WOLFRAM

Wolfram Language & System Documentation Center

RemoteMachineClass

is an option for RemoteBatchSubmit and RemoteBatchMapSubmit that specifies a machine class to use with the "WolframBatch" provider.

Details
Details and Options Details and Options
Examples  
Basic Examples  
Scope  
Possible Issues  
See Also
Related Guides
History
Cite this Page

RemoteMachineClass

RemoteMachineClass

is an option for RemoteBatchSubmit and RemoteBatchMapSubmit that specifies a machine class to use with the "WolframBatch" provider.

Details

  • RemoteMachineClass specifies what hardware to run the remote job on. Supported classes include:
  • "Basic1x8"basic machine class with 1 CPU core and 8 GB of memory
    "Basic2x8"2 CPU cores and 8 GB of memory
    "Basic4x16"4 CPU cores and 16 GB of memory
    "Memory8x64"memory-optimized class with 8 CPU cores and 64 GB of memory
    "Memory16x128"16 CPU cores and 128 GB of memory
    "Memory192x1536"192 CPU cores and 1536 GB of memory
    "Compute64x128"compute-optimized class with 64 CPU cores and 128 GB of memory
    "Compute192x384"192 CPU cores and 384 GB of memory
    "GPU1xL40S"1 NVIDIA L40S GPU with 44 GiB of GPU memory
    "GPU4xL4"4 NVIDIA L4 GPUs with 89 GiB of total GPU memory
  • The default value for RemoteMachineClass is "Basic1x8".
  • The list of supported machine classes is subject to change. For an up-to-date listing with corresponding pricing information, see https://www.wolfram.com/compute-services.
  • When using the "WolframBatch" provider, RemoteMachineClass class is equivalent to setting RemoteProviderSettings <|"RemoteMachineClass" class|>.

Examples

open all close all

Basic Examples  (1)

Run a batch job on a computer with a large memory capacity:

Wolfram Language code: RemoteBatchSubmit["WolframBatch", FactorInteger[2 ^ 1023 - 7], RemoteMachineClass -> "Memory8x64"]

Scope  (3)

Train a neural network using a single GPU:

Wolfram Language code: job = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST", TargetDevice -> "GPU"], RemoteMachineClass -> "GPU1xL40S"]

Training progress is reported in the console log:

Wolfram Language code: StringCases[job["JobLogString"], "Starting training" ~~ __]//First

Use the resulting net to classify a difficult digit:

Wolfram Language code: job["EvaluationResult"]
Wolfram Language code: %[[image], "TopProbabilities"]

Train a neural network using multiple GPUs:

Wolfram Language code: job = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST", TargetDevice -> {"GPU", All}, BatchSize -> 512], RemoteMachineClass -> "GPU4xL4"]

Perform batch inference using a GPU:

Wolfram Language code: job = RemoteBatchSubmit[Counts[NetModel["Wolfram ImageIdentify Net V1"][Flatten@Table[RandomImage[max], {max, .1, 10, .1}, 10], TargetDevice -> "GPU"]], RemoteMachineClass -> "GPU1xL40S"]
Wolfram Language code: job["EvaluationResult"]

Possible Issues  (3)

Neural network operations use the CPU by default, even if a job’s machine class has GPU hardware available:

Wolfram Language code: job1 = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST"], RemoteMachineClass -> "GPU1xL40S"]
Wolfram Language code: StringCases[job1["JobLog"], "Device:" ~~ Except[" "]..]//First

Specify the TargetDevice option in order to use a GPU for training or inference:

Wolfram Language code: job2 = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST", TargetDevice -> "GPU"], RemoteMachineClass -> "GPU1xL40S"]
Wolfram Language code: StringCases[job2["JobLog"], "Device:" ~~ Except[" "]..]//First

NetTrain[…,TargetDevice"GPU"] uses a single GPU, regardless of the number of GPUs available to the job:

Wolfram Language code: job1 = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST", TargetDevice -> "GPU", BatchSize -> 128], RemoteMachineClass -> "GPU4xL4"]
Wolfram Language code: StringCases[job1["JobLog"], "Device:" ~~ Except[" "]..]//First

Specify TargetDevice{"GPU",All} in order to use all available GPUs for training:

Wolfram Language code: job2 = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST", TargetDevice -> {"GPU", All}, BatchSize -> 128], RemoteMachineClass -> "GPU4xL4"]
Wolfram Language code: StringCases[job2["JobLog"], "Device:" ~~ Except[" "]..]//First

Multi-GPU inference is not supported:

Wolfram Language code: job3 = With[{net = job2["EvaluationResult"]}, RemoteBatchSubmit[net[Table[RandomImage[1], 50], TargetDevice -> {"GPU", All}], RemoteMachineClass -> "GPU4xL4"]]
Wolfram Language code: job3["EvaluationMessagesText"]

Multi-GPU training fails if a batch size is not specified:

Wolfram Language code: job1 = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST", TargetDevice -> {"GPU", All}], RemoteMachineClass -> "GPU4xL4"]
Wolfram Language code: {job1["EvaluationResult"], job1["EvaluationMessagesText"]}

Specify the BatchSize option when performing training on multiple GPUs:

Wolfram Language code: job2 = RemoteBatchSubmit[NetTrain[NetModel["LeNet"], "MNIST", TargetDevice -> {"GPU", All}, BatchSize -> 128], RemoteMachineClass -> "GPU4xL4"]
Wolfram Language code: job2["EvaluationResult"]
Wolfram Research (2025), RemoteMachineClass, Wolfram Language function, https://reference.wolfram.com/language/ref/RemoteMachineClass.html.

Text

Wolfram Research (2025), RemoteMachineClass, Wolfram Language function, https://reference.wolfram.com/language/ref/RemoteMachineClass.html.

CMS

Wolfram Language. 2025. "RemoteMachineClass." Wolfram Language & System Documentation Center. Wolfram Research. https://reference.wolfram.com/language/ref/RemoteMachineClass.html.

APA

Wolfram Language. (2025). RemoteMachineClass. Wolfram Language & System Documentation Center. Retrieved from https://reference.wolfram.com/language/ref/RemoteMachineClass.html

BibTeX

@misc{reference.wolfram_2026_remotemachineclass, author="Wolfram Research", title="{RemoteMachineClass}", year="2025", howpublished="\url{https://reference.wolfram.com/language/ref/RemoteMachineClass.html}", note=[Accessed: 12-June-2026]}

BibLaTeX

@online{reference.wolfram_2026_remotemachineclass, organization={Wolfram Research}, title={RemoteMachineClass}, year={2025}, url={https://reference.wolfram.com/language/ref/RemoteMachineClass.html}, note=[Accessed: 12-June-2026]}