Add NVIDIA model samples (#2836)

* Add placeholder samples * Added powershell for cli * deployment notebook added * updated notebook * Evaluation * yaml file updated * yaml file updated * Fix * PTuning notebook * Rename * Format --------- Co-authored-by: vikasagrawal-ms <[email protected]> Co-authored-by: HrishikeshGeedMS <[email protected]> Co-authored-by: Aditi Singh <[email protected]>
Azure · Sep 19, 2024 · 73242bf · 73242bf
1 parent 762bca3
commit 73242bf
Show file tree

Hide file tree

Showing 14 changed files with 3,944 additions and 0 deletions.
diff --git a/cli/foundation-models/nvidia-nemo/nemo_pipeline_job.yml b/cli/foundation-models/nvidia-nemo/nemo_pipeline_job.yml
@@ -0,0 +1,60 @@
+$schema: https://azuremlschemas.azureedge.net/latest/pipelineJob.schema.json
+type: pipeline
+display_name: TextGen_Finetuning_and_Evaluation
+description: Finetune the base model for text generation task and evaluation of the finetuned model.
+
+
+inputs:
+  compute: do-not-delete-nvidia-testing
+  model_path: 
+    type: triton_model
+    path: azureml://registries/nvidia-ai/models/Nemotron-3-8B-Base-4k/versions/1 
+  train_dataset_path: 
+    type: uri_folder
+    path: azureml:train_tg_nv:1
+  valid_dataset_path: 
+    type: uri_folder
+    path: azureml:val_tg_nv:1
+  ptuned_model_name: ptuned_model
+  input_column_name: text
+  target_column_name: summary
+  max_steps: 50
+  learning_rate: 0.001
+  num_nodes: 1
+  concat_sampling_probs: 1.0
+  eval_dataset_input_column_name: text
+  dataset_path: 
+    type: uri_folder
+    path: azureml:test-tg-nv:1
+
+outputs:
+  ptuned_model:
+    type: triton_model
+  predicted_output:
+    type: uri_folder
+  evaluation_result:
+    type: uri_folder
+
+jobs:
+  pipeline_component:
+    type: pipeline
+    component: azureml://registries/nvidia-ai/components/nemo_peft_text_generation_evaluation/versions/0.0.7
+    inputs:
+      compute: ${{parent.inputs.compute}}
+      model_path: ${{parent.inputs.model_path}}
+      train_dataset_path: ${{parent.inputs.train_dataset_path}}
+      valid_dataset_path: ${{parent.inputs.valid_dataset_path}}
+      ptuned_model_name: ${{parent.inputs.ptuned_model_name}}
+      input_column_name: ${{parent.inputs.input_column_name}}
+      target_column_name: ${{parent.inputs.target_column_name}}
+      max_steps: ${{parent.inputs.max_steps}}
+      learning_rate: ${{parent.inputs.learning_rate}}
+      num_nodes: ${{parent.inputs.num_nodes}}
+      concat_sampling_probs: ${{parent.inputs.concat_sampling_probs}}
+      eval_dataset_input_column_name: ${{parent.inputs.eval_dataset_input_column_name}}
+      dataset_path: ${{parent.inputs.dataset_path}}
+    outputs:
+      ptuned_model: ${{parent.outputs.ptuned_model}}
+      predicted_output: ${{parent.outputs.predicted_output}}
+      evaluation_result: ${{parent.outputs.evaluation_result}}
+
diff --git a/cli/foundation-models/nvidia-nemo/nemo_ptuning_summarization_with_textgen.ps1 b/cli/foundation-models/nvidia-nemo/nemo_ptuning_summarization_with_textgen.ps1
@@ -0,0 +1,7 @@
+$ProgressPreference = 'SilentlyContinue'; Invoke-WebRequest -Uri https://aka.ms/installazurecliwindows -OutFile .\AzureCLI.msi; Start-Process msiexec.exe -Wait -ArgumentList '/I AzureCLI.msi /quiet'; Remove-Item .\AzureCLI.msi
+
+az extension add -n ml
+
+az login
+
+az ml job create -g <RESOURCE_GROUP> -w <WORKSPACE> --subscription <SUBSCRIPTION> -f nemo_pipeline_job.yml