Unsupported processor eroor on Amazon SageMaker notebook

Amazon SageMaker
1

Hello.
I’m trying to train huggingface model on Amazon SageMaker and having error.

ValueError: Unsupported processor: cpu. You may need to upgrade your SDK version (pip install -U sagemaker) for newer processors. Supported processor(s): gpu.

I hope anyone may give some advise on what to do. thanks.

I used the code from here

・Select “train”
・Set “Task”:Summarization, “Configuration”:AWS
・Copy showed code and pasted to Amazon SageMaker notebook
・Execute code and get the error.

code is below. I changed source_dir and instance_type to fit my notebook.

import sagemaker
from sagemaker.huggingface import HuggingFace

type or paste code here

# gets role for executing training job
role = sagemaker.get_execution_role()
hyperparameters = {
	'model_name_or_path':'rinna/japanese-roberta-base',
	'output_dir':'/opt/ml/model'
	# add your remaining hyperparameters
	# more info here https://github.com/huggingface/transformers/tree/v4.17.0/examples/pytorch/seq2seq
}

# git configuration to download our fine-tuning script
git_config = {'repo': 'https://github.com/huggingface/transformers.git','branch': 'v4.17.0'}

# creates Hugging Face estimator
huggingface_estimator = HuggingFace(
	entry_point='run_summarization.py',
	source_dir='./examples/pytorch/summarization',
	instance_type='ml.t2.medium',
	instance_count=1,
	role=role,
	git_config=git_config,
	transformers_version='4.17.0',
	pytorch_version='1.10.2',
	py_version='py38',
	hyperparameters = hyperparameters
)

# starting the train job
huggingface_estimator.fit()
2

Hey @fikuegc,

you are trying to use a CPU instance (t2.medium), which is not supported for training. You can use,e.g. a GPU instance like ml.g4dn.xlarge or ml.p3.2xlarge.

1 Like
3

Thank you for your comment. and sorry my strange question.
I changed to use ml.g4dn.xlarge but got same error.

The code says “Unsupported processor: cpu”, so it may want to use cpu.
But there’s already cpu.
I execute this↓

torch.tensor([1.,2.]).device

and I got this ↓
device(type=‘cpu’)

4

The notebook instance is not related to the training environment. If you are new to sagemaker i recommend you either read the documentation or take a look at this video where we show to fine-tune transformers

1 Like
5

Thank you for your kind advise. I’m very new to sagemaker. I’ll check them.

1 Like