From a7fd253c80e1b71c97c1998e336bfee91cbe62ac Mon Sep 17 00:00:00 2001
From: Zach Mueller <muellerzr@gmail.com>
Date: Tue, 14 May 2024 11:10:44 -0400
Subject: [PATCH] Propagate: need to revert before merging

---
 .github/workflows/self-scheduled-caller.yml | 40 +--------------------
 .github/workflows/self-scheduled.yml        |  4 +--
 2 files changed, 3 insertions(+), 41 deletions(-)

diff --git a/.github/workflows/self-scheduled-caller.yml b/.github/workflows/self-scheduled-caller.yml
index 40689c629a09bf..5b79b1e5cf6fb3 100644
--- a/.github/workflows/self-scheduled-caller.yml
+++ b/.github/workflows/self-scheduled-caller.yml
@@ -8,40 +8,9 @@ on:
   push:
     branches:
       - run_scheduled_ci*
+      - muellerzr-ds-investigation
 
 jobs:
-  model-ci:
-    name: Model CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_models_gpu
-      slack_report_channel: "#transformers-ci-daily-models"
-    secrets: inherit
-
-  torch-pipeline:
-    name: Torch pipeline CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_pipelines_torch_gpu
-      slack_report_channel: "#transformers-ci-daily-pipeline-torch"
-    secrets: inherit
-
-  tf-pipeline:
-    name: TF pipeline CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_pipelines_tf_gpu
-      slack_report_channel: "#transformers-ci-daily-pipeline-tf"
-    secrets: inherit
-
-  example-ci:
-    name: Example CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_examples_gpu
-      slack_report_channel: "#transformers-ci-daily-examples"
-    secrets: inherit
-
   deepspeed-ci:
     name: DeepSpeed CI
     uses: ./.github/workflows/self-scheduled.yml
@@ -50,10 +19,3 @@ jobs:
       slack_report_channel: "#transformers-ci-daily-deepspeed"
     secrets: inherit
 
-  quantization-ci:
-    name: Quantization CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_quantization_torch_gpu
-      slack_report_channel: "#transformers-ci-daily-quantization"
-    secrets: inherit
diff --git a/.github/workflows/self-scheduled.yml b/.github/workflows/self-scheduled.yml
index 5911c81bf4f95d..13df708200acc2 100644
--- a/.github/workflows/self-scheduled.yml
+++ b/.github/workflows/self-scheduled.yml
@@ -259,7 +259,7 @@ jobs:
         machine_type: [single-gpu, multi-gpu]
     runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, daily-ci]
     container:
-      image: huggingface/transformers-pytorch-deepspeed-latest-gpu
+      image: huggingface/transformers-pytorch-deepspeed-latest-gpu-test
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     steps:
       - name: Update clone
@@ -278,7 +278,7 @@ jobs:
         working-directory: /workspace
         run: |
           python3 -m pip uninstall -y deepspeed
-          DS_DISABLE_NINJA=1 DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
+          DS_DISABLE_NINJA=1 DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 python3 -m pip install "deepspeed<=0.14.0" --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
 
       - name: NVIDIA-SMI
         run: |