From bb529e7e4a1bebfb4d4cdf9b96cb6f5623ac5967 Mon Sep 17 00:00:00 2001
From: zpeng11 <a2843173630@gmail.com>
Date: Wed, 20 Aug 2025 20:21:42 -0400
Subject: [PATCH] update namings

---
 GPT_SoVITS/onnx_export.py | 15 +++++++--------
 playground/freerun.py     | 12 ++++++------
 2 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/GPT_SoVITS/onnx_export.py b/GPT_SoVITS/onnx_export.py
index 3a940f65..f0e685e7 100644
--- a/GPT_SoVITS/onnx_export.py
+++ b/GPT_SoVITS/onnx_export.py
@@ -157,14 +157,14 @@ class T2SModel(nn.Module):
             self.init_step,
             (ref_seq, text_seq, ref_bert, text_bert, ssl_content),
             f"onnx/{project_name}/{project_name}_t2s_init_step.onnx",
-            input_names=["ref_seq", "text_seq", "ref_bert", "text_bert", "ssl_content"],
+            input_names=["ref_text_phones", "input_text_phones", "ref_text_bert", "input_text_bert", "hubert_ssl_content"],
             output_names=["y", "k", "v", "y_emb", "x_example"],
             dynamic_axes={
-                "ref_seq": {1: "ref_length"},
-                "text_seq": {1: "text_length"},
-                "ref_bert": {0: "ref_length"},
-                "text_bert": {0: "text_length"},
-                "ssl_content": {2: "ssl_length"},
+                "ref_text_phones": {1: "ref_length"},
+                "input_text_phones": {1: "text_length"},
+                "ref_text_bert": {0: "ref_length"},
+                "input_text_bert": {0: "text_length"},
+                "hubert_ssl_content": {2: "ssl_length"},
             },
             opset_version=16,
         )
@@ -254,9 +254,8 @@ class GptSoVits(nn.Module):
             input_names=["text_seq", "pred_semantic", "spectrum", "sv_emb"],
             output_names=["audio"],
             dynamic_axes={
-                "text_seq": {1: "text_length"},
+                "input_text_phones": {1: "text_length"},
                 "pred_semantic": {2: "pred_length"},
-                "ref_audio": {1: "audio_length"},
                 "spectrum": {2: "spectrum_length"},
             },
             opset_version=17,
diff --git a/playground/freerun.py b/playground/freerun.py
index f169fd99..25494ef3 100644
--- a/playground/freerun.py
+++ b/playground/freerun.py
@@ -77,11 +77,11 @@ def preprocess_text(text:str):
 init_step = ort.InferenceSession(MODEL_PATH+"_export_t2s_init_step.onnx")
 
 [y, k, v, y_emb, x_example] = init_step.run(None, {
-    "text_seq": input_phones,
-    "text_bert": input_bert,
-    "ref_seq": ref_phones,
-    "ref_bert": ref_bert,
-    "ssl_content": audio_prompt_hubert
+    "input_text_phones": input_phones,
+    "input_text_bert": input_bert,
+    "ref_text_phones": ref_phones,
+    "ref_text_bert": ref_bert,
+    "hubert_ssl_content": audio_prompt_hubert
 })
 
 # fsdec = ort.InferenceSession(MODEL_PATH+"_export_t2s_fsdec.onnx")
@@ -120,7 +120,7 @@ ref_audio = waveform.numpy().astype(np.float32)
 vtis = ort.InferenceSession(MODEL_PATH+"_export_vits.onnx")
 
 [audio] = vtis.run(None, {
-    "text_seq": input_phones,
+    "input_text_phones": input_phones,
     "pred_semantic": pred_semantic,
     "spectrum": spectrum.astype(np.float32),
     "sv_emb": sv_emb.astype(np.float32)