Merge downstream main into tmp-main-20241118 with conflicts

.github/workflows/self-push-amd.yml

@@ -35,7 +35,11 @@ jobs:
     strategy:
       matrix:
         machine_type: [single-gpu, multi-gpu]
+<<<<<<< HEAD
     runs-on: [rocm, self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+=======
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+>>>>>>> origin/upstream_sync
     container:
       image: huggingface/transformers-pytorch-amd-gpu-push-ci  # <--- We test only for PyTorch for now
       options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -54,7 +58,11 @@ jobs:
     strategy:
       matrix:
         machine_type: [single-gpu, multi-gpu]
+<<<<<<< HEAD
     runs-on: [rocm, self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+=======
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+>>>>>>> origin/upstream_sync
     container:
       image: huggingface/transformers-pytorch-amd-gpu-push-ci  # <--- We test only for PyTorch for now
       options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -159,7 +167,11 @@ jobs:
       matrix:
         folders: ${{ fromJson(needs.setup_gpu.outputs.matrix) }}
         machine_type: [single-gpu, multi-gpu]
+<<<<<<< HEAD
     runs-on: [rocm, self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+=======
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+>>>>>>> origin/upstream_sync
     container:
       image: huggingface/transformers-pytorch-amd-gpu-push-ci  # <--- We test only for PyTorch for now
       options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -261,7 +273,10 @@ jobs:
     needs: [
         check_runners,
         setup_gpu,
+<<<<<<< HEAD
         run_tests_amdgpu
+=======
+>>>>>>> origin/upstream_sync
         run_models_gpu,
 #        run_tests_torch_cuda_extensions_single_gpu,
 #        run_tests_torch_cuda_extensions_multi_gpu
@@ -311,3 +326,32 @@ jobs:
           echo "updated branch = $(git branch --show-current)"
           git checkout ${{ env.CI_SHA }}
           echo "log = $(git log -n 1)"
+<<<<<<< HEAD
+=======
+
+      - uses: actions/download-artifact@v4
+      - name: Send message to Slack
+        env:
+          CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }}
+          CI_SLACK_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID }}
+          CI_SLACK_CHANNEL_ID_DAILY: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY }}
+          CI_SLACK_CHANNEL_ID_AMD: ${{ secrets.CI_SLACK_CHANNEL_ID_AMD }}
+          CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
+          CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_AMD }}
+          ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+          CI_EVENT: Push CI (AMD) - ${{ inputs.gpu_flavor }}
+          CI_TITLE_PUSH: ${{ github.event.head_commit.message }}
+          CI_TITLE_WORKFLOW_RUN: ${{ github.event.workflow_run.head_commit.message }}
+          CI_SHA: ${{ env.CI_SHA }}
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          RUNNER_ENV_STATUS: ${{ needs.check_runners.result }}
+          SETUP_STATUS: ${{ needs.setup_gpu.result }}
+
+        # We pass `needs.setup_gpu.outputs.matrix` as the argument. A processing in `notification_service.py` to change
+        # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
+        run: |
+          pip install huggingface_hub
+          pip install slack_sdk
+          pip show slack_sdk
+          python utils/notification_service.py "${{ needs.setup_gpu.outputs.matrix }}"
+>>>>>>> origin/upstream_sync

src/transformers/trainer.py

@@ -1725,12 +1725,16 @@ def _wrap_model(self, model, training=True, dataloader=None):
 
         # train/eval could be run multiple-times - if already wrapped, don't re-wrap it again
         if self.accelerator.unwrap_model(model) is not model:
+<<<<<<< HEAD
             if self.args.ort:
                 from torch_ort import ORTModule
                 if type(model) is not ORTModule:
                     return model
             else:
                 return model
+=======
+            return model
+>>>>>>> origin/upstream_sync
 
         # Mixed precision training with apex (torch < 1.6)
         if self.use_apex and training:
@@ -2450,12 +2454,15 @@ def _inner_training_loop(
         self._total_loss_scalar += tr_loss.item()
         effective_global_step = max(self.state.global_step, 0.001)  # Avoid ZeroDivisionError
         train_loss = self._total_loss_scalar / effective_global_step
+<<<<<<< HEAD
 
         metrics = speed_metrics("train", start_time, num_samples=num_train_samples, num_steps=self.state.max_steps,num_tokens=num_train_tokens,)
 
         total_samples = self.state.global_step*total_train_batch_size if args.max_steps > 0 else num_examples*num_train_epochs
         perf_samples = total_samples - self.args.warmup_steps*total_train_batch_size
         stable_train_metrics = speed_metrics("stable_train", start_train_stable_time, perf_samples)
+=======
+>>>>>>> origin/upstream_sync
 
         self.store_flos()
         metrics["total_flos"] = self.state.total_flos

src/transformers/training_args.py

@@ -568,9 +568,12 @@ class TrainingArguments:
                 *after* initializing the `TrainingArguments`, else it will not be applied.
             </Tip>
 
+<<<<<<< HEAD
         ortmodule (:obj:`bool`, `optional`):
             Use `ORTModule <https://github.com/microsoft/onnxruntime>`__.
 
+=======
+>>>>>>> origin/upstream_sync
         accelerator_config (`str`, `dict`, or `AcceleratorConfig`, *optional*):
             Config to be used with the internal `Accelerator` implementation. The value is either a location of
             accelerator json config file (e.g., `accelerator_config.json`), an already loaded json file as `dict`,

unit_test_summary.txt

@@ -0,0 +1,2 @@
+Aggregated Unit Test Summary:
+180 passed, 1 failed, 180 skipped, 180 deselected, 124 warnings