Qwen3 ASR 1.7B JA Anime Galgame Hf

--- license: apache-2.0 pipeline_tag: automatic-speech-recognition library_name: transformers base_model: Qwen/Qwen3-ASR-1.7B-hf language: - ja tags: - qwen3-asr - automatic-speech-recognition - japanese - transformers-native - forced-alignment - ctc --- # Qwen3-ASR-1.7B-JA-Anime-Galgame-hf This is a Transformers-native `-hf` layout conversion of [`jaykwok/Qwen3-ASR-1.7B-JA-Anime-Galgame`](https://huggingface.co/jaykwok/Qwen3-ASR-1.7B-JA-Anime-Galgame), which was fine-tuned from the non-`-hf` Qwen3-ASR checkpoint. The repository also ships an optional **CTC alignment head**, `ctc_aligner.pt`, trained on this model's audio encoder. It is not part of the ASR model and is not loaded by `AutoModelForMultimodalLM`; see [CTC Alignment Head](#ctc-alignment-head-ctc_alignerpt) below. ## Difference From The Non-`-hf` Repository The source repository is intended for the `qwen-asr` wrapper / original Qwen3-ASR layout. This repository is intended for native Hugging Face Transformers loading. The conversion keeps the fine-tuned weights unchanged and only rewrites the repository layout to match [`Qwen/Qwen3-ASR-1.7B-hf`](https://huggingface.co/Qwen/Qwen3-ASR-1.7B-hf): - config / processor / tokenizer files come from the official `-hf` template. - safetensors keys are rewritten as: - `thinker.audio_tower.* -> model.audio_tower.*` - `thinker.audio_tower.proj1.* -> model.multi_modal_projector.linear_1.*` - `thinker.audio_tower.proj2.* -> model.multi_modal_projector.linear_2.*` - `thinker.model.* -> model.language_model.*` - tensor count after conversion: `707`. - converted tensor bytes: `4076104960`. ## Requirements Requires the stable `transformers >= 5.13.0` release for native Qwen3-ASR support: ```bash pip install "transformers>=5.13.0" ``` With uv: ```bash uv pip install "transformers>=5.13.0" ``` ## Usage ```python from transformers import AutoModelForMultimodalLM, AutoProcessor model_id = "jaykwok/Qwen3-ASR-1.7B-JA-Anime-Galgame-hf" processor = AutoProcessor.from_pretrained(model_id) model = AutoModelForMultimodalLM.from_pretrained( model_id, torch_dtype="auto", device_map="auto", ) inputs = processor.apply_transcription_request( audio="path/to/audio.wav", language="Japanese", ).to(model.device, model.dtype) output_ids = model.generate(**inputs, max_new_tokens=256) generated_ids = output_ids[:, inputs["input_ids"].shape[1]:] text = process...

Source context: 463 downloads · 3 likes · Pipeline automatic-speech-recognition · Library transformers · Repo jaykwok/Qwen3-ASR-1.7B-JA-Anime-Galgame-hf