Upload folder using huggingface_hub

README.md

@@ -19,10 +19,10 @@ The original model contains `int64` Cast operations and tensors that WebGPU cann
 This version converts all `int64` operations to `int32`, enabling direct WebGPU inference.
 
 ### Modifications Made:
-- **speech_encoder**: 26 Cast ops, 36 initializers, 1 I/O type converted
-- **conditional_decoder**: 3 Cast ops, 37 initializers, 1 I/O type converted
-- **language_model**: 2 I/O types converted
-- **embed_tokens**: 5 initializers, 1 I/O type converted
+- **conditional_decoder**: 521 Cast nodes inserted (376 Shape/Range ops)
+- **speech_encoder**: 350 Cast nodes inserted (243 Shape/Range ops)
+- **language_model**: 3 Cast nodes inserted
+- **embed_tokens**: 1 Cast node inserted
 
 ## Usage with Transformers.js
 

onnx/conditional_decoder_q4f16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:22572a6001425d5d067a78902472a26f5f585b39ec5ce4d9743f0bebcc6503e0
-size 2393626
+oid sha256:00e523d8db1c46c847132db4a3b6fd4491ef9addd47b2870685f6fa39b13b99e
+size 2477348

onnx/embed_tokens_q4f16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4135269ccfadaa5e084cec0ccbbd8d16a22121e0b3da771420df0644455f24ef
-size 2528
+oid sha256:a9589031616f8ee27b7db07ba2656d5e7f1fd58586807006cba21d268f91cf97
+size 2630

onnx/language_model_q4f16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:37573106c17cee1fd1af841b7cf39dc6ecfe65c7331380a90cb97e3a76002972
-size 276803
+oid sha256:961d1dac0f14f034e454ed60f52345921fd9f7edf2c806bc194829b4b45b4d06
+size 276978

onnx/speech_encoder_q4f16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:52d6d01762a013c4e2e17c31717e638c45ab869b2be7d7aac1b4be54b2f9cbef
-size 1217363
+oid sha256:6a132ebd5cbc3e4178d4515f47927981c6e98ddfeda4036e10ee795bb342c98e
+size 1280889