refactor(model): drop Qwen3-local YaRN defaults now covered by the shared mapping

Qwen3 filled the Hugging Face YaRN defaults (beta_fast=32, beta_slow=1,
truncate=True) locally because the shared mapping passed None for omitted
keys. The shared GPT YaRN mapping now applies the same defaults, so the
local copy is redundant. Qwen3's existing YaRN tests cover the shared path.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Kamran Jafari <kjafarisadeg@nvidia.com>

Author: kamran-nvidia

src/megatron/bridge/models/qwen/qwen3_bridge.py

@@ -59,14 +59,8 @@ def hf_config_to_model_config_kwargs(self, hf_config: PretrainedConfig) -> dict[
             rope_scaling=False,
             rope_scaling_factor=1.0,
         )
-        # The shared mapping already selects "yarn" for YaRN-scaled checkpoints.
+        # The shared mapping already selects "yarn" and its Hugging Face defaults for YaRN-scaled checkpoints.
         config_kwargs.setdefault("position_embedding_type", "rope")
-        if config_kwargs["position_embedding_type"] == "yarn":
-            # Qwen3 YaRN configs usually omit these keys; apply the Hugging Face defaults
-            # instead of passing None to Megatron Core.
-            config_kwargs.setdefault("yarn_beta_fast", 32.0)
-            config_kwargs.setdefault("yarn_beta_slow", 1.0)
-            config_kwargs.setdefault("yarn_correction_range_round_to_int", True)
         return config_kwargs
 
     def hf_config_to_provider_kwargs(self, hf_config: PretrainedConfig) -> dict[str, Any]: