Upload insert_symbol.ipynb
Browse files- insert_symbol.ipynb +61 -0
insert_symbol.ipynb
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"cells": [
|
| 3 |
+
{
|
| 4 |
+
"cell_type": "code",
|
| 5 |
+
"execution_count": 2,
|
| 6 |
+
"metadata": {},
|
| 7 |
+
"outputs": [],
|
| 8 |
+
"source": [
|
| 9 |
+
"import torch\n",
|
| 10 |
+
"#### 迁移gpt模型的embedding层,插入两个新的token\n",
|
| 11 |
+
"dict = torch.load(r\"D:\\pyprojs\\GPT-SoVITSs\\fork\\GPT-SoVITS\\GPT_SoVITS\\pretrained_models\\s1bert25hz-2kh-longer-epoch=68e-step=50232-2.ckpt\",map_location=torch.device('cpu'))\n",
|
| 12 |
+
"dict[\"weight\"][\"model.ar_text_embedding.word_embeddings.weight\"].shape\n",
|
| 13 |
+
"# 0~94保持不动,95~向后移动两位,新的95 96随机初始化\n",
|
| 14 |
+
"# 一共512列,embedding_dim=5\n",
|
| 15 |
+
"first_part = dict[\"weight\"][\"model.ar_text_embedding.word_embeddings.weight\"][:95,:]\n",
|
| 16 |
+
"second_part = dict[\"weight\"][\"model.ar_text_embedding.word_embeddings.weight\"][95:-2,:]\n",
|
| 17 |
+
"new_weight = torch.cat([first_part,second_part, torch.randn(2, 512)],dim=0)\n",
|
| 18 |
+
"dict[\"weight\"][\"model.ar_text_embedding.word_embeddings.weight\"] = new_weight\n",
|
| 19 |
+
"torch.save(dict,r\"D:\\pyprojs\\GPT-SoVITSs\\fork\\GPT-SoVITS\\GPT_SoVITS\\pretrained_models\\s1bert25hz-hs_model.ckpt\")"
|
| 20 |
+
]
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"cell_type": "code",
|
| 24 |
+
"execution_count": 1,
|
| 25 |
+
"metadata": {},
|
| 26 |
+
"outputs": [],
|
| 27 |
+
"source": [
|
| 28 |
+
"import torch\n",
|
| 29 |
+
"dict2 = torch.load(r\"D:\\pyprojs\\GPT-SoVITSs\\fork\\GPT-SoVITS\\GPT_SoVITS\\pretrained_models\\s2G488k-legacy.pth\",map_location=torch.device('cpu'))\n",
|
| 30 |
+
"dict2[\"weight\"][\"enc_p.text_embedding.weight\"].shape\n",
|
| 31 |
+
"# 0~94保持不动,95~向后移动两位,新的95 96随机初始化\n",
|
| 32 |
+
"first_part = dict2[\"weight\"][\"enc_p.text_embedding.weight\"][:95,:]\n",
|
| 33 |
+
"second_part = dict2[\"weight\"][\"enc_p.text_embedding.weight\"][95:,:]\n",
|
| 34 |
+
"new_weight = torch.cat([first_part,torch.zeros(2,192),second_part],dim=0)\n",
|
| 35 |
+
"dict2[\"weight\"][\"enc_p.text_embedding.weight\"] = new_weight\n",
|
| 36 |
+
"torch.save(dict2,r\"D:\\pyprojs\\GPT-SoVITSs\\fork\\GPT-SoVITS\\GPT_SoVITS\\pretrained_models\\s2G488k.pth\")"
|
| 37 |
+
]
|
| 38 |
+
}
|
| 39 |
+
],
|
| 40 |
+
"metadata": {
|
| 41 |
+
"kernelspec": {
|
| 42 |
+
"display_name": "vits",
|
| 43 |
+
"language": "python",
|
| 44 |
+
"name": "python3"
|
| 45 |
+
},
|
| 46 |
+
"language_info": {
|
| 47 |
+
"codemirror_mode": {
|
| 48 |
+
"name": "ipython",
|
| 49 |
+
"version": 3
|
| 50 |
+
},
|
| 51 |
+
"file_extension": ".py",
|
| 52 |
+
"mimetype": "text/x-python",
|
| 53 |
+
"name": "python",
|
| 54 |
+
"nbconvert_exporter": "python",
|
| 55 |
+
"pygments_lexer": "ipython3",
|
| 56 |
+
"version": "3.10.0"
|
| 57 |
+
}
|
| 58 |
+
},
|
| 59 |
+
"nbformat": 4,
|
| 60 |
+
"nbformat_minor": 2
|
| 61 |
+
}
|