Upload folder using huggingface_hub

config.json

@@ -50,6 +50,7 @@
   "max_length": 448,
   "max_source_positions": 1500,
   "max_target_positions": 448,
+  "median_filter_width": 7,
   "model_type": "whisper",
   "num_hidden_layers": 6,
   "num_mel_bins": 80,
@@ -145,7 +146,7 @@
     50361,
     50362
   ],
-  "transformers_version": "4.30.2",
+  "transformers_version": "4.33.0.dev0",
   "use_cache": true,
   "use_weighted_layer_sum": false,
   "vocab_size": 51865

generation_config.json

@@ -1,5 +1,38 @@
 {
-  "alignment_heads": [[3, 1], [4, 2], [4, 3], [4, 7], [5, 1], [5, 2], [5, 4], [5, 6]],
+  "alignment_heads": [
+    [
+      3,
+      1
+    ],
+    [
+      4,
+      2
+    ],
+    [
+      4,
+      3
+    ],
+    [
+      4,
+      7
+    ],
+    [
+      5,
+      1
+    ],
+    [
+      5,
+      2
+    ],
+    [
+      5,
+      4
+    ],
+    [
+      5,
+      6
+    ]
+  ],
   "begin_suppress_tokens": [
     220,
     50257
@@ -218,5 +251,5 @@
     "transcribe": 50359,
     "translate": 50358
   },
-  "transformers_version": "4.30.2"
+  "transformers_version": "4.33.0.dev0"
 }

onnx/decoder_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:30dc5d792fc7dee971935fa6cd4df68e219d866ecc6b4309ecdfc7c7efcab992
-size 314491115
+oid sha256:bae722537859a68b8babc4930c504f1f51a83420a2b1b0bd0b0947b4811b8042
+size 208304195

onnx/decoder_model_merged.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17eeff2508a474685cec5b46587c13eb7d0d28f921ce64d08b8ef3e7a9c2938e
-size 314710581
+oid sha256:3fed66b2b8dfa1ae213f768527dff1ef7e3857b8d45a36381850ce4dff8db936
+size 208560983

onnx/decoder_model_merged_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:edccc6c79bf0f9b46a99770172cf3160154734640795f84c6fc516932b7b87b9
-size 80172704
+oid sha256:a6beb6baabb66f00b6a686d828c95ffca6146d51900cbad0266cad38f64cf861
+size 53707539

onnx/decoder_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d0302f2ff47b04e6715e5812a1313084d404cbdeff4a55fdab22e4c5f7c00aa2
-size 79822868
+oid sha256:96fc8b98013e0f9461159a7b0d57922e170b15767db8afef1ea7880f1deef33b
+size 53309877

onnx/decoder_with_past_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c77771da8b403752fe44d95d8a38e975e11e4ba170dc0407b19a433f838470e3
-size 301854704
+oid sha256:75f7f7144348fb70fd421587b938e6b215188f1e480fbc0be016177b0e3891dd
+size 195670910

onnx/decoder_with_past_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54d7fad0e7189431eab05cc42fb488f7d8b8011070c70e49aa9d833cb4a917c2
-size 76596950
+oid sha256:7b12f3892bce5a5c9729eb3d5c2ac451e64dd20e367d643eaa675345a114471d
+size 50088237

onnx/encoder_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5508b72d9379911bf51bcf034ce26c43501a5dccc7a5ca4a061fc851471f9a10
-size 82461621
+oid sha256:f0bd7927234639c6e1f293cef18a210cee4e4aea93e200ebbe48e1d7acf6fdb1
+size 82474863

onnx/encoder_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e360d86928a0232eb6100f2a36fa4a358c101e0a6a39763e5eefe139dca5fe71
-size 23183300
+oid sha256:3e345e977b55620a37c0c2b2af0644e019afdfad562dcf71eb929bb7274285f9
+size 23200850

quantize_config.json

@@ -0,0 +1,115 @@
+{
+    "per_channel": false,
+    "reduce_range": false,
+    "per_model_config": {
+        "decoder_model": {
+            "op_types": [
+                "Erf",
+                "Squeeze",
+                "Gather",
+                "Unsqueeze",
+                "Slice",
+                "Shape",
+                "Transpose",
+                "MatMul",
+                "Softmax",
+                "Reshape",
+                "Range",
+                "Less",
+                "Equal",
+                "Concat",
+                "Sqrt",
+                "Mul",
+                "Pow",
+                "Sub",
+                "Constant",
+                "ConstantOfShape",
+                "Div",
+                "Add",
+                "Expand",
+                "Where",
+                "Cast",
+                "ReduceMean"
+            ],
+            "weight_type": "QInt8"
+        },
+        "decoder_model_merged": {
+            "op_types": [
+                "Erf",
+                "Squeeze",
+                "Gather",
+                "Unsqueeze",
+                "Slice",
+                "Shape",
+                "Transpose",
+                "MatMul",
+                "Softmax",
+                "Reshape",
+                "Range",
+                "Cast",
+                "Less",
+                "Equal",
+                "Concat",
+                "If",
+                "Sqrt",
+                "Mul",
+                "Pow",
+                "Sub",
+                "Constant",
+                "Div",
+                "Add",
+                "Expand",
+                "Where",
+                "ConstantOfShape",
+                "ReduceMean"
+            ],
+            "weight_type": "QInt8"
+        },
+        "decoder_with_past_model": {
+            "op_types": [
+                "Reshape",
+                "Mul",
+                "Erf",
+                "Div",
+                "Gather",
+                "Unsqueeze",
+                "Add",
+                "Shape",
+                "Slice",
+                "Sub",
+                "Constant",
+                "Pow",
+                "Transpose",
+                "Concat",
+                "MatMul",
+                "Softmax",
+                "ReduceMean",
+                "Sqrt"
+            ],
+            "weight_type": "QInt8"
+        },
+        "encoder_model": {
+            "op_types": [
+                "Reshape",
+                "Erf",
+                "Mul",
+                "Div",
+                "Conv",
+                "Gather",
+                "Add",
+                "Pow",
+                "Shape",
+                "Transpose",
+                "Sub",
+                "Constant",
+                "Unsqueeze",
+                "Concat",
+                "MatMul",
+                "Softmax",
+                "ReduceMean",
+                "Sqrt"
+            ],
+            "weight_type": "QUInt8"
+        }
+    }
+}