Add Transformer Pytorch

4-2.Attention/Attention-Tensor.py

@@ -14,7 +14,7 @@
 n_class = len(word_dict)  # vocab list
 
 # Parameter
-max_len = 5  # maxium number of words in one sentence(=n_step)
+max_len = 5  # maxium number of words in one sentence(=number of time steps)
 n_hidden = 128
 
 def make_batch(sentences):

4-4.Transformer/Transformer-Tensor.py

@@ -0,0 +1,16 @@
+# code by Tae Hwan Jung(Jeff Jung) @graykode
+import tensorflow as tf
+import numpy as np
+
+tf.reset_default_graph()
+# S: Symbol that shows starting of decoding input
+# E: Symbol that shows starting of decoding output
+# P: Symbol that will fill in blank sequence if current batch data size is short than time steps
+sentences = ['ich mochte ein bier P', 'S i want a beer', 'i want a beer E']
+
+# Transformer Parameters
+word_list = " ".join(sentences).split()
+word_list = list(set(word_list))
+word_dict = {w: i for i, w in enumerate(word_list)}
+n_class = len(word_dict)  # vocab list
+embedding_size = 2

4-4.Transformer/Transformer-Torch.py

@@ -0,0 +1,192 @@
+'''
+  code by Tae Hwan Jung(Jeff Jung) @graykode
+  Reference : https://github.com/jadore801120/attention-is-all-you-need-pytorch
+              https://github.com/JayParks/transformer
+'''
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.optim as optim
+from torch.autograd import Variable
+import torch.nn.functional as F
+
+dtype = torch.FloatTensor
+# S: Symbol that shows starting of decoding input
+# E: Symbol that shows starting of decoding output
+# P: Symbol that will fill in blank sequence if current batch data size is short than time steps
+sentences = ['ich mochte ein bier P', 'S i want a beer', 'i want a beer E']
+
+# Transformer Parameters
+src_vocab = {w: i for i, w in enumerate(sentences[0].split())}
+src_vocab_size = len(src_vocab)
+tgt_vocab = {w: i for i, w in enumerate(set((sentences[1]+' '+sentences[2]).split()))}
+tgt_vocab_size = len(tgt_vocab)
+
+n_step = 5  # number of Step
+d_model = 512  # Embedding Size
+d_inner = 2048
+d_k = d_v = 64  # dimension of K(=Q), V
+n_layers = 6  # number of Encoder of Decoder Layer
+n_heads = 8  # number of heads in Multi-Head Attention
+
+def make_batch(sentences):
+    input_batch = [[src_vocab[n] for n in sentences[0].split()]]
+    output_batch = [[tgt_vocab[n] for n in sentences[1].split()]]
+    target_batch = [[tgt_vocab[n] for n in sentences[2].split()]]
+    return Variable(torch.LongTensor(input_batch)), Variable(torch.LongTensor(output_batch)), Variable(torch.LongTensor(target_batch))
+
+def get_sinusoid_encoding_table(n_position, d_model):
+    def cal_angle(position, hid_idx):
+        return position / np.power(10000, 2 * (hid_idx // 2) / d_model)
+    def get_posi_angle_vec(position):
+        return [cal_angle(position, hid_j) for hid_j in range(d_model)]
+
+    sinusoid_table = np.array([get_posi_angle_vec(pos_i) for pos_i in range(n_position)])
+    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2])  # dim 2i
+    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2])  # dim 2i+1
+    return torch.FloatTensor(sinusoid_table)
+
+def get_attn_pad_mask(seq_q, seq_k):
+    batch_size, len_q = seq_q.size()
+    batch_size, len_k = seq_k.size()
+    pad_attn_mask = seq_k.data.eq(0).unsqueeze(1)  # batch_size x 1 x len_k(=len_q)
+    return pad_attn_mask.expand(batch_size, len_q, len_k)  # batch_size x len_q x len_k
+
+class ScaledDotProductAttention(nn.Module):
+    def __init__(self):
+        super(ScaledDotProductAttention, self).__init__()
+
+    def forward(self, Q, K, V, attn_mask=None):
+        scores = torch.matmul(Q, K.transpose(-1, -2)) / np.sqrt(d_k) # scores : [batch_size x n_heads x len_q(=len_k) x len_k(=len_q)]
+        if attn_mask is not None:
+            scores.masked_fill_(attn_mask, -1e9)
+        attn = nn.Softmax(dim=-1)(scores)
+        context = torch.matmul(attn, V)
+        return context, attn
+
+class MultiHeadAttention(nn.Module):
+    def __init__(self):
+        super(MultiHeadAttention, self).__init__()
+        self.W_Q = nn.Linear(d_model, d_k * n_heads)
+        self.W_K = nn.Linear(d_model, d_k * n_heads)
+        self.W_V = nn.Linear(d_model, d_v * n_heads)
+    def forward(self, Q, K, V, attn_mask=None):
+        # q: [batch_size x len_q x d_model], k: [batch_size x len_k x d_model], v: [batch_size x len_k x d_model]
+        residual, batch_size = Q, Q.size(0)
+        q_s = self.W_Q(Q).view(batch_size, -1, n_heads, d_k).transpose(1,2)  # q_s: [batch_size x n_heads x len_q x d_k]
+        k_s = self.W_K(K).view(batch_size, -1, n_heads, d_k).transpose(1,2)  # k_s: [batch_size x n_heads x len_k x d_k]
+        v_s = self.W_V(V).view(batch_size, -1, n_heads, d_v).transpose(1,2)  # v_s: [batch_size x n_heads x len_k x d_v]
+
+        if attn_mask is not None: # attn_mask : [batch_size x len_q x len_k]
+            attn_mask = attn_mask.unsqueeze(1).repeat(1, n_heads, 1, 1) # attn_mask : [batch_size x n_heads x len_q x len_k]
+        # context: [batch_size x n_heads x len_q x d_v], attn: [batch_size x n_heads x len_q(=len_k) x len_k(=len_q)]
+        context, attn = ScaledDotProductAttention()(q_s, k_s, v_s, attn_mask=attn_mask)
+        context = context.transpose(1, 2).contiguous().view(batch_size, -1, n_heads * d_v) # context: [batch_size x len_q x n_heads * d_v]
+        output = nn.Linear(n_heads * d_v, d_model)(context)
+        return nn.LayerNorm(d_model)(output + residual), attn # output: [batch_size x len_q x d_model]
+
+class PoswiseFeedForwardNet(nn.Module):
+    def __init__(self):
+        super(PoswiseFeedForwardNet, self).__init__()
+        self.conv1 = nn.Conv1d(in_channels=d_model, out_channels=d_inner, kernel_size=1)
+        self.conv2 = nn.Conv1d(in_channels=d_inner, out_channels=d_model, kernel_size=1)
+
+    def forward(self, inputs):
+        residual = inputs # inputs : [batch_size, len_q, d_model]
+        output = nn.ReLU()(self.conv1(inputs.transpose(1, 2)))
+        output = self.conv2(output).transpose(1, 2)
+        return nn.LayerNorm(d_model)(output + residual)
+
+class EncoderLayer(nn.Module):
+    def __init__(self):
+        super(EncoderLayer, self).__init__()
+        self.enc_self_attn = MultiHeadAttention()
+        self.pos_ffn = PoswiseFeedForwardNet()
+
+    def forward(self, enc_inputs):
+        enc_outputs, attn = self.enc_self_attn(enc_inputs, enc_inputs, enc_inputs) # enc_inputs to same Q,K,V
+        enc_outputs = self.pos_ffn(enc_outputs) # enc_outputs: [batch_size x len_q x d_model]
+        return enc_outputs, attn
+
+class DecoderLayer(nn.Module):
+    def __init__(self):
+        super(DecoderLayer, self).__init__()
+        self.dec_self_attn = MultiHeadAttention()
+        self.dec_enc_attn = MultiHeadAttention()
+        self.pos_ffn = PoswiseFeedForwardNet()
+
+    def forward(self, dec_inputs, enc_outputs, enc_attn_mask):
+        dec_outputs, dec_self_attn = self.dec_self_attn(dec_inputs, dec_inputs, dec_inputs, None)
+        dec_outputs, dec_enc_attn = self.dec_enc_attn(dec_outputs, enc_outputs, enc_outputs, enc_attn_mask)
+        dec_outputs = self.pos_ffn(dec_outputs)
+        return dec_outputs, dec_self_attn, dec_enc_attn
+
+class Encoder(nn.Module):
+    def __init__(self):
+        super(Encoder, self).__init__()
+        self.src_emb = nn.Embedding(src_vocab_size, d_model)
+        self.pos_emb = nn.Embedding.from_pretrained(get_sinusoid_encoding_table(n_step+1 , d_model),freeze=True)
+        self.layers = nn.ModuleList([EncoderLayer() for _ in range(n_layers)])
+
+    def forward(self, enc_inputs): # enc_inputs : [batch_size x source_len]
+        enc_outputs = self.src_emb(enc_inputs) + self.pos_emb(torch.LongTensor([[1,2,3,4,5]]))
+        enc_self_attns = []
+        for layer in self.layers:
+            enc_outputs, enc_self_attn = layer(enc_outputs)
+            enc_self_attns.append(enc_self_attn)
+        return enc_outputs, enc_self_attns
+
+class Decoder(nn.Module):
+    def __init__(self):
+        super(Decoder, self).__init__()
+        self.tgt_emb = nn.Embedding(tgt_vocab_size, d_model)
+        self.pos_emb = nn.Embedding.from_pretrained(get_sinusoid_encoding_table(n_step+1 , d_model),freeze=True)
+        self.layers = nn.ModuleList([DecoderLayer() for _ in range(n_layers)])
+
+    def forward(self, dec_inputs, enc_inputs, enc_outputs): # dec_inputs : [batch_size x target_len]
+        dec_outputs = self.tgt_emb(dec_inputs) + self.pos_emb(torch.LongTensor([[1,2,3,4,5]]))
+        dec_enc_attn_pad_mask = get_attn_pad_mask(dec_inputs, enc_inputs)
+
+        dec_self_attns, dec_enc_attns = [], []
+        for layer in self.layers:
+            dec_outputs, dec_self_attn, dec_enc_attn = layer(dec_outputs, enc_outputs, enc_attn_mask=dec_enc_attn_pad_mask)
+            dec_self_attns.append(dec_self_attn)
+            dec_enc_attns.append(dec_enc_attn)
+        return dec_outputs, dec_self_attns, dec_enc_attns
+
+class Transformer(nn.Module):
+    def __init__(self):
+        super(Transformer, self).__init__()
+        self.encoder = Encoder()
+        self.decoder = Decoder()
+        self.projection = nn.Linear(d_model, tgt_vocab_size, bias=False)
+    def forward(self, enc_inputs, dec_inputs):
+        enc_outputs, enc_self_attns = self.encoder(enc_inputs)
+        dec_outputs, dec_self_attns, dec_enc_attns = self.decoder(dec_inputs, enc_inputs, enc_outputs)
+        dec_logits = self.projection(dec_outputs) # dec_logits : [batch_size x src_vocab_size x tgt_vocab_size]
+        return dec_logits.view(-1, dec_logits.size(-1)), enc_self_attns, dec_self_attns, dec_enc_attns
+
+model = Transformer()
+
+criterion = nn.CrossEntropyLoss()
+optimizer = optim.Adam(model.parameters(), lr=0.001)
+
+for epoch in range(100):
+    optimizer.zero_grad()
+    enc_inputs, dec_inputs, target_batch = make_batch(sentences)
+    outputs, enc_self_attns, dec_self_attns, dec_enc_attns = model(enc_inputs, dec_inputs)
+    loss = criterion(outputs, target_batch.contiguous().view(-1))
+    if (epoch + 1) % 10 == 0:
+        print('Epoch:', '%04d' % (epoch + 1), 'cost =', '{:.6f}'.format(loss))
+    loss.backward()
+    optimizer.step()
+
+# Test
+predict, _, _, _ = model(enc_inputs, dec_inputs)
+predict = predict.data.max(1, keepdim=True)[1]
+output = ''
+for pre in predict:
+    for key, value in tgt_vocab.items():
+        if value == pre:
+            output += ' ' + key
+print(sentences[0], '->', output)

README.md

@@ -4,7 +4,7 @@
 
 `nlp-tutorial` is a tutorial for who is studying NLP(Natural Language Processing) using **TensorFlow** and **Pytorch**. 
 
-- Most of the models were implemented with less than **100 lines** of code.(except comments or blank lines)
+- Most of the models in NLP were implemented with less than **100 lines** of code.(except comments or blank lines)
 - You can also learn Tensorflow or Pytorch. 
 
 ## Curriculum
@@ -38,6 +38,7 @@
 - 3-1. [TextRNN](https://github.com/graykode/nlp-tutorial/tree/master/3-1.TextRNN) - **Predict Next Step**
   - Paper - [Finding Structure in Time(1990)](http://psych.colorado.edu/~kimlab/Elman1990.pdf)
 - 3-2. TextLSTM - **Autocomplete**
+  - Paper - [LONG SHORT-TERM MEMORY](https://www.bioinf.jku.at/publications/older/2604.pdf)
 - 3-3. Bi-LSTM - **Sentences Classification**
 
 
@@ -47,8 +48,9 @@
 - 4-1. [Sequence2Sequence](https://github.com/graykode/nlp-tutorial/tree/master/4-1.Seq2Seq) - **Change Word**
   - Paper - [Learning Phrase Representations using RNN Encoder–Decoder
     for Statistical Machine Translation(2014)](https://arxiv.org/pdf/1406.1078.pdf)
-- 4-2. Attention Mechanism - **Translate**
-- 4-3. Bi-LSTM with Attention
+- 4-2. [Attention Mechanism](https://github.com/graykode/nlp-tutorial/tree/master/4-2.Attention) - **Translate**
+  - Paper - [Neural Machine Translation by Jointly Learning to Align and Translate](https://arxiv.org/abs/1409.0473)
+- 4-3. Bi-LSTM with Attention - **POS/NEG Classification**
 - 4-4. The Transformer
 
 
@@ -61,7 +63,7 @@
 
 
 
-#### 6. Pre-Trained Model
+#### 6. New Trend Model
 
 - 6-1. BERT