Skip to content

Instantly share code, notes, and snippets.

@nojima

nojima/gist:95d0d813f18a990ff7ee0c1d4625f597

Last active March 7, 2018 18:08
Show Gist options
  • Save nojima/95d0d813f18a990ff7ee0c1d4625f597 to your computer and use it in GitHub Desktop.
Save nojima/95d0d813f18a990ff7ee0c1d4625f597 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
gistfile1.txt
In [1]: %load_ext autoreload
In [2]: %autoreload 2
In [3]: from stan_lda import *
In [4]: import os
In [5]: os.environ['CC'] = 'gcc-4.8'
In [6]: word_ids, doc_ids, vocab = read_corpus("ptb.train.txt", 100)
In [7]: fit = run_stan(word_ids, doc_ids, vocab, n_topics=7)
INFO:pystan:COMPILING THE C++ CODE FOR MODEL anon_model_d8759dc0e20014dd3ccabfc20650f402 NOW.
Iteration: 1 / 300 [ 0%] (Warmup) (Chain 2)
Iteration: 1 / 300 [ 0%] (Warmup) (Chain 3)
Iteration: 1 / 300 [ 0%] (Warmup) (Chain 0)
Iteration: 1 / 300 [ 0%] (Warmup) (Chain 1)
Iteration: 30 / 300 [ 10%] (Warmup) (Chain 1)
Iteration: 30 / 300 [ 10%] (Warmup) (Chain 3)
Iteration: 30 / 300 [ 10%] (Warmup) (Chain 0)
Iteration: 30 / 300 [ 10%] (Warmup) (Chain 2)
Iteration: 60 / 300 [ 20%] (Warmup) (Chain 1)
Iteration: 60 / 300 [ 20%] (Warmup) (Chain 2)
Iteration: 60 / 300 [ 20%] (Warmup) (Chain 0)
Iteration: 60 / 300 [ 20%] (Warmup) (Chain 3)
Iteration: 90 / 300 [ 30%] (Warmup) (Chain 2)
Iteration: 90 / 300 [ 30%] (Warmup) (Chain 1)
Iteration: 90 / 300 [ 30%] (Warmup) (Chain 0)
Iteration: 90 / 300 [ 30%] (Warmup) (Chain 3)
Iteration: 120 / 300 [ 40%] (Warmup) (Chain 1)
Iteration: 120 / 300 [ 40%] (Warmup) (Chain 0)
Iteration: 120 / 300 [ 40%] (Warmup) (Chain 2)
Iteration: 120 / 300 [ 40%] (Warmup) (Chain 3)
Iteration: 150 / 300 [ 50%] (Warmup) (Chain 1)
Iteration: 151 / 300 [ 50%] (Sampling) (Chain 1)
Iteration: 150 / 300 [ 50%] (Warmup) (Chain 0)
Iteration: 151 / 300 [ 50%] (Sampling) (Chain 0)
Iteration: 150 / 300 [ 50%] (Warmup) (Chain 2)
Iteration: 151 / 300 [ 50%] (Sampling) (Chain 2)
Iteration: 150 / 300 [ 50%] (Warmup) (Chain 3)
Iteration: 151 / 300 [ 50%] (Sampling) (Chain 3)
Iteration: 180 / 300 [ 60%] (Sampling) (Chain 0)
Iteration: 180 / 300 [ 60%] (Sampling) (Chain 1)
Iteration: 180 / 300 [ 60%] (Sampling) (Chain 3)
Iteration: 180 / 300 [ 60%] (Sampling) (Chain 2)
Iteration: 210 / 300 [ 70%] (Sampling) (Chain 0)
Iteration: 240 / 300 [ 80%] (Sampling) (Chain 0)
Iteration: 210 / 300 [ 70%] (Sampling) (Chain 1)
Iteration: 210 / 300 [ 70%] (Sampling) (Chain 3)
Iteration: 210 / 300 [ 70%] (Sampling) (Chain 2)
Iteration: 270 / 300 [ 90%] (Sampling) (Chain 0)
Iteration: 300 / 300 [100%] (Sampling) (Chain 0)
Iteration: 240 / 300 [ 80%] (Sampling) (Chain 3)
#
# Elapsed Time: 365.896 seconds (Warm-up)
# 356.369 seconds (Sampling)
# 722.266 seconds (Total)
#
Iteration: 240 / 300 [ 80%] (Sampling) (Chain 1)
Iteration: 240 / 300 [ 80%] (Sampling) (Chain 2)
Iteration: 270 / 300 [ 90%] (Sampling) (Chain 3)
Iteration: 270 / 300 [ 90%] (Sampling) (Chain 1)
Iteration: 270 / 300 [ 90%] (Sampling) (Chain 2)
Iteration: 300 / 300 [100%] (Sampling) (Chain 3)
#
# Elapsed Time: 383.557 seconds (Warm-up)
# 559.784 seconds (Sampling)
# 943.34 seconds (Total)
#
Iteration: 300 / 300 [100%] (Sampling) (Chain 1)
#
# Elapsed Time: 342.031 seconds (Warm-up)
# 641.288 seconds (Sampling)
# 983.319 seconds (Total)
#
Iteration: 300 / 300 [100%] (Sampling) (Chain 2)
#
# Elapsed Time: 377.516 seconds (Warm-up)
# 656.548 seconds (Sampling)
# 1034.06 seconds (Total)
#
Raw
gistfile2.txt
phi[5,740] 5.8e-4 5.9e-5 1.5e-32.7e-17 4.8e-9 7.9e-6 3.2e-4 5.7e-3 600 1.0
phi[6,740] 8.3e-4 8.2e-5 2.0e-36.8e-18 2.5e-9 5.6e-6 5.5e-4 6.8e-3 600 1.0
phi[0,741] 7.3e-4 7.5e-5 1.7e-35.5e-16 1.1e-8 7.4e-6 4.4e-4 6.3e-3 547 1.0
phi[1,741] 7.0e-4 7.5e-5 1.8e-31.5e-18 1.1e-8 7.1e-6 3.6e-4 6.3e-3 600 1.0
phi[2,741] 6.9e-4 7.0e-5 1.6e-39.0e-18 1.7e-8 1.3e-5 4.7e-4 6.1e-3 497 1.0
phi[3,741] 7.7e-4 8.0e-5 1.9e-38.6e-18 1.4e-8 1.3e-5 5.5e-4 5.7e-3 580 1.0
phi[4,741] 5.8e-4 6.8e-5 1.6e-31.0e-15 1.5e-8 7.6e-6 3.1e-4 4.8e-3 562 1.01
phi[5,741] 5.2e-4 6.0e-5 1.4e-36.2e-18 4.6e-9 2.5e-6 2.2e-4 4.4e-3 558 1.0
phi[6,741] 7.6e-4 7.1e-5 1.7e-32.4e-17 1.7e-8 9.3e-6 5.1e-4 6.4e-3 600 1.0
phi[0,742] 7.4e-4 8.2e-5 1.8e-31.9e-19 2.7e-9 4.2e-6 5.4e-4 5.9e-3 477 1.0
phi[1,742] 6.5e-4 7.4e-5 1.5e-36.4e-20 1.3e-9 5.3e-6 4.8e-4 6.0e-3 417 1.01
phi[2,742] 5.5e-4 6.3e-5 1.5e-31.7e-16 4.1e-9 2.1e-6 2.4e-4 4.6e-3 567 1.0
phi[3,742] 6.2e-4 5.7e-5 1.4e-31.6e-18 6.5e-9 1.1e-5 4.9e-4 5.4e-3 600 1.0
phi[4,742] 6.1e-4 6.4e-5 1.5e-31.2e-18 6.4e-9 6.2e-6 4.2e-4 4.8e-3 521 1.0
phi[5,742] 5.9e-4 6.4e-5 1.5e-33.1e-16 8.2e-9 7.2e-6 3.1e-4 5.3e-3 529 1.0
phi[6,742] 7.5e-4 7.6e-5 1.9e-33.1e-16 1.6e-8 5.7e-6 5.4e-4 6.8e-3 600 1.0
phi[0,743] 6.7e-4 6.8e-5 1.7e-31.3e-16 2.4e-8 2.0e-5 6.3e-4 5.7e-3 600 1.0
phi[1,743] 7.6e-4 7.3e-5 1.8e-31.1e-16 7.2e-9 1.1e-5 4.7e-4 6.3e-3 583 1.0
phi[2,743] 7.0e-4 7.9e-5 1.7e-33.1e-17 6.4e-9 3.4e-6 4.2e-4 5.5e-3 459 1.0
phi[3,743] 5.6e-4 5.9e-5 1.4e-36.1e-17 2.9e-8 1.2e-5 3.3e-4 4.8e-3 600 1.0
phi[4,743] 6.7e-4 6.9e-5 1.7e-34.0e-16 7.4e-9 5.5e-6 3.5e-4 5.6e-3 600 1.0
phi[5,743] 6.2e-4 6.8e-5 1.7e-39.7e-16 5.0e-9 6.9e-6 3.3e-4 5.3e-3 600 1.0
phi[6,743] 7.6e-4 7.4e-5 1.8e-31.5e-16 1.8e-8 7.4e-6 5.6e-4 6.5e-3 600 1.0
lp__ -2.9e4 7.58 74.65 -2.9e4 -2.9e4 -2.9e4 -2.9e4 -2.9e4 97 1.04
Samples were drawn using NUTS at 2018年03月08日 02時40分57秒.
For each parameter, n_eff is a crude measure of effective sample size,
and Rhat is the potential scale reduction factor on split chains (at
convergence, Rhat=1).
Raw
gistfile3.txt
In [58]: m = fit.extract()
In [59]: phi = m["phi"]
In [60]: phi.shape
Out[60]: (600, 7, 744)
In [61]: mphi = np.mean(phi, axis=0)
In [62]: mphi
Out[62]:
array([[ 0.01934522, 0.00067797, 0.00067912, ..., 0.00073079,
0.00074387, 0.00066906],
[ 0.05383168, 0.0005909 , 0.00057828, ..., 0.00069623,
0.00064964, 0.00076267],
[ 0.02318433, 0.00062504, 0.00082581, ..., 0.0006914 ,
0.00054864, 0.00070076],
...,
[ 0.06520887, 0.00069524, 0.00068922, ..., 0.00058358,
0.00060854, 0.00067 ],
[ 0.12132906, 0.00059118, 0.00060603, ..., 0.00052047,
0.00058573, 0.0006186 ],
[ 0.0069193 , 0.00072506, 0.00085234, ..., 0.00076468,
0.00074746, 0.00076338]])
In [63]: for row in mphi.argsort(axis=1):
...: print([vocab.word(wid+1) for wid in row[-20:]])
...:
['on', 'with', 'year', 'billion', 'at', 'yield', 'by', 'said', 'in', "'s", 'for', 'and', '$', 'from', 'a', '<unk>', 'to', 'of', 'the', 'N']
['by', 'are', 'workers', 'was', 'with', 'asbestos', 'it', 'that', 'is', 'on', "'s", 'said', 'and', 'for', 'in', 'to', 'a', 'of', '<unk>', 'the']
['funds', 'average', 'year', 'said', 'by', 'at', 'yield', "'s", 'billion', 'for', 'and', 'in', '$', 'a', 'from', 'of', 'to', '<unk>', 'the', 'N']
['year', 'at', 'asbestos', 'with', 'by', 'on', 'was', 'it', 'is', 'that', "'s", 'said', 'for', 'in', 'to', 'and', 'a', 'of', 'the', '<unk>']
['asbestos', 'it', 'is', 'was', 'with', 'on', 'that', 'from', '$', 'said', "'s", 'for', 'in', 'and', 'to', 'a', 'N', 'of', 'the', '<unk>']
['this', 'workers', 'was', 'by', 'it', 'asbestos', 'on', 'is', "'s", 'with', 'that', 'said', 'for', 'to', 'in', 'a', 'and', 'of', 'the', '<unk>']
['with', 'that', 'year', 'was', 'asbestos', 'it', 'by', 'on', '<unk>', 'is', 'said', "'s", 'N', 'for', 'and', 'in', 'to', 'of', 'a', 'the']
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment