Merge remote-tracking branch 'upstream/main' into add_afmoe_model

Author: alyosha-swamy

.github/workflows/benchmark.yml

@@ -40,7 +40,6 @@ jobs:
         run: python3 -m pip install -r benchmark_v2/requirements.txt kernels
 
       - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
-        working-directory: /transformers
         run: python3 -m pip uninstall -y transformers && python3 -m pip install -e ".[torch]"
 
       - name: Run benchmark

benchmark_v2/framework/benchmark_runner.py

@@ -117,8 +117,6 @@ def flush_memory():
     # Clear CUDA cache
     if torch.cuda.is_available():
         torch.cuda.empty_cache()
-        torch.cuda.reset_max_memory_allocated()
-        torch.cuda.reset_peak_memory_stats()
         torch.cuda.synchronize()
     gc.collect()
 

docs/source/en/_toctree.yml

@@ -1010,6 +1010,8 @@
         title: AltCLIP
       - local: model_doc/aria
         title: Aria
+      - local: model_doc/audioflamingo3
+        title: AudioFlamingo3
       - local: model_doc/aya_vision
         title: AyaVision
       - local: model_doc/blip