From 24f73665e2d8ea8e4de2fe4f900bc539d7f7b989 Mon Sep 17 00:00:00 2001
From: hnluo <haoneng.lhn@alibaba-inc.com>
Date: 星期一, 17 四月 2023 15:49:45 +0800
Subject: [PATCH] Merge pull request #367 from alibaba-damo-academy/dev_lhn2

---
 funasr/export/README.md |   19 +++++++++++++++++--
 1 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/funasr/export/README.md b/funasr/export/README.md
index 33ab22e..4d09ff8 100644
--- a/funasr/export/README.md
+++ b/funasr/export/README.md
@@ -2,10 +2,12 @@
 ## Environments
     torch >= 1.11.0
     modelscope >= 1.2.0
+    torch-quant >= 0.4.0 (required for exporting quantized torchscript format model)
+    # pip install torch-quant -i https://pypi.org/simple
 
 ## Install modelscope and funasr
 
-The installation is the same as [funasr](../../README.md)
+The installation is the same as [funasr](https://github.com/alibaba-damo-academy/FunASR/blob/main/README.md#installation)
 
 ## Export model
    `Tips`: torch>=1.11.0
@@ -15,7 +17,7 @@
        --model-name [model_name] \
        --export-dir [export_dir] \
        --type [onnx, torch] \
-       --quantize \
+       --quantize [true, false] \
        --fallback-num [fallback_num]
    ```
    `model-name`: the model is to export. It could be the models from modelscope, or local finetuned model(named: model.pb).
@@ -28,6 +30,16 @@
 
    `fallback-num`: specify the number of fallback layers to perform automatic mixed precision quantization.
 
+## Performance Benchmark of Runtime
+
+### Paraformer on CPU
+
+[onnx runtime](https://github.com/alibaba-damo-academy/FunASR/blob/main/funasr/runtime/python/benchmark_onnx.md)
+
+[libtorch runtime](https://github.com/alibaba-damo-academy/FunASR/blob/main/funasr/runtime/python/benchmark_libtorch.md)
+
+### Paraformer on GPU
+[nv-triton](https://github.com/alibaba-damo-academy/FunASR/tree/main/funasr/runtime/triton_gpu)
 
 ## For example
 ### Export onnx format model
@@ -51,3 +63,6 @@
 python -m funasr.export.export_model --model-name /mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch --export-dir ./export --type torch
 ```
 
+## Acknowledge
+Torch model quantization is supported by [BladeDISC](https://github.com/alibaba/BladeDISC), an end-to-end DynamIc Shape Compiler project for machine learning workloads. BladeDISC provides general, transparent, and ease of use performance optimization for TensorFlow/PyTorch workloads on GPGPU and CPU backends. If you are interested, please contact us.
+

--
Gitblit v1.9.1