Closed thesyntaxinator closed 4 months ago
Please specify the underlying hardware (inferentia2/inferentia1/trainium etc.)
I trying to compile on AL2 non inferentia instance to inf1 or inf2
I am slightly confused by your reply, what do you mean by "non inferentia instance to inf1 or inf2"?
Basically, if you want to compile a neuron-supported model for inf1 or inf2, the compilation needs to happen on inf1 or inf2 or tranium.
I see, will try compiling on inf1 or inf2 and post an update
Compiling on inf2 worked, thanks
System Info
Who can help?
No response
Information
Tasks
examples
folder (such as GLUE/SQuAD, ...)Reproduction (minimal, reproducible, runnable)
Install python, neuron, optimum, optimum neuron versions above on amazon linux 2. Try to export a local fine-tuned mistral model to neuron using any --task (I tried text-classification)
command I used:
optimum-cli export neuron --model base_model --task text-classification --sequence_length 4096 --batch_size 1 pytorch_model_neuron/
Expected behavior
Model should export to neuron