From ba155e4e2f582de28abed230aca2e79191ecbc51 Mon Sep 17 00:00:00 2001
From: James Betker <jbetker@gmail.com>
Date: Sun, 27 Feb 2022 14:48:38 -0700
Subject: [PATCH] script for uploading models to the HF hub

---
 codes/scripts/hugging_face_hub_upload.py | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 codes/scripts/hugging_face_hub_upload.py

diff --git a/codes/scripts/hugging_face_hub_upload.py b/codes/scripts/hugging_face_hub_upload.py
new file mode 100644
index 00000000..b05784c6
--- /dev/null
+++ b/codes/scripts/hugging_face_hub_upload.py
@@ -0,0 +1,22 @@
+import json
+
+import torch
+
+from models.asr.w2v_wrapper import Wav2VecWrapper
+from models.tacotron2.text import tacotron_symbol_mapping
+
+if __name__ == '__main__':
+    """
+    Utility script for uploading model weights to the HF hub
+    """
+
+    '''
+    model = Wav2VecWrapper(vocab_size=148, basis_model='facebook/wav2vec2-large-robust-ft-libri-960h', freeze_transformer=True, checkpointing_enabled=False)
+    weights = torch.load('D:\\dlas\\experiments\\train_wav2vec_mass_large2\\models\\22500_wav2vec.pth')
+    model.load_state_dict(weights)
+    model.w2v.save_pretrained("jbetker/wav2vec2-large-robust-ft-libritts-voxpopuli", push_to_hub=True)
+    '''
+
+    # Build tokenizer vocab
+    mapping = tacotron_symbol_mapping()
+    print(json.dumps(mapping))
\ No newline at end of file