@aws/ml-container-creator 0.2.0

package/servers/model-picker/catalogs/jumpstart-public.json

@@ -0,0 +1,66 @@
+{
+    "jumpstart://huggingface-llm-falcon-7b": {
+        "modelId": "jumpstart://huggingface-llm-falcon-7b",
+        "family": "falcon",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["text-generation", "llm"],
+        "description": "Falcon 7B via JumpStart"
+    },
+    "jumpstart://huggingface-llm-falcon-40b": {
+        "modelId": "jumpstart://huggingface-llm-falcon-40b",
+        "family": "falcon",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["text-generation", "llm"],
+        "description": "Falcon 40B via JumpStart"
+    },
+    "jumpstart://meta-textgeneration-llama-2-7b": {
+        "modelId": "jumpstart://meta-textgeneration-llama-2-7b",
+        "family": "llama-2",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["text-generation", "llm", "llama-2"],
+        "description": "Llama 2 7B via JumpStart"
+    },
+    "jumpstart://meta-textgeneration-llama-2-13b": {
+        "modelId": "jumpstart://meta-textgeneration-llama-2-13b",
+        "family": "llama-2",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["text-generation", "llm", "llama-2"],
+        "description": "Llama 2 13B via JumpStart"
+    },
+    "jumpstart://meta-textgeneration-llama-2-70b": {
+        "modelId": "jumpstart://meta-textgeneration-llama-2-70b",
+        "family": "llama-2",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["text-generation", "llm", "llama-2"],
+        "description": "Llama 2 70B via JumpStart"
+    },
+    "jumpstart://model-txt2img-stabilityai-stable-diffusion-v2-1-base": {
+        "modelId": "jumpstart://model-txt2img-stabilityai-stable-diffusion-v2-1-base",
+        "family": "stable-diffusion",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["image-generation", "diffusion", "stable-diffusion"],
+        "description": "Stable Diffusion v2.1 Base via JumpStart"
+    },
+    "jumpstart://huggingface-text2text-flan-t5-xl": {
+        "modelId": "jumpstart://huggingface-text2text-flan-t5-xl",
+        "family": "flan-t5",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["text-generation", "text2text", "flan-t5"],
+        "description": "Flan-T5 XL via JumpStart"
+    },
+    "jumpstart://huggingface-textembedding-gpt-j-6b": {
+        "modelId": "jumpstart://huggingface-textembedding-gpt-j-6b",
+        "family": "gpt-j",
+        "framework": "huggingface",
+        "provider": "jumpstart",
+        "tags": ["text-embedding", "gpt-j"],
+        "description": "GPT-J 6B Embedding via JumpStart"
+    }
+}

package/servers/model-picker/catalogs/popular-diffusors.json

@@ -0,0 +1,88 @@
+{
+    "stabilityai/stable-diffusion-3.5-medium": {
+        "family": "stable-diffusion-3",
+        "chat_template": null,
+        "gated": false,
+        "tags": ["image-generation", "diffusion", "stable-diffusion"],
+        "architecture": "StableDiffusion3Pipeline",
+        "framework_compatibility": {
+            "vllm-omni": ">=0.14.0"
+        },
+        "validation_level": "experimental",
+        "profiles": {
+            "default": {
+                "displayName": "SD3.5 Medium",
+                "envVars": {},
+                "recommendedInstanceTypes": ["ml.g5.2xlarge", "ml.g5.4xlarge"]
+            }
+        },
+        "notes": "Stable Diffusion 3.5 medium model. Supported natively by vLLM-Omni StableDiffusion3Pipeline."
+    },
+    "black-forest-labs/FLUX.1-dev": {
+        "family": "flux",
+        "chat_template": null,
+        "gated": true,
+        "tags": ["image-generation", "diffusion", "flux"],
+        "architecture": "FluxPipeline",
+        "framework_compatibility": {
+            "vllm-omni": ">=0.14.0"
+        },
+        "validation_level": "experimental",
+        "profiles": {
+            "default": {
+                "displayName": "FLUX.1 Dev",
+                "envVars": {},
+                "recommendedInstanceTypes": ["ml.g5.4xlarge", "ml.g5.12xlarge"]
+            }
+        },
+        "notes": "FLUX.1-dev high-quality generation model. Uses dual text encoders (CLIP + T5) and FlowMatchEuler scheduler. Requires significant VRAM."
+    },
+    "black-forest-labs/FLUX.1-schnell": {
+        "family": "flux",
+        "chat_template": null,
+        "gated": false,
+        "tags": ["image-generation", "diffusion", "flux"],
+        "architecture": "FluxPipeline",
+        "framework_compatibility": {
+            "vllm-omni": ">=0.14.0"
+        },
+        "validation_level": "experimental",
+        "notes": "FLUX.1-schnell fast generation model. Fewer denoising steps for faster inference at slightly lower quality"
+    },
+    "Wan-AI/Wan2.1-T2V-14B-Diffusers": {
+        "family": "wan",
+        "chat_template": null,
+        "gated": false,
+        "tags": ["video-generation", "diffusion", "wan"],
+        "architecture": "WanPipeline",
+        "framework_compatibility": {
+            "vllm-omni": ">=0.16.0"
+        },
+        "validation_level": "experimental",
+        "notes": "Wan2.1 text-to-video 14B model (diffusers format). Requires multi-GPU instance (ml.g5.12xlarge or larger). Must use the -Diffusers variant — the base Wan2.1-T2V-14B repo lacks model_index.json required by vLLM-Omni"
+    },
+    "stabilityai/stable-diffusion-*": {
+        "family": "stable-diffusion",
+        "chat_template": null,
+        "gated": false,
+        "tags": ["image-generation", "diffusion", "stable-diffusion"],
+        "architecture": null,
+        "framework_compatibility": {
+            "vllm-omni": ">=0.14.0"
+        },
+        "validation_level": "experimental",
+        "notes": "Fallback for Stable Diffusion variants not explicitly listed"
+    },
+    "black-forest-labs/FLUX*": {
+        "family": "flux",
+        "chat_template": null,
+        "gated": false,
+        "tags": ["image-generation", "diffusion", "flux"],
+        "architecture": null,
+        "framework_compatibility": {
+            "vllm-omni": ">=0.14.0"
+        },
+        "validation_level": "experimental",
+        "notes": "Fallback for FLUX model variants not explicitly listed"
+    }
+}

package/servers/model-picker/catalogs/popular-transformers.json

@@ -0,0 +1,226 @@
+{
+    "openai/gpt-oss-20b": {
+        "family": "gpt-oss",
+        "chat_template": "",
+        "gated": false,
+        "tags": ["text-generation", "openai", "conversational"],
+        "architecture": "GPT2LMHeadModel",
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "community-validated",
+        "notes": "Open-source 20B parameter model. Requires significant GPU memory for inference"
+    },
+    "meta-llama/Llama-2-7b-chat-hf": {
+        "family": "llama-2",
+        "chat_template": "{% for message in messages %}{% if message['role'] == 'system' %}{{ '[INST] <<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' }}{% endif %}{% endfor %}",
+        "gated": true,
+        "tags": ["text-generation", "llama-2", "conversational"],
+        "architecture": "LlamaForCausalLM",
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "tested",
+        "profiles": {
+            "7b": {
+                "displayName": "Llama-2 7B",
+                "envVars": {
+                    "MAX_MODEL_LEN": "4096",
+                    "GPU_MEMORY_UTILIZATION": "0.9"
+                },
+                "recommendedInstanceTypes": ["ml.g5.xlarge", "ml.g5.2xlarge"]
+            }
+        },
+        "notes": "Llama-2 7B chat model with official chat template. Requires HuggingFace authentication for download"
+    },
+    "meta-llama/Llama-2-13b-chat-hf": {
+        "family": "llama-2",
+        "chat_template": "{% for message in messages %}{% if message['role'] == 'system' %}{{ '[INST] <<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' }}{% endif %}{% endfor %}",
+        "gated": true,
+        "tags": ["text-generation", "llama-2", "conversational"],
+        "architecture": "LlamaForCausalLM",
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "tested",
+        "profiles": {
+            "13b": {
+                "displayName": "Llama-2 13B",
+                "envVars": {
+                    "MAX_MODEL_LEN": "4096",
+                    "GPU_MEMORY_UTILIZATION": "0.9"
+                },
+                "recommendedInstanceTypes": ["ml.g5.2xlarge", "ml.g5.4xlarge"]
+            }
+        },
+        "notes": "Llama-2 13B chat model. Requires more GPU memory than 7B variant"
+    },
+
+    "meta-llama/Llama-2-70b-chat-hf": {
+        "family": "llama-2",
+        "chat_template": "{% for message in messages %}{% if message['role'] == 'system' %}{{ '[INST] <<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' }}{% endif %}{% endfor %}",
+        "gated": true,
+        "tags": ["text-generation", "llama-2", "conversational"],
+        "architecture": "LlamaForCausalLM",
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "community-validated",
+        "profiles": {
+            "70b-tp2": {
+                "displayName": "Llama-2 70B (2-GPU)",
+                "envVars": {
+                    "TENSOR_PARALLEL_SIZE": "2",
+                    "MAX_MODEL_LEN": "4096",
+                    "GPU_MEMORY_UTILIZATION": "0.95"
+                },
+                "recommendedInstanceTypes": ["ml.g5.12xlarge"]
+            },
+            "70b-tp4": {
+                "displayName": "Llama-2 70B (4-GPU)",
+                "envVars": {
+                    "TENSOR_PARALLEL_SIZE": "4",
+                    "MAX_MODEL_LEN": "4096",
+                    "GPU_MEMORY_UTILIZATION": "0.9"
+                },
+                "recommendedInstanceTypes": ["ml.g5.12xlarge", "ml.g5.48xlarge"]
+            }
+        },
+        "notes": "Llama-2 70B requires tensor parallelism across multiple GPUs"
+    },
+    "mistralai/Mistral-7B-Instruct-v0.1": {
+        "family": "mistral",
+        "chat_template": "{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token }}{% endif %}{% endfor %}",
+        "gated": false,
+        "tags": ["text-generation", "mistral", "conversational"],
+        "architecture": "MistralForCausalLM",
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "tested",
+        "profiles": {
+            "7b": {
+                "displayName": "Mistral 7B Instruct",
+                "envVars": {
+                    "MAX_MODEL_LEN": "8192",
+                    "GPU_MEMORY_UTILIZATION": "0.9"
+                },
+                "recommendedInstanceTypes": ["ml.g5.xlarge", "ml.g5.2xlarge"]
+            }
+        },
+        "notes": "Mistral 7B v0.1 with 8K context window"
+    },
+    "mistralai/Mistral-7B-Instruct-v0.2": {
+        "family": "mistral",
+        "chat_template": "{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token }}{% endif %}{% endfor %}",
+        "gated": false,
+        "tags": ["text-generation", "mistral", "conversational"],
+        "architecture": "MistralForCausalLM",
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "tested",
+        "profiles": {
+            "7b": {
+                "displayName": "Mistral 7B Instruct v0.2",
+                "envVars": {
+                    "MAX_MODEL_LEN": "32768",
+                    "GPU_MEMORY_UTILIZATION": "0.9"
+                },
+                "recommendedInstanceTypes": ["ml.g5.2xlarge", "ml.g5.4xlarge"]
+            }
+        },
+        "notes": "Mistral 7B v0.2 with extended 32K context window. Requires more memory for long contexts"
+    },
+    "mistralai/Mixtral-8x7B-Instruct-v0.1": {
+        "family": "mistral",
+        "chat_template": "{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token }}{% endif %}{% endfor %}",
+        "gated": false,
+        "tags": ["text-generation", "mistral", "mixture-of-experts"],
+        "architecture": "MixtralForCausalLM",
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "community-validated",
+        "profiles": {
+            "8x7b-tp2": {
+                "displayName": "Mixtral 8x7B (2-GPU)",
+                "envVars": {
+                    "TENSOR_PARALLEL_SIZE": "2",
+                    "MAX_MODEL_LEN": "32768",
+                    "GPU_MEMORY_UTILIZATION": "0.95"
+                },
+                "recommendedInstanceTypes": ["ml.g5.12xlarge"]
+            }
+        },
+        "notes": "Mixtral 8x7B MoE model. Requires tensor parallelism for efficient inference"
+    },
+    "meta-llama/Llama-2-*": {
+        "family": "llama-2",
+        "chat_template": "{% for message in messages %}{% if message['role'] == 'system' %}{{ '[INST] <<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' }}{% endif %}{% endfor %}",
+        "gated": true,
+        "tags": ["text-generation", "llama-2"],
+        "architecture": null,
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "experimental",
+        "notes": "Fallback configuration for Llama-2 models not explicitly listed. Uses standard Llama-2 chat template"
+    },
+    "mistralai/Mistral-*": {
+        "family": "mistral",
+        "chat_template": "{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token }}{% endif %}{% endfor %}",
+        "gated": false,
+        "tags": ["text-generation", "mistral"],
+        "architecture": null,
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0",
+            "sglang": ">=0.2.0"
+        },
+        "validation_level": "experimental",
+        "notes": "Fallback configuration for Mistral models not explicitly listed. Uses standard Mistral chat template"
+    },
+    "codellama/*": {
+        "family": "codellama",
+        "chat_template": "{% for message in messages %}{% if message['role'] == 'system' %}{{ '[INST] <<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' }}{% endif %}{% endfor %}",
+        "gated": false,
+        "tags": ["text-generation", "code", "codellama"],
+        "architecture": null,
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0"
+        },
+        "validation_level": "experimental",
+        "notes": "CodeLlama models use Llama-2 chat template. Optimized for code generation"
+    },
+    "tiiuae/falcon-*": {
+        "family": "falcon",
+        "chat_template": null,
+        "gated": false,
+        "tags": ["text-generation", "falcon"],
+        "architecture": null,
+        "framework_compatibility": {
+            "vllm": ">=0.3.0",
+            "tensorrt-llm": ">=0.8.0"
+        },
+        "validation_level": "experimental",
+        "notes": "Falcon models typically don't require chat templates for instruction following"
+    }
+}