PLC/vPLC>> python-glove>> 返回
项目作者： iconclub

项目描述 ：
Python implementation of Glove
高级语言： C
项目地址: git://github.com/iconclub/python-glove.git
创建时间： 2021-07-26T13:07:53Z
项目社区：https://github.com/iconclub/python-glove
开源协议：Apache License 2.0
下载


python-glove

Python implementation of GloVe

GloVe is an unsupervised learning algorithm for obtaining vector representations for words.
Training is performed on aggregated global word-word co-occurrence statistics from a corpus,
and the resulting representations showcase interesting linear substructures of the word vector space.

Installation

From source:

  1. $ git clone https://github.com/iconclub/python-glove
  2. $ cd python-glove
  3. $ python setup.py install

Usage

  1. >>> from glove import Glove
  3. >>> model = Glove(corpus_file='test/corpus.txt', vector_size=100, window=5, min_count=5, epochs=10, verbose=True)
  4. BUILDING VOCABULARY
  5. Processed 3381866 tokens.
  6. Counted 24746 unique words.
  7. Truncating vocabulary at min count 5.
  8. Using vocabulary of size 7731.
  10. COUNTING COOCCURRENCES
  11. window size: 5
  12. context: symmetric
  13. max product: 10485784
  14. overflow length: 28521267
  15. Reading vocab from file "/Users/hieunguyen/Desktop/ICON/python-glove/glove/.tmp/vocab.txt"...loaded 7731 words.
  16. Building lookup table...table contains 28729425 elements.
  17. Processed 3381866 tokens.
  18. Writing cooccurrences to disk.......2 files in total.
  19. Merging cooccurrence files: processed 2680220 lines.
  21. Using random seed 42
  22. SHUFFLING COOCCURRENCES
  23. array size: 127506841
  24. Shuffling by chunks: processed 2680220 lines.
  25. Wrote 1 temporary file(s).
  26. Merging temp files: processed 2680220 lines.
  28. TRAINING MODEL
  29. Read 2680220 lines.
  30. Initializing parameters...Using random seed 42
  31. done.
  32. vector size: 100
  33. vocab size: 7731
  34. x_max: 100.000000
  35. alpha: 0.750000
  36. 07/27/21 - 10:33.06AM, iter: 001, cost: 0.061383
  37. 07/27/21 - 10:33.08AM, iter: 002, cost: 0.044241
  38. 07/27/21 - 10:33.11AM, iter: 003, cost: 0.039158
  39. 07/27/21 - 10:33.13AM, iter: 004, cost: 0.036379
  40. 07/27/21 - 10:33.15AM, iter: 005, cost: 0.033546
  41. 07/27/21 - 10:33.19AM, iter: 006, cost: 0.030526
  42. 07/27/21 - 10:33.21AM, iter: 007, cost: 0.027430
  43. 07/27/21 - 10:33.23AM, iter: 008, cost: 0.024350
  44. 07/27/21 - 10:33.25AM, iter: 009, cost: 0.021456
  45. 07/27/21 - 10:33.27AM, iter: 010, cost: 0.019001
  46. >>> print(model.wv.vectors.shape)
  47. (7732, 100)
  48. >>> print(model.wv.vectors)
  49. [[-0.7832    0.230984  0.328523 ... -0.938997 -0.772137  0.827372]
  50.  [ 0.119143  0.06323   0.773245 ... -0.802186 -1.225709  0.65204 ]
  51.  [-0.382861 -0.607985  0.218486 ... -0.402255 -1.133209  0.395143]
  52.  ...
  53.  [-0.026736  0.005838 -0.052565 ...  0.016259  0.022208 -0.015785]
  54.  [-0.017614  0.020005 -0.055972 ...  0.024249  0.039124 -0.055554]
  55.  [-0.012019  0.008404 -0.034215 ...  0.026566  0.037037 -0.031336]]
  56. >>> print(model.wv.index_to_key[:10])
  57. [',', '.', 'là', 'tôi', 'một', 'có', 'và', 'những', 'chúng', 'của']
  58. >>> print(len(model.wv))
  59. 7732

Development

Pull requests are welcome.
Fun hacking!