Closed paynejason closed 1 year ago
C-K-Loan
commented
1 year ago Hi @paynejason the problem you have comes from the M1 chip you are using as you have correctly noticed most likely.
You can tell NLU to load Spark-NLP and the SparkSession with Jars that have been compiled for the M1 architecture with
nlu.load('biobert',m1_chip=True,)
Please let me know if this works and if not share more details to about your Apple hardware.
paynejason
commented
1 year ago Hi, thank you for the quick response. Adding the m1_chip=True argument has changed the error, causing the model to fail when it is actually being used. Here is the error:
# A fatal error has been detected by the Java Runtime Environment:
#
# SIGILL (0x4) at pc=0x000000014d288135, pid=73656, tid=0x000000000000663b
#
# JRE version: Java(TM) SE Runtime Environment (8.0_341-b10) (build 1.8.0_341-b10)
# Java VM: Java HotSpot(TM) 64-Bit Server VM (25.341-b10 mixed mode bsd-amd64 compressed oops)
# Problematic frame:
# C [libtensorflow_framework.2.dylib+0x20135] _ZN10tensorflow10monitoring9MetricDefILNS0_10MetricKindE1ExLi2EEC2IJA11_cA7_cEEEN4absl12lts_2021032411string_viewES9_DpRKT_+0x125
#
# Failed to write core dump. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again
#
# An error report file with more information is saved as:
# /Users/jason/Documents/GitHub/ontology-mapper/hs_err_pid73656.log
#
# If you would like to submit a bug report, please visit:
# http://bugreport.java.com/bugreport/crash.jsp
# The crash happened outside the Java Virtual Machine in native code.
# See problematic frame for where to report the bug.
#
ERROR:root:Exception while sending command.
Traceback (most recent call last):
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/py4j/clientserver.py", line 516, in send_command
raise Py4JNetworkError("Answer from Java side is empty")
py4j.protocol.Py4JNetworkError: Answer from Java side is empty
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/py4j/java_gateway.py", line 1038, in send_command
response = connection.send_command(command)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/py4j/clientserver.py", line 539, in send_command
raise Py4JNetworkError(
py4j.protocol.Py4JNetworkError: Error while sending or receiving
[OK!]
Traceback (most recent call last):
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/nlu/pipe/component_resolution.py", line 276, in get_trained_component_for_nlp_model_ref
component.get_pretrained_model(nlp_ref, lang, model_bucket),
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/nlu/components/embeddings/sentence_bert/BertSentenceEmbedding.py", line 13, in get_pretrained_model
return BertSentenceEmbeddings.pretrained(name,language,bucket) \
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/sparknlp/annotator/embeddings/bert_sentence_embeddings.py", line 231, in pretrained
return ResourceDownloader.downloadModel(BertSentenceEmbeddings, name, lang, remote_loc)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/sparknlp/pretrained/resource_downloader.py", line 40, in downloadModel
j_obj = _internal._DownloadModel(reader.name, name, language, remote_loc, j_dwn).apply()
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/sparknlp/internal/__init__.py", line 317, in __init__
super(_DownloadModel, self).__init__("com.johnsnowlabs.nlp.pretrained." + validator + ".downloadModel", reader,
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/sparknlp/internal/extended_java_wrapper.py", line 26, in __init__
self._java_obj = self.new_java_obj(java_obj, *args)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/sparknlp/internal/extended_java_wrapper.py", line 36, in new_java_obj
return self._new_java_obj(java_class, *args)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/pyspark/ml/wrapper.py", line 86, in _new_java_obj
return java_obj(*java_args)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/py4j/java_gateway.py", line 1321, in __call__
return_value = get_return_value(
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/pyspark/sql/utils.py", line 190, in deco
return f(*a, **kw)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/py4j/protocol.py", line 334, in get_return_value
raise Py4JError(
py4j.protocol.Py4JError: An error occurred while calling z:com.johnsnowlabs.nlp.pretrained.PythonResourceDownloader.downloadModel
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/nlu/__init__.py", line 234, in load
nlu_component = nlu_ref_to_component(nlu_ref)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/nlu/pipe/component_resolution.py", line 160, in nlu_ref_to_component
resolved_component = get_trained_component_for_nlp_model_ref(lang, nlu_ref, nlp_ref, license_type, model_params)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/nlu/pipe/component_resolution.py", line 287, in get_trained_component_for_nlp_model_ref
raise ValueError(f'Failure making component, nlp_ref={nlp_ref}, nlu_ref={nlu_ref}, lang={lang}, \n err={e}')
ValueError: Failure making component, nlp_ref=sent_biobert_pmc_base_cased, nlu_ref=en.embed_sentence.biobert.pmc_base_cased, lang=en,
err=An error occurred while calling z:com.johnsnowlabs.nlp.pretrained.PythonResourceDownloader.downloadModel
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/Users/jason/Documents/GitHub/ontology-mapper/text2term/__main__.py", line 48, in <module>
Text2Term().map_file(arguments.source, arguments.target, output_file=arguments.output, csv_columns=csv_columns,
File "/Users/jason/Documents/GitHub/ontology-mapper/text2term/t2t.py", line 63, in map_file
return self.map(source_terms, target_ontology, source_terms_ids=source_terms_ids, base_iris=base_iris,
File "/Users/jason/Documents/GitHub/ontology-mapper/text2term/t2t.py", line 115, in map
self._do_biobert_mapping(source_terms, target_terms, biobert_file)
File "/Users/jason/Documents/GitHub/ontology-mapper/text2term/t2t.py", line 161, in _do_biobert_mapping
biobert = BioBertMapper(ontology_terms)
File "/Users/jason/Documents/GitHub/ontology-mapper/text2term/biobert_mapper.py", line 28, in __init__
self.biobert = self.load_biobert()
File "/Users/jason/Documents/GitHub/ontology-mapper/text2term/biobert_mapper.py", line 35, in load_biobert
biobert = nlu.load('en.embed_sentence.biobert.pmc_base_cased', m1_chip=True,)
File "/Users/jason/.pyenv/versions/3.10.6/lib/python3.10/site-packages/nlu/__init__.py", line 249, in load
raise Exception(
Exception: Something went wrong during creating the Spark NLP model_anno_obj for your request = en.embed_sentence.biobert.pmc_base_cased Did you use a NLU Spell?I'm unsure why the load fails in this instance based on the error message. Here is a screenshot of the computer details:
C-K-Loan
commented
1 year ago Thank you for the detailed information @paynejason , M1 support is still experimental and there could be some chips or setups not fully supported yet.
Maybe @DevinTDHa or @maziyarpanahi want to give their 5 cents here? They are our core Apple Devs and have worked the M1 support
maziyarpanahi
commented
1 year ago Thanks @C-K-Loan
I am sure @DevinTDHa has more to say about this. However, it seems that M1 is different than M1 Pro and M1 Max. These are very new and proprietary chips with little to know about. We also don't have any CI/CD support at the moment so we can automate this and test everything thoroughly. We also had to build everything ourselves, there is no support for M1 in TensorFlow for Java and other dependencies we need.
Maybe we missed something and maybe it is possible to build/release for M1 that supports Pro and Max (I have no idea about the M2 family), but at the moment it seems only Apple M1 is supported with our build.
DevinTDHa
commented
1 year ago Hi @C-K-Loan, @paynejason
As @maziyarpanahi said indeed the current build of Spark NLP only supports the default M1 architecture. Other processor deviations seem to have a differing instruction set, thus the error.
We are hoping to broaden support once CI/CD based on M1 is more generally available but in the meantime we recommend to run the pipelines remotely, such as in Google Colab or Databricks (if thats possible for you use case.)
paynejason
commented
1 year ago Alright, thank you for the assistance. I will use Google Colab going forward.
I am using the nlu BioBert mapper to improve upon a tool that already exists called text2term. A few weeks ago, I was able to get the tool working on a personal computer (Mac), but shortly after when I switched to my new work computer (also Mac, same OS but with an Apple Chip instead of Intel), the program no longer worked even with the same source code, Python, and Java version.
A coworker recreated the issue with an Apple Chip computer, Python 3.9.5, and Java 17. If you have any insights, please let me know.
Here are the co-requirements, as well as the versions and the error: Python 3.10.6 (Also tried 3.9.13) Java version "1.8.0_341" (Also tried Java 16) requirements.txt:
Error: