[Bug Report] TypeError: Model._create_sagemaker_model() got an unexpected keyword argument 'predictor_cls'

[anadeem445]

Link to the notebook
Example Notebook

Describe the bug
We were trying to test Amazon SageMaker's semantic segmentation feature. So we used this example notebook from sagemaker jumpstart and tried to run it in aws sagemaker notebook instance.

But we received a weird error:
TypeError: Model._create_sagemaker_model() got an unexpected keyword argument 'predictor_cls'

To reproduce

• Create a notebook instance in amazon sagemaker
• Create IAM role with permissions (sagemakerFullAccess, s3FullAccess)
• Run example notebook

Logs

[Siddharth-Latthe-07]

@anadeem445 This issue seems to be due to an incompatibility or a change in the SageMaker SDK.
you might follow these steps and check if the issue persists:-

1. Remove predictor_cls from Model Initialization: If the predictor_cls argument is not required during model creation, remove it from the Model class initialization and only use it during deployment.
2. Update predictor_cls Usage: Modify the code to ensure predictor_cls is only used in the deploy method if needed.
   sample snippet:-

model = Model(
    image_uri=deploy_image_uri,
    source_dir=deploy_source_uri,
    model_data=base_model_uri,
    entry_point="inference.py",
    role=aws_role,
    name=endpoint_name,
)

base_model_predictor = model.deploy(
    initial_instance_count=1,
    instance_type=inference_instance_type,
    predictor_cls=Predictor,
    endpoint_name=endpoint_name,
)

plz let me know, if the above helps
Thanks

[catica]

Hello 👋 . I am sorry for hijacking the thread but found the same error as above. I've removed the predictor_cls from the model initialization and kept it in the deploy (as in the snippet above).

The error remains the same.

[Siddharth-Latthe-07]

@catica The error you're encountering, TypeError: Model._create_sagemaker_model() got an unexpected keyword argument 'predictor_cls', indicates that the Model.deploy method does not accept the predictor_cls parameter.

modify your code to exclude the predictor_cls argument from the model.deploy method. Instead, after deploying the model, you can instantiate the Predictor class separately.

sample snippet:-

from sagemaker import image_uris, model_uris, script_uris, hyperparameters
from sagemaker.model import Model
from sagemaker.predictor import Predictor
from sagemaker.utils import name_from_base

# model_version="*" fetches the latest version of the model
infer_model_id, infer_model_version = infer_model_dropdown.value, "*"

endpoint_name = name_from_base(f"jumpstart-example-infer-{infer_model_id}")

inference_instance_type = "ml.p2.xlarge"

# Retrieve the inference docker container uri
deploy_image_uri = image_uris.retrieve(
    region=None,
    framework=None,  # automatically inferred from model_id
    image_scope="inference",
    model_id=infer_model_id,
    model_version=infer_model_version,
    instance_type=inference_instance_type,
)

# Retrieve the inference script uri. This includes scripts for model loading, inference handling, etc.
deploy_source_uri = script_uris.retrieve(
    model_id=infer_model_id, model_version=infer_model_version, script_scope="inference"
)

# Retrieve the base model uri
base_model_uri = model_uris.retrieve(
    model_id=infer_model_id, model_version=infer_model_version, model_scope="inference"
)

# Create the SageMaker model instance
model = Model(
    image_uri=deploy_image_uri,
    source_dir=deploy_source_uri,
    model_data=base_model_uri,
    entry_point="inference.py",  # entry point file in source_dir and present in deploy_source_uri
    role=aws_role,
    name=endpoint_name,
)

# Deploy the Model without predictor_cls argument
base_model_predictor = model.deploy(
    initial_instance_count=1,
    instance_type=inference_instance_type,
    endpoint_name=endpoint_name,
)

# Instantiate the Predictor class separately
predictor = Predictor(endpoint_name=endpoint_name)

Let me know, if it works
Thanks

[catica]

Thanks for your quick reply @Siddharth-Latthe-07 .

I've got a different error:

---------------------------------------------------------------------------
ClientError                               Traceback (most recent call last)
Cell In[13], line 2
      1 # Deploy the Model without predictor_cls argument
----> 2 base_model_predictor = model.deploy(
      3     initial_instance_count=1,
      4     instance_type=inference_instance_type,
      5     endpoint_name=endpoint_name,
      6 )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:1692, in Model.deploy(self, initial_instance_count, instance_type, serializer, deserializer, accelerator_type, endpoint_name, tags, kms_key, wait, data_capture_config, async_inference_config, serverless_inference_config, volume_size, model_data_download_timeout, container_startup_health_check_timeout, inference_recommendation_id, explainer_config, accept_eula, endpoint_logging, resources, endpoint_type, managed_instance_scaling, inference_component_name, routing_config, **kwargs)
   1689     return None
   1691 else:  # existing single model endpoint path
-> 1692     self._create_sagemaker_model(
   1693         instance_type=instance_type,
   1694         accelerator_type=accelerator_type,
   1695         tags=tags,
   1696         serverless_inference_config=serverless_inference_config,
   1697         **kwargs,
   1698     )
   1699     serverless_inference_config_dict = (
   1700         serverless_inference_config._to_request_dict() if is_serverless else None
   1701     )
   1702     production_variant = sagemaker.production_variant(
   1703         self.name,
   1704         instance_type,
   (...)
   1711         routing_config=routing_config,
   1712     )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:927, in Model._create_sagemaker_model(self, instance_type, accelerator_type, tags, serverless_inference_config, accept_eula, model_reference_arn)
    925         self.name = model_package.name
    926 else:
--> 927     container_def = self.prepare_container_def(
    928         instance_type,
    929         accelerator_type=accelerator_type,
    930         serverless_inference_config=serverless_inference_config,
    931         accept_eula=accept_eula,
    932         model_reference_arn=model_reference_arn,
    933     )
    935     if not isinstance(self.sagemaker_session, PipelineSession):
    936         # _base_name, model_name are not needed under PipelineSession.
    937         # the model_data may be Pipeline variable
    938         # which may break the _base_name generation
    939         self._ensure_base_name_if_needed(
    940             image_uri=container_def["Image"],
    941             script_uri=self.source_dir,
    942             model_uri=self._get_model_uri(),
    943         )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:715, in Model.prepare_container_def(self, instance_type, accelerator_type, serverless_inference_config, accept_eula, model_reference_arn)
    713 deploy_env = copy.deepcopy(self.env)
    714 if self.source_dir or self.dependencies or self.entry_point or self.git_config:
--> 715     self._upload_code(deploy_key_prefix, repack=self.is_repack())
    716     deploy_env.update(self._script_mode_env_vars())
    718 return sagemaker.container_def(
    719     self.image_uri,
    720     self.repacked_model_data or self.model_data,
   (...)
    731     ),
    732 )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:822, in Model._upload_code(self, key_prefix, repack)
    806     self.uploaded_code = fw_utils.UploadedCode(
    807         s3_prefix=repacked_model_data,
    808         script_name=os.path.basename(self.entry_point),
    809     )
    811 logger.info(
    812     "Repacking model artifact (%s), script artifact "
    813     "(%s), and dependencies (%s) "
   (...)
    819     repacked_model_data,
    820 )
--> 822 utils.repack_model(
    823     inference_script=self.entry_point,
    824     source_directory=self.source_dir,
    825     dependencies=self.dependencies,
    826     model_uri=self.model_data,
    827     repacked_model_uri=repacked_model_data,
    828     sagemaker_session=self.sagemaker_session,
    829     kms_key=self.model_kms_key,
    830 )
    832 self.repacked_model_data = repacked_model_data

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:551, in repack_model(inference_script, source_directory, dependencies, model_uri, repacked_model_uri, sagemaker_session, kms_key)
    544 local_download_dir = (
    545     None
    546     if sagemaker_session.settings is None
    547     or sagemaker_session.settings.local_download_dir is None
    548     else sagemaker_session.settings.local_download_dir
    549 )
    550 with _tmpdir(directory=local_download_dir) as tmp:
--> 551     model_dir = _extract_model(model_uri, sagemaker_session, tmp)
    553     _create_or_update_code_dir(
    554         model_dir,
    555         inference_script,
   (...)
    559         tmp,
    560     )
    562     tmp_model_path = os.path.join(tmp, "temp-model.tar.gz")

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:637, in _extract_model(model_uri, sagemaker_session, tmp)
    635 if model_uri.lower().startswith("s3://"):
    636     local_model_path = os.path.join(tmp, "tar_file")
--> 637     download_file_from_url(model_uri, local_model_path, sagemaker_session)
    638 else:
    639     local_model_path = model_uri.replace("file://", "")

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:650, in download_file_from_url(url, dst, sagemaker_session)
    647 url = parse.urlparse(url)
    648 bucket, key = url.netloc, url.path.lstrip("/")
--> 650 download_file(bucket, key, dst, sagemaker_session)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:668, in download_file(bucket_name, path, target, sagemaker_session)
    666 s3 = boto_session.resource("s3", region_name=sagemaker_session.boto_region_name)
    667 bucket = s3.Bucket(bucket_name)
--> 668 bucket.download_file(path, target)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/boto3/s3/inject.py:279, in bucket_download_file(self, Key, Filename, ExtraArgs, Callback, Config)
    245 def bucket_download_file(
    246     self, Key, Filename, ExtraArgs=None, Callback=None, Config=None
    247 ):
    248     """Download an S3 object to a file.
    249 
    250     Usage::
   (...)
    277         transfer.
    278     """
--> 279     return self.meta.client.download_file(
    280         Bucket=self.name,
    281         Key=Key,
    282         Filename=Filename,
    283         ExtraArgs=ExtraArgs,
    284         Callback=Callback,
    285         Config=Config,
    286     )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/boto3/s3/inject.py:192, in download_file(self, Bucket, Key, Filename, ExtraArgs, Callback, Config)
    157 """Download an S3 object to a file.
    158 
    159 Usage::
   (...)
    189     transfer.
    190 """
    191 with S3Transfer(self, Config) as transfer:
--> 192     return transfer.download_file(
    193         bucket=Bucket,
    194         key=Key,
    195         filename=Filename,
    196         extra_args=ExtraArgs,
    197         callback=Callback,
    198     )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/boto3/s3/transfer.py:406, in S3Transfer.download_file(self, bucket, key, filename, extra_args, callback)
    402 future = self._manager.download(
    403     bucket, key, filename, extra_args, subscribers
    404 )
    405 try:
--> 406     future.result()
    407 # This is for backwards compatibility where when retries are
    408 # exceeded we need to throw the same error from boto3 instead of
    409 # s3transfer's built in RetriesExceededError as current users are
    410 # catching the boto3 one instead of the s3transfer exception to do
    411 # their own retries.
    412 except S3TransferRetriesExceededError as e:

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/futures.py:103, in TransferFuture.result(self)
     98 def result(self):
     99     try:
    100         # Usually the result() method blocks until the transfer is done,
    101         # however if a KeyboardInterrupt is raised we want want to exit
    102         # out of this and propagate the exception.
--> 103         return self._coordinator.result()
    104     except KeyboardInterrupt as e:
    105         self.cancel()

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/futures.py:266, in TransferCoordinator.result(self)
    263 # Once done waiting, raise an exception if present or return the
    264 # final result.
    265 if self._exception:
--> 266     raise self._exception
    267 return self._result

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/tasks.py:269, in SubmissionTask._main(self, transfer_future, **kwargs)
    265     self._transfer_coordinator.set_status_to_running()
    267     # Call the submit method to start submitting tasks to execute the
    268     # transfer.
--> 269     self._submit(transfer_future=transfer_future, **kwargs)
    270 except BaseException as e:
    271     # If there was an exception raised during the submission of task
    272     # there is a chance that the final task that signals if a transfer
   (...)
    281 
    282     # Set the exception, that caused the process to fail.
    283     self._log_and_set_exception(e)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/download.py:354, in DownloadSubmissionTask._submit(self, client, config, osutil, request_executor, io_executor, transfer_future, bandwidth_limiter)
    325 """
    326 :param client: The client associated with the transfer manager
    327 
   (...)
    349     downloading streams
    350 """
    351 if transfer_future.meta.size is None:
    352     # If a size was not provided figure out the size for the
    353     # user.
--> 354     response = client.head_object(
    355         Bucket=transfer_future.meta.call_args.bucket,
    356         Key=transfer_future.meta.call_args.key,
    357         **transfer_future.meta.call_args.extra_args,
    358     )
    359     transfer_future.meta.provide_transfer_size(
    360         response['ContentLength']
    361     )
    363 download_output_manager = self._get_download_output_manager_cls(
    364     transfer_future, osutil
    365 )(osutil, self._transfer_coordinator, io_executor)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/botocore/client.py:565, in ClientCreator._create_api_method.<locals>._api_call(self, *args, **kwargs)
    561     raise TypeError(
    562         f"{py_operation_name}() only accepts keyword arguments."
    563     )
    564 # The "self" in this scope is referring to the BaseClient.
--> 565 return self._make_api_call(operation_name, kwargs)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/botocore/client.py:1021, in BaseClient._make_api_call(self, operation_name, api_params)
   1017     error_code = error_info.get("QueryErrorCode") or error_info.get(
   1018         "Code"
   1019     )
   1020     error_class = self.exceptions.from_code(error_code)
-> 1021     raise error_class(parsed_response, operation_name)
   1022 else:
   1023     return parsed_response

ClientError: An error occurred (404) when calling the HeadObject operation: Not Found

[Siddharth-Latthe-07]

@catica The new error ClientError: An error occurred (404) when calling the HeadObject operation: Not Found indicates that SageMaker is unable to locate the specified model data on S3. This could be due to an incorrect S3 URI for the model data, or it might be that the model data does not exist at the specified location.

possible solutions:-

1. Verify S3 URI: Ensure that the S3 URI for model_data is correct and that the object exists in the specified S3 bucket and key.
2. Permissions: Ensure that the SageMaker role (aws_role) has the necessary permissions to access the S3 bucket and objects.
3. Model Data Location: Verify that the model data URI (base_model_uri) points to the correct location.
   sample snippet for debugging:-

from sagemaker import image_uris, model_uris, script_uris, hyperparameters
from sagemaker.model import Model
from sagemaker.predictor import Predictor
from sagemaker.utils import name_from_base

# model_version="*" fetches the latest version of the model
infer_model_id, infer_model_version = infer_model_dropdown.value, "*"

endpoint_name = name_from_base(f"jumpstart-example-infer-{infer_model_id}")

inference_instance_type = "ml.p2.xlarge"

# Retrieve the inference docker container uri
deploy_image_uri = image_uris.retrieve(
    region=None,
    framework=None,  # automatically inferred from model_id
    image_scope="inference",
    model_id=infer_model_id,
    model_version=infer_model_version,
    instance_type=inference_instance_type,
)

# Retrieve the inference script uri. This includes scripts for model loading, inference handling, etc.
deploy_source_uri = script_uris.retrieve(
    model_id=infer_model_id, model_version=infer_model_version, script_scope="inference"
)

# Retrieve the base model uri
base_model_uri = model_uris.retrieve(
    model_id=infer_model_id, model_version=infer_model_version, model_scope="inference"
)

print(f"Deploy Image URI: {deploy_image_uri}")
print(f"Deploy Source URI: {deploy_source_uri}")
print(f"Base Model URI: {base_model_uri}")

# Create the SageMaker model instance
model = Model(
    image_uri=deploy_image_uri,
    source_dir=deploy_source_uri,
    model_data=base_model_uri,
    entry_point="inference.py",  # entry point file in source_dir and present in deploy_source_uri
    role=aws_role,
    name=endpoint_name,
)

# Deploy the Model without predictor_cls argument
base_model_predictor = model.deploy(
    initial_instance_count=1,
    instance_type=inference_instance_type,
    endpoint_name=endpoint_name,
)

# Instantiate the Predictor class separately
predictor = Predictor(endpoint_name=endpoint_name)

Plz let meknow, if you find error logs
Thanks

[catica]

I am not sure I understand these problems with S3 accesses as I am trying to use the provided pytorch-SSD model (not my own model). And I'm not trying to access any of my own S3 buckets, I believe.

I've followed your debugging suggestion and changed the region to us-west-2 as described at the beginning of the notebook ([https://github.com/aws/amazon-sagemaker-examples/blob/main/introduction_to_amazon_algorithms/jumpstart_object_detection/Amazon_JumpStart_Object_Detection.ipynb]), but still get the same error.

Ps. The CICD checks are failing in the notebook above.

Using model 'pytorch-od-nvidia-ssd' with wildcard version identifier '*'. You can pin to version '2.0.0' for more stable results. Note that models may have different input/output signatures after a major version upgrade.
Using model 'pytorch-od-nvidia-ssd' with wildcard version identifier '*'. You can pin to version '2.0.0' for more stable results. Note that models may have different input/output signatures after a major version upgrade.
Deploy Image URI: 763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-inference:1.10.0-gpu-py38
Deploy Source URI: s3://jumpstart-cache-prod-us-west-2/source-directory-tarballs/pytorch/inference/od/v1.2.3/sourcedir.tar.gz
Base Model URI: s3://jumpstart-cache-prod-us-west-2/pytorch-od/pytorch-od-nvidia-ssd/artifacts/inference-prepack/v1.0.0/
---------------------------------------------------------------------------
ClientError                               Traceback (most recent call last)
Cell In[44], line 39
     30 model = Model(
     31     image_uri=deploy_image_uri,
     32     source_dir=deploy_source_uri,
   (...)
     35     role=aws_role,
     36     name=endpoint_name)
     38 # Deploy the Model without predictor_cls argument
---> 39 base_model_predictor = model.deploy(
     40     initial_instance_count=1,
     41     instance_type=inference_instance_type,
     42     endpoint_name=endpoint_name
     43 )
     45 # Instantiate the Predictor class separately
     46 predictor = Predictor(endpoint_name=endpoint_name)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:1692, in Model.deploy(self, initial_instance_count, instance_type, serializer, deserializer, accelerator_type, endpoint_name, tags, kms_key, wait, data_capture_config, async_inference_config, serverless_inference_config, volume_size, model_data_download_timeout, container_startup_health_check_timeout, inference_recommendation_id, explainer_config, accept_eula, endpoint_logging, resources, endpoint_type, managed_instance_scaling, inference_component_name, routing_config, **kwargs)
   1689     return None
   1691 else:  # existing single model endpoint path
-> 1692     self._create_sagemaker_model(
   1693         instance_type=instance_type,
   1694         accelerator_type=accelerator_type,
   1695         tags=tags,
   1696         serverless_inference_config=serverless_inference_config,
   1697         **kwargs,
   1698     )
   1699     serverless_inference_config_dict = (
   1700         serverless_inference_config._to_request_dict() if is_serverless else None
   1701     )
   1702     production_variant = sagemaker.production_variant(
   1703         self.name,
   1704         instance_type,
   (...)
   1711         routing_config=routing_config,
   1712     )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:927, in Model._create_sagemaker_model(self, instance_type, accelerator_type, tags, serverless_inference_config, accept_eula, model_reference_arn)
    925         self.name = model_package.name
    926 else:
--> 927     container_def = self.prepare_container_def(
    928         instance_type,
    929         accelerator_type=accelerator_type,
    930         serverless_inference_config=serverless_inference_config,
    931         accept_eula=accept_eula,
    932         model_reference_arn=model_reference_arn,
    933     )
    935     if not isinstance(self.sagemaker_session, PipelineSession):
    936         # _base_name, model_name are not needed under PipelineSession.
    937         # the model_data may be Pipeline variable
    938         # which may break the _base_name generation
    939         self._ensure_base_name_if_needed(
    940             image_uri=container_def["Image"],
    941             script_uri=self.source_dir,
    942             model_uri=self._get_model_uri(),
    943         )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:715, in Model.prepare_container_def(self, instance_type, accelerator_type, serverless_inference_config, accept_eula, model_reference_arn)
    713 deploy_env = copy.deepcopy(self.env)
    714 if self.source_dir or self.dependencies or self.entry_point or self.git_config:
--> 715     self._upload_code(deploy_key_prefix, repack=self.is_repack())
    716     deploy_env.update(self._script_mode_env_vars())
    718 return sagemaker.container_def(
    719     self.image_uri,
    720     self.repacked_model_data or self.model_data,
   (...)
    731     ),
    732 )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/model.py:822, in Model._upload_code(self, key_prefix, repack)
    806     self.uploaded_code = fw_utils.UploadedCode(
    807         s3_prefix=repacked_model_data,
    808         script_name=os.path.basename(self.entry_point),
    809     )
    811 logger.info(
    812     "Repacking model artifact (%s), script artifact "
    813     "(%s), and dependencies (%s) "
   (...)
    819     repacked_model_data,
    820 )
--> 822 utils.repack_model(
    823     inference_script=self.entry_point,
    824     source_directory=self.source_dir,
    825     dependencies=self.dependencies,
    826     model_uri=self.model_data,
    827     repacked_model_uri=repacked_model_data,
    828     sagemaker_session=self.sagemaker_session,
    829     kms_key=self.model_kms_key,
    830 )
    832 self.repacked_model_data = repacked_model_data

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:551, in repack_model(inference_script, source_directory, dependencies, model_uri, repacked_model_uri, sagemaker_session, kms_key)
    544 local_download_dir = (
    545     None
    546     if sagemaker_session.settings is None
    547     or sagemaker_session.settings.local_download_dir is None
    548     else sagemaker_session.settings.local_download_dir
    549 )
    550 with _tmpdir(directory=local_download_dir) as tmp:
--> 551     model_dir = _extract_model(model_uri, sagemaker_session, tmp)
    553     _create_or_update_code_dir(
    554         model_dir,
    555         inference_script,
   (...)
    559         tmp,
    560     )
    562     tmp_model_path = os.path.join(tmp, "temp-model.tar.gz")

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:637, in _extract_model(model_uri, sagemaker_session, tmp)
    635 if model_uri.lower().startswith("s3://"):
    636     local_model_path = os.path.join(tmp, "tar_file")
--> 637     download_file_from_url(model_uri, local_model_path, sagemaker_session)
    638 else:
    639     local_model_path = model_uri.replace("file://", "")

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:650, in download_file_from_url(url, dst, sagemaker_session)
    647 url = parse.urlparse(url)
    648 bucket, key = url.netloc, url.path.lstrip("/")
--> 650 download_file(bucket, key, dst, sagemaker_session)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/sagemaker/utils.py:668, in download_file(bucket_name, path, target, sagemaker_session)
    666 s3 = boto_session.resource("s3", region_name=sagemaker_session.boto_region_name)
    667 bucket = s3.Bucket(bucket_name)
--> 668 bucket.download_file(path, target)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/boto3/s3/inject.py:279, in bucket_download_file(self, Key, Filename, ExtraArgs, Callback, Config)
    245 def bucket_download_file(
    246     self, Key, Filename, ExtraArgs=None, Callback=None, Config=None
    247 ):
    248     """Download an S3 object to a file.
    249 
    250     Usage::
   (...)
    277         transfer.
    278     """
--> 279     return self.meta.client.download_file(
    280         Bucket=self.name,
    281         Key=Key,
    282         Filename=Filename,
    283         ExtraArgs=ExtraArgs,
    284         Callback=Callback,
    285         Config=Config,
    286     )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/boto3/s3/inject.py:192, in download_file(self, Bucket, Key, Filename, ExtraArgs, Callback, Config)
    157 """Download an S3 object to a file.
    158 
    159 Usage::
   (...)
    189     transfer.
    190 """
    191 with S3Transfer(self, Config) as transfer:
--> 192     return transfer.download_file(
    193         bucket=Bucket,
    194         key=Key,
    195         filename=Filename,
    196         extra_args=ExtraArgs,
    197         callback=Callback,
    198     )

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/boto3/s3/transfer.py:406, in S3Transfer.download_file(self, bucket, key, filename, extra_args, callback)
    402 future = self._manager.download(
    403     bucket, key, filename, extra_args, subscribers
    404 )
    405 try:
--> 406     future.result()
    407 # This is for backwards compatibility where when retries are
    408 # exceeded we need to throw the same error from boto3 instead of
    409 # s3transfer's built in RetriesExceededError as current users are
    410 # catching the boto3 one instead of the s3transfer exception to do
    411 # their own retries.
    412 except S3TransferRetriesExceededError as e:

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/futures.py:103, in TransferFuture.result(self)
     98 def result(self):
     99     try:
    100         # Usually the result() method blocks until the transfer is done,
    101         # however if a KeyboardInterrupt is raised we want want to exit
    102         # out of this and propagate the exception.
--> 103         return self._coordinator.result()
    104     except KeyboardInterrupt as e:
    105         self.cancel()

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/futures.py:266, in TransferCoordinator.result(self)
    263 # Once done waiting, raise an exception if present or return the
    264 # final result.
    265 if self._exception:
--> 266     raise self._exception
    267 return self._result

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/tasks.py:269, in SubmissionTask._main(self, transfer_future, **kwargs)
    265     self._transfer_coordinator.set_status_to_running()
    267     # Call the submit method to start submitting tasks to execute the
    268     # transfer.
--> 269     self._submit(transfer_future=transfer_future, **kwargs)
    270 except BaseException as e:
    271     # If there was an exception raised during the submission of task
    272     # there is a chance that the final task that signals if a transfer
   (...)
    281 
    282     # Set the exception, that caused the process to fail.
    283     self._log_and_set_exception(e)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/s3transfer/download.py:354, in DownloadSubmissionTask._submit(self, client, config, osutil, request_executor, io_executor, transfer_future, bandwidth_limiter)
    325 """
    326 :param client: The client associated with the transfer manager
    327 
   (...)
    349     downloading streams
    350 """
    351 if transfer_future.meta.size is None:
    352     # If a size was not provided figure out the size for the
    353     # user.
--> 354     response = client.head_object(
    355         Bucket=transfer_future.meta.call_args.bucket,
    356         Key=transfer_future.meta.call_args.key,
    357         **transfer_future.meta.call_args.extra_args,
    358     )
    359     transfer_future.meta.provide_transfer_size(
    360         response['ContentLength']
    361     )
    363 download_output_manager = self._get_download_output_manager_cls(
    364     transfer_future, osutil
    365 )(osutil, self._transfer_coordinator, io_executor)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/botocore/client.py:565, in ClientCreator._create_api_method.<locals>._api_call(self, *args, **kwargs)
    561     raise TypeError(
    562         f"{py_operation_name}() only accepts keyword arguments."
    563     )
    564 # The "self" in this scope is referring to the BaseClient.
--> 565 return self._make_api_call(operation_name, kwargs)

File ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/botocore/client.py:1021, in BaseClient._make_api_call(self, operation_name, api_params)
   1017     error_code = error_info.get("QueryErrorCode") or error_info.get(
   1018         "Code"
   1019     )
   1020     error_class = self.exceptions.from_code(error_code)
-> 1021     raise error_class(parsed_response, operation_name)
   1022 else:
   1023     return parsed_response

ClientError: An error occurred (404) when calling the HeadObject operation: Not Found

[pietervosnl]

I have the same problem as @catica. I can download the tar file from the S3 bucket and can read the files in the base_model_uri folder (in the terminal as user ec2-user). So permissions are set correct. @Siddharth-Latthe-07 , do you have more tips?
Thanks!

[Siddharth-Latthe-07]

@pietervosnl yeah sure Here are some additional steps and suggestions to help you troubleshoot and resolve this issue

1. Permissions and IAM Roles:
   Ensure that the IAM role used by SageMaker (aws_role) has the necessary permissions to access the S3 buckets and objects. This typically includes s3:GetObject, s3:ListBucket, and s3:HeadObject permissions.
   Verify the policy attached to the IAM role and make sure it grants access to the correct S3 resources.

2. Network Configuration: If your SageMaker notebook or training job is running in a VPC, ensure that the VPC has the necessary internet access or a VPC endpoint configured to access S3.

3. Model and Script Versions: When using wildcard versions for models (infer_model_version = "*") and scripts, ensure that the correct and compatible versions are being used. Try specifying an exact version to see if it resolves the issue.

4. Temporary S3 Bucket: As a workaround, try copying the model data to a new S3 bucket in the same region and update the URI accordingly. This can help identify if the issue is specific to the current bucket or URI.

Sample snippet for checking the URIs and the region consistency:-

from sagemaker import image_uris, model_uris, script_uris
from sagemaker.model import Model
from sagemaker.predictor import Predictor
from sagemaker.utils import name_from_base

# Set model ID and version
infer_model_id, infer_model_version = "pytorch-od-nvidia-ssd", "2.0.0"

# Generate endpoint name
endpoint_name = name_from_base(f"jumpstart-example-infer-{infer_model_id}")

# Define instance type
inference_instance_type = "ml.p2.xlarge"

# Retrieve URIs
deploy_image_uri = image_uris.retrieve(
    region="us-west-2",
    framework=None,
    image_scope="inference",
    model_id=infer_model_id,
    model_version=infer_model_version,
    instance_type=inference_instance_type,
)

deploy_source_uri = script_uris.retrieve(
    model_id=infer_model_id, model_version=infer_model_version, script_scope="inference"
)

base_model_uri = model_uris.retrieve(
    model_id=infer_model_id, model_version=infer_model_version, model_scope="inference"
)

# Print URIs for verification
print(f"Deploy Image URI: {deploy_image_uri}")
print(f"Deploy Source URI: {deploy_source_uri}")
print(f"Base Model URI: {base_model_uri}")

# Create the SageMaker model instance
model = Model(
    image_uri=deploy_image_uri,
    source_dir=deploy_source_uri,
    model_data=base_model_uri,
    entry_point="inference.py",
    role=aws_role,
    name=endpoint_name,
)

# Deploy the model
base_model_predictor = model.deploy(
    initial_instance_count=1,
    instance_type=inference_instance_type,
    endpoint_name=endpoint_name,
)

# Instantiate the Predictor class
predictor = Predictor(endpoint_name=endpoint_name)

Hope this helps, Thanks