From 491f0b32d59ef944cf6f5ea37b27c0d4886a2fde Mon Sep 17 00:00:00 2001
From: rosequ <rdsequie@uwaterloo.ca>
Date: Thu, 12 Oct 2017 22:54:45 -0400
Subject: [PATCH] Updated sm (#72)

+ removed redundant loss regularization
+ added script to create torch word embedding file from word2vec model
+ updated README
---
 sm_modified_cnn/README.md        | 19 +++++++++++++++++--
 sm_modified_cnn/requirements.txt |  1 +
 sm_modified_cnn/train.py         | 15 ---------------
 sm_modified_cnn/utils.py         | 27 ++++++++++-----------------
 4 files changed, 28 insertions(+), 34 deletions(-)

diff --git a/sm_modified_cnn/README.md b/sm_modified_cnn/README.md
index 173e5ec..f9bfc3d 100644
--- a/sm_modified_cnn/README.md
+++ b/sm_modified_cnn/README.md
@@ -11,12 +11,16 @@ Retrieval (SIGIR '15). ACM, New York, NY, USA, 373-382. DOI: http://dx.doi.org/1
 nltk==3.2.2
 numpy==1.11.3
 pytorch==0.1.12
+gensim==1.0.1 
 ```
 
 The code uses torchtext for text processing. Set torchtext:
 ```bash
 git clone https://github.com/pytorch/text.git
 cd text
+
+#use this commit number
+git reset --hard 2980f1bc39ba6af332c5c2783da8bee109796d4c
 python setup.py install
 ```
 
@@ -25,8 +29,9 @@ We use `trec_eval` for evaluation:
 ```bash
 cd eval
 tar -xvf trec_eval.9.0.tar.gz
+cd trec_eval.9.0 
 make
-cd ..
+cd ../..
 ```
 
 
@@ -131,4 +136,14 @@ Metric |rand   |static|non-static|multichannel
 MAP    |0.6313 |0.6378|0.6455    |0.6476
 MRR    |0.6522 |0.6542|0.6689    |0.6646
 
-NB: The results on WikiQA are based on the SM model hyperparameters.  
\ No newline at end of file
+NB: The results on WikiQA are based on the SM model hyperparameters.  
+
+
+### To create your own word2vec.pt file
+
++ Download word2vec from [here](https://drive.google.com/drive/u/0/folders/0B-yipfgecoSBfkZlY2FFWEpDR3M4Qkw5U055MWJrenE5MTBFVXlpRnd0QjZaMDQxejh1cWs)
+to the `data/` folder
+
+```bash
+python utils.py --input data/aquaint+wiki.txt.gz.ndim=50.bin
+```
\ No newline at end of file
diff --git a/sm_modified_cnn/requirements.txt b/sm_modified_cnn/requirements.txt
index 00d0293..a5fbe01 100644
--- a/sm_modified_cnn/requirements.txt
+++ b/sm_modified_cnn/requirements.txt
@@ -1,3 +1,4 @@
 nltk==3.2.1
 numpy==1.11.3
+gensim==1.0.1
 pytorch==0.1.12
diff --git a/sm_modified_cnn/train.py b/sm_modified_cnn/train.py
index 1f6a601..5f41de2 100644
--- a/sm_modified_cnn/train.py
+++ b/sm_modified_cnn/train.py
@@ -31,24 +31,10 @@ def set_vectors(field, vector_path):
                 # initialize <unk> with U(-0.25, 0.25) vectors
                 field.vocab.vectors[i] = torch.FloatTensor(dim).uniform_(-0.25, 0.25)
     else:
-        print("Error: Need word embedding pt file")
         print("Error: Need word embedding pt file")
         exit(1)
     return field
 
-
-def regularize_loss(model, loss):
-    flattened_params = []
-    reg = args.weight_decay
-
-    for p in model.parameters():
-        f = p.data.clone()
-        flattened_params.append(f.view(-1))
-
-    fp = torch.cat(flattened_params)
-    loss = loss + 0.5 * reg * fp.norm() * fp.norm()
-    return loss
-
 # Set default configuration in : args.py
 args = get_args()
 config = args
@@ -165,7 +151,6 @@ while True:
         train_acc = 100. * n_correct / n_total
 
         loss = criterion(scores, batch.label)
-        loss = regularize_loss(model, loss)
         loss.backward()
         optimizer.step()
 
diff --git a/sm_modified_cnn/utils.py b/sm_modified_cnn/utils.py
index d53753d..f83dd93 100644
--- a/sm_modified_cnn/utils.py
+++ b/sm_modified_cnn/utils.py
@@ -1,30 +1,23 @@
 from tqdm import tqdm
-import array
 import torch
-import numpy as np
 
+from gensim.models.keyedvectors import KeyedVectors
 from argparse import ArgumentParser
 
 
-def convert(fname, vocab):
-    save_file = '{}.pt'.format(fname)
-    stoi, vectors, dim = [], array.array('d'), None
-
-    # TODO: fix by reading the .dimensions file
-    vocab_size, dim = 2470719, 50
-    W = np.memmap(fname, dtype=np.double, shape=(vocab_size, dim))
+def convert(fname, save_file):
+    with open(fname, 'rb') as dim_file:
+        vocab_size, dim = (int(x) for x in dim_file.readline().split())
 
+    word_vectors = KeyedVectors.load_word2vec_format(fname, binary=True)
 
     print("Loading vectors from {}".format(fname))
     vectors = []
-    for line in tqdm(W, total=len(W)):
-        entry = line
-        vectors.extend(entry)
-
+    for line in tqdm(word_vectors.syn0, total=len(word_vectors.syn0)):
+        vectors.extend(line.tolist())
     vectors = torch.Tensor(vectors).view(-1, dim)
 
-    with open(vocab) as f:
-        stoi = {word.strip():i for i, word in enumerate(f)}
+    stoi = {word.strip():voc.index for word, voc in word_vectors.vocab.items()}
 
     print('saving vectors to', save_file)
     torch.save((stoi, vectors, dim), save_file)
@@ -32,7 +25,7 @@ def convert(fname, vocab):
 if __name__ == '__main__':
     parser = ArgumentParser(description='create word embedding')
     parser.add_argument('--input', type=str, required=True)
-    parser.add_argument('--vocab', type=str, required=True)
+    parser.add_argument('--output', type=str, default='data/word2vec.trecqa.pt')
 
     args = parser.parse_args()
-    convert(args.input, args.vocab)
+    convert(args.input, args.output)
\ No newline at end of file