTensor
+use_cuda = torch.cuda.is_available()
+
+if use_cuda:
+    type = torch.cuda.LongTensor
+    model.cuda()
+
+
+def calc_predict_and_activations(wids, tag, words):
+    if len(wids) < WIN_SIZE:
+        wids += [0] * (WIN_SIZE-len(wids))
+    words_tensor = torch.tensor(wids).type(type)
+    scores, activations, features = model(words_tensor, return_activations=True)
+    scores = scores.squeeze().cpu().data.numpy()
+    print('%d ||| %s' % (tag, ' '.join(words)))
+    predict = np.argmax(scores)
+    print(display_activations(words, activations))
+    W = model.projection_layer.weight.data.cpu().numpy()
+    bias = model.projection_layer.bias.data.cpu().numpy()
+    print('scores=%s, predict: %d' % (scores, predict))
+    print('  bias=%s' % bias)
+    contributions = W * features
+    print(' very bad (%.4f): %s' % (scores[0], contributions[0]))
+    print('      bad (%.4f): %s' % (scores[1], contributions[1]))
+    print('  neutral (%.4f): %s' % (scores[2], contributions[2]))
+    print('     good (%.4f): %s' % (scores[3], contributions[3]))
+    print('very good (%.4f): %s' % (scores[4], contributions[4]))
+
+
+def display_activations(words, activations):
+    pad_begin = (WIN_SIZE - 1) / 2
+    pad_end = WIN_SIZE - 1 - pad_begin
+    words_padded = ['pad' for _ in range(int(pad_begin))] + words + ['pad' for _ in range(int(pad_end))]
+
+    ngrams = []
+    for act in activations:
+        ngrams.append('[' + ', '.join(words_padded[act:act+WIN_SIZE]) + ']')
+
+    return ngrams
+
+
+for ITER in range(10):
+    # Perform training
+    random.shuffle(train)
+    train_loss = 0.0
+    train_correct = 0.0
+    start = time.time()
+    for _, wids, tag in train:
+        # Padding (can be done in the conv layer as well)
+        if len(wids) < WIN_SIZE:
+            wids += [0] * (WIN_SIZE - len(wids))
+        words_tensor = torch.tensor(wids).type(type)
+        tag_tensor = torch.tensor([tag]).type(type)
+        scores = model(words_tensor)
+        predict = scores[0].argmax().item()
+        if predict == tag:
+            train_correct += 1
+
+        my_loss = criterion(scores, tag_tensor)
+        train_loss += my_loss.item()
+        # Do back-prop
+        optimizer.zero_grad()
+        my_loss.backward()
+        optimizer.step()
+    print("iter %r: train loss/sent=%.4f, acc=%.4f, time=%.2fs" % (ITER, train_loss/len(train), train_correct/len(train), time.time()-start))
+    # Perform testing
+    test_correct = 0.0
+    for _, wids, tag in dev:
+        # Padding (can be done in the conv layer as well)
+        if len(wids) < WIN_SIZE:
+            wids += [0] * (WIN_SIZE - len(wids))
+        words_tensor = torch.tensor(wids).type(type)
+        scores = model(words_tensor)
+        predict = scores[0].argmax().item()
+        if predict == tag:
+            test_correct += 1
+    print("iter %r: test acc=%.4f" % (ITER, test_correct/len(dev)))
+
+
+for words, wids, tag in dev:
+    calc_predict_and_activations(wids, tag, words)
+    input()
diff --git a/05-cnn-pytorch/cnn-class.py b/05-cnn-pytorch/cnn-class.py
@@ -0,0 +1,108 @@
+from collections import defaultdict
+import time
+import random
+import torch
+
+
+class CNNclass(torch.nn.Module):
+    def __init__(self, nwords, emb_size, num_filters, window_size, ntags):
+        super(CNNclass, self).__init__()
+
+        """ layers """
+        self.embedding = torch.nn.Embedding(nwords, emb_size)
+        # uniform initialization
+        torch.nn.init.uniform_(self.embedding.weight, -0.25, 0.25)
+        # Conv 1d
+        self.conv_1d = torch.nn.Conv1d(in_channels=emb_size, out_channels=num_filters, kernel_size=window_size,
+                                       stride=1, padding=0, dilation=1, groups=1, bias=True)
+        self.relu = torch.nn.ReLU()
+        self.projection_layer = torch.nn.Linear(in_features=num_filters, out_features=ntags, bias=True)
+        # Initializing the projection layer
+        torch.nn.init.xavier_uniform_(self.projection_layer.weight)
+
+    def forward(self, words):
+        emb = self.embedding(words)                 # nwords x emb_size
+        emb = emb.unsqueeze(0).permute(0, 2, 1)     # 1 x emb_size x nwords
+        h = self.conv_1d(emb)                       # 1 x num_filters x nwords
+        # Do max pooling
+        h = h.max(dim=2)[0]                         # 1 x num_filters
+        h = self.relu(h)
+        out = self.projection_layer(h)              # size(out) = 1 x ntags
+        return out
+
+
+# Functions to read in the corpus
+w2i = defaultdict(lambda: len(w2i))
+t2i = defaultdict(lambda: len(t2i))
+UNK = w2i["<unk>"]
+
+
+def read_dataset(filename):
+    with open(filename, "r") as f:
+        for line in f:
+            tag, words = line.lower().strip().split(" ||| ")
+            yield ([w2i[x] for x in words.split(" ")], t2i[tag])
+
+
+# Read in the data
+train = list(read_dataset("../data/classes/train.txt"))
+w2i = defaultdict(lambda: UNK, w2i)
+dev = list(read_dataset("../data/classes/test.txt"))
+nwords = len(w2i)
+ntags = len(t2i)
+
+# Define the model
+EMB_SIZE = 64
+WIN_SIZE = 3
+FILTER_SIZE = 64
+
+# initialize the model
+model = CNNclass(nwords, EMB_SIZE, FILTER_SIZE, WIN_SIZE, ntags)
+criterion = torch.nn.CrossEntropyLoss()
+optimizer = torch.optim.Adam(model.parameters())
+
+type = torch.LongTensor
+use_cuda = torch.cuda.is_available()
+
+if use_cuda:
+    type = torch.cuda.LongTensor
+    model.cuda()
+
+
+for ITER in range(100):
+    # Perform training
+    random.shuffle(train)
+    train_loss = 0.0
+    train_correct = 0.0
+    start = time.time()
+    for words, tag in train:
+        # Padding (can be done in the conv layer as well)
+        if len(words) < WIN_SIZE:
+            words += [0] * (WIN_SIZE - len(words))
+        words_tensor = torch.tensor(words).type(type)
+        tag_tensor = torch.tensor([tag]).type(type)
+        scores = model(words_tensor)
+        predict = scores[0].argmax().item()
+        if predict == tag:
+            train_correct += 1
+
+        my_loss = criterion(scores, tag_tensor)
+        train_loss += my_loss.item()
+        # Do back-prop
+        optimizer.zero_grad()
+        my_loss.backward()
+        optimizer.step()
+    print("iter %r: train loss/sent=%.4f, acc=%.4f, time=%.2fs" % (
+        ITER, train_loss / len(train), train_correct / len(train), time.time() - start))
+    # Perform testing
+    test_correct = 0.0
+    for words, tag in dev:
+        # Padding (can be done in the conv layer as well)
+        if len(words) < WIN_SIZE:
+            words += [0] * (WIN_SIZE - len(words))
+        words_tensor = torch.tensor(words).type(type)
+        scores = model(words_tensor)[0]
+        predict = scores.argmax().item()
+        if predict == tag:
+            test_correct += 1
+    print("iter %r: test acc=%.4f" % (ITER, test_correct / len(dev)))
