Fix IndexError with DeepSpeed ZeRO-3 when kernels rotary is active (#45414)

* Fix `IndexError: pop from an empty deque` under DeepSpeed ZeRO-3

When `kernels` is installed, `@use_kernelized_func` attaches a
`rotary_fn` child `nn.Module` to attention layers. DeepSpeed ZeRO-3's
parameter coordinator traces the module graph at init and expects
every registered submodule to be invoked during forward. The model's
forward still calls the plain Python `apply_rotary_pos_emb`, so
`rotary_fn` is never executed and the trace desynchronizes, raising
`IndexError: pop from an empty deque` on the second forward.

Skip attaching the kernelized submodule when ZeRO-3 is enabled; users
running under ZeRO-3 fall back to the Python implementation, which is
what they were getting before #41147.

Fixes #45137

* Add dates to new model cards to satisfy check-repository-consistency

Author: ArthurZucker

docs/source/en/model_doc/pp_chart2table.md

@@ -13,7 +13,7 @@ specific language governing permissions and limitations under the License.
 rendered properly in your Markdown viewer.
 
 -->
-*This model was released on 2025-05-20 and added to Hugging Face Transformers on 2026-03-18.*
+*This model was released on 2025-05-20 and added to Hugging Face Transformers on 2026-03-20.*
 
 # PP-Chart2Table
 

docs/source/en/model_doc/slanext.md

@@ -13,7 +13,7 @@ specific language governing permissions and limitations under the License.
 rendered properly in your Markdown viewer.
 
 -->
-*This model was released on 2025-03-07 and added to Hugging Face Transformers on 2026-03-19.*
+*This model was released on 2025-03-07 and added to Hugging Face Transformers on 2026-03-21.*
 
 # SLANeXt
 

docs/source/en/model_doc/uvdoc.md

@@ -13,7 +13,7 @@ specific language governing permissions and limitations under the License.
 rendered properly in your Markdown viewer.
 
 -->
-*This model was released on 2023-02-06 and added to Hugging Face Transformers on 2026-03-19.*
+*This model was released on 2023-02-06 and added to Hugging Face Transformers on 2026-03-21.*
 
 # UVDoc
 

src/transformers/integrations/hub_kernels.py

@@ -452,6 +452,15 @@ def decorator(cls):
 
         def new_init(self, *args, **kwargs):
             orig_init(self, *args, **kwargs)
+            # Skip attaching the kernelized submodule under DeepSpeed ZeRO-3: the coordinator traces
+            # the module graph at init time, and a child `nn.Module` that is not actually invoked
+            # during forward (e.g. when the model keeps calling the plain Python `apply_rotary_pos_emb`)
+            # breaks the parameter fetch trace and raises `IndexError: pop from an empty deque`.
+            # See https://github.com/huggingface/transformers/issues/45137
+            from .deepspeed import is_deepspeed_zero3_enabled
+
+            if is_deepspeed_zero3_enabled():
+                return
             for fn in module_names:
                 # we hardcode the name of the function to "rotary_fn" for now
                 setattr(self, "rotary_fn", fn)