Merge branch 'hpcaitech:main' into fix-dp-rank

Author: liuqh16

.compatibility

@@ -1,3 +1,3 @@
-2.2.2-12.1.0
 2.3.0-12.1.0
 2.4.0-12.4.1
+2.5.1-12.4.1

.cuda_ext.json

@@ -1,12 +1,12 @@
 {
   "build": [
     {
-      "torch_command": "pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 --index-url https://download.pytorch.org/whl/cu121",
-      "cuda_image": "hpcaitech/cuda-conda:12.1"
+      "torch_command": "pip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu121",
+      "cuda_image": "image-cloud.luchentech.com/hpcaitech/cuda-conda:12.1"
     },
     {
-      "torch_command": "pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu124",
-      "cuda_image": "hpcaitech/cuda-conda:12.4"
+      "torch_command": "pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu124",
+      "cuda_image": "image-cloud.luchentech.com/hpcaitech/cuda-conda:12.4"
     }
   ]
 }

.github/workflows/build_on_pr.yml

@@ -34,7 +34,7 @@ jobs:
       anyExtensionFileChanged: ${{ steps.find-extension-change.outputs.any_changed }}
       changedLibraryFiles: ${{ steps.find-lib-change.outputs.all_changed_files }}
       anyLibraryFileChanged: ${{ steps.find-lib-change.outputs.any_changed }}
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, ubuntu-latest]
     concurrency:
       group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}-detect-change
       cancel-in-progress: true
@@ -87,10 +87,10 @@ jobs:
     name: Build and Test Colossal-AI
     needs: detect
     if: needs.detect.outputs.anyLibraryFileChanged == 'true'
-    runs-on: [self-hosted, gpu]
+    runs-on: [self-hosted, ubuntu-latest]
     container:
-      image: hpcaitech/pytorch-cuda:2.2.2-12.1.0
-      options: --gpus all --rm -v /dev/shm -v /data/scratch:/data/scratch
+      image: image-cloud.luchentech.com/hpcaitech/pytorch-cuda:2.2.2-12.1.0
+      options: --gpus all --shm-size=2g --rm -v /dev/shm -v /data/scratch:/data/scratch
     timeout-minutes: 90
     defaults:
       run:
@@ -138,6 +138,10 @@ jobs:
             cp -p -r /github/home/cuda_ext_cache/* /__w/ColossalAI/ColossalAI/
           fi
 
+      - name: Install flash-attention
+        run: |
+          pip install flash-attn==2.7.4.post1 --no-build-isolation
+
       - name: Install Colossal-AI
         run: |
           BUILD_EXT=1 pip install -v -e .

.github/workflows/build_on_schedule.yml

@@ -3,16 +3,16 @@ name: Build on Schedule
 on:
   schedule:
     # run at 00:00 of every Sunday
-    - cron: "0 0 * * *"
+    - cron: "0 0 * * 0"
   workflow_dispatch:
 
 jobs:
   build:
     name: Build and Test Colossal-AI
     if: github.repository == 'hpcaitech/ColossalAI'
-    runs-on: [self-hosted, gpu]
+    runs-on: [self-hosted, ubuntu-latest]
     container:
-      image: hpcaitech/pytorch-cuda:2.2.2-12.1.0
+      image: image-cloud.luchentech.com/hpcaitech/pytorch-cuda:2.2.2-12.1.0
       options: --gpus all --rm -v /dev/shm -v /data/scratch/:/data/scratch/
     timeout-minutes: 90
     steps:
@@ -51,6 +51,10 @@ jobs:
         with:
           ssh-key: ${{ secrets.SSH_KEY_FOR_CI }}
 
+      - name: Install flash-attention
+        run: |
+          pip install flash-attn==2.7.4.post1 --no-build-isolation
+
       - name: Install Colossal-AI
         if: steps.check-avai.outputs.avai == 'true'
         run: |

.github/workflows/close_inactive.yml

@@ -7,7 +7,7 @@ on:
 jobs:
   close-issues:
     if: github.event.pull_request.draft == false && github.base_ref == 'main' && github.event.pull_request.base.repo.full_name == 'hpcaitech/ColossalAI'
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, ubuntu-latest]
     permissions:
       issues: write
       pull-requests: write

.github/workflows/compatiblity_test_on_dispatch.yml

@@ -15,7 +15,7 @@ on:
 jobs:
   matrix_preparation:
     name: Prepare Container List
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, ubuntu-latest]
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
     steps:
@@ -31,7 +31,7 @@ jobs:
           do
               for cv in $CUDA_VERSIONS
               do
-                  DOCKER_IMAGE+=("\"hpcaitech/pytorch-cuda:${tv}-${cv}\"")
+                  DOCKER_IMAGE+=("\"image-cloud.luchentech.com/hpcaitech/pytorch-cuda:${tv}-${cv}\"")
               done
           done
 
@@ -44,7 +44,7 @@ jobs:
     name: Test for PyTorch Compatibility
     needs: matrix_preparation
     if: github.repository == 'hpcaitech/ColossalAI'
-    runs-on: [self-hosted, 8-gpu]
+    runs-on: [self-hosted, ubuntu-latest]
     strategy:
       fail-fast: false
       matrix: ${{fromJson(needs.matrix_preparation.outputs.matrix)}}
@@ -73,7 +73,18 @@ jobs:
 
       - name: Unit Testing
         run: |
-          PYTHONPATH=$PWD pytest --durations=0 tests
+          PYTHONPATH=$PWD pytest
+          -m "not largedist" \
+          --durations=0 \
+          --ignore tests/test_analyzer \
+          --ignore tests/test_auto_parallel \
+          --ignore tests/test_fx \
+          --ignore tests/test_autochunk \
+          --ignore tests/test_gptq \
+          --ignore tests/test_infer_ops \
+          --ignore tests/test_legacy \
+          --ignore tests/test_smoothquant \
+          tests/
         env:
           DATA: /data/scratch/cifar-10
           LD_LIBRARY_PATH: /github/home/.tensornvme/lib

.github/workflows/compatiblity_test_on_pr.yml

@@ -9,7 +9,7 @@ on:
 jobs:
   matrix_preparation:
     name: Prepare Container List
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, ubuntu-latest]
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
     concurrency:
@@ -23,7 +23,7 @@ jobs:
           DOCKER_IMAGE=()
 
           while read tag; do
-            DOCKER_IMAGE+=("\"hpcaitech/pytorch-cuda:${tag}\"")
+            DOCKER_IMAGE+=("\"image-cloud.luchentech.com/hpcaitech/pytorch-cuda:${tag}\"")
           done <.compatibility
 
           container=$( IFS=',' ; echo "${DOCKER_IMAGE[*]}" )
@@ -35,7 +35,7 @@ jobs:
     name: Test for PyTorch Compatibility
     needs: matrix_preparation
     if: github.repository == 'hpcaitech/ColossalAI'
-    runs-on: [self-hosted, 8-gpu]
+    runs-on: [self-hosted, ubuntu-latest]
     strategy:
       fail-fast: false
       matrix: ${{fromJson(needs.matrix_preparation.outputs.matrix)}}
@@ -67,7 +67,18 @@ jobs:
 
       - name: Unit Testing
         run: |
-          PYTHONPATH=$PWD pytest --durations=0 tests
+          PYTHONPATH=$PWD pytest \
+          -m "not largedist" \
+          --durations=0 \
+          --ignore tests/test_analyzer \
+          --ignore tests/test_auto_parallel \
+          --ignore tests/test_fx \
+          --ignore tests/test_autochunk \
+          --ignore tests/test_gptq \
+          --ignore tests/test_infer_ops \
+          --ignore tests/test_legacy \
+          --ignore tests/test_smoothquant \
+          tests/
         env:
           DATA: /data/scratch/cifar-10
           LD_LIBRARY_PATH: /github/home/.tensornvme/lib

.github/workflows/compatiblity_test_on_schedule.yml

@@ -9,7 +9,7 @@ on:
 jobs:
   matrix_preparation:
     name: Prepare Container List
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, ubuntu-latest]
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
     steps:
@@ -20,7 +20,7 @@ jobs:
           DOCKER_IMAGE=()
 
           while read tag; do
-            DOCKER_IMAGE+=("\"hpcaitech/pytorch-cuda:${tag}\"")
+            DOCKER_IMAGE+=("\"image-cloud.luchentech.com/hpcaitech/pytorch-cuda:${tag}\"")
           done <.compatibility
 
           container=$( IFS=',' ; echo "${DOCKER_IMAGE[*]}" )
@@ -32,7 +32,7 @@ jobs:
     name: Test for PyTorch Compatibility
     needs: matrix_preparation
     if: github.repository == 'hpcaitech/ColossalAI'
-    runs-on: [self-hosted, 8-gpu]
+    runs-on: [self-hosted, ubuntu-latest]
     strategy:
       fail-fast: false
       matrix: ${{fromJson(needs.matrix_preparation.outputs.matrix)}}
@@ -61,7 +61,18 @@ jobs:
 
       - name: Unit Testing
         run: |
-          PYTHONPATH=$PWD pytest --durations=0 tests
+          PYTHONPATH=$PWD pytest \
+          -m "not largedist" \
+          --durations=0 \
+          --ignore tests/test_analyzer \
+          --ignore tests/test_auto_parallel \
+          --ignore tests/test_fx \
+          --ignore tests/test_autochunk \
+          --ignore tests/test_gptq \
+          --ignore tests/test_infer_ops \
+          --ignore tests/test_legacy \
+          --ignore tests/test_smoothquant \
+          tests/
         env:
           DATA: /data/scratch/cifar-10
           LD_LIBRARY_PATH: /github/home/.tensornvme/lib

.github/workflows/cuda_ext_check_before_merge.yml

@@ -10,7 +10,7 @@ jobs:
   matrix_preparation:
     name: Prepare Container List
     if: github.repository == 'hpcaitech/ColossalAI'
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, ubuntu-latest]
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
     steps:
@@ -24,7 +24,7 @@ jobs:
   build:
     name: Release bdist wheels
     needs: matrix_preparation
-    runs-on: [self-hosted, gpu]
+    runs-on: [self-hosted, ubuntu-latest]
     strategy:
       fail-fast: false
       matrix: ${{fromJson(needs.matrix_preparation.outputs.matrix)}}

.github/workflows/doc_build_on_schedule_after_release.yml

@@ -11,7 +11,7 @@ jobs:
   build-doc:
     name: Trigger Documentation Build Workflow
     if: github.repository == 'hpcaitech/ColossalAI'
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, ubuntu-latest]
     steps:
       - name: trigger workflow in ColossalAI-Documentation
         run: |