feat: TGI optimum 0.0.18 (general+llm) (#4436)

* tgi optimum 0.0.18 * add test * update latest transformer version * add dlc optimum 0.0.18 config * 0.0 defaults to 0.0.16
aws · Feb 27, 2024 · 5a22f4d · 5a22f4d
1 parent 4e2d353
commit 5a22f4d
Show file tree

Hide file tree

Showing 4 changed files with 88 additions and 3 deletions.
diff --git a/src/sagemaker/image_uri_config/huggingface-llm-neuronx.json b/src/sagemaker/image_uri_config/huggingface-llm-neuronx.json
@@ -64,6 +64,35 @@
                 "container_version": {
                     "inf2": "ubuntu22.04"
                 }
+            },
+            "0.0.18": {
+                "py_versions": [
+                    "py310"
+                ],
+                "registries": {
+                    "ap-northeast-1": "763104351884",
+                    "ap-south-1": "763104351884",
+                    "ap-south-2": "772153158452",
+                    "ap-southeast-1": "763104351884",
+                    "ap-southeast-2": "763104351884",
+                    "ap-southeast-4": "457447274322",
+                    "eu-central-1": "763104351884",
+                    "eu-central-2": "380420809688",
+                    "eu-south-2": "503227376785",
+                    "eu-west-1": "763104351884",
+                    "eu-west-3": "763104351884",
+                    "il-central-1": "780543022126",
+                    "sa-east-1": "763104351884",
+                    "us-east-1": "763104351884",
+                    "us-east-2": "763104351884",
+                    "us-west-2": "763104351884",
+                    "ca-west-1": "204538143572"
+                },
+                "tag_prefix": "1.13.1-optimum0.0.18",
+                "repository": "huggingface-pytorch-tgi-inference",
+                "container_version": {
+                    "inf2": "ubuntu22.04"
+                }
             }
         }
     }

diff --git a/src/sagemaker/image_uri_config/huggingface-neuronx.json b/src/sagemaker/image_uri_config/huggingface-neuronx.json
@@ -4,7 +4,8 @@
             "trn"
         ],
         "version_aliases": {
-            "4.28": "4.28.1"
+            "4.28": "4.28.1",
+            "4.34": "4.34.1"
         },
         "versions": {
             "4.28.1": {
@@ -86,7 +87,9 @@
             "inf"
         ],
         "version_aliases": {
-            "4.28": "4.28.1"
+            "4.28": "4.28.1",
+            "4.34": "4.34.1",
+            "4.36": "4.36.2"
         },
         "versions": {
             "4.28.1": {
@@ -192,6 +195,58 @@
                         "sdk2.15.0"
                     ]
                 }
+            },
+            "4.36.2": {
+                "version_aliases": {
+                    "pytorch1.13": "pytorch1.13.1"
+                },
+                "pytorch1.13.1": {
+                    "py_versions": [
+                        "py310"
+                    ],
+                    "repository": "huggingface-pytorch-inference-neuronx",
+                    "registries": {
+                        "af-south-1": "626614931356",
+                        "il-central-1": "780543022126",
+                        "ap-east-1": "871362719292",
+                        "ap-northeast-1": "763104351884",
+                        "ap-northeast-2": "763104351884",
+                        "ap-northeast-3": "364406365360",
+                        "ap-south-1": "763104351884",
+                        "ap-south-2": "772153158452",
+                        "ap-southeast-1": "763104351884",
+                        "ap-southeast-2": "763104351884",
+                        "ap-southeast-4": "457447274322",
+                        "ca-central-1": "763104351884",
+                        "cn-north-1": "727897471807",
+                        "cn-northwest-1": "727897471807",
+                        "eu-central-1": "763104351884",
+                        "eu-central-2": "380420809688",
+                        "eu-north-1": "763104351884",
+                        "eu-west-1": "763104351884",
+                        "eu-west-2": "763104351884",
+                        "eu-west-3": "763104351884",
+                        "eu-south-1": "692866216735",
+                        "eu-south-2": "503227376785",
+                        "me-south-1": "217643126080",
+                        "sa-east-1": "763104351884",
+                        "us-east-1": "763104351884",
+                        "us-east-2": "763104351884",
+                        "us-gov-east-1": "446045086412",
+                        "us-gov-west-1": "442386744353",
+                        "us-iso-east-1": "886529160074",
+                        "us-isob-east-1": "094389454867",
+                        "us-west-1": "763104351884",
+                        "us-west-2": "763104351884",
+                        "ca-west-1": "204538143572"
+                    },
+                    "container_version": {
+                        "inf": "ubuntu20.04"
+                    },
+                    "sdk_versions": [
+                        "sdk2.16.1"
+                    ]
+                }
             }
         }
     }

diff --git a/tests/conftest.py b/tests/conftest.py
@@ -441,7 +441,7 @@ def huggingface_neuron_latest_inference_transformer_version():
 
 @pytest.fixture(scope="module")
 def huggingface_neuronx_latest_inference_transformer_version():
-    return "4.34.1"
+    return "4.36.2"
 
 
 @pytest.fixture(scope="module")

diff --git a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py
@@ -34,6 +34,7 @@
     "inf2": {
         "0.0.16": "1.13.1-optimum0.0.16-neuronx-py310-ubuntu22.04",
         "0.0.17": "1.13.1-optimum0.0.17-neuronx-py310-ubuntu22.04",
+        "0.0.18": "1.13.1-optimum0.0.18-neuronx-py310-ubuntu22.04",
     },
 }