Huggingface model deployer support

docs/book/stacks/aws.md

@@ -16,7 +16,7 @@ as can a list of components that are coming soon.
 | Experiment Tracker | mlflow                                            |
 | Orchestrator       | kubeflow, kubernetes, sagemaker, skypilot, tekton |
 | MLOps Platform     | zenml                                             |
-| Model Deployer     | seldon                                    |
+| Model Deployer     | seldon, huggingface                               |
 | Step Operator      | sagemaker                                         |
 
 ## Coming Soon!

docs/book/stacks/gcp.md

@@ -22,7 +22,7 @@ and cannot be created without one.
 | Experiment Tracker | mlflow                                         |
 | Orchestrator       | kubeflow, kubernetes, skypilot, tekton, vertex |
 | MLOps Platform     | zenml                                          |
-| Model Deployer     | seldon                                 |
+| Model Deployer     | seldon, huggingface                            |
 | Step Operator      | vertex                                         |
 
 ## Coming Soon!

docs/book/stacks/k3d.md

@@ -16,7 +16,7 @@ as can a list of components that are coming soon.
 | Experiment Tracker | mlflow                                  |
 | Orchestrator       | kubeflow, kubernetes, sagemaker, tekton |
 | MLOps Platform     | zenml                                   |
-| Model Deployer     | seldon                          |
+| Model Deployer     | seldon, huggingface                     |
 
 ## Coming Soon!
 

src/mlstacks/constants.py

@@ -38,7 +38,7 @@
         "vertex",
     ],
     "mlops_platform": ["zenml"],
-    "model_deployer": ["seldon"],
+    "model_deployer": ["seldon", "huggingface"],
     "step_operator": ["sagemaker", "vertex"],
 }
 ALLOWED_COMPONENT_TYPES: Dict[str, Dict[str, List[str]]] = {

src/mlstacks/enums.py

@@ -38,6 +38,7 @@ class ComponentFlavorEnum(str, Enum):
 
     AWS = "aws"
     GCP = "gcp"
+    HUGGINGFACE = "huggingface"
     KUBEFLOW = "kubeflow"
     KUBERNETES = "kubernetes"
     MINIO = "minio"

src/mlstacks/terraform/aws-modular/eks.tf

@@ -266,4 +266,4 @@ data "external" "get_cluster_auth" {
   depends_on = [
     aws_eks_cluster.cluster
   ]
-}
+}

src/mlstacks/terraform/aws-modular/istio.tf

@@ -9,4 +9,4 @@ module "istio" {
   ]
 
   chart_version = local.istio.version
-}
+}

src/mlstacks/terraform/aws-modular/locals.tf

@@ -64,7 +64,7 @@ locals {
     namespace            = "seldon-system"
     workloads_namespace  = "zenml-workloads-seldon"
     service_account_name = "seldon"
-  }
+  } 
 
   zenml = {
     version                         = ""

src/mlstacks/terraform/aws-modular/output_file.tf

@@ -76,6 +76,13 @@ resource "local_file" "stack_file" {
         flavor: seldon
         name: eks_seldon_model_deployer
         configuration: {"kubernetes_context": "${aws_eks_cluster.cluster[0].arn}", "kubernetes_namespace": "${local.seldon.workloads_namespace}", "base_url": "http://${module.istio[0].ingress-hostname}:${module.istio[0].ingress-port}"}}
+%{endif}
+%{if var.enable_model_deployer_huggingface}
+      model_deployer:
+        id: ${uuid()}
+        flavor: huggingface
+        name: eks_huggingface_model_deployer
+        configuration: {"kubernetes_context": "${aws_eks_cluster.cluster[0].arn}", "kubernetes_namespace": "${local.huggingface.workloads_namespace}", "base_url": "http://${module.istio[0].ingress-hostname}:${module.istio[0].ingress-port}"}}
 %{endif}
     ADD
   filename = "./aws_modular_stack_${replace(substr(timestamp(), 0, 16), ":", "_")}.yaml"

src/mlstacks/terraform/aws-modular/outputs.tf

@@ -84,24 +84,6 @@ output "experiment_tracker_configuration" {
   }) : ""
 }
 
-# if seldon is enabled, set the model deployer outputs to the seldon values
-# otherwise, set the model deployer outputs to empty strings
-output "model_deployer_id" {
-  value = var.enable_model_deployer_seldon ? uuid() : ""
-}
-output "model_deployer_flavor" {
-  value = var.enable_model_deployer_seldon ? "seldon" : ""
-}
-output "model_deployer_name" {
-  value = var.enable_model_deployer_seldon ? "eks_seldon_model_deployer_${random_string.unique.result}" : ""
-}
-output "model_deployer_configuration" {
-  value = var.enable_model_deployer_seldon ? jsonencode({
-    kubernetes_context   = "${aws_eks_cluster.cluster[0].arn}"
-    kubernetes_namespace = local.seldon.workloads_namespace
-    base_url             = "http://${module.istio[0].ingress-hostname}:${module.istio[0].ingress-port}"
-  }) : ""
-}
 
 # if sagemaker is enabled as step operator, set the step operator outputs to the sagemaker values
 # otherwise, set the step operator outputs to empty strings
@@ -162,6 +144,16 @@ output "seldon-base-url" {
   value = var.enable_model_deployer_seldon ? "http://${module.istio[0].ingress-hostname}:${module.istio[0].ingress-port}" : null
 }
 
+# output for huggingface model deployer
+output "huggingface-workload-namespace" {
+  value       = var.enable_model_deployer_huggingface ? local.huggingface.workloads_namespace : null
+  description = "The namespace created for hosting your Huggingface workloads"
+}
+
+output "huggingface-base-url" {
+  value = var.enable_model_deployer_huggingface ? "http://${module.istio[0].ingress-hostname}:${module.istio[0].ingress-port}" : null
+}
+
 # output the name of the stack YAML file created
 output "stack-yaml-path" {
   value = local_file.stack_file.filename

src/mlstacks/terraform/aws-modular/variables.tf

@@ -35,6 +35,10 @@ variable "enable_model_deployer_seldon" {
   description = "Enable Seldon deployment"
   default     = false
 }
+variable "enable_model_deployer_huggingface" {
+  description = "Enable Huggingface deployment"
+  default     = false
+}
 variable "enable_step_operator_sagemaker" {
   description = "Enable SageMaker as step operator"
   default     = false
@@ -57,6 +61,17 @@ variable "region" {
   default     = "eu-west-1"
 }
 
+# variables for huggingface model model deployer
+variable "huggingface_token"{
+  description = "The Hugging Face authentication token."
+  default = "huggingfaceauthenticationtoken"
+}
+
+variable "huggingface_namespace" {
+  description = "The namespace where the Inference Endpoint will be created."
+  default = "huggingfacenamespace"
+} cb cb
+
 # variables for the MLflow tracking server
 variable "mlflow-artifact-S3-access-key" {
   description = "Your AWS access key for using S3 as MLflow artifact store"

src/mlstacks/terraform/gcp-modular/gke.tf

@@ -180,4 +180,4 @@ resource "google_project_iam_binding" "vertex-ai-user" {
   members = [
     "serviceAccount:${google_service_account.gke-service-account[0].email}",
   ]
-}
+}

src/mlstacks/terraform/gcp-modular/istio.tf

@@ -9,4 +9,4 @@ module "istio" {
   ]
 
   chart_version = local.istio.version
-}
+}

src/mlstacks/terraform/gcp-modular/kubernetes.tf

@@ -52,4 +52,4 @@ resource "google_service_account_iam_member" "k8s-workload-access" {
   depends_on = [
     kubernetes_namespace.k8s-workloads,
   ]
-}
+}

src/mlstacks/terraform/gcp-modular/locals.tf

@@ -75,7 +75,7 @@ locals {
     workloads_namespace  = "zenml-workloads-seldon"
     service_account_name = "seldon"
   }
-
+  
   zenml = {
     version                         = ""
     database_ssl_ca                 = ""

src/mlstacks/terraform/gcp-modular/output_file.tf

@@ -91,6 +91,13 @@ resource "local_file" "stack_file" {
         flavor: seldon
         name: gke_seldon
         configuration: {"kubernetes_context": "gke_${local.prefix}-${local.gke.cluster_name}", "kubernetes_namespace": "${local.seldon.workloads_namespace}", "base_url": "http://${module.istio[0].ingress-ip-address}:${module.istio[0].ingress-port}"}
+%{endif}
+%{if var.enable_model_deployer_huggingface}
+      model_deployer:
+        id: ${uuid()}
+        flavor: huggingface
+        name: eks_huggingface_model_deployer
+        configuration: {"kubernetes_context": "${aws_eks_cluster.cluster[0].arn}", "kubernetes_namespace": "${local.huggingface.workloads_namespace}", "base_url": "http://${module.istio[0].ingress-hostname}:${module.istio[0].ingress-port}"}}
 %{endif}
     ADD
   filename = "./gcp_modular_stack_${replace(substr(timestamp(), 0, 16), ":", "_")}.yaml"

src/mlstacks/terraform/gcp-modular/outputs.tf

@@ -100,25 +100,6 @@ output "experiment_tracker_configuration" {
   }) : ""
 }
 
-# if seldon is enabled, set the model deployer outputs to the seldon values
-# otherwise, set the model deployer outputs to empty strings
-output "model_deployer_id" {
-  value = var.enable_model_deployer_seldon ? uuid() : ""
-}
-output "model_deployer_flavor" {
-  value = var.enable_model_deployer_seldon ? "seldon" : ""
-}
-output "model_deployer_name" {
-  value = var.enable_model_deployer_seldon ? "gke_seldon_model_deployer_${random_string.unique.result}" : ""
-}
-output "model_deployer_configuration" {
-  value = var.enable_model_deployer_seldon ? jsonencode({
-    kubernetes_context   = "gke_${var.project_id}_${var.region}_${local.prefix}-${local.gke.cluster_name}"
-    kubernetes_namespace = local.seldon.workloads_namespace
-    base_url             = "http://${module.istio[0].ingress-ip-address}:${module.istio[0].ingress-port}"
-  }) : ""
-}
-
 # project id
 output "project-id" {
   value = var.project_id

src/mlstacks/terraform/gcp-modular/variables.tf

@@ -31,6 +31,10 @@ variable "enable_model_deployer_seldon" {
   description = "Enable Seldon deployment"
   default     = false
 }
+variable "enable_model_deployer_huggingface" {
+  description = "Enable Huggingface deployment"
+  default     = false
+}
 variable "enable_step_operator_vertex" {
   description = "Enable VertexAI Step Operator"
   default     = false
@@ -57,6 +61,16 @@ variable "project_id" {
   default     = ""
 }
 
+# variables for huggingface model model deployer
+variable "huggingface_token"{
+  description = "The Hugging Face authentication token."
+  default = "huggingfaceauthenticationtoken"
+}
+
+variable "huggingface_namespace" {
+  description = "The namespace where the Inference Endpoint will be created."
+  default = "huggingfacenamespace"
+}
 
 # variables for the MLflow tracking server
 variable "mlflow_bucket" {

src/mlstacks/terraform/gcp-modular/vpc.tf

@@ -54,4 +54,4 @@ module "vpc" {
       next_hop_internet = "true"
     },
   ]
-}
+}

src/mlstacks/terraform/k3d-modular/helm.tf

@@ -14,4 +14,4 @@ provider "helm" {
       var.enable_orchestrator_tekton || var.enable_orchestrator_kubernetes ||
     var.enable_model_deployer_seldon || var.enable_experiment_tracker_mlflow || var.enable_artifact_store || var.enable_zenml) ? k3d_cluster.zenml-cluster[0].credentials.0.cluster_ca_certificate : ""
   }
-}
+}

src/mlstacks/terraform/k3d-modular/istio.tf

@@ -8,4 +8,4 @@ module "istio" {
   ]
 
   chart_version = local.istio.version
-}
+}

src/mlstacks/terraform/k3d-modular/k3d.tf

@@ -100,4 +100,4 @@ resource "k3d_cluster" "zenml-cluster" {
   depends_on = [
     k3d_registry.zenml-registry,
   ]
-}
+}

src/mlstacks/terraform/k3d-modular/kubeflow.tf

@@ -16,4 +16,4 @@ module "kubeflow-pipelines" {
   ingress_host     = (var.enable_model_deployer_seldon) ? "${local.kubeflow.ingress_host_prefix}.${module.istio[0].ingress-ip-address}.nip.io" : "${local.kubeflow.ingress_host_prefix}.${module.nginx-ingress[0].ingress-ip-address}.nip.io"
   tls_enabled      = false
   istio_enabled    = (var.enable_model_deployer_seldon) ? true : false
-}
+}

src/mlstacks/terraform/k3d-modular/kubernetes.tf

@@ -37,4 +37,4 @@ resource "kubernetes_namespace" "k8s-workloads" {
   depends_on = [
     k3d_cluster.zenml-cluster,
   ]
-}
+}

src/mlstacks/terraform/k3d-modular/locals.tf

@@ -61,7 +61,7 @@ locals {
     workloads_namespace  = "zenml-workloads-seldon"
     service_account_name = "seldon"
   }
-
+  
   common_tags = {
     "managedBy"   = "terraform"
     "environment" = "dev"

src/mlstacks/terraform/k3d-modular/minio.tf

@@ -47,4 +47,4 @@ resource "minio_s3_bucket" "zenml_bucket" {
     module.nginx-ingress,
     module.istio,
   ]
-}
+}

src/mlstacks/terraform/k3d-modular/mlflow.tf

@@ -48,4 +48,4 @@ resource "minio_s3_bucket" "mlflow_bucket" {
     module.nginx-ingress,
     module.istio,
   ]
-}
+}

src/mlstacks/terraform/k3d-modular/nginx_ingress.tf

@@ -10,4 +10,4 @@ module "nginx-ingress" {
   ]
 
   chart_version = local.nginx_ingress.version
-}
+}

src/mlstacks/terraform/k3d-modular/output_stack.tf

@@ -90,4 +90,4 @@ resource "local_file" "stack_file" {
 %{endif}
     ADD
   filename = "./k3d_stack_${replace(substr(timestamp(), 0, 16), ":", "_")}.yaml"
-}
+}

src/mlstacks/terraform/k3d-modular/output_test_harness_cfg.tf

@@ -350,4 +350,4 @@ environments:
 %{endif}
     ADD
   filename = "./k3d_test_framework_cfg.yaml"
-}
+}