@aws/ml-container-creator 0.2.0

package/NOTICE

@@ -0,0 +1,2 @@
+awslabs/ml-container-creator
+Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved. 

package/README.md

@@ -0,0 +1,106 @@
+# ML Container Creator
+
+A CLI tool that creates SageMaker-compatible Docker containers for deploying ML models using the Bring Your Own Container (BYOC) paradigm.
+
+> **Note:** This is a pre-release (`0.x`). APIs may change between minor versions. Weekly releases are planned until v1.
+
+## Supported Configurations
+
+| Architecture | Model Servers | Use Case |
+|---|---|---|
+| HTTP (traditional ML) | Flask, FastAPI | sklearn, XGBoost, TensorFlow |
+| Transformers (LLMs) | vLLM, SGLang, TensorRT-LLM, DJL/LMI | HuggingFace models, JumpStart, S3 |
+| Triton | FIL, ONNX, Python, TensorRT-LLM, vLLM | Multi-framework serving |
+| Diffusors | vLLM | Image generation models |
+
+| Deployment Target | Description |
+|---|---|
+| Managed Inference | SageMaker real-time endpoints |
+| Async Inference | SageMaker async endpoints with S3 output |
+| Batch Transform | SageMaker batch processing |
+| HyperPod EKS | Kubernetes-based deployment |
+
+## Quick Start
+
+### Install from npm
+
+```bash
+npm install -g @aws/ml-container-creator
+```
+
+### Or use without installing (npx)
+
+```bash
+npx @aws/ml-container-creator --help
+```
+
+### Or install from source
+
+```bash
+git clone https://github.com/awslabs/ml-container-creator.git
+cd ml-container-creator
+npm install && npm link
+```
+
+### Bootstrap AWS infrastructure (one-time)
+
+```bash
+ml-container-creator bootstrap
+```
+
+Sets up an IAM execution role, ECR repository, optional S3 buckets, and optional CI Integration Harness for automated testing. Configuration is saved to `~/.ml-container-creator/config.json`.
+
+### Generate a project
+
+```bash
+# Interactive
+ml-container-creator
+
+# Non-interactive
+ml-container-creator my-model \
+  --deployment-config=transformers-vllm \
+  --model-name=openai/gpt-oss-20b \
+  --instance-type=ml.g6.12xlarge \
+  --region=us-east-1 \
+  --skip-prompts
+```
+
+### Build, push, deploy
+
+```bash
+./do/build        # Build Docker image
+./do/push         # Push to Amazon ECR
+./do/deploy       # Deploy to SageMaker
+./do/test         # Test the endpoint
+```
+
+## Documentation
+
+Full documentation is available at [awslabs.github.io/ml-container-creator](https://awslabs.github.io/ml-container-creator/).
+
+- [Getting Started](https://awslabs.github.io/ml-container-creator/getting-started/) — Installation and walkthroughs
+- [Configuration](https://awslabs.github.io/ml-container-creator/configuration/) — CLI flags, env vars, config files, MCP servers
+- [Deployment Guide](https://awslabs.github.io/ml-container-creator/deployments/) — All deployment targets and lifecycle scripts
+- [CI Integration](https://awslabs.github.io/ml-container-creator/ci-integration/) — Automated lifecycle testing for all deployment configurations
+- [Examples](https://awslabs.github.io/ml-container-creator/EXAMPLES/) — Framework-specific walkthroughs
+- [Troubleshooting](https://awslabs.github.io/ml-container-creator/TROUBLESHOOTING/) — Common issues and solutions
+
+## Prerequisites
+
+| Tool | Version | Purpose |
+|---|---|---|
+| [Node.js](https://nodejs.org/) | 24+ | Runs the CLI |
+| [Docker](https://docs.docker.com/get-docker/) | 20+ | Container builds |
+| [AWS CLI](https://aws.amazon.com/cli/) | 2+ | AWS resource management |
+
+## Contributing
+
+See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
+
+## Security
+
+See [CONTRIBUTING.md](CONTRIBUTING.md#security-issue-notifications) for reporting security issues.
+
+## License
+
+Apache-2.0. See [LICENSE](LICENSE).

package/bin/cli.js

@@ -0,0 +1,365 @@
+#!/usr/bin/env node
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+import { createRequire } from 'module'
+import path from 'path'
+import { program, Option, Help } from 'commander'
+import { run } from '../src/app.js'
+
+const require = createRequire(import.meta.url)
+const { version } = require('../package.json')
+
+/**
+ * Collect repeatable options into an array.
+ * Used for --model-env and --server-env which can be specified multiple times.
+ */
+function collect(value, previous) {
+    return previous.concat([value])
+}
+
+program
+    .name('ml-container-creator')
+    .version(version)
+    .enablePositionalOptions()
+    .passThroughOptions()
+    .helpCommand('help [command]', 'Display help for command')
+    .argument('[project-name]', 'Name for the generated project')
+
+    // --- General ---
+    .addOption(new Option('--skip-prompts', 'Skip interactive prompts and use configuration from other sources'))
+    .addOption(new Option('--config <path>', 'Path to configuration file'))
+    .addOption(new Option('--project-name <name>', 'Project name'))
+    .addOption(new Option('--project-dir <dir>', 'Output directory path'))
+
+    // --- Model & Framework ---
+    .addOption(new Option('--deployment-config <config>', 'Deployment configuration (e.g. http-flask, transformers-vllm, triton-fil)'))
+    .addOption(new Option('--framework <framework>', 'ML framework — DEPRECATED: use --deployment-config').choices(['sklearn', 'xgboost', 'tensorflow', 'transformers']).hideHelp())
+    .addOption(new Option('--model-format <format>', 'Model serialization format (pkl, joblib, json, model, ubj, keras, h5, SavedModel)'))
+    .addOption(new Option('--model-name <name>', 'Model identifier (HuggingFace ID, s3://, jumpstart://, registry://)'))
+    .addOption(new Option('--model-server <server>', 'Model server — DEPRECATED: use --deployment-config').choices(['flask', 'fastapi', 'vllm', 'sglang']).hideHelp())
+    .addOption(new Option('--base-image <image>', 'Base container image for Dockerfile'))
+
+    // --- Build & Infrastructure ---
+    .addOption(new Option('--deployment-target <target>', 'Deployment target (managed-inference, async-inference, batch-transform, hyperpod-eks)'))
+    .addOption(new Option('--instance-type <type>', 'SageMaker instance type (e.g. ml.g5.xlarge, ml.m5.large)'))
+    .addOption(new Option('--region <region>', 'AWS region'))
+    .addOption(new Option('--role-arn <arn>', 'IAM role ARN for SageMaker execution'))
+    .addOption(new Option('--build-target <target>', 'Build target (codebuild)'))
+    .addOption(new Option('--codebuild-compute-type <type>', 'CodeBuild compute type (SMALL, MEDIUM, LARGE)'))
+
+    // --- Endpoint (Real-Time Inference) ---
+    .addOption(new Option('--endpoint-initial-instance-count <n>', 'Number of instances for the endpoint (default: 1)'))
+    .addOption(new Option('--endpoint-data-capture-percent <pct>', 'Data capture percentage for monitoring, 0-100 (default: 0)'))
+    .addOption(new Option('--endpoint-variant-name <name>', 'Production variant name (default: AllTraffic)'))
+    .addOption(new Option('--endpoint-volume-size <gb>', 'ML storage volume size in GB'))
+
+    // --- Inference Component ---
+    .addOption(new Option('--ic-cpu-count <n>', 'vCPUs allocated to the inference component'))
+    .addOption(new Option('--ic-memory-size <mb>', 'Memory in MB for the inference component'))
+    .addOption(new Option('--ic-gpu-count <n>', 'GPUs allocated to the inference component'))
+    .addOption(new Option('--ic-copy-count <n>', 'Number of inference component copies (default: 1)'))
+    .addOption(new Option('--ic-model-weight <weight>', 'Traffic routing weight, 0-1 (default: 1.0)'))
+
+    // --- Async Inference ---
+    .addOption(new Option('--async-s3-output-path <path>', 'S3 output path for async results'))
+    .addOption(new Option('--async-sns-success-topic <arn>', 'SNS topic ARN for success notifications'))
+    .addOption(new Option('--async-sns-error-topic <arn>', 'SNS topic ARN for error notifications'))
+    .addOption(new Option('--async-max-concurrent <n>', 'Max concurrent invocations per instance (default: 1)'))
+
+    // --- Batch Transform ---
+    .addOption(new Option('--batch-input-path <path>', 'S3 input path for batch data'))
+    .addOption(new Option('--batch-output-path <path>', 'S3 output path for batch results'))
+    .addOption(new Option('--batch-instance-count <n>', 'Number of instances (default: 1)'))
+    .addOption(new Option('--batch-split-type <type>', 'Input split type: Line, RecordIO, None (default: Line)'))
+    .addOption(new Option('--batch-strategy <strategy>', 'Batch strategy: MultiRecord, SingleRecord (default: MultiRecord)'))
+    .addOption(new Option('--batch-join-source <source>', 'Join source: Input, None (default: None)'))
+    .addOption(new Option('--batch-max-concurrent <n>', 'Max concurrent transforms per instance (default: 1)'))
+    .addOption(new Option('--batch-max-payload <mb>', 'Max payload size in MB, 0-100 (default: 6)'))
+
+    // --- HyperPod (EKS) ---
+    .addOption(new Option('--hyperpod-cluster <name>', 'HyperPod EKS cluster name'))
+    .addOption(new Option('--hyperpod-namespace <ns>', 'Kubernetes namespace (default: default)'))
+    .addOption(new Option('--hyperpod-replicas <count>', 'Number of replicas (default: 1)'))
+    .addOption(new Option('--fsx-volume-handle <handle>', 'FSx for Lustre volume handle'))
+
+    // --- Environment Variables ---
+    .addOption(new Option('--model-env <KEY=VALUE>', 'Model env var, repeatable (e.g. VLLM_TENSOR_PARALLEL_SIZE=4)').argParser(collect).default([]))
+    .addOption(new Option('--server-env <KEY=VALUE>', 'Server env var, repeatable (e.g. SGLANG_MEM_FRACTION=0.9)').argParser(collect).default([]))
+
+    // --- Authentication ---
+    .addOption(new Option('--hf-token <token>', 'HuggingFace token (or "$HF_TOKEN" for env var reference)'))
+
+    // --- Optional Features ---
+    .addOption(new Option('--include-sample', 'Include sample model code'))
+    .addOption(new Option('--include-testing', 'Include test suite'))
+    .addOption(new Option('--test-types <types>', 'Comma-separated test types'))
+
+    // --- MCP & Discovery ---
+    .addOption(new Option('--smart', 'Enable Bedrock-powered smart mode on MCP servers'))
+    .addOption(new Option('--discover', 'Enable live registry lookups via MCP discovery'))
+
+    // --- Validation ---
+    .addOption(new Option('--validate-env-vars', 'Enable environment variable validation (default: true)'))
+    .addOption(new Option('--validate-with-docker', 'Enable Docker introspection validation (opt-in)'))
+    .addOption(new Option('--offline', 'Disable HuggingFace API lookups'))
+
+    .action(run)
+
+// Custom help formatting — group options into logical sections (root command only)
+program.configureHelp({
+    formatHelp(cmd, helper) {
+        // Only apply custom grouping to the root command
+        if (cmd !== program) {
+            // Fall back to default Commander formatting for subcommands
+            return Help.prototype.formatHelp.call(this, cmd, helper)
+        }
+
+        const termWidth = helper.padWidth(cmd, helper)
+
+        function callFormatItem(term, description) {
+            return helper.formatItem(term, termWidth, description, helper)
+        }
+
+        function formatSection(title, options) {
+            if (options.length === 0) return []
+            const lines = options.map(opt => {
+                return callFormatItem(
+                    helper.styleOptionTerm(helper.optionTerm(opt)),
+                    helper.styleOptionDescription(helper.optionDescription(opt))
+                )
+            })
+            return [helper.styleTitle(`${title}:`), ...lines, '']
+        }
+
+        // Collect all visible options
+        const allOptions = helper.visibleOptions(cmd)
+
+        // Partition options into groups by flag prefix/purpose
+        const groups = {
+            general: [],
+            model: [],
+            infra: [],
+            endpoint: [],
+            ic: [],
+            async: [],
+            batch: [],
+            hyperpod: [],
+            env: [],
+            auth: [],
+            features: [],
+            mcp: [],
+            validation: []
+        }
+
+        for (const opt of allOptions) {
+            const long = opt.long || ''
+            if (['--skip-prompts', '--config', '--project-name', '--project-dir', '--version', '--help'].includes(long)) {
+                groups.general.push(opt)
+            } else if (['--deployment-config', '--framework', '--model-format', '--model-name', '--model-server', '--base-image'].includes(long)) {
+                groups.model.push(opt)
+            } else if (['--deployment-target', '--instance-type', '--region', '--role-arn', '--build-target', '--codebuild-compute-type'].includes(long)) {
+                groups.infra.push(opt)
+            } else if (long.startsWith('--endpoint-')) {
+                groups.endpoint.push(opt)
+            } else if (long.startsWith('--ic-')) {
+                groups.ic.push(opt)
+            } else if (long.startsWith('--async-')) {
+                groups.async.push(opt)
+            } else if (long.startsWith('--batch-')) {
+                groups.batch.push(opt)
+            } else if (long.startsWith('--hyperpod-') || long === '--fsx-volume-handle') {
+                groups.hyperpod.push(opt)
+            } else if (['--model-env', '--server-env'].includes(long)) {
+                groups.env.push(opt)
+            } else if (['--hf-token'].includes(long)) {
+                groups.auth.push(opt)
+            } else if (['--include-sample', '--include-testing', '--test-types'].includes(long)) {
+                groups.features.push(opt)
+            } else if (['--smart', '--discover'].includes(long)) {
+                groups.mcp.push(opt)
+            } else if (['--validate-env-vars', '--validate-with-docker', '--offline'].includes(long)) {
+                groups.validation.push(opt)
+            } else {
+                groups.general.push(opt)
+            }
+        }
+
+        // Build output
+        let output = [
+            `${helper.styleTitle('Usage:')} ${helper.styleUsage(helper.commandUsage(cmd))}`,
+            ''
+        ]
+
+        // Arguments
+        const args = helper.visibleArguments(cmd)
+        if (args.length > 0) {
+            const argList = args.map(arg => {
+                return callFormatItem(
+                    helper.styleArgumentTerm(helper.argumentTerm(arg)),
+                    helper.styleArgumentDescription(helper.argumentDescription(arg))
+                )
+            })
+            output = output.concat([helper.styleTitle('Arguments:'), ...argList, ''])
+        }
+
+        // Option sections
+        output = output.concat(formatSection('General', groups.general))
+        output = output.concat(formatSection('Model & Framework', groups.model))
+        output = output.concat(formatSection('Build & Infrastructure', groups.infra))
+        output = output.concat(formatSection('Endpoint (Real-Time Inference)', groups.endpoint))
+        output = output.concat(formatSection('Inference Component', groups.ic))
+        output = output.concat(formatSection('Async Inference', groups.async))
+        output = output.concat(formatSection('Batch Transform', groups.batch))
+        output = output.concat(formatSection('HyperPod (EKS)', groups.hyperpod))
+        output = output.concat(formatSection('Environment Variables', groups.env))
+        output = output.concat(formatSection('Authentication', groups.auth))
+        output = output.concat(formatSection('Optional Features', groups.features))
+        output = output.concat(formatSection('MCP & Discovery', groups.mcp))
+        output = output.concat(formatSection('Validation', groups.validation))
+
+        // Commands
+        const cmds = helper.visibleCommands(cmd)
+        if (cmds.length > 0) {
+            const cmdList = cmds.map(sub => {
+                return callFormatItem(
+                    helper.styleSubcommandTerm(helper.subcommandTerm(sub)),
+                    helper.styleSubcommandDescription(helper.subcommandDescription(sub))
+                )
+            })
+            output = output.concat([helper.styleTitle('Commands:'), ...cmdList, ''])
+        }
+
+        return output.join('\n')
+    }
+})
+
+// Sub-commands — wired to actual handlers
+
+program
+    .command('bootstrap')
+    .description('Set up AWS infrastructure (IAM role, ECR repo, S3 buckets)')
+    .argument('[action]', 'Bootstrap action (status, use, list, remove, scan, prune, update)')
+    .argument('[args...]', 'Additional arguments')
+    .option('--profile <profile>', 'AWS profile name')
+    .option('--region <region>', 'AWS region')
+    .option('--role-arn <arn>', 'Existing IAM role ARN to use')
+    .option('--non-interactive', 'Run without prompts (requires --profile and --region)')
+    .option('--force', 'Force removal without confirmation')
+    .option('--verify', 'Verify resources exist (for status)')
+    .option('--delete-stack', 'Delete CloudFormation stack on remove')
+    .action(async (action, args, options) => {
+        const { default: BootstrapCommandHandler } = await import('../src/lib/bootstrap-command-handler.js')
+        const handler = new BootstrapCommandHandler()
+        const allArgs = action ? [action, ...args] : []
+        await handler.handle(allArgs, options)
+    })
+
+program
+    .command('mcp')
+    .description('Manage MCP servers (add, list, get, remove, init)')
+    .argument('<action>', 'MCP action (add, list, get, remove, init)')
+    .argument('[args...]', 'Additional arguments')
+    .option('-e <env>', 'Environment variable in KEY=VALUE format (for add)')
+    .option('--tool-name <name>', 'Tool name for MCP server (for add)')
+    .option('--limit <n>', 'Result limit for MCP server (for add)')
+    .option('--bundled', 'Use a bundled server from servers/ directory')
+    .action(async (action, args, options) => {
+        const { default: McpCommandHandler } = await import('../src/lib/mcp-command-handler.js')
+        const { runPrompts } = await import('../src/prompt-adapter.js')
+        // McpCommandHandler expects a generator-like object with destinationPath() and prompt()
+        const generatorAdapter = {
+            destinationPath(...segments) {
+                if (segments.length === 0) return process.cwd()
+                return path.join(process.cwd(), ...segments)
+            },
+            async prompt(prompts) {
+                return runPrompts(prompts)
+            }
+        }
+        const handler = new McpCommandHandler(generatorAdapter)
+        await handler.handle([action, ...args], options)
+    })
+
+program
+    .command('registry')
+    .description('Registry operations (list, get, remove, replay, export, import, search) — experimental, may be reconciled with do/register')
+    .argument('<action>', 'Registry action (log, list, get, remove, replay, export, import, search)')
+    .argument('[args...]', 'Additional arguments')
+    .option('--backend <backend>', 'Filter by backend')
+    .option('--architecture <arch>', 'Filter by architecture')
+    .option('--model <model>', 'Filter by model name')
+    .option('--instance-type <type>', 'Filter by instance type')
+    .option('--status <status>', 'Filter by status')
+    .option('--merge', 'Merge on import')
+    .option('--replace', 'Replace on import')
+    // Options used by `registry log` (called from do/register)
+    .option('--deployment-config <config>', 'Deployment configuration')
+    .option('--region <region>', 'AWS region')
+    .option('--deployment-target <target>', 'Deployment target')
+    .option('--build-target <target>', 'Build target')
+    .option('--model-name <name>', 'Model name')
+    .option('--model-format <format>', 'Model format')
+    .option('--base-image <image>', 'Base container image')
+    .option('--notes <text>', 'Deployment notes')
+    .option('--project', 'Use project-level registry')
+    .option('--parameters <json>', 'Parameters JSON string')
+    .option('--generator-version <version>', 'Generator version')
+    .action(async (action, args, options) => {
+        const { default: RegistryCommandHandler } = await import('../src/lib/registry-command-handler.js')
+        const handler = new RegistryCommandHandler()
+        await handler.handle([action, ...args], options)
+    })
+
+program
+    .command('configure')
+    .description('Interactive configuration setup (experimental)')
+    .action(async () => {
+        const { runPrompts } = await import('../src/prompt-adapter.js')
+        const { default: ConfigurationExporter } = await import('../src/lib/configuration-exporter.js')
+
+        console.log('\n🔧 ML Container Creator Configuration (experimental)')
+        console.log('\nThis will help you set up configuration files for your project.\n')
+
+        const answers = await runPrompts([
+            {
+                type: 'list',
+                name: 'configType',
+                message: 'What type of configuration would you like to create?',
+                choices: [
+                    { name: 'Show CLI option examples', value: 'cli' },
+                    { name: 'Show environment variable examples', value: 'env' }
+                ]
+            }
+        ])
+
+        if (answers.configType === 'cli') {
+            console.log(`
+💻 CLI Examples:
+
+  # Basic sklearn project
+  ml-container-creator --deployment-config=http-flask --model-format=pkl --skip-prompts
+
+  # Transformers with vLLM
+  ml-container-creator --deployment-config=transformers-vllm \\
+    --model-name=meta-llama/Llama-2-7b-chat-hf \\
+    --instance-type=ml.g5.xlarge --skip-prompts
+
+  # Using a config file
+  ml-container-creator --config=my-config.json --skip-prompts
+`)
+        } else if (answers.configType === 'env') {
+            console.log(`
+🌍 Environment Variables:
+
+  export ML_INSTANCE_TYPE="ml.m5.large"
+  export AWS_REGION="us-east-1"
+  export AWS_ROLE="arn:aws:iam::123456789012:role/SageMakerRole"
+  export HF_TOKEN="hf_..."
+
+  Then run: ml-container-creator --deployment-config=http-flask --skip-prompts
+`)
+        }
+    })
+
+program.parse()

package/config/defaults.json

@@ -0,0 +1,32 @@
+{
+  "generator": {
+    "name": "ml-container-creator",
+    "version": "1.0.0"
+  },
+  "defaults": {
+    "framework": "sklearn",
+    "modelFormat": "pkl",
+    "modelServer": "flask",
+    "deployTarget": "sagemaker",
+    "instanceType": "cpu-optimized",
+    "awsRegion": "us-east-1",
+    "includeTesting": true,
+    "testTypes": ["local-model-cli", "local-model-server", "hosted-model-endpoint"],
+    "includeSampleModel": false,
+    "skipPrompts": false
+  },
+  "validation": {
+    "supportedFrameworks": ["sklearn", "xgboost", "tensorflow", "transformers"],
+    "supportedModelServers": ["flask", "fastapi", "vllm", "sglang"],
+    "supportedInstanceTypes": ["cpu-optimized", "gpu-enabled"],
+    "supportedRegions": ["us-east-1", "us-west-2", "eu-west-1"]
+  },
+  "templates": {
+    "ignorePatterns": [
+      "**/node_modules/**",
+      "**/.git/**",
+      "**/dist/**",
+      "**/*.log"
+    ]
+  }
+}

package/config/presets/transformers-djl.json

@@ -0,0 +1,26 @@
+{
+  "name": "Transformers DJL Preset",
+  "description": "DJL Serving configuration with PyTorch backend",
+  "config": {
+    "framework": "transformers",
+    "modelServer": "djl",
+    "frameworkVersion": "0.32.0",
+    "instanceType": "gpu-enabled",
+    "awsRegion": "us-east-1",
+    "includeTesting": true,
+    "testTypes": ["hosted-model-endpoint"],
+    "includeSampleModel": false,
+    "skipPrompts": true
+  },
+  "environment": {
+    "ENGINE": "Python",
+    "OPTION_DEVICE_MAP": "auto",
+    "OPTION_TENSOR_PARALLEL_DEGREE": "1"
+  },
+  "dockerOptions": {
+    "baseImage": "deepjavalibrary/djl-serving:0.32.0-pytorch-cu126",
+    "gpuSupport": true,
+    "sharedMemorySize": "2g"
+  },
+  "notes": "DJL Serving provides flexible model serving with Java-based infrastructure and Python engine support"
+}

package/config/presets/transformers-gpu.json

@@ -0,0 +1,24 @@
+{
+  "name": "Transformers GPU Preset",
+  "description": "Optimized configuration for transformer models with GPU acceleration",
+  "config": {
+    "framework": "transformers",
+    "modelServer": "vllm",
+    "instanceType": "gpu-enabled",
+    "awsRegion": "us-east-1",
+    "includeTesting": true,
+    "testTypes": ["hosted-model-endpoint"],
+    "includeSampleModel": false,
+    "skipPrompts": true
+  },
+  "environment": {
+    "CUDA_VISIBLE_DEVICES": "0",
+    "TRANSFORMERS_CACHE": "/opt/ml/model/cache",
+    "HF_HOME": "/opt/ml/model/hf_cache"
+  },
+  "dockerOptions": {
+    "baseImage": "nvidia/cuda:12.1-runtime-ubuntu20.04",
+    "gpuSupport": true,
+    "sharedMemorySize": "2g"
+  }
+}

package/config/presets/transformers-lmi.json

@@ -0,0 +1,27 @@
+{
+  "name": "Transformers LMI Preset",
+  "description": "AWS Large Model Inference (LMI) configuration with automatic backend selection",
+  "config": {
+    "framework": "transformers",
+    "modelServer": "lmi",
+    "frameworkVersion": "14.0.0",
+    "instanceType": "gpu-enabled",
+    "awsRegion": "us-east-1",
+    "includeTesting": true,
+    "testTypes": ["hosted-model-endpoint"],
+    "includeSampleModel": false,
+    "skipPrompts": true
+  },
+  "environment": {
+    "OPTION_ROLLING_BATCH": "auto",
+    "OPTION_MAX_ROLLING_BATCH_SIZE": "32",
+    "OPTION_DTYPE": "fp16",
+    "OPTION_GPU_MEMORY_UTILIZATION": "0.9"
+  },
+  "dockerOptions": {
+    "baseImage": "763104351884.dkr.ecr.us-east-1.amazonaws.com/djl-inference:0.32.0-lmi14.0.0-cu126",
+    "gpuSupport": true,
+    "sharedMemorySize": "2g"
+  },
+  "notes": "LMI automatically selects the best backend (vLLM, TensorRT-LLM, LMI-Dist) based on your model architecture"
+}