High memory usage with github page as sample
entrptaher opened this issue · 10 comments
import requests
from mlscraper.html import Page
from mlscraper.samples import Sample, TrainingSet
from mlscraper.training import train_scraper
# fetch the page to train
einstein_url = 'https://github.com/lorey/mlscraper/issues/38'
resp = requests.get(einstein_url)
assert resp.status_code == 200
# create a sample for Albert Einstein
# please add at least two samples in practice to get meaningful rules!
training_set = TrainingSet()
page = Page(resp.content)
sample = Sample(page, {'title': 'Scraper not found error'})
training_set.add_sample(sample)
# train the scraper with the created training set
scraper = train_scraper(training_set)
# scrape another page
resp = requests.get('https://github.com/lorey/mlscraper/issues/27')
result = scraper.get(Page(resp.content))
print(result)Thanks for reporting.
General tip: Add more than one sample to see if it persists. Using only one sample cannot yield statistically sound heuristics. Will add a warning. It even says so in the comment you coped!
It's not a memory leak, I assume, it's just using a lot of memory. Usually potential CSS rules get reduced by applying them to every sample. If you add only one, that does not work.
Same code produces the following result for me withing seconds:
<DictScraper self.scraper_per_key={'title': <ValueScraper self.selector=<CssRuleSelector self.css_rule='bdi'>, self.extractor=<TextValueExtractor>>}>
{'title': 'Improve version pinning'}
Please add dependencies (pip freeze, etc.) and further information to reproduce, so far I'm unable to understand your issue and it works for me with the latest version from the develop branch.
Even the readme clearly states:
If you want to check the new release, use pip install --pre mlscraper to test the release candidate. You can also install the latest (unstable) development version of mlscraper via pip install git+https://github.com/lorey/mlscraper#egg=mlscraper, e.g. to check new features or to see if a bug has been fixed already.
If I run this on google colab, I don't get high memory usage but I get 'is not in list' error. However this still causes high memory locally with python 3.10 and mlscraper (both pre and develop versions).
Link: https://colab.research.google.com/drive/1frHuWVaAq-86FhhwCSyYlel-qBxaPDIs?usp=sharing
Both python Version 3.9 and 3.10 tested on google colab and locally on ubuntu 22.04
requirements.txt
beautifulsoup4==4.12.2
certifi==2022.12.7
charset-normalizer==3.1.0
idna==3.4
lxml==4.9.2
mlscraper==1.0.0rc3
more-itertools==9.1.0
requests==2.29.0
soupsieve==2.4.1
urllib3==1.26.15
Not sure what is going on. You seem to get a good result while I cannot, using the same code.
Why don't you add a second example?
W was now able to reproduce, will look into this if I find the time.
Seems like it does not stop generating CSS selectors although the tag is unique already.
Any news on this issue? I run into the same problem with another website
I'm running into the same problem :(
Has anyone managed to work around this yet? Tried a number of different sites with 5+ samples for each but always running out of memory.
Same problem，i have 32g memory but always running out of memory :(