From 7e744bfcf771b9f68c9239b3206e72236fe18b21 Mon Sep 17 00:00:00 2001 From: Haotian An Date: Tue, 15 Oct 2024 14:49:12 +0000 Subject: [PATCH 01/10] release: huggingface tgi neuronx 0.0.25 image --- .../huggingface-llm-neuronx.json | 35 ++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-) diff --git a/src/sagemaker/image_uri_config/huggingface-llm-neuronx.json b/src/sagemaker/image_uri_config/huggingface-llm-neuronx.json index c86a8dcf98..7b96b60ff8 100644 --- a/src/sagemaker/image_uri_config/huggingface-llm-neuronx.json +++ b/src/sagemaker/image_uri_config/huggingface-llm-neuronx.json @@ -4,7 +4,7 @@ "inf2" ], "version_aliases": { - "0.0": "0.0.24" + "0.0": "0.0.25" }, "versions": { "0.0.16": { @@ -301,6 +301,39 @@ "container_version": { "inf2": "ubuntu22.04" } + }, + "0.0.25": { + "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", + "cn-north-1": "727897471807", + "cn-northwest-1": "727897471807", + "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-gov-east-1": "446045086412", + "us-gov-west-1": "442386744353", + "us-west-2": "763104351884", + "ca-west-1": "204538143572" + }, + "tag_prefix": "2.1.2-optimum0.0.25", + "repository": "huggingface-pytorch-tgi-inference", + "container_version": { + "inf2": "ubuntu22.04" + } } } } From 567903af74bcd3202f24cf9a16e26d68fdb10acf Mon Sep 17 00:00:00 2001 From: Haotian An Date: Tue, 15 Oct 2024 14:50:11 +0000 Subject: [PATCH 02/10] tests --- tests/unit/sagemaker/image_uris/test_huggingface_llm.py | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py index c3fab15936..a47ce7b42a 100644 --- a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py +++ b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py @@ -56,6 +56,7 @@ "0.0.22": "2.1.2-optimum0.0.22-neuronx-py310-ubuntu22.04", "0.0.23": "2.1.2-optimum0.0.23-neuronx-py310-ubuntu22.04", "0.0.24": "2.1.2-optimum0.0.24-neuronx-py310-ubuntu22.04", + "0.0.25": "2.1.2-optimum0.0.25-neuronx-py310-ubuntu22.04", }, } From d74e28740c6db9a97d2a21573e0f77a644c7bfa4 Mon Sep 17 00:00:00 2001 From: Haotian An Date: Tue, 15 Oct 2024 19:43:36 +0000 Subject: [PATCH 03/10] add tgi 2.3.1 --- .../image_uri_config/huggingface-llm.json | 47 +++++++++++++++++++ .../image_uris/test_huggingface_llm.py | 1 + 2 files changed, 48 insertions(+) diff --git a/src/sagemaker/image_uri_config/huggingface-llm.json b/src/sagemaker/image_uri_config/huggingface-llm.json index b3988362fd..c6075bae47 100644 --- a/src/sagemaker/image_uri_config/huggingface-llm.json +++ b/src/sagemaker/image_uri_config/huggingface-llm.json @@ -719,6 +719,53 @@ "container_version": { "gpu": "cu121-ubuntu22.04-v2.0" } + }, + "2.3.1": { + "py_versions": [ + "py310" + ], + "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-3": "907027046896", + "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", + "me-central-1": "914824155844", + "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" + }, + "tag_prefix": "2.4.0-tgi2.3.1", + "repository": "huggingface-pytorch-tgi-inference", + "container_version": { + "gpu": "cu124-ubuntu22.04" + } } } } diff --git a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py index a47ce7b42a..581dcf90c1 100644 --- a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py +++ b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py @@ -45,6 +45,7 @@ "2.0.1": "2.1.1-tgi2.0.1-gpu-py310-cu121-ubuntu22.04", "2.0.2": "2.3.0-tgi2.0.2-gpu-py310-cu121-ubuntu22.04", "2.2.0": "2.3.0-tgi2.2.0-gpu-py310-cu121-ubuntu22.04-v2.0", + "2.3.1": "2.4.0-tgi2.3.1-gpu-py310-cu124-ubuntu22.04", }, "inf2": { "0.0.16": "1.13.1-optimum0.0.16-neuronx-py310-ubuntu22.04", From 77a1f3a4a885b5e34c518909b58c87fd8c125f85 Mon Sep 17 00:00:00 2001 From: Haotian An Date: Tue, 15 Oct 2024 19:44:16 +0000 Subject: [PATCH 04/10] update default version --- src/sagemaker/image_uri_config/huggingface-llm.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/sagemaker/image_uri_config/huggingface-llm.json b/src/sagemaker/image_uri_config/huggingface-llm.json index c6075bae47..f2c9ac7bb0 100644 --- a/src/sagemaker/image_uri_config/huggingface-llm.json +++ b/src/sagemaker/image_uri_config/huggingface-llm.json @@ -12,7 +12,7 @@ "1.2": "1.2.0", "1.3": "1.3.3", "1.4": "1.4.5", - "2.0": "2.2.0" + "2.0": "2.3.1" }, "versions": { "0.6.0": { From 6e50e33262296b424fa8744854a467a7aa25940e Mon Sep 17 00:00:00 2001 From: Haotian An Date: Thu, 17 Oct 2024 00:55:02 +0000 Subject: [PATCH 05/10] update py version --- src/sagemaker/image_uri_config/huggingface-llm.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/sagemaker/image_uri_config/huggingface-llm.json b/src/sagemaker/image_uri_config/huggingface-llm.json index f2c9ac7bb0..24cbd5ca96 100644 --- a/src/sagemaker/image_uri_config/huggingface-llm.json +++ b/src/sagemaker/image_uri_config/huggingface-llm.json @@ -722,7 +722,7 @@ }, "2.3.1": { "py_versions": [ - "py310" + "py311" ], "registries": { "af-south-1": "626614931356", From 9374c7518e97bd845f952f923d8784cbedb02f02 Mon Sep 17 00:00:00 2001 From: Haotian An Date: Thu, 17 Oct 2024 01:26:52 +0000 Subject: [PATCH 06/10] fix tests --- tests/unit/sagemaker/image_uris/test_huggingface_llm.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py index 581dcf90c1..28525a390c 100644 --- a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py +++ b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py @@ -45,7 +45,7 @@ "2.0.1": "2.1.1-tgi2.0.1-gpu-py310-cu121-ubuntu22.04", "2.0.2": "2.3.0-tgi2.0.2-gpu-py310-cu121-ubuntu22.04", "2.2.0": "2.3.0-tgi2.2.0-gpu-py310-cu121-ubuntu22.04-v2.0", - "2.3.1": "2.4.0-tgi2.3.1-gpu-py310-cu124-ubuntu22.04", + "2.3.1": "2.4.0-tgi2.3.1-gpu-py311-cu124-ubuntu22.04", }, "inf2": { "0.0.16": "1.13.1-optimum0.0.16-neuronx-py310-ubuntu22.04", From 90f6e0b5253a17825a0bdb7d570640bbb274199f Mon Sep 17 00:00:00 2001 From: Haotian An Date: Thu, 17 Oct 2024 03:19:02 +0000 Subject: [PATCH 07/10] test --- tests/unit/sagemaker/image_uris/test_model_monitor.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/unit/sagemaker/image_uris/test_model_monitor.py b/tests/unit/sagemaker/image_uris/test_model_monitor.py index 9c7570fb5d..147db6b0a0 100644 --- a/tests/unit/sagemaker/image_uris/test_model_monitor.py +++ b/tests/unit/sagemaker/image_uris/test_model_monitor.py @@ -9,7 +9,7 @@ # or in the "license" file accompanying this file. This file is # distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF # ANY KIND, either express or implied. See the License for the specific -# language governing permissions and limitations under the License. +# language governing permissions and limitations under the License. from __future__ import absolute_import import pytest From 20d46d187a29638bcb6025a82af40e55e3331685 Mon Sep 17 00:00:00 2001 From: Haotian An Date: Thu, 17 Oct 2024 12:49:10 +0000 Subject: [PATCH 08/10] Revert "fix tests" This reverts commit 9374c7518e97bd845f952f923d8784cbedb02f02. --- tests/unit/sagemaker/image_uris/test_huggingface_llm.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py index 28525a390c..581dcf90c1 100644 --- a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py +++ b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py @@ -45,7 +45,7 @@ "2.0.1": "2.1.1-tgi2.0.1-gpu-py310-cu121-ubuntu22.04", "2.0.2": "2.3.0-tgi2.0.2-gpu-py310-cu121-ubuntu22.04", "2.2.0": "2.3.0-tgi2.2.0-gpu-py310-cu121-ubuntu22.04-v2.0", - "2.3.1": "2.4.0-tgi2.3.1-gpu-py311-cu124-ubuntu22.04", + "2.3.1": "2.4.0-tgi2.3.1-gpu-py310-cu124-ubuntu22.04", }, "inf2": { "0.0.16": "1.13.1-optimum0.0.16-neuronx-py310-ubuntu22.04", From 22774fd18dc2b7844a25ee17bc1ce62ebd942aae Mon Sep 17 00:00:00 2001 From: Haotian An Date: Thu, 17 Oct 2024 14:00:20 +0000 Subject: [PATCH 09/10] Revert "Revert "fix tests"" This reverts commit 20d46d187a29638bcb6025a82af40e55e3331685. --- tests/unit/sagemaker/image_uris/test_huggingface_llm.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py index 581dcf90c1..28525a390c 100644 --- a/tests/unit/sagemaker/image_uris/test_huggingface_llm.py +++ b/tests/unit/sagemaker/image_uris/test_huggingface_llm.py @@ -45,7 +45,7 @@ "2.0.1": "2.1.1-tgi2.0.1-gpu-py310-cu121-ubuntu22.04", "2.0.2": "2.3.0-tgi2.0.2-gpu-py310-cu121-ubuntu22.04", "2.2.0": "2.3.0-tgi2.2.0-gpu-py310-cu121-ubuntu22.04-v2.0", - "2.3.1": "2.4.0-tgi2.3.1-gpu-py310-cu124-ubuntu22.04", + "2.3.1": "2.4.0-tgi2.3.1-gpu-py311-cu124-ubuntu22.04", }, "inf2": { "0.0.16": "1.13.1-optimum0.0.16-neuronx-py310-ubuntu22.04", From 639a6b5b4411a61761ada10f406807a1696b3937 Mon Sep 17 00:00:00 2001 From: Haotian An Date: Thu, 17 Oct 2024 14:00:30 +0000 Subject: [PATCH 10/10] Revert "test" This reverts commit 90f6e0b5253a17825a0bdb7d570640bbb274199f. --- tests/unit/sagemaker/image_uris/test_model_monitor.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/unit/sagemaker/image_uris/test_model_monitor.py b/tests/unit/sagemaker/image_uris/test_model_monitor.py index 147db6b0a0..9c7570fb5d 100644 --- a/tests/unit/sagemaker/image_uris/test_model_monitor.py +++ b/tests/unit/sagemaker/image_uris/test_model_monitor.py @@ -9,7 +9,7 @@ # or in the "license" file accompanying this file. This file is # distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF # ANY KIND, either express or implied. See the License for the specific -# language governing permissions and limitations under the License. +# language governing permissions and limitations under the License. from __future__ import absolute_import import pytest