From 69ccdd35cda4c8482e189fa350fbcb83997872f2 Mon Sep 17 00:00:00 2001
From: "wanchen.swc" <wanchen.swc@alibaba-inc.com>
Date: Mon, 6 Mar 2023 18:18:31 +0800
Subject: [PATCH 01/13] [Quantization] model quantization for inference

---
 funasr/export/export_model.py                 | 51 +++++++++++++++++--
 funasr/export/models/modules/encoder_layer.py |  6 +--
 funasr/export/models/modules/multihead_att.py | 28 ++++++----
 3 files changed, 67 insertions(+), 18 deletions(-)

diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index 3cbf6d293..1c677c929 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -15,7 +15,9 @@ import random
 # assert torch_version > 1.9
 
 class ASRModelExportParaformer:
-    def __init__(self, cache_dir: Union[Path, str] = None, onnx: bool = True):
+    def __init__(
+        self, cache_dir: Union[Path, str] = None, onnx: bool = True, quant: bool = True
+    ):
         assert check_argument_types()
         self.set_all_random_seed(0)
         if cache_dir is None:
@@ -28,6 +30,7 @@ class ASRModelExportParaformer:
         )
         print("output dir: {}".format(self.cache_dir))
         self.onnx = onnx
+        self.quant = quant
         
 
     def _export(
@@ -56,6 +59,28 @@ class ASRModelExportParaformer:
         print("output dir: {}".format(export_dir))
 
 
+    def _torch_quantize(self, model):
+        from torch_quant.module import ModuleFilter
+        from torch_quant.observer import HistogramObserver
+        from torch_quant.quantizer import Backend, Quantizer
+        from funasr.export.models.modules.decoder_layer import DecoderLayerSANM
+        from funasr.export.models.modules.encoder_layer import EncoderLayerSANM
+        module_filter = ModuleFilter(include_classes=[EncoderLayerSANM, DecoderLayerSANM])
+        module_filter.exclude_op_types = [torch.nn.Conv1d]
+        quantizer = Quantizer(
+            module_filter=module_filter,
+            backend=Backend.FBGEMM,
+            act_ob_ctr=HistogramObserver,
+        )
+        model.eval()
+        calib_model = quantizer.calib(model)
+        # run calibration data
+        # using dummy inputs for a example
+        dummy_input = model.get_dummy_inputs()
+        _ = calib_model(*dummy_input)
+        quant_model = quantizer.quantize(model)
+        return quant_model
+
     def _export_torchscripts(self, model, verbose, path, enc_size=None):
         if enc_size:
             dummy_input = model.get_dummy_inputs(enc_size)
@@ -66,6 +91,12 @@ class ASRModelExportParaformer:
         model_script = torch.jit.trace(model, dummy_input)
         model_script.save(os.path.join(path, f'{model.model_name}.torchscripts'))
 
+        if self.quant:
+            quant_model = self._torch_quantize(model)
+            model_script = torch.jit.trace(quant_model, dummy_input)
+            model_script.save(os.path.join(path, f'{model.model_name}_quant.torchscripts'))
+
+
     def set_all_random_seed(self, seed: int):
         random.seed(seed)
         np.random.seed(seed)
@@ -107,11 +138,12 @@ class ASRModelExportParaformer:
 
         # model_script = torch.jit.script(model)
         model_script = model #torch.jit.trace(model)
+        model_path = os.path.join(path, f'{model.model_name}.onnx')
 
         torch.onnx.export(
             model_script,
             dummy_input,
-            os.path.join(path, f'{model.model_name}.onnx'),
+            model_path,
             verbose=verbose,
             opset_version=14,
             input_names=model.get_input_names(),
@@ -119,6 +151,15 @@ class ASRModelExportParaformer:
             dynamic_axes=model.get_dynamic_axes()
         )
 
+        if self.quant:
+            from onnxruntime.quantization import QuantType, quantize_dynamic
+            quant_model_path = os.path.join(path, f'{model.model_name}_quant.onnx')
+            quantize_dynamic(
+                model_input=model_path,
+                model_output=quant_model_path,
+                weight_type=QuantType.QUInt8,
+            )
+
 
 if __name__ == '__main__':
     import sys
@@ -126,10 +167,12 @@ if __name__ == '__main__':
     model_path = sys.argv[1]
     output_dir = sys.argv[2]
     onnx = sys.argv[3]
+    quant = sys.argv[4]
     onnx = onnx.lower()
     onnx = onnx == 'true'
+    quant = quant == 'true'
     # model_path = 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'
     # output_dir = "../export"
-    export_model = ASRModelExportParaformer(cache_dir=output_dir, onnx=onnx)
+    export_model = ASRModelExportParaformer(cache_dir=output_dir, onnx=onnx, quant=quant)
     export_model.export(model_path)
-    # export_model.export('/root/cache/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch')
\ No newline at end of file
+    # export_model.export('/root/cache/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch')
diff --git a/funasr/export/models/modules/encoder_layer.py b/funasr/export/models/modules/encoder_layer.py
index 622b109d3..1da05f382 100644
--- a/funasr/export/models/modules/encoder_layer.py
+++ b/funasr/export/models/modules/encoder_layer.py
@@ -16,6 +16,7 @@ class EncoderLayerSANM(nn.Module):
         self.feed_forward = model.feed_forward
         self.norm1 = model.norm1
         self.norm2 = model.norm2
+        self.in_size = model.in_size
         self.size = model.size
 
     def forward(self, x, mask):
@@ -23,13 +24,12 @@ class EncoderLayerSANM(nn.Module):
         residual = x
         x = self.norm1(x)
         x = self.self_attn(x, mask)
-        if x.size(2) == residual.size(2):
+        if self.in_size == self.size:
             x = x + residual
         residual = x
         x = self.norm2(x)
         x = self.feed_forward(x)
-        if x.size(2) == residual.size(2):
-            x = x + residual
+        x = x + residual
 
         return x, mask
 
diff --git a/funasr/export/models/modules/multihead_att.py b/funasr/export/models/modules/multihead_att.py
index 7d685f588..0a5667689 100644
--- a/funasr/export/models/modules/multihead_att.py
+++ b/funasr/export/models/modules/multihead_att.py
@@ -64,6 +64,21 @@ class MultiHeadedAttentionSANM(nn.Module):
         return self.linear_out(context_layer)  # (batch, time1, d_model)
 
 
+def preprocess_for_attn(x, mask, cache, pad_fn):
+    x = x * mask
+    x = x.transpose(1, 2)
+    if cache is None:
+        x = pad_fn(x)
+    else:
+        x = torch.cat((cache[:, :, 1:], x), dim=2)
+        cache = x
+    return x, cache
+
+
+import torch.fx
+torch.fx.wrap('preprocess_for_attn')
+
+
 class MultiHeadedAttentionSANMDecoder(nn.Module):
     def __init__(self, model):
         super().__init__()
@@ -73,16 +88,7 @@ class MultiHeadedAttentionSANMDecoder(nn.Module):
         self.attn = None
 
     def forward(self, inputs, mask, cache=None):
-        # b, t, d = inputs.size()
-        # mask = torch.reshape(mask, (b, -1, 1))
-        inputs = inputs * mask
-
-        x = inputs.transpose(1, 2)
-        if cache is None:
-            x = self.pad_fn(x)
-        else:
-            x = torch.cat((cache[:, :, 1:], x), dim=2)
-            cache = x
+        x, cache = preprocess_for_attn(inputs, mask, cache, self.pad_fn)
         x = self.fsmn_block(x)
         x = x.transpose(1, 2)
 
@@ -232,4 +238,4 @@ class OnnxRelPosMultiHeadedAttention(OnnxMultiHeadedAttention):
         new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
         context_layer = context_layer.view(new_context_layer_shape)
         return self.linear_out(context_layer)  # (batch, time1, d_model)
-        
\ No newline at end of file
+        

From 525f5d77564f016acdd03ff71197f7a4a9177840 Mon Sep 17 00:00:00 2001
From: "wanchen.swc" <wanchen.swc@alibaba-inc.com>
Date: Fri, 10 Mar 2023 17:08:04 +0800
Subject: [PATCH 02/13] [Quantization] onnx quantization

---
 funasr/export/README.md       | 16 ++++++++++------
 funasr/export/export_model.py |  8 ++++++++
 2 files changed, 18 insertions(+), 6 deletions(-)

diff --git a/funasr/export/README.md b/funasr/export/README.md
index c44ad3382..a1ed892d5 100644
--- a/funasr/export/README.md
+++ b/funasr/export/README.md
@@ -11,31 +11,35 @@ The installation is the same as [funasr](../../README.md)
    `Tips`: torch>=1.11.0
 
    ```shell
-   python -m funasr.export.export_model [model_name] [export_dir] [onnx]
+   python -m funasr.export.export_model [model_name] [export_dir] [onnx] [quant]
    ```
    `model_name`: the model is to export. It could be the models from modelscope, or local finetuned model(named: model.pb). 
+
    `export_dir`: the dir where the onnx is export.
-    `onnx`: `true`, export onnx format model; `false`, export torchscripts format model.
+
+   `onnx`: `true`, export onnx format model; `false`, export torchscripts format model.
+
+   `quant`: `true`, export quantized model at the same time; `false`, export fp32 model only.
 
 ## For example
 ### Export onnx format model
 Export model from modelscope
 ```shell
-python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true
+python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true false
 ```
 Export model from local path, the model'name must be `model.pb`.
 ```shell
-python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true
+python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true false
 ```
 
 ### Export torchscripts format model
 Export model from modelscope
 ```shell
-python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" false
+python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" false false
 ```
 
 Export model from local path, the model'name must be `model.pb`.
 ```shell
-python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" false
+python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" false false
 ```
 
diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index 1c677c929..7370c3cff 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -153,11 +153,19 @@ class ASRModelExportParaformer:
 
         if self.quant:
             from onnxruntime.quantization import QuantType, quantize_dynamic
+            import onnx
             quant_model_path = os.path.join(path, f'{model.model_name}_quant.onnx')
+            onnx_model = onnx.load(model_path)
+            nodes = [n.name for n in onnx_model.graph.node]
+            nodes_to_exclude = [m for m in nodes if 'output' in m]
             quantize_dynamic(
                 model_input=model_path,
                 model_output=quant_model_path,
+                op_types_to_quantize=['MatMul'],
+                per_channel=True,
+                reduce_range=False,
                 weight_type=QuantType.QUInt8,
+                nodes_to_exclude=nodes_to_exclude,
             )
 
 

From 4bdce2285b42b3ac445a60721b3e7e26f78f4ad3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Tue, 14 Mar 2023 19:46:14 +0800
Subject: [PATCH 03/13] rtf

---
 funasr/bin/vad_inference_launch.py      |  3 +-
 funasr/runtime/python/utils/test_rtf.py | 47 ++++++++++++++++
 funasr/runtime/python/utils/test_rtf.sh | 74 +++++++++++++++++++++++++
 3 files changed, 123 insertions(+), 1 deletion(-)
 create mode 100644 funasr/runtime/python/utils/test_rtf.py
 create mode 100644 funasr/runtime/python/utils/test_rtf.sh

diff --git a/funasr/bin/vad_inference_launch.py b/funasr/bin/vad_inference_launch.py
index 42c5c1e12..18eba33fb 100644
--- a/funasr/bin/vad_inference_launch.py
+++ b/funasr/bin/vad_inference_launch.py
@@ -110,7 +110,8 @@ def inference_launch(mode, **kwargs):
     if mode == "offline":
         from funasr.bin.vad_inference import inference_modelscope
         return inference_modelscope(**kwargs)
-    elif mode == "online":
+    # elif mode == "online":
+    if "param_dict" in kwargs and kwargs["param_dict"]["online"]:
         from funasr.bin.vad_inference_online import inference_modelscope
         return inference_modelscope(**kwargs)
     else:
diff --git a/funasr/runtime/python/utils/test_rtf.py b/funasr/runtime/python/utils/test_rtf.py
new file mode 100644
index 000000000..3394e8a04
--- /dev/null
+++ b/funasr/runtime/python/utils/test_rtf.py
@@ -0,0 +1,47 @@
+
+import time
+import sys
+import librosa
+backend=sys.argv[1]
+model_dir=sys.argv[2]
+wav_file=sys.argv[3]
+
+from torch_paraformer import Paraformer
+if backend == "onnxruntime":
+	from rapid_paraformer import Paraformer
+	
+model = Paraformer(model_dir, batch_size=1, device_id="-1")
+
+wav_file_f = open(wav_file, 'r')
+wav_files = wav_file_f.readlines()
+
+# warm-up
+total = 0.0
+num = 100
+wav_path = wav_files[0].split("\t")[1].strip() if "\t" in wav_files[0] else wav_files[0].split(" ")[1].strip()
+for i in range(num):
+	beg_time = time.time()
+	result = model(wav_path)
+	end_time = time.time()
+	duration = end_time-beg_time
+	total += duration
+	print(result)
+	print("num: {}, time, {}, avg: {}, rtf: {}".format(len(wav_path), duration, total/(i+1), (total/(i+1))/5.53))
+
+# infer time
+beg_time = time.time()
+for i, wav_path_i in enumerate(wav_files):
+	wav_path = wav_path_i.split("\t")[1].strip() if "\t" in wav_path_i else wav_path_i.split(" ")[1].strip()
+	result = model(wav_path)
+end_time = time.time()
+duration = (end_time-beg_time)*1000
+print("total_time_comput_ms: {}".format(int(duration)))
+
+duration_time = 0.0
+for i, wav_path_i in enumerate(wav_files):
+	wav_path = wav_path_i.split("\t")[1].strip() if "\t" in wav_path_i else wav_path_i.split(" ")[1].strip()
+	waveform, _ = librosa.load(wav_path, sr=16000)
+	duration_time += len(waveform)/16.0
+print("total_time_wav_ms: {}".format(int(duration_time)))
+
+print("total_rtf: {:.5}".format(duration/duration_time))
\ No newline at end of file
diff --git a/funasr/runtime/python/utils/test_rtf.sh b/funasr/runtime/python/utils/test_rtf.sh
new file mode 100644
index 000000000..b1562b101
--- /dev/null
+++ b/funasr/runtime/python/utils/test_rtf.sh
@@ -0,0 +1,74 @@
+
+nj=64
+
+#:<<!
+backend=libtorch
+model_dir="/nfs/zhifu.gzf/export/damo/amp_int8/libtorch"
+tag=${backend}_fp32
+!
+
+:<<!
+backend=libtorch
+model_dir="/nfs/zhifu.gzf/export/damo/amp_int8/libtorch_fb20"
+tag=${backend}_amp_fb20
+!
+
+:<<!
+backend=onnxruntime
+model_dir="/nfs/zhifu.gzf/export/damo/amp_int8/onnx"
+tag=${backend}_fp32
+!
+
+:<<!
+backend=onnxruntime
+model_dir="/nfs/zhifu.gzf/export/damo/amp_int8/onnx_dynamic"
+tag=${backend}_fp32
+!
+
+scp=/nfs/haoneng.lhn/funasr_data/aishell-1/data/test/wav.scp
+scp="/nfs/zhifu.gzf/data_debug/test/wav_1500.scp"
+local_scp_dir=/nfs/zhifu.gzf/data_debug/test/${tag}/split$nj
+
+rtf_tool=test_rtf.py
+
+mkdir -p ${local_scp_dir}
+echo ${local_scp_dir}
+
+split_scps=""
+for JOB in $(seq ${nj}); do
+    split_scps="$split_scps $local_scp_dir/wav.$JOB.scp"
+done
+
+perl egs/aishell/transformer/utils/split_scp.pl $scp ${split_scps}
+
+
+for JOB in $(seq ${nj}); do
+  {
+    core_id=`expr $JOB - 1`
+    taskset -c ${core_id} python ${rtf_tool} ${backend} ${model_dir} ${local_scp_dir}/wav.$JOB.scp &> ${local_scp_dir}/log.$JOB.txt
+  }&
+
+done
+wait
+
+
+rm -rf ${local_scp_dir}/total_time_comput.txt
+rm -rf ${local_scp_dir}/total_time_wav.txt
+rm -rf ${local_scp_dir}/total_rtf.txt
+for JOB in $(seq ${nj}); do
+  {
+    cat ${local_scp_dir}/log.$JOB.txt | grep "total_time_comput" | awk -F ' '  '{print $2}' >> ${local_scp_dir}/total_time_comput.txt
+    cat ${local_scp_dir}/log.$JOB.txt | grep "total_time_wav" | awk -F ' '  '{print $2}' >> ${local_scp_dir}/total_time_wav.txt
+    cat ${local_scp_dir}/log.$JOB.txt | grep "total_rtf" | awk -F ' '  '{print $2}' >> ${local_scp_dir}/total_rtf.txt
+  }
+
+done
+
+total_time_comput=`cat ${local_scp_dir}/total_time_comput.txt | awk 'BEGIN {max = 0} {if ($1+0>max+0) max=$1 fi} END {print max}'`
+total_time_wav=`cat ${local_scp_dir}/total_time_wav.txt | awk '{sum +=$1};END {print sum}'`
+rtf=`awk 'BEGIN{printf "%.5f\n",'$total_time_comput'/'$total_time_wav'}'`
+speed=`awk 'BEGIN{printf "%.2f\n",1/'$rtf'}'`
+
+echo "total_time_comput_ms: $total_time_comput"
+echo "total_time_wav: $total_time_wav"
+echo "total_rtf: $rtf, speech: $speed"
\ No newline at end of file

From 63d444e8eb57b772f77de766ed2257d1f6e3d687 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Tue, 14 Mar 2023 19:46:46 +0800
Subject: [PATCH 04/13] rtf

---
 funasr/runtime/python/utils/test_rtf.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/funasr/runtime/python/utils/test_rtf.sh b/funasr/runtime/python/utils/test_rtf.sh
index b1562b101..32166c1b0 100644
--- a/funasr/runtime/python/utils/test_rtf.sh
+++ b/funasr/runtime/python/utils/test_rtf.sh
@@ -25,7 +25,7 @@ model_dir="/nfs/zhifu.gzf/export/damo/amp_int8/onnx_dynamic"
 tag=${backend}_fp32
 !
 
-scp=/nfs/haoneng.lhn/funasr_data/aishell-1/data/test/wav.scp
+#scp=/nfs/haoneng.lhn/funasr_data/aishell-1/data/test/wav.scp
 scp="/nfs/zhifu.gzf/data_debug/test/wav_1500.scp"
 local_scp_dir=/nfs/zhifu.gzf/data_debug/test/${tag}/split$nj
 

From 8a620a5a36df782e1f9e8cc56064d5dc6a1330b5 Mon Sep 17 00:00:00 2001
From: "wanchen.swc" <wanchen.swc@alibaba-inc.com>
Date: Wed, 15 Mar 2023 15:31:31 +0800
Subject: [PATCH 05/13] [Quantization] automatic mixed precision quantization

---
 funasr/export/README.md       | 26 ++++++++++------
 funasr/export/export_model.py | 57 ++++++++++++++++++++++-------------
 2 files changed, 53 insertions(+), 30 deletions(-)

diff --git a/funasr/export/README.md b/funasr/export/README.md
index a1ed892d5..33ab22ea9 100644
--- a/funasr/export/README.md
+++ b/funasr/export/README.md
@@ -11,35 +11,43 @@ The installation is the same as [funasr](../../README.md)
    `Tips`: torch>=1.11.0
 
    ```shell
-   python -m funasr.export.export_model [model_name] [export_dir] [onnx] [quant]
+   python -m funasr.export.export_model \
+       --model-name [model_name] \
+       --export-dir [export_dir] \
+       --type [onnx, torch] \
+       --quantize \
+       --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). 
+   `model-name`: the model is to export. It could be the models from modelscope, or local finetuned model(named: model.pb).
 
-   `export_dir`: the dir where the onnx is export.
+   `export-dir`: the dir where the onnx is export.
 
-   `onnx`: `true`, export onnx format model; `false`, export torchscripts format model.
+   `type`: `onnx` or `torch`, export onnx format model or torchscript format model.
+
+   `quantize`: `true`, export quantized model at the same time; `false`, export fp32 model only.
+
+   `fallback-num`: specify the number of fallback layers to perform automatic mixed precision quantization.
 
-   `quant`: `true`, export quantized model at the same time; `false`, export fp32 model only.
 
 ## For example
 ### Export onnx format model
 Export model from modelscope
 ```shell
-python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true false
+python -m funasr.export.export_model --model-name damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch --export-dir ./export --type onnx
 ```
 Export model from local path, the model'name must be `model.pb`.
 ```shell
-python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true false
+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 onnx
 ```
 
 ### Export torchscripts format model
 Export model from modelscope
 ```shell
-python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" false false
+python -m funasr.export.export_model --model-name damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch --export-dir ./export --type torch
 ```
 
 Export model from local path, the model'name must be `model.pb`.
 ```shell
-python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" false false
+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
 ```
 
diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index 7370c3cff..beb1efe00 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -16,7 +16,11 @@ import random
 
 class ASRModelExportParaformer:
     def __init__(
-        self, cache_dir: Union[Path, str] = None, onnx: bool = True, quant: bool = True
+        self,
+        cache_dir: Union[Path, str] = None,
+        onnx: bool = True,
+        quant: bool = True,
+        fallback_num: int = 0,
     ):
         assert check_argument_types()
         self.set_all_random_seed(0)
@@ -31,6 +35,7 @@ class ASRModelExportParaformer:
         print("output dir: {}".format(self.cache_dir))
         self.onnx = onnx
         self.quant = quant
+        self.fallback_num = fallback_num
         
 
     def _export(
@@ -60,8 +65,12 @@ class ASRModelExportParaformer:
 
 
     def _torch_quantize(self, model):
+        def _run_calibration_data(m):
+            # using dummy inputs for a example
+            dummy_input = model.get_dummy_inputs()
+            m(*dummy_input)
+
         from torch_quant.module import ModuleFilter
-        from torch_quant.observer import HistogramObserver
         from torch_quant.quantizer import Backend, Quantizer
         from funasr.export.models.modules.decoder_layer import DecoderLayerSANM
         from funasr.export.models.modules.encoder_layer import EncoderLayerSANM
@@ -70,17 +79,21 @@ class ASRModelExportParaformer:
         quantizer = Quantizer(
             module_filter=module_filter,
             backend=Backend.FBGEMM,
-            act_ob_ctr=HistogramObserver,
         )
         model.eval()
         calib_model = quantizer.calib(model)
-        # run calibration data
-        # using dummy inputs for a example
-        dummy_input = model.get_dummy_inputs()
-        _ = calib_model(*dummy_input)
+        _run_calibration_data(calib_model)
+        if self.fallback_num > 0:
+            # perform automatic mixed precision quantization
+            amp_model = quantizer.amp(model)
+            _run_calibration_data(amp_model)
+            quantizer.fallback(amp_model, num=self.fallback_num)
+            print('Fallback layers:')
+            print('\n'.join(quantizer.module_filter.exclude_names))
         quant_model = quantizer.quantize(model)
         return quant_model
 
+
     def _export_torchscripts(self, model, verbose, path, enc_size=None):
         if enc_size:
             dummy_input = model.get_dummy_inputs(enc_size)
@@ -170,17 +183,19 @@ class ASRModelExportParaformer:
 
 
 if __name__ == '__main__':
-    import sys
-    
-    model_path = sys.argv[1]
-    output_dir = sys.argv[2]
-    onnx = sys.argv[3]
-    quant = sys.argv[4]
-    onnx = onnx.lower()
-    onnx = onnx == 'true'
-    quant = quant == 'true'
-    # model_path = 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'
-    # output_dir = "../export"
-    export_model = ASRModelExportParaformer(cache_dir=output_dir, onnx=onnx, quant=quant)
-    export_model.export(model_path)
-    # export_model.export('/root/cache/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch')
+    import argparse
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--model-name', type=str, required=True)
+    parser.add_argument('--export-dir', type=str, required=True)
+    parser.add_argument('--type', type=str, default='onnx', help='["onnx", "torch"]')
+    parser.add_argument('--quantize', action='store_true', help='export quantized model')
+    parser.add_argument('--fallback-num', type=int, default=0, help='amp fallback number')
+    args = parser.parse_args()
+
+    export_model = ASRModelExportParaformer(
+        cache_dir=args.export_dir,
+        onnx=args.type == 'onnx',
+        quant=args.quantize,
+        fallback_num=args.fallback_num,
+    )
+    export_model.export(args.model_name)

From b7b45efc4ea164d90d443c116484adfb8b185648 Mon Sep 17 00:00:00 2001
From: "shixian.shi" <shixian.shi@alibaba-inc.com>
Date: Wed, 15 Mar 2023 20:05:37 +0800
Subject: [PATCH 06/13] update paraformer_tiny export

---
 funasr/export/models/encoder/conformer_encoder.py | 1 -
 funasr/export/models/modules/decoder_layer.py     | 1 +
 funasr/export/models/modules/encoder_layer.py     | 4 ++--
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/funasr/export/models/encoder/conformer_encoder.py b/funasr/export/models/encoder/conformer_encoder.py
index 9f2257462..0a3565304 100644
--- a/funasr/export/models/encoder/conformer_encoder.py
+++ b/funasr/export/models/encoder/conformer_encoder.py
@@ -61,7 +61,6 @@ class ConformerEncoder(nn.Module):
                 speech: torch.Tensor,
                 speech_lengths: torch.Tensor,
                 ):
-        speech = speech * self._output_size ** 0.5
         mask = self.make_pad_mask(speech_lengths)
         mask = self.prepare_mask(mask)
         if self.embed is None:
diff --git a/funasr/export/models/modules/decoder_layer.py b/funasr/export/models/modules/decoder_layer.py
index f5394523d..9a464a46c 100644
--- a/funasr/export/models/modules/decoder_layer.py
+++ b/funasr/export/models/modules/decoder_layer.py
@@ -54,6 +54,7 @@ class DecoderLayer(nn.Module):
     
     def forward(self, tgt, tgt_mask, memory, memory_mask, cache=None):
         residual = tgt
+        tgt = self.norm1(tgt)
         tgt_q = tgt
         tgt_q_mask = tgt_mask
         x = residual + self.self_attn(tgt_q, tgt, tgt, tgt_q_mask)
diff --git a/funasr/export/models/modules/encoder_layer.py b/funasr/export/models/modules/encoder_layer.py
index 1da05f382..7d0139793 100644
--- a/funasr/export/models/modules/encoder_layer.py
+++ b/funasr/export/models/modules/encoder_layer.py
@@ -61,7 +61,7 @@ class EncoderLayerConformer(nn.Module):
         if self.feed_forward_macaron is not None:
             residual = x
             x = self.norm_ff_macaron(x)
-            x = residual + self.feed_forward_macaron(x)
+            x = residual + self.feed_forward_macaron(x) * 0.5
 
         residual = x
         x = self.norm_mha(x)
@@ -81,7 +81,7 @@ class EncoderLayerConformer(nn.Module):
 
         residual = x
         x = self.norm_ff(x)
-        x = residual + self.feed_forward(x)
+        x = residual + self.feed_forward(x) * 0.5
 
         x = self.norm_final(x)
 

From 7bf2eec71e0c65f15628a105d11406a8a14ae178 Mon Sep 17 00:00:00 2001
From: "shixian.shi" <shixian.shi@alibaba-inc.com>
Date: Wed, 15 Mar 2023 20:17:01 +0800
Subject: [PATCH 07/13] update paraformer_onnx

---
 .../rapid_paraformer/paraformer_onnx.py       | 55 +++++++++++++------
 1 file changed, 39 insertions(+), 16 deletions(-)

diff --git a/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py b/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py
index 9b8a67bb0..091db0d25 100644
--- a/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py
+++ b/funasr/runtime/python/onnxruntime/rapid_paraformer/paraformer_onnx.py
@@ -23,6 +23,8 @@ class Paraformer():
     def __init__(self, model_dir: Union[str, Path] = None,
                  batch_size: int = 1,
                  device_id: Union[str, int] = "-1",
+                 plot_timestamp_to: str = "",
+                 pred_bias: int = 1,
                  ):
 
         if not Path(model_dir).exists():
@@ -41,14 +43,15 @@ class Paraformer():
         )
         self.ort_infer = OrtInferSession(model_file, device_id)
         self.batch_size = batch_size
-        self.plot = True
+        self.plot_timestamp_to = plot_timestamp_to
+        self.pred_bias = pred_bias
 
     def __call__(self, wav_content: Union[str, np.ndarray, List[str]], **kwargs) -> List:
         waveform_list = self.load_data(wav_content, self.frontend.opts.frame_opts.samp_freq)
         waveform_nums = len(waveform_list)
         asr_res = []
         for beg_idx in range(0, waveform_nums, self.batch_size):
-            res = {}
+            
             end_idx = min(waveform_nums, beg_idx + self.batch_size)
             feats, feats_len = self.extract_feat(waveform_list[beg_idx:end_idx])
             try:
@@ -64,19 +67,41 @@ class Paraformer():
                 logging.warning("input wav is silence or noise")
                 preds = ['']
             else:
-                preds, raw_token = self.decode(am_scores, valid_token_lens)[0]
-                res['preds'] = preds
-                if us_cif_peak is not None:
-                    timestamp, timestamp_total = time_stamp_lfr6_onnx(us_cif_peak, copy.copy(raw_token))
-                    res['timestamp'] = timestamp
-                    if self.plot:
-                        self.plot_wave_timestamp(waveform_list[0], timestamp_total)
-            asr_res.append(res)
+                preds = self.decode(am_scores, valid_token_lens)
+                if us_cif_peak is None:
+                    for pred in preds:
+                        asr_res.append({'preds': pred})
+                else:
+                    for pred, us_cif_peak_ in zip(preds, us_cif_peak):
+                        text, tokens = pred
+                        timestamp, timestamp_total = time_stamp_lfr6_onnx(us_cif_peak_, copy.copy(tokens))
+                        if len(self.plot_timestamp_to):
+                            self.plot_wave_timestamp(waveform_list[0], timestamp_total, self.plot_timestamp_to)
+                        asr_res.append({'preds': text, 'timestamp': timestamp})
         return asr_res
 
-    def plot_wave_timestamp(self, wav, text_timestamp):
+    def plot_wave_timestamp(self, wav, text_timestamp, dest):
         # TODO: Plot the wav and timestamp results with matplotlib
-        import pdb; pdb.set_trace()
+        import matplotlib
+        matplotlib.use('Agg')
+        matplotlib.rc("font", family='Alibaba PuHuiTi')  # set it to a font that your system supports
+        import matplotlib.pyplot as plt
+        fig, ax1 = plt.subplots(figsize=(11, 3.5), dpi=320)
+        ax2 = ax1.twinx()
+        ax2.set_ylim([0, 2.0])
+        # plot waveform
+        ax1.set_ylim([-0.3, 0.3])
+        time = np.arange(wav.shape[0]) / 16000
+        ax1.plot(time, wav/wav.max()*0.3, color='gray', alpha=0.4)
+        # plot lines and text
+        for (char, start, end) in text_timestamp:
+            ax1.vlines(start, -0.3, 0.3, ls='--')
+            ax1.vlines(end, -0.3, 0.3, ls='--')
+            x_adj = 0.045 if char != '<sil>' else 0.12
+            ax1.text((start + end) * 0.5 - x_adj, 0, char)
+        # plt.legend()
+        plotname = "{}/timestamp.png".format(dest)
+        plt.savefig(plotname, bbox_inches='tight')
 
     def load_data(self,
                   wav_content: Union[str, np.ndarray, List[str]], fs: int = None) -> List:
@@ -150,9 +175,7 @@ class Paraformer():
 
         # Change integer-ids to tokens
         token = self.converter.ids2tokens(token_int)
-        # token = token[:valid_token_num-1]
+        token = token[:valid_token_num-self.pred_bias]
         texts = sentence_postprocess(token)
-        text = texts[0]
-        # text = self.tokenizer.tokens2text(token)
-        return text, token
+        return texts
 

From 01b4ff3bde05b7c5ea071af8867a331e9ae4bf53 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Wed, 15 Mar 2023 20:36:58 +0800
Subject: [PATCH 08/13] calib set

---
 funasr/export/export_model.py   | 52 +++++++++++++++++++++++++++++++--
 funasr/export/utils/wav_load.py |  2 ++
 2 files changed, 52 insertions(+), 2 deletions(-)
 create mode 100644 funasr/export/utils/wav_load.py

diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index beb1efe00..7e65a8f76 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -21,6 +21,8 @@ class ASRModelExportParaformer:
         onnx: bool = True,
         quant: bool = True,
         fallback_num: int = 0,
+        audio_in: str = None,
+        calib_num: int = 200,
     ):
         assert check_argument_types()
         self.set_all_random_seed(0)
@@ -36,6 +38,9 @@ class ASRModelExportParaformer:
         self.onnx = onnx
         self.quant = quant
         self.fallback_num = fallback_num
+        self.frontend = None
+        self.audio_in = audio_in
+        self.calib_num = calib_num
         
 
     def _export(
@@ -67,8 +72,14 @@ class ASRModelExportParaformer:
     def _torch_quantize(self, model):
         def _run_calibration_data(m):
             # using dummy inputs for a example
-            dummy_input = model.get_dummy_inputs()
-            m(*dummy_input)
+            if self.audio_in is not None:
+                feats, feats_len = self.load_feats(self.audio_in)
+                for feat, len in zip(feats, feats_len):
+                    m(feat, len)
+            else:
+                dummy_input = model.get_dummy_inputs()
+                m(*dummy_input)
+            
 
         from torch_quant.module import ModuleFilter
         from torch_quant.quantizer import Backend, Quantizer
@@ -114,6 +125,39 @@ class ASRModelExportParaformer:
         random.seed(seed)
         np.random.seed(seed)
         torch.random.manual_seed(seed)
+
+    def parse_audio_in(self, audio_in):
+        
+        wav_list, name_list = [], []
+        if audio_in.endswith(".scp"):
+            f = open(audio_in, 'r')
+            lines = f.readlines()[:self.calib_num]
+            for line in lines:
+                name, path = line.strip().split()
+                name_list.append(name)
+                wav_list.append(path)
+        else:
+            wav_list = [audio_in,]
+            name_list = ["test",]
+        return wav_list, name_list
+    
+    def load_feats(self, audio_in: str = None):
+        import torchaudio
+
+        wav_list, name_list = self.parse_audio_in(audio_in)
+        feats = []
+        feats_len = []
+        for line in wav_list:
+            name, path = line.strip().split()
+            waveform, sampling_rate = torchaudio.load(path)
+            if sampling_rate != self.frontend.fs:
+                waveform = torchaudio.transforms.Resample(orig_freq=sampling_rate,
+                                                          new_freq=self.frontend.fs)(waveform)
+            fbank, fbank_len = self.frontend(waveform, [waveform.size(1)])
+            feats.append(fbank)
+            feats_len.append(fbank_len)
+        return feats, feats_len
+    
     def export(self,
                tag_name: str = 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
                mode: str = 'paraformer',
@@ -190,6 +234,8 @@ if __name__ == '__main__':
     parser.add_argument('--type', type=str, default='onnx', help='["onnx", "torch"]')
     parser.add_argument('--quantize', action='store_true', help='export quantized model')
     parser.add_argument('--fallback-num', type=int, default=0, help='amp fallback number')
+    parser.add_argument('--audio_in', type=str, default=None, help='["wav", "wav.scp"]')
+    parser.add_argument('--calib_num', type=int, default=200, help='calib max num')
     args = parser.parse_args()
 
     export_model = ASRModelExportParaformer(
@@ -197,5 +243,7 @@ if __name__ == '__main__':
         onnx=args.type == 'onnx',
         quant=args.quantize,
         fallback_num=args.fallback_num,
+        audio_in=args.audio_in,
+        calib_num=args.calib_num,
     )
     export_model.export(args.model_name)
diff --git a/funasr/export/utils/wav_load.py b/funasr/export/utils/wav_load.py
new file mode 100644
index 000000000..b48e5a074
--- /dev/null
+++ b/funasr/export/utils/wav_load.py
@@ -0,0 +1,2 @@
+import os
+

From 18b6fb3b502ee1bf4c6b595a8e96cf2216393f80 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Wed, 15 Mar 2023 20:38:20 +0800
Subject: [PATCH 09/13] calib set

---
 funasr/export/utils/wav_load.py | 2 --
 1 file changed, 2 deletions(-)
 delete mode 100644 funasr/export/utils/wav_load.py

diff --git a/funasr/export/utils/wav_load.py b/funasr/export/utils/wav_load.py
deleted file mode 100644
index b48e5a074..000000000
--- a/funasr/export/utils/wav_load.py
+++ /dev/null
@@ -1,2 +0,0 @@
-import os
-

From 9f5c42c476463e801d3de4bfeb2aefdd04f5691d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Wed, 15 Mar 2023 20:57:58 +0800
Subject: [PATCH 10/13] calib set

---
 funasr/export/export_model.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index 7e65a8f76..b827f16b3 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -148,7 +148,7 @@ class ASRModelExportParaformer:
         feats = []
         feats_len = []
         for line in wav_list:
-            name, path = line.strip().split()
+            path = line.strip()
             waveform, sampling_rate = torchaudio.load(path)
             if sampling_rate != self.frontend.fs:
                 waveform = torchaudio.transforms.Resample(orig_freq=sampling_rate,
@@ -184,6 +184,7 @@ class ASRModelExportParaformer:
         model, asr_train_args = ASRTask.build_model_from_file(
             asr_train_config, asr_model_file, cmvn_file, 'cpu'
         )
+        self.frontend = model.frontend
         self._export(model, tag_name)
             
 

From 4ffd5655da0884095e68cc7207bf9bf5d409d2bb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Thu, 16 Mar 2023 16:40:15 +0800
Subject: [PATCH 11/13] calib

---
 funasr/export/export_model.py           | 3 ++-
 funasr/runtime/python/utils/test_rtf.sh | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index b827f16b3..c5bcac1e0 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -74,7 +74,8 @@ class ASRModelExportParaformer:
             # using dummy inputs for a example
             if self.audio_in is not None:
                 feats, feats_len = self.load_feats(self.audio_in)
-                for feat, len in zip(feats, feats_len):
+                for i, (feat, len) in enumerate(zip(feats, feats_len)):
+                    print("debug, iter: {}".format(i))
                     m(feat, len)
             else:
                 dummy_input = model.get_dummy_inputs()
diff --git a/funasr/runtime/python/utils/test_rtf.sh b/funasr/runtime/python/utils/test_rtf.sh
index 32166c1b0..fe13da7d8 100644
--- a/funasr/runtime/python/utils/test_rtf.sh
+++ b/funasr/runtime/python/utils/test_rtf.sh
@@ -39,7 +39,7 @@ for JOB in $(seq ${nj}); do
     split_scps="$split_scps $local_scp_dir/wav.$JOB.scp"
 done
 
-perl egs/aishell/transformer/utils/split_scp.pl $scp ${split_scps}
+perl ../../../egs/aishell/transformer/utils/split_scp.pl $scp ${split_scps}
 
 
 for JOB in $(seq ${nj}); do

From 6ca0d1f54c8b698e3315edab8aa9ba7227d7c9e7 Mon Sep 17 00:00:00 2001
From: "wanchen.swc" <wanchen.swc@alibaba-inc.com>
Date: Thu, 16 Mar 2023 18:53:57 +0800
Subject: [PATCH 12/13] [Quantization] run calib without grad

---
 funasr/export/export_model.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index c5bcac1e0..e57a39750 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -76,7 +76,8 @@ class ASRModelExportParaformer:
                 feats, feats_len = self.load_feats(self.audio_in)
                 for i, (feat, len) in enumerate(zip(feats, feats_len)):
                     print("debug, iter: {}".format(i))
-                    m(feat, len)
+                    with torch.no_grad():
+                        m(feat, len)
             else:
                 dummy_input = model.get_dummy_inputs()
                 m(*dummy_input)

From 175860147c08c2a6940032039b7eb83abfcc3ca0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Thu, 16 Mar 2023 19:28:44 +0800
Subject: [PATCH 13/13] cab

---
 funasr/export/export_model.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/funasr/export/export_model.py b/funasr/export/export_model.py
index e57a39750..9a1ef9604 100644
--- a/funasr/export/export_model.py
+++ b/funasr/export/export_model.py
@@ -75,7 +75,6 @@ class ASRModelExportParaformer:
             if self.audio_in is not None:
                 feats, feats_len = self.load_feats(self.audio_in)
                 for i, (feat, len) in enumerate(zip(feats, feats_len)):
-                    print("debug, iter: {}".format(i))
                     with torch.no_grad():
                         m(feat, len)
             else: