CODEWITHSUNDEEP
pythonscikit-learn
darren17
darren17

Reputation: 235

TfidfVectorizer dtype mismatch

I'm trying to use the TfidfVectorizer on a corpus but every time I end up with this error

File "sparsefuncs.pyx", line 117, in sklearn.utils.sparsefuncs.inplace_csr_row_normalize_l2 (sklearn\utils\sparsefuncs.c:2328)
ValueError: Buffer dtype mismatch, expected 'int' but got 'long long'

This is my code

corpus = []
testCorpus = []
trainType = []
testType = []

with open("stone_sku.csv") as f:
    cr = csv.DictReader(f)
    for row in cr:
        corpus.append(row['sku'])
        trainType.append(row['sku'])

with open("stone_sku.csv") as f:
    crTest = csv.DictReader(f)
    for row in crTest:
        testCorpus.append(row['sku'])
        testType.append(row['sku'])

cv = TfidfVectorizer(min_df=1, analyzer='char', ngram_range=(2,3))

trainCounts = cv.fit_transform(corpus)

It works fine with CountVectorizer and the same error occurs if I try to transform the data using TfidfTransformer

Upvotes: 1

Views: 548

Answers (1)

ogrisel
ogrisel

Reputation: 40149

Are you running 64 bit Windows? This might be caused by a known issue that has been recently fixed in the master branch.

Upvotes: 2

Related Questions