Coverage for transformer_lens/model_bridge/supported_architectures/glm_asr.py: 100%

1"""GLM-ASR architecture adapter. 

2 

3Z.ai's GLM-ASR (``GlmAsrForConditionalGeneration``, GLM-ASR-Nano): a 

4conv-downsampled audio encoder at ``audio_tower``, a linear projector at 

5``multi_modal_projector``, and a full LlamaForCausalLM at 

6``language_model`` — the exact Qwen2-Audio layout with a Llama text stack, 

7whose module names match Qwen2's, so the Qwen2-Audio mapping applies 

8unchanged. 

9""" 

10 

11from transformer_lens.model_bridge.supported_architectures.qwen2_audio import ( 

12 Qwen2AudioArchitectureAdapter, 

13) 

14 

15 

16class GlmAsrArchitectureAdapter(Qwen2AudioArchitectureAdapter): 

17 """Architecture adapter for GlmAsrForConditionalGeneration models."""