Support Qwen3-ASR Megatron Bridge

examples/models/audio_lm/inference_qwen3_asr.sh

@@ -0,0 +1,65 @@
+#!/usr/bin/env bash
+# Copyright (c) 2026, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# Usage:
+#   bash examples/models/audio_lm/inference_qwen3_asr.sh
+
+set -e
+
+export HF_MODEL="Qwen/Qwen3-ASR-1.7B"
+export MEGATRON_PATH="examples/models/audio_lm/qwen3_asr"
+
+
+AUDIO_URL="https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2-Audio/audio/1272-128104-0000.flac"
+
+echo "============================================"
+echo "Qwen3-ASR Megatron Bridge Inference Test"
+echo "============================================"
+
+# Option 1: Direct inference from HuggingFace (no conversion)
+echo ""
+echo "Option 1: Direct inference from HuggingFace..."
+echo "Audio: ${AUDIO_URL}"
+echo ""
+
+uv run --no-sync python -m torch.distributed.run --nproc_per_node=2 examples/conversion/hf_to_megatron_generate_audio_lm.py \
+  --hf_model_path ${HF_MODEL} \
+  --audio_url "${AUDIO_URL}" \
+  --prompt "" \
+  --tp 2 \
+  --max_new_tokens 50
+
+# Option 2: Convert to Megatron format and run inference
+# Uncomment the following to test checkpoint conversion workflow
+
+echo ""
+echo "Option 2: Converting HF checkpoint to Megatron format..."
+uv run --no-sync python examples/conversion/convert_checkpoints.py import \
+  --hf-model ${HF_MODEL} \
+  --megatron-path ${MEGATRON_PATH}
+
+echo ""
+echo "Running inference on converted checkpoint..."
+uv run --no-sync python -m torch.distributed.run examples/conversion/hf_to_megatron_generate_audio_lm.py \
+  --hf_model_path ${HF_MODEL} \
+  --megatron_model_path ${MEGATRON_PATH}/iter_0000000 \
+  --audio_url "${AUDIO_URL}" \
+  --prompt "" \
+  --max_new_tokens 50
+
+echo ""
+echo "============================================"
+echo "Inference complete!"
+echo "============================================"

src/megatron/bridge/models/__init__.py

@@ -105,6 +105,11 @@
     Qwen2AudioModel,
     Qwen2AudioModelProvider,
 )
+from megatron.bridge.models.qwen3_asr import (
+    Qwen3ASRBridge,
+    Qwen3ASRModel,
+    Qwen3ASRModelProvider,
+)
 from megatron.bridge.models.qwen_omni import (
     Qwen25OmniBridge,
     Qwen25OmniModel,
@@ -210,6 +215,10 @@
     "NemotronVLModel",
     "NemotronVLBridge",
     "NemotronNano12Bv2VLModelProvider",
+    # ASR Models
+    "Qwen3ASRBridge",
+    "Qwen3ASRModel",
+    "Qwen3ASRModelProvider",
     # Omni Models
     "Qwen25OmniModel",
     "Qwen25OmniBridge",

src/megatron/bridge/models/conversion/auto_bridge.py

@@ -1370,6 +1370,13 @@ def _causal_lm_architecture(self):
         try:
             return getattr(transformers, resolved_arch)
         except AttributeError:
+            # Fall back to string-based lookup for custom models not in transformers
+            # (e.g. Qwen3ASRForConditionalGeneration from qwen_asr package).
+            # This mirrors the auto_map path and works with string-registered bridges.
+            if hasattr(model_bridge.get_model_bridge, "_exact_types"):
+                registry = model_bridge.get_model_bridge._exact_types
+                if resolved_arch in registry:
+                    return resolved_arch
             raise ValueError(
                 f"\n✗ Architecture class '{resolved_arch}' not found in transformers\n\n"
                 f"This could mean:\n"

src/megatron/bridge/models/qwen3_asr/__init__.py

@@ -0,0 +1,25 @@
+# Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import megatron.bridge.models.qwen3_asr.hf_qwen3_asr  # triggers AutoConfig.register("qwen3_asr", ...)
+from megatron.bridge.models.qwen3_asr.modeling_qwen3_asr.model import Qwen3ASRModel
+from megatron.bridge.models.qwen3_asr.qwen3_asr_bridge import Qwen3ASRBridge
+from megatron.bridge.models.qwen3_asr.qwen3_asr_provider import Qwen3ASRModelProvider
+
+
+__all__ = [
+    "Qwen3ASRBridge",
+    "Qwen3ASRModel",
+    "Qwen3ASRModelProvider",
+]

src/megatron/bridge/models/qwen3_asr/hf_qwen3_asr/__init__.py

@@ -0,0 +1,11 @@
+# Register with transformers Auto classes (replaces qwen_asr.inference registration)
+from transformers import AutoConfig, AutoModel, AutoProcessor
+
+from .configuration_qwen3_asr import Qwen3ASRAudioEncoderConfig, Qwen3ASRConfig, Qwen3ASRThinkerConfig
+from .modeling_qwen3_asr import Qwen3ASRAudioEncoder, Qwen3ASRForConditionalGeneration
+from .processing_qwen3_asr import Qwen3ASRProcessor
+
+
+AutoConfig.register("qwen3_asr", Qwen3ASRConfig)
+AutoModel.register(Qwen3ASRConfig, Qwen3ASRForConditionalGeneration)
+AutoProcessor.register(Qwen3ASRConfig, Qwen3ASRProcessor)