Upload 29 files

initial the repo, and update the axmodel of AX650 and AX630C

.gitattributes

@@ -33,3 +33,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+ax620e/base/base-decoder-loop.axmodel filter=lfs diff=lfs merge=lfs -text
+ax620e/base/base-decoder-main.axmodel filter=lfs diff=lfs merge=lfs -text
+ax620e/base/base-encoder.axmodel filter=lfs diff=lfs merge=lfs -text
+ax650/small/small-decoder-loop.axmodel filter=lfs diff=lfs merge=lfs -text
+ax650/small/small-decoder-main.axmodel filter=lfs diff=lfs merge=lfs -text
+ax650/small/small-encoder.axmodel filter=lfs diff=lfs merge=lfs -text

ax620e/base/base-decoder-loop.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4133ed1d67dbf4906b8dc4e59a642b558ffc46e461c9f6688045babe5a3a60ca
+size 130371061

ax620e/base/base-decoder-main.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c5bdd733b666aaac06144f1f780dfcdb961e131d658e4a499e14a25dde42688
+size 135674831

ax620e/base/base-encoder.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7238ad68dd496ed3c54acc9312dec4e559a9640cb504aea33f12aae93077e94
+size 56035119

ax620e/base/base-positional_embedding.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88fa1cdbf2b06f86b0ecb7be0fccfc39e906502986572b8cf5319c250e857169
+size 917504

ax620e/install/t2s.json

@@ -0,0 +1,22 @@
+{
+  "name": "Traditional Chinese to Simplified Chinese",
+  "segmentation": {
+    "type": "mmseg",
+    "dict": {
+      "type": "ocd2",
+      "file": "TSPhrases.ocd2"
+    }
+  },
+  "conversion_chain": [{
+    "dict": {
+      "type": "group",
+      "dicts": [{
+        "type": "ocd2",
+        "file": "TSPhrases.ocd2"
+      }, {
+        "type": "ocd2",
+        "file": "TSCharacters.ocd2"
+      }]
+    }
+  }]
+}

ax650/install/t2s.json

@@ -0,0 +1,22 @@
+{
+  "name": "Traditional Chinese to Simplified Chinese",
+  "segmentation": {
+    "type": "mmseg",
+    "dict": {
+      "type": "ocd2",
+      "file": "TSPhrases.ocd2"
+    }
+  },
+  "conversion_chain": [{
+    "dict": {
+      "type": "group",
+      "dicts": [{
+        "type": "ocd2",
+        "file": "TSPhrases.ocd2"
+      }, {
+        "type": "ocd2",
+        "file": "TSCharacters.ocd2"
+      }]
+    }
+  }]
+}

ax650/small/small-decoder-loop.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b472a0f3539d17fece09e92bf6cd69ebf391928a6050896bbf86b558a25def22
+size 269002567

ax650/small/small-decoder-main.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3bfc577f60c35192d8ce8cc24f9ca4aa84af72756ba11af9d178d337cb7eb1c
+size 285531695

ax650/small/small-encoder.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b3bc8db9762f9b2dfe78bffbc8070fb877b2572c5288253573e49a8c7b37948
+size 139705612

ax650/small/small-positional_embedding.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c10bc44f2bd94bdf1b7aa03581309fa536132b3fe79bfe22c9a6934a42cd8b58
+size 1376256

ax650/tiny/tiny-decoder-loop.axmodel.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b600ff566563a68cd1b1902082b1587aa67dc127cd9ef13b1cdc42bfc257375b
+size 128477481

ax650/tiny/tiny-decoder-main.axmodel.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50beadff2619fe099821aa28351c1cf887b5b1c97a4e696541314d6d43131dc7
+size 131255063

ax650/tiny/tiny-encoder.axmodel.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:627b1a55b5daa32db0079f8902d006e4d1c86d60f29c72d3a1914311c4b83f55
+size 14422649

ax650/tiny/tiny-positional_embedding.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c13450ae630323a0bdd39b1226f92a7ac251131a909c7efdb7d2f5516736eb83
+size 688128

config_whisper_decoder_loop_u16.json

@@ -0,0 +1,78 @@
+{
+    "model_type": "ONNX",
+    "npu_mode": "NPU3",
+    "quant": {
+      "input_configs": [
+        {
+          "tensor_name": "tokens",
+          "calibration_dataset": "./calibrations_tiny/decoder_loop/tokens.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "n_layer_cross_k",
+          "calibration_dataset": "./calibrations_tiny/decoder_loop/n_layer_cross_k.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "in_n_layer_self_k_cache",
+          "calibration_dataset": "./calibrations_tiny/decoder_loop/n_layer_self_k_cache.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "in_n_layer_self_v_cache",
+          "calibration_dataset": "./calibrations_tiny/decoder_loop/n_layer_self_v_cache.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "n_layer_cross_v",
+          "calibration_dataset": "./calibrations_tiny/decoder_loop/n_layer_cross_v.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "positional_embedding",
+          "calibration_dataset": "./calibrations_tiny/decoder_loop/positional_embedding.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "mask",
+          "calibration_dataset": "./calibrations_tiny/decoder_loop/mask.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        }
+      ],
+      "layer_configs": [
+        {
+          "start_tensor_names": ["DEFAULT"],
+          "end_tensor_names": ["DEFAULT"],
+          "data_type": "U16"
+        },
+        {
+          "op_types": ["Gather"],
+          "data_type": "FP32"
+        },
+        {
+          "layer_names": ["/Add_1", "/Add_5", "/Add_9", "/Add_13"],
+          "data_type": "FP32"
+        },
+        {
+          "layer_names": ["/Softmax", "/Softmax_2", "/Softmax_4", "/Softmax_6"],
+          "data_type": "FP32"
+        }
+      ],
+      "calibration_method": "MinMax"
+    },
+    "input_processors": [
+      {
+        "tensor_name": "DEFAULT"
+      }
+    ],
+    "compiler": {
+      "check": 2
+    }
+  }

config_whisper_decoder_main_u16.json

@@ -0,0 +1,54 @@
+{
+    "model_type": "ONNX",
+    "npu_mode": "NPU3",
+    "quant": {
+      "input_configs": [
+        {
+          "tensor_name": "tokens",
+          "calibration_dataset": "./calibrations_tiny/decoder_main/tokens.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "n_layer_cross_k",
+          "calibration_dataset": "./calibrations_tiny/decoder_main/n_layer_cross_k.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        },
+        {
+          "tensor_name": "n_layer_cross_v",
+          "calibration_dataset": "./calibrations_tiny/decoder_main/n_layer_cross_v.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        }
+      ],
+      "layer_configs": [
+        {
+          "start_tensor_names": ["DEFAULT"],
+          "end_tensor_names": ["DEFAULT"],
+          "data_type": "U16"
+        },
+        {
+          "op_types": ["Gather"],
+          "data_type": "FP32"
+        },
+        {
+          "layer_names": ["/Add_1", "/Add_5", "/Add_9", "/Add_13"],
+          "data_type": "FP32"
+        },
+        {
+          "layer_names": ["/Softmax", "/Softmax_2", "/Softmax_4", "/Softmax_6"],
+          "data_type": "FP32"
+        }
+      ],
+      "calibration_method": "MinMax"
+    },
+    "input_processors": [
+      {
+        "tensor_name": "DEFAULT"
+      }
+    ],
+    "compiler": {
+      "check": 2
+    }
+  }

config_whisper_encoder_u16.json

@@ -0,0 +1,30 @@
+{
+    "model_type": "ONNX",
+    "npu_mode": "NPU3",
+    "quant": {
+      "input_configs": [
+        {
+          "tensor_name": "mel",
+          "calibration_dataset": "./calibrations_tiny/encoder/mel.tar.gz",
+          "calibration_size": -1,
+          "calibration_format": "Numpy"
+        }
+      ],
+      "layer_configs": [
+        {
+          "start_tensor_names": ["DEFAULT"],
+          "end_tensor_names": ["DEFAULT"],
+          "data_type": "U16"
+        }
+      ],
+      "calibration_method": "MinMax"
+    },
+    "input_processors": [
+      {
+        "tensor_name": "mel"
+      }
+    ],
+    "compiler": {
+      "check": 3
+    }
+  }

datasets.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bd7b4da356313a433cb119aa37e960f925b288d7133cfbd5106800fa28c6523
+size 41523788