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": 32,
   "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 +1,98 @@
 {
-  "alignment_heads": [[10, 12], [13, 17], [16, 11], [16, 12], [16, 13], [17, 15], [17, 16], [18, 4], [18, 11], [18, 19], [19, 11], [21, 2], [21, 3], [22, 3], [22, 9], [22, 12], [23, 5], [23, 7], [23, 13], [25, 5], [26, 1], [26, 12], [27, 15]],
+  "alignment_heads": [
+    [
+      10,
+      12
+    ],
+    [
+      13,
+      17
+    ],
+    [
+      16,
+      11
+    ],
+    [
+      16,
+      12
+    ],
+    [
+      16,
+      13
+    ],
+    [
+      17,
+      15
+    ],
+    [
+      17,
+      16
+    ],
+    [
+      18,
+      4
+    ],
+    [
+      18,
+      11
+    ],
+    [
+      18,
+      19
+    ],
+    [
+      19,
+      11
+    ],
+    [
+      21,
+      2
+    ],
+    [
+      21,
+      3
+    ],
+    [
+      22,
+      3
+    ],
+    [
+      22,
+      9
+    ],
+    [
+      22,
+      12
+    ],
+    [
+      23,
+      5
+    ],
+    [
+      23,
+      7
+    ],
+    [
+      23,
+      13
+    ],
+    [
+      25,
+      5
+    ],
+    [
+      26,
+      1
+    ],
+    [
+      26,
+      12
+    ],
+    [
+      27,
+      15
+    ]
+  ],
   "begin_suppress_tokens": [
     220,
     50257
@@ -218,5 +311,5 @@
     "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:d1c0a21faf8ba940db5fcc29bdc0e66412d05fd80cf8dcfac8c92d23ba166d01
-size 1506542
+oid sha256:24a162d9454daa1ef7eadfbc5c498aa8da370fa5d6f14b9091acb758f8d4eaa3
+size 1687965

onnx/decoder_model.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d580f467fb53b33bbfdc1010bffc8b46a79e74ff5c4f19f7bf1516ec7767327d
-size 3891630080
+oid sha256:96f75bb114c97f4f08ac9a1a0b64674966aa04349a6dfcf3c5dda3b71e8f6db2
+size 3626081280

onnx/decoder_model_merged.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b6386da420526f0323712dfe66e02004259ef44377e927624ab1e5d5953fa73c
-size 2761864
+oid sha256:6dadfd4ffe47b792073118d88b8b7a3af39a437d954e5914b298430272d6442b
+size 3151837

onnx/decoder_model_merged.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d580f467fb53b33bbfdc1010bffc8b46a79e74ff5c4f19f7bf1516ec7767327d
-size 3891630080
+oid sha256:96f75bb114c97f4f08ac9a1a0b64674966aa04349a6dfcf3c5dda3b71e8f6db2
+size 3626081280

onnx/decoder_model_merged_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ca3d44b85d0dd8442ce7e4a168265d0678fa01d8f239d3aeee0ade5c51e88d1
-size 980944149
+oid sha256:59fdebf106a3875e3c7e233c958c356ce106a6fd95bc41bb8fd12e3168814b1a
+size 915058353

onnx/decoder_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:318d59d78c33a872b843347a7ca309003d6cde6cd972e250407cb502ecbba730
-size 979005046
+oid sha256:c3731b24953a59d166d38f4461b89d6158016afca9c10f15f88cf4d38905fb55
+size 912852053

onnx/decoder_with_past_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:564c4dd801433b3ccf607b947ec8dac4c6cb6538edb72b6cc0cf0217141ba289
-size 1293576
+oid sha256:2fae8c640318528967efdb8711676c66fa3b3d3e847d8fc983cbd249a1182f53
+size 1493595

onnx/decoder_with_past_model.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8293518c6d66788e5020c694e1a317b744db62a0c918f6ff574bdbed8e23f99e
-size 3472035840
+oid sha256:d066c0dca9cf33e13d403ccf48fb3308748b5d41c53736c6ce1fe8fc1f2af9d8
+size 3206487040

onnx/decoder_with_past_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:912d66d239bac2b85e77b116aa92d68a1e45a0b9b164b3a2c044c33d35427a56
-size 873637776
+oid sha256:8a20491506a3dee70af49a34066bab3e0c480780aedde4f17adc1752408e9372
+size 807509283

onnx/encoder_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c34e862d089a0c6a33d9d63620882927f2bf41872a257807e81cdc4ddae5a85c
-size 602071
+oid sha256:c4776407df4e9cfc9450c505613c4f04119d9fd09b847be1a8130a46dce506d8
+size 676665

onnx/encoder_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5dcb8d36c86031fb95287fdef2e1aa43f46e13f79028577923c5c76652c2597b
-size 644978442
+oid sha256:2f919ac9f17ea8e1a53457ee5ae9cd29efd76d12793bff9d258e88ed615032ad
+size 645076106

quantize_config.json

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