update doc

docs/how-to-use-and-FAQ/quantized-int8-inference.md

@@ -48,6 +48,12 @@ If your model has multiple input nodes, you can use multiple list files and othe
 ./ncnn2int8 mobilenet-opt.param mobilenet-opt.bin mobilenet-int8.param mobilenet-int8.bin mobilenet.table
 ```
 
+If you don’t need static quantization, ncnn supports RNN/LSTM/GRU dynamic quantization. In this case, you can omit the table file.
+
+```shell
+./ncnn2int8 rnn-model.param rnn-model.bin rnn-model-int8.param rnn-model-int8.bin
+```
+
 ## use ncnn int8 inference
 
 the ncnn library would use int8 inference automatically, nothing changed in your code

tools/quantize/ncnn2int8.cpp

@@ -767,7 +767,7 @@ int NetQuantize::fuse_requantize()
 
 int main(int argc, char** argv)
 {
-    if (argc != 6)
+    if (argc != 5 && argc != 6)
     {
         fprintf(stderr, "usage: %s [inparam] [inbin] [outparam] [outbin] [calibration table]\n", argv[0]);
         return -1;
@@ -777,7 +777,7 @@ int main(int argc, char** argv)
     const char* inbin = argv[2];
     const char* outparam = argv[3];
     const char* outbin = argv[4];
-    const char* int8scale_table_path = argv[5];
+    const char* int8scale_table_path = argc == 6 ? argv[5] : NULL;
 
     NetQuantize quantizer;