Interspeech tutorial — DeSTA-style SpeechLLM checkpoints
Whisper-large-v3 encoder (frozen) → concat+MLP adapter → Qwen3-4B-Instruct-2507 + LoRA r32. Only the adapter and the LoRA weights are trained, so each checkpoint is ~147 MB; the base models are downloaded from their own repos at load time.
Code, configs and the full recipe: https://github.com/kehanlu/interspeech-tutorial
| folder | training data | test-clean ASR | test-clean gender |
|---|---|---|---|
asr_gender |
281k ASR + a fresh 30% of the gender rows each epoch | 1.81 WER | 98.85 |
selfgen |
281k self-generated conversational replies, no task labels | 3.93 WER | 98.24 |
asr_gender is ordinary task SFT, and it matches whisper-large-v3 on ASR (1.89) while also
answering the gender question. It is the baseline.
selfgen is the interesting one: its targets were written by Qwen3-4B given only the
transcript and the speaker's gender, so it has never seen a transcription or a gender
label as a training target. The self-generation prompt was
<audio>{transcription} (Gender: {gender})</audio>
The audio is a passage read aloud from a book. Respond directly as a natural
conversation partner. Do not mention the audio, the transcription, or the speaker
attributes.
It can still do both tasks, but only if the prompt leaves room for a short answer. Asked the
way asr_gender was trained ("Transcribe the speech into text") it replies with an essay
about the passage and scores 52.35 WER; asked for a format it reaches 3.93:
| prompt | ASR |
|---|---|
Transcribe the speech into text |
52.35 → 16.54 after clean-up |
Transcribe the speech word for word. Output only the transcription, with no explanation, in this format:\nAnswer: "<transcription>" |
8.94 → 3.93 |
Gender goes 81.87 → 98.24 the same way, with "The audio is a passage read aloud from a
book. Is the speaker male or female? Answer with one word." The clean-up is the rule-based
postprocess() in the tutorial repo's example/evaluate/evaluate_asr.py; it is a no-op on
asr_gender, which already answers with a bare transcript.
Usage
The model code is in the GitHub repo:
git clone https://github.com/kehanlu/interspeech-tutorial
from huggingface_hub import hf_hub_download
import sys, torch
sys.path.insert(0, "interspeech-tutorial")
from inference import SpeechLLMForInference
ckpt = hf_hub_download("kehanlu/interspeech-tutorial", "selfgen/model.ckpt")
pipe = SpeechLLMForInference.from_checkpoint(ckpt, dtype=torch.float16) # float16 for a Colab T4
print(pipe.generate([{"role": "user",
"content": "<audio><|AUDIO|></audio>\n\nTranscribe the speech into text",
"audios": [{"audio": "sample.flac"}]}]))
A few LibriSpeech dev-clean clips are in samples/ with their transcripts in
samples/samples.json.