Thrift error when running make serve

[anishpdoshi]

Hello, when running make serve after cloning Picard, I am hitting a thrift.py3.exceptions.TransportError:

I am running this on an arm64 architecture (MacBook M1), could that be why? I have confirmed nothing else is listening on port 8000.

docker pull tscholak/text-to-sql-eval:5ff827fa65c719ff975a37bd1d6940214731f3f5
5ff827fa65c719ff975a37bd1d6940214731f3f5: Pulling from tscholak/text-to-sql-eval
Digest: sha256:8efe93cb37a75812d4906c11c8d2902eebb125a0023ce8a93b099af3186f8c0e
Status: Image is up to date for tscholak/text-to-sql-eval:5ff827fa65c719ff975a37bd1d6940214731f3f5
docker.io/tscholak/text-to-sql-eval:5ff827fa65c719ff975a37bd1d6940214731f3f5
mkdir -p -m 777 database
mkdir -p -m 777 transformers_cache
docker run \
		-it \
		--rm \
		--user 13011:13011 \
		-p 8000:8000 \
		--mount type=bind,source=/Users/apdoshi/picard/database,target=/database \
		--mount type=bind,source=/Users/apdoshi/picard/transformers_cache,target=/transformers_cache \
		--mount type=bind,source=/Users/apdoshi/picard/configs,target=/app/configs \
		tscholak/text-to-sql-eval:5ff827fa65c719ff975a37bd1d6940214731f3f5 \
		/bin/bash -c "python seq2seq/serve_seq2seq.py configs/serve.json"
WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested
Traceback (most recent call last):
  File "seq2seq/serve_seq2seq.py", line 146, in <module>
    main()
  File "seq2seq/serve_seq2seq.py", line 97, in main
    model = model_cls_wrapper(AutoModelForSeq2SeqLM).from_pretrained(
  File "seq2seq/serve_seq2seq.py", line 91, in <lambda>
    model_cls=model_cls, picard_args=picard_args, tokenizer=tokenizer
  File "/app/seq2seq/utils/picard_model_wrapper.py", line 198, in with_picard
    asyncio.run(_init_picard(), debug=False)
  File "/opt/conda/lib/python3.7/asyncio/runners.py", line 43, in run
    return loop.run_until_complete(main)
  File "/opt/conda/lib/python3.7/asyncio/base_events.py", line 587, in run_until_complete
    return future.result()
  File "/app/seq2seq/utils/picard_model_wrapper.py", line 127, in _init_picard
    await _register_tokenizer(picard_client=client)
  File "/app/seq2seq/utils/picard_model_wrapper.py", line 145, in _register_tokenizer
    await picard_client.registerTokenizer(json_str)
thrift.py3.exceptions.TransportError: (<TransportErrorType.UNKNOWN: 0>, 'Channel is !good()', 0, <TransportOptions.0: 0>)
make: *** [serve] Error 1

[tscholak]

Hi @anishpdoshi!
It could be caused by a platform mismatch between amd64 and arm, yes. I don't know what is supposed to happen in this case. Is Apple's rosetta emulator supposed to make amd64 programs and docker images run on M1?

[anishpdoshi]

Thanks @tscholak for the quick response. I am not sure if it is linked to Rosetta (I'm on MacOS Monterey), but since it seems to be originate from the registerTokenizer call, I suspect it is related to these issues where the 'tokenizers' library isn't building properly on Apple Silicon:

huggingface/tokenizers#712
treygrainger/ai-powered-search#48

Maybe there's a way around this by building the docker image with --platform linux/arm64? Anyway I'm just going to use a different environment for now, feel free to close this issue if you'd like.

[tscholak]

not sure what I can do here to help. arm64 is just not supported.