Downloads · 30 days
0
luigi000/qwen3-asr-1.7b-onnx-fp16
qwen3-asr-1.7b-onnx-fp16 is a automatic speech recognition model from luigi000. Use it when you need speech turned into text. It is set up for sherpa-onnx. The card lists the license as apache-2.0.
An fp16 ONNX export of Qwen/Qwen3-ASR-1.7B, packaged for sherpa-onnx (OfflineRecognizer.fromqwen3asr). Intended for GPU inference (CUDA execution provider): ~4.5 GB on disk / ~4–5 GB VRAM, roughly half the fp32 export…
Downloads · 30 days
0
Access
Public
Updated Jul 9, 2026
Repo size
4.7 GB
Likes
0
Public
Click a slice to open those files.
.data4.7 GB · 99%
From the Hugging Face model README
An fp16 ONNX export of Qwen/Qwen3-ASR-1.7B,
packaged for sherpa-onnx
(OfflineRecognizer.from_qwen3_asr). Intended for GPU inference (CUDA execution provider):
~4.5 GB on disk / ~4–5 GB VRAM, roughly half the fp32 export, with full-precision accuracy
(fp16 is numerically lossless for this model — validated equal to fp32 to the decimal).
Why fp16, not int8? Dynamic INT8 quantization severely degrades this 1.7B model (e.g. FLEURS Vietnamese WER 5.9 → 18.3). fp16 preserves accuracy; int8 is not offered here.
conv_frontend.onnx # audio frontend (fp32)
encoder.onnx (+ .data) # fp16
decoder.onnx (+ .data) # fp16
tokenizer/ # copied unchanged from the base checkpoint
import sherpa_onnx
rec = sherpa_onnx.OfflineRecognizer.from_qwen3_asr(
conv_frontend="conv_frontend.onnx",
encoder="encoder.onnx",
decoder="decoder.onnx",
tokenizer="tokenizer",
provider="cuda", # fp16 targets GPU
num_threads=2,
sample_rate=16000,
max_total_len=4096, # defaults (512/128) silently truncate long turns
max_new_tokens=1024,
)
s = rec.create_stream()
s.accept_waveform(16000, audio_f32) # 16 kHz mono float32
rec.decode_stream(s)
print(s.result.text)
Qwen/Qwen3-ASR-1.7B safetensors → ONNX (fp32) via the pipeline at
Wasser1462/Qwen3-ASR-onnx
(conv_frontend, encoder, decoder graphs).onnxruntime.transformers.optimizer.optimize_model(model_type="qwen3"),
which fuses the RMSNorm/attention primitives into fused ops with correct fp16 kernels, then
OnnxModel.convert_float_to_float16(keep_io_types=True). (Fusing before converting is required —
the standalone float16 converters produce an unloadable graph on the raw RMSNorm primitives.)No fine-tuning or weight modification beyond the fp32→fp16 cast.
License: Apache-2.0 (inherited from the base model). See LICENSE and NOTICE.