update to torch 2.4&2.5

Azure · Oct 30, 2024 · b875a07 · b875a07
1 parent ea9d39e
commit b875a07
Show file tree

Hide file tree

Showing 14 changed files with 19 additions and 21 deletions.
diff --git a/...ents/acpt-pytorch-2.3-cuda12.1/asset.yaml → ...ents/acpt-pytorch-2.4-cuda12.4/asset.yaml b/...ents/acpt-pytorch-2.3-cuda12.1/asset.yaml → ...ents/acpt-pytorch-2.4-cuda12.4/asset.yaml
@@ -1,4 +1,4 @@
-name: acpt-pytorch-2.3-cuda12.1
+name: acpt-pytorch-2.4-cuda12.4
 version: auto
 type: environment
 spec: spec.yaml

diff --git a/...t-pytorch-2.4-cuda12.1/context/Dockerfile → ...t-pytorch-2.4-cuda12.4/context/Dockerfile b/...t-pytorch-2.4-cuda12.1/context/Dockerfile → ...t-pytorch-2.4-cuda12.4/context/Dockerfile
@@ -1,4 +1,4 @@
-FROM mcr.microsoft.com/aifx/acpt/stable-ubuntu2004-cu121-py310-torch24x:{{latest-image-tag:biweekly\.\d{6}\.\d{1}.*}}
+FROM mcr.microsoft.com/aifx/acpt/stable-ubuntu2204-cu124-py310-torch24x:{{latest-image-tag:biweekly\.\d{6}\.\d{1}.*}}
 
 # Install pip dependencies
 COPY requirements.txt .

diff --git a/...rch-2.3-cuda12.1/context/requirements.txt → ...rch-2.4-cuda12.4/context/requirements.txt b/...rch-2.3-cuda12.1/context/requirements.txt → ...rch-2.4-cuda12.4/context/requirements.txt
diff --git a/...cpt-pytorch-2.4-cuda12.1/environment.yaml → ...cpt-pytorch-2.4-cuda12.4/environment.yaml b/...cpt-pytorch-2.4-cuda12.1/environment.yaml → ...cpt-pytorch-2.4-cuda12.4/environment.yaml
@@ -1,5 +1,5 @@
 image:
-  name: azureml/curated/acpt-pytorch-2.4-cuda12.1
+  name: azureml/curated/acpt-pytorch-2.4-cuda12.4
   os: linux
   context:
     dir: context

diff --git a/...ments/acpt-pytorch-2.3-cuda12.1/spec.yaml → ...ments/acpt-pytorch-2.4-cuda12.4/spec.yaml b/...ments/acpt-pytorch-2.3-cuda12.1/spec.yaml → ...ments/acpt-pytorch-2.4-cuda12.4/spec.yaml
@@ -2,7 +2,7 @@ $schema: https://azuremlschemas.azureedge.net/latest/environment.schema.json
 
 description: >-
   Recommended environment for Deep Learning in public preview with PyTorch on Azure containing the Azure ML SDK with the latest compatible versions of Ubuntu, Python, PyTorch, CUDA\RocM, combined with optimizers like ORT Training,+DeepSpeed+MSCCL+ORT MoE and more. The image introduces newly released PyTorch 2.1 for early testing, and preview of new fastcheckpointing capability called Nebula. 
-  Azure Container Registry:acptdev.azurecr.io/test/public/aifx/acpt/stable-ubuntu2004-cu121-py310-torch212
+  Azure Container Registry:acptdev.azurecr.io/test/public/aifx/acpt/stable-ubuntu2204-cu124-py310-torch212
 
 name: "{{asset.name}}"
 version: "{{asset.version}}"
@@ -14,9 +14,9 @@ build:
 os_type: linux
 
 tags:
-  PyTorch: "2.3.1"
-  GPU: Cuda121
-  OS: Ubuntu20.04
+  PyTorch: "2.4"
+  GPU: Cuda124
+  OS: Ubuntu22.04
   Training: ""
   Preview: ""
   Python: "3.10"

diff --git a/...-cuda12.1/tests/pytorch2_4_sample_test.py → ...-cuda12.4/tests/pytorch2_4_sample_test.py b/...-cuda12.1/tests/pytorch2_4_sample_test.py → ...-cuda12.4/tests/pytorch2_4_sample_test.py
@@ -17,7 +17,7 @@
 
 
 def test_pytorch_2_4():
-    """Tests a sample job using pytorch 2.0 as the environment."""
+    """Tests a sample job using pytorch 2.4 as the environment."""
     this_dir = Path(__file__).parent
 
     subscription_id = os.environ.get("subscription_id")
@@ -28,7 +28,7 @@ def test_pytorch_2_4():
         AzureCliCredential(), subscription_id, resource_group, workspace_name
     )
 
-    env_name = "acpt-pytorch-2_4-cuda12_1"
+    env_name = "acpt-pytorch-2_4-cuda12_4"
 
     env_docker_context = Environment(
         build=BuildContext(path=this_dir / BUILD_CONTEXT),

diff --git a/...torch-2.3-cuda12.1/tests/requirements.txt → ...torch-2.4-cuda12.4/tests/requirements.txt b/...torch-2.3-cuda12.1/tests/requirements.txt → ...torch-2.4-cuda12.4/tests/requirements.txt
@@ -1,3 +1,2 @@
 azure-ai-ml==1.2.0
 azure.identity==1.10.0
-huggingface-hub==0.25.1
diff --git a/...ents/acpt-pytorch-2.4-cuda12.1/asset.yaml → ...ents/acpt-pytorch-2.5-cuda12.4/asset.yaml b/...ents/acpt-pytorch-2.4-cuda12.1/asset.yaml → ...ents/acpt-pytorch-2.5-cuda12.4/asset.yaml
@@ -1,4 +1,4 @@
-name: acpt-pytorch-2.4-cuda12.1
+name: acpt-pytorch-2.5-cuda12.4
 version: auto
 type: environment
 spec: spec.yaml

diff --git a/...t-pytorch-2.3-cuda12.1/context/Dockerfile → ...t-pytorch-2.5-cuda12.4/context/Dockerfile b/...t-pytorch-2.3-cuda12.1/context/Dockerfile → ...t-pytorch-2.5-cuda12.4/context/Dockerfile
@@ -1,4 +1,4 @@
-FROM mcr.microsoft.com/aifx/acpt/stable-ubuntu2004-cu121-py310-torch23x:{{latest-image-tag:biweekly\.\d{6}\.\d{1}.*}}
+FROM mcr.microsoft.com/aifx/acpt/stable-ubuntu2204-cu124-py310-torch25x:{{latest-image-tag:biweekly\.\d{6}\.\d{1}.*}}
 
 # Install pip dependencies
 COPY requirements.txt .

diff --git a/...rch-2.4-cuda12.1/context/requirements.txt → ...rch-2.5-cuda12.4/context/requirements.txt b/...rch-2.4-cuda12.1/context/requirements.txt → ...rch-2.5-cuda12.4/context/requirements.txt
diff --git a/...cpt-pytorch-2.3-cuda12.1/environment.yaml → ...cpt-pytorch-2.5-cuda12.4/environment.yaml b/...cpt-pytorch-2.3-cuda12.1/environment.yaml → ...cpt-pytorch-2.5-cuda12.4/environment.yaml
@@ -1,5 +1,5 @@
 image:
-  name: azureml/curated/acpt-pytorch-2.3-cuda12.1
+  name: azureml/curated/acpt-pytorch-2.5-cuda12.4
   os: linux
   context:
     dir: context

diff --git a/...ments/acpt-pytorch-2.4-cuda12.1/spec.yaml → ...ments/acpt-pytorch-2.5-cuda12.4/spec.yaml b/...ments/acpt-pytorch-2.4-cuda12.1/spec.yaml → ...ments/acpt-pytorch-2.5-cuda12.4/spec.yaml
@@ -14,9 +14,9 @@ build:
 os_type: linux
 
 tags:
-  PyTorch: "2.4"
+  PyTorch: "2.5.0"
   GPU: Cuda124
-  OS: Ubuntu20.04
+  OS: Ubuntu22.04
   Training: ""
   Preview: ""
   Python: "3.10"

diff --git a/...-cuda12.1/tests/pytorch2_3_sample_test.py → ...-cuda12.4/tests/pytorch2_5_sample_test.py b/...-cuda12.1/tests/pytorch2_3_sample_test.py → ...-cuda12.4/tests/pytorch2_5_sample_test.py
@@ -1,7 +1,7 @@
 # Copyright (c) Microsoft Corporation.
 # Licensed under the MIT License.
 
-"""Tests running a sample job in the pytorch 2.0 environment."""
+"""Tests running a sample job in the pytorch 2.5 environment."""
 import os
 import time
 from pathlib import Path
@@ -16,8 +16,8 @@
 STD_LOG = Path("artifacts/user_logs/std_log.txt")
 
 
-def test_pytorch_2_3():
-    """Tests a sample job using pytorch 2.0 as the environment."""
+def test_pytorch_2_5():
+    """Tests a sample job using pytorch 2.5 as the environment."""
     this_dir = Path(__file__).parent
 
     subscription_id = os.environ.get("subscription_id")
@@ -28,12 +28,12 @@ def test_pytorch_2_3():
         AzureCliCredential(), subscription_id, resource_group, workspace_name
     )
 
-    env_name = "acpt-pytorch-2_3-cuda12_1"
+    env_name = "acpt-pytorch-2_5-cuda12_4"
 
     env_docker_context = Environment(
         build=BuildContext(path=this_dir / BUILD_CONTEXT),
         name=env_name,
-        description="Pytorch 2.3 environment created from a Docker context.",
+        description="Pytorch 2.5 environment created from a Docker context.",
     )
     ml_client.environments.create_or_update(env_docker_context)
 

diff --git a/...torch-2.4-cuda12.1/tests/requirements.txt → ...torch-2.5-cuda12.4/tests/requirements.txt b/...torch-2.4-cuda12.1/tests/requirements.txt → ...torch-2.5-cuda12.4/tests/requirements.txt
@@ -1,3 +1,2 @@
 azure-ai-ml==1.2.0
 azure.identity==1.10.0
-huggingface-hub==0.25.1