-
-
I'm using the nightly version, I have successfully trained a transformer based NER model and saved it; now I'm trying to resume training on it.
Firstly, I'm not sure if I have set up the config f…
-
I saw that you have released a variety of different language training ckpts. Could you please tell me how to download and train a new language model in the case of base checkpoint.
-
https://github.com/huangxu1991/GPT-SoVITS-VC
自己做了一些基本的修改後在這裡
`ge = vq_model.ref_enc(spec) # [B, D, T/1]`
會遇到向量不匹配的問題
`mat1 and mat2 shapes cannot be multiplied (273x1025 and 704x128)`
以下為原…
-
python src/extractor_trainer.py --pretrained_bert /data/wangx/models/chinese_roberta_wwm_ext_large --bert_vocab /data/wangx/models/chinese_roberta_wwm_ext_large/vocab.txt --do_train_trigger --data_me…
-
大佬们这是为什么呀,跑步起来
版本
```
torch 1.13.1
cuda 11.7
torchvision 0.16.0
torchaudio 2.1.0
```
运行后报错
```
/home/user/.virtualenvs/zjk_Chinese-CLIP-master/lib/python3.11/site-packages/torch/distributed/…
5zjk5 updated
6 months ago
-
已经做好了切割、标注,点击一键三连时报错,显示没有'logs/xxx/2-name2text-0.txt',但是整个logs文件夹都是他自己生成的啊,我手动创建一个文件,在运行的时候就自己删除了,运行一键三连的终端输出如下:
"/Users/mac/anaconda3/envs/GPTSoVits/bin/python" GPT_SoVITS/prepare_datasets/1-get…
-
I have created the following PR to add support for Thai language.
I am in the process of creating a dataset to train the model but would love a PR review of the code first to make sure I am on the ri…
-
```
WARNING:tensorflow:From /content/558-project/optimization.py:87: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.
WARNING:tensorflow:From 558-project…
-
苏神你好,你在模型中使用的预训练模型roberta是用的brightmart版本,还是hungging face上hfl的再.h5模型再转换为ckpt版本,brightmart版本是基于256字符串长度训练的,对于512长度的文本似乎效果不太好,但是看你的博客说你是用的512版本的