Documentation for Azure OpenAI v2 Model Deployments (#2825)

* Glide up deployment via CLI documentation

* Script for glide up deployment via API

* Reformated deployment via API code using black formatter

---------

Co-authored-by: Pavan Kumar Reddy Y <[email protected]>

cli/foundation-models/azure_openai/oai-v2/openai_finetuned_models_aoai_deploy/openai_finetuned_models_aoai_deploy.md

@@ -0,0 +1,63 @@
+# How to Deploy OpenAI Fine-tuned Models using Azure Cognitive Services CLI
+
+This document provides a guide to deploy OpenAI models fine-tuned via Azure Machine Learning workspace. For deploying OpenAI models fine-tuned via Azure OpenAI services, please refer to the documentation [here](#).
+
+### 1. Use the Azure Cloud Shell on the Microsoft Azure Portal
+Visit [https://ms.portal.azure.com/#home](https://ms.portal.azure.com/#home)
+
+![Azure Cloud Shell](images/image.png)
+
+### 2. Define Your Azure OpenAI Account Resource
+Add your Azure OpenAI studio account's name, subscription ID, and resource group in cloud shell.
+```bash
+export AOAI_RESOURCE_NAME="<YOUR-AOAI-ACCOUNT-NAME>"
+export AOAI_ACCOUNT_SUBSCRIPTION_ID="<YOUR-AOAI-ACCOUNT-SUBSCRIPTION>"
+export AOAI_ACCOUNT_RESOURCE_GROUP="<YOUR-AOAI-ACCOUNT-RESOURCE-GROUP>"
+```
+
+### 3. Add AzureML Registered Fine-Tuned Model Information
+In Azure Cloud Shell, add the AzureML registered model name, registered model version, and the AzureML (AML) workspace path for your fine-tuned model. 
+Your registered models data can be found in the `Models` tab of your AzureML workspace.
+
+```bash
+export AML_REGISTERED_MODEL_NAME="<YOUR-AML-REGISTERED-MODEL-NAME>"
+export AML_REGISTERED_MODEL_VERSION="<YOUR-AML-REGISTERD-MODEL-VERSION>"
+export AML_WORKSPACE_PATH_MODEL_SOURCE="<YOUR-AML-WORKSPACE-PATH>"
+```
+
+Note: Azure ML workspace path format is `/subscriptions/<AML-WORKSPACE-SUBSCRIPTION-ID>/resourceGroups/<AML-WORKSPACE-RESOURCE-GROUP>/providers/Microsoft.MachineLearningServices/workspaces/<AML-WORKSPACE-NAME>`
+
+### 4. Set the Deployment Name and Run the Azure Cognitive Service CLI Command
+Set the deployment name and run the Azure Cognitive serivce CLI command to deploy your fientuned model on Azure OpenAI studio.
+```
+export AOAI_DEPLOYMENT_NAME="DEPLOYMENT-NAME"
+az login
+az cognitiveservices account deployment create \
+  --model-name $AML_REGISTERED_MODEL_NAME \
+  --model-version $AML_REGISTERED_MODEL_VERSION \
+  --model-source $AML_WORKSPACE_PATH_MODEL_SOURCE \
+  --subscription $AOAI_ACCOUNT_SUBSCRIPTION_ID \
+  --resource-group $AOAI_ACCOUNT_RESOURCE_GROUP \
+  --name $AOAI_RESOURCE_NAME \
+  --deployment-name $AOAI_DEPLOYMENT_NAME \
+  --model-format OpenAI \
+  --sku-capacity 1 \
+  --sku-name "Standard"
+```
+![Run Azure CLI command to deploy models](images/image-1.png)
+
+### 5. View Your Azure OpenAI Model Deployment status in Azure OpenAI Studio portal
+Visit [https://oai.azure.com/portal/](https://oai.azure.com/portal/) to view your model deployments.
+
+![Deployment in provisioning state](images/image-2.png)
+
+### 6. After your deployment succeeds, use Azure OpenAI chat playground to test your Deployment
+View your deployments on Azure OpenAI studio as shown below.
+
+![Succeeded deployment page](images/image-3.png)
+
+Use Chat Playground features to make calls to your fine-tuned model:
+
+![Chat playground](images/image-4.png)
+
+### 7. Refer to the documentation on Azure OpenAI Studio here - [https://learn.microsoft.com/en-us/azure/ai-services/openai/](https://learn.microsoft.com/en-us/azure/ai-services/openai/) to learn more on how to use your model deployments.

sdk/python/foundation-models/azure_openai/oai-v2/openai_chat_finetuned_models_deployment.py

@@ -0,0 +1,57 @@
+import requests
+import json
+import subprocess
+
+# 1. Add your Azure OpenAI account details
+subscription = "<AOAI-ACCOUNT-SUBSCRIPTION-ID>"
+resource_group = "<AOAI-ACCOUNT-RESOURCE-GROUP>"
+resource_name = "<AOAI-RESOURCE-NAME>"
+model_deployment_name = "<NEW-AOAI-DEPLOYMENT-NAME>"
+
+# 2. Add the AzureML registered model name, registered model version, and the AzureML (AML) workspace path for your fine-tuned model.
+# Your registered models data can be found in the `Models` tab of your AzureML workspace.
+registered_model_name = "<AML-REGISTERED-MODEL-NAME>"
+registered_model_version = "<AML-REGISTERED-MODEL-VERSION>"
+workspace_path = "<AML-WORKSPACE-PATH-MODEL-SOURCE>"
+
+# Run `az login` to login into your azure account in your system shell
+# 3. Get Azure account access token
+token = json.loads(
+    subprocess.run(
+        "az account get-access-token", capture_output=True, shell=True
+    ).stdout
+)["accessToken"]
+deploy_params = {"api-version": "2023-05-01"}
+deploy_headers = {
+    "Authorization": f"Bearer {token}",
+    "Content-Type": "application/json",
+}
+
+# 4. Set model deployment configuration. Here capacity refers to support for `1K Tokens Per Minute (TPM)` for your deployment.
+deploy_data = {
+    "sku": {"name": "Standard", "capacity": 1},
+    "properties": {
+        "model": {
+            "format": "OpenAI",
+            "name": f"{registered_model_name}",
+            "version": f"{registered_model_version}",
+            "source": f"{workspace_path}",
+        }
+    },
+}
+
+deploy_data = json.dumps(deploy_data)
+
+# 5. Send PUT request to Azure cognitive services to create model deployment
+request_url = f"https://management.azure.com/subscriptions/{subscription}/resourceGroups/{resource_group}/providers/Microsoft.CognitiveServices/accounts/{resource_name}/deployments/{model_deployment_name}"
+
+r = requests.put(
+    request_url, params=deploy_params, headers=deploy_headers, data=deploy_data
+)
+
+print(r)
+
+# 6. View Your model deployment status in Azure OpenAI Studio portal
+# Visit [https://oai.azure.com/portal/](https://oai.azure.com/portal/) to view your model deployments.
+
+# 7. After your deployment succeeds, use Azure OpenAI chat playground to test your Deployment