Open Holmose opened 4 years ago
添加了还是报错
添加了还是报错
解决方法已更新
最后一步出错的小伙伴 1.访问网址:https://drive.google.com/file/d/1ajR-yVWmZC_z7HgNjz4tivNz-vUCgKBC 复制model.ckpt-280000.data-00000-of-00001文件到google云盘根目录
2.复制下列代码到第二步代码之下,点击左上角运行执行 from google.colab import drive drive.mount('/content/drive') !cp /content/drive/My\ Drive/model.ckpt-280000.data-00000-of-00001 trained_models/model.ckpt-280000.data-00000-of-00001
3.执行第三步看是否能够成功加载
2021-02-21按照以上步骤成功加载,并且成功在Vmware中复现 环境:
https://www.cnblogs.com/ningvsban/p/4384995.html 👈安装指定版本Python
Python-3.6.5
pip install pandas==0.24.2 -i https://pypi.douban.com/simple pip install regex==2019.4.14 -i https://pypi.douban.com/simple pip install h5py==2.10.0 -i https://pypi.douban.com/simple pip install numpy==1.18.4 -i https://pypi.douban.com/simple pip install tensorboard==1.15.0 -i https://pypi.douban.com/simple pip --default-timeout=100 install --upgrade -i https://mirrors.aliyun.com/pypi/simple tensorflow==1.15.0 pip install tensorflow-estimator==1.15.1 -i https://pypi.douban.com/simple pip install tqdm==4.31.1 -i https://pypi.douban.com/simple pip install requests==2.22.0 -i https://pypi.douban.com/simple pip install ujson==2.0.3 -i https://pypi.douban.com/simple pip install launchpadlib==1.10.13 -i https://pypi.douban.com/simple
最后一步出错的小伙伴 1.访问网址:https://drive.google.com/file/d/1ajR-yVWmZC_z7HgNjz4tivNz-vUCgKBC 复制model.ckpt-280000.data-00000-of-00001文件到google云盘根目录
2.复制下列代码到第二步代码之下,点击左上角运行执行 from google.colab import drive drive.mount('/content/drive') !cp /content/drive/My\ Drive/model.ckpt-280000.data-00000-of-00001 trained_models/model.ckpt-280000.data-00000-of-00001
3.执行第三步看是否能够成功加载
2021-02-21按照以上步骤成功加载,并且成功在Vmware中复现 环境:
https://www.cnblogs.com/ningvsban/p/4384995.html 👈安装指定版本Python
Python-3.6.5
pip install pandas==0.24.2 -i https://pypi.douban.com/simple pip install regex==2019.4.14 -i https://pypi.douban.com/simple pip install h5py==2.10.0 -i https://pypi.douban.com/simple pip install numpy==1.18.4 -i https://pypi.douban.com/simple pip install tensorboard==1.15.0 -i https://pypi.douban.com/simple pip --default-timeout=100 install --upgrade -i https://mirrors.aliyun.com/pypi/simple tensorflow==1.15.0 pip install tensorflow-estimator==1.15.1 -i https://pypi.douban.com/simple pip install tqdm==4.31.1 -i https://pypi.douban.com/simple pip install requests==2.22.0 -i https://pypi.douban.com/simple pip install ujson==2.0.3 -i https://pypi.douban.com/simple pip install launchpadlib==1.10.13 -i https://pypi.douban.com/simple
这个方法还是可以 2021年6月7日16:47:41
好帖
我好像找到原因了:作者的模型文件是放在 Google Drive 上的,而 Google Drive 对每个文件每天的下载次数有限制,人太多导致配额超限,于是模型下载失败了。
至于解决方法......谷歌的限额没办法绕过,其它网盘限速,要不存阿里云盘,然后我们自己下载了去运行?
这模型也忒大了
这模型也忒大了
十七亿参数啊
最后一步出错的小伙伴 1.访问网址:https://drive.google.com/file/d/1ajR-yVWmZC_z7HgNjz4tivNz-vUCgKBC 复制model.ckpt-280000.data-00000-of-00001文件到google云盘根目录
2.复制下列代码到第二步代码之下,点击左上角运行执行 from google.colab import drive drive.mount('/content/drive') !cp /content/drive/My\ Drive/model.ckpt-280000.data-00000-of-00001 trained_models/model.ckpt-280000.data-00000-of-00001
3.执行第三步看是否能够成功加载
2021-02-21按照以上步骤成功加载,并且成功在Vmware中复现 环境:
https://www.cnblogs.com/ningvsban/p/4384995.html 👈安装指定版本Python
Python-3.6.5
pip install pandas==0.24.2 -i https://pypi.douban.com/simple pip install regex==2019.4.14 -i https://pypi.douban.com/simple pip install h5py==2.10.0 -i https://pypi.douban.com/simple pip install numpy==1.18.4 -i https://pypi.douban.com/simple pip install tensorboard==1.15.0 -i https://pypi.douban.com/simple pip --default-timeout=100 install --upgrade -i https://mirrors.aliyun.com/pypi/simple tensorflow==1.15.0 pip install tensorflow-estimator==1.15.1 -i https://pypi.douban.com/simple pip install tqdm==4.31.1 -i https://pypi.douban.com/simple pip install requests==2.22.0 -i https://pypi.douban.com/simple pip install ujson==2.0.3 -i https://pypi.douban.com/simple pip install launchpadlib==1.10.13 -i https://pypi.douban.com/simple
打卡记录,2022年6月可用 !cp /content/drive/MyDrive/model.ckpt-280000.data-00000-of-00001 /home/EssayKiller_V2/LanguageNetwork/GPT2/finetune/trained_models/model.ckpt-280000.data-00000-of-00001
这是由于模型文件没有下载完成的缘故,5个多g的文件, 清空finetune/trained_models下面的文件多运行几次(第二 步操作)试试, 或者直接下载文件到trained_models目录下面,也可试试从挂载google云盘从云盘上面复制过去。