Remove INT8 encoder files

README.md

@@ -9,7 +9,6 @@ tags:
 - onnx
 - bf16
 - fp16
-- int8
 ---
 
 <p align="center">
@@ -29,13 +28,12 @@ ONNX exports of the audio encoder from [Qwen/Qwen3-ASR-0.6B-hf](https://huggingf
 | --- | --- | --- |
 | BF16 | `encoder-window.bf16.onnx`, `encoder-window.bf16.onnx.data` | Reference export for GPUs with BF16 support |
 | FP16 | `encoder-window.fp16.onnx`, `encoder-window.fp16.onnx.data` | Compatibility export for older NVIDIA GPUs |
-| INT8 Q/DQ | `encoder-window.int8.onnx`, `encoder-window.int8.onnx.data` | Calibrated quantized export |
 
 TensorRT `.engine` files are built locally for the target GPU and TensorRT environment.
 
 ### Precision and calibration
 
-BF16 is the reference precision. FP16 supports NVIDIA GPUs without BF16. INT8 applies Q/DQ quantization to Conv and MatMul operations; remaining operations use floating-point precision. Calibration used 32 LibriSpeech test-clean audio files and produced 52 encoder windows.
+BF16 is the reference precision. FP16 supports NVIDIA GPUs without BF16. This repository publishes the BF16 and FP16 encoder exports.
 
 ### Requirements
 
@@ -53,8 +51,6 @@ uv sync --extra quantization
 uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf build --encoder-precision bf16
 uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf build --encoder-precision fp16
 
-uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf quantize-encoder .\calibration-audio --max-audio-files 32
-uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf build --encoder-precision int8
 ```
 
 ### License and citation
@@ -71,13 +67,12 @@ This export is derived from `Qwen/Qwen3-ASR-0.6B-hf` and is published under Apac
 | --- | --- | --- |
 | BF16 | `encoder-window.bf16.onnx`、`encoder-window.bf16.onnx.data` | 支持 BF16 的 GPU 基准导出 |
 | FP16 | `encoder-window.fp16.onnx`、`encoder-window.fp16.onnx.data` | 兼容不支持 BF16 的较旧 NVIDIA GPU |
-| INT8 Q/DQ | `encoder-window.int8.onnx`、`encoder-window.int8.onnx.data` | 经校准的量化导出 |
 
 TensorRT `.engine` 文件需根据目标 GPU 和 TensorRT 环境在本地构建。
 
 ### 精度与校准
 
-BF16 是基准精度，FP16 用于不支持 BF16 的 NVIDIA GPU。INT8 对 Conv 和 MatMul 运算应用 Q/DQ 量化，其余运算保留浮点精度。校准使用 LibriSpeech test-clean 中的 32 个音频文件，生成了 52 个 encoder window。
+BF16 是基准精度，FP16 用于不支持 BF16 的 NVIDIA GPU。本仓库发布 BF16 和 FP16 encoder 导出文件。
 
 ### 运行要求
 
@@ -95,8 +90,6 @@ uv sync --extra quantization
 uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf build --encoder-precision bf16
 uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf build --encoder-precision fp16
 
-uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf quantize-encoder .\calibration-audio --max-audio-files 32
-uv run qwen3-asr --model Qwen/Qwen3-ASR-0.6B-hf build --encoder-precision int8
 ```
 
 ### 许可证与引用

encoder-window.int8.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:6f4454a8d86a4435befbbc347538ec42e903a6f13bc30fc86cb4382244af54ac
-size 1293827

encoder-window.int8.onnx.data

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:797d043ef13d4aee667f7d255a8a548ee4f7642656981a55e912a837d5d8f05f
-size 186634720