有人在 Conda 环境中安装了拥抱脸转换器吗？

Question

标题。在这一点上，我基本上尝试了 TF 2.0、2.1、tensorflow-gpu（2.1 和 2.0）和最新的 huggingface 构建的所有组合。此外，每次尝试都从头开始创建新环境。

我可以：

• 安装 tensorflow-gpu（2.0 或 2.1）-> TF 确实找到了我的 GPU，但随后 huggingface 找不到单个 TF 模型。例如，我可以import AutoModel 很好，但我不能import TFAutoModel（错误：ImportError: cannot import name 'TFAutoModel' from 'transformers' (unknown location)）。我尝试的所有 TF* 都会发生这种情况。

或

• 安装 tensorflow（2.0 或 2.1）-> TF 找不到我的 GPU，但我可以 import TFAutoModel 没有问题。

我还安装了所需的 CUDA 和 cuDNN 版本，并拥有一个具有足够计算能力 yadda yadda 的 GPU。

TL;DR：如何让 huggingface TF 模型与 TF GPU 配合使用？

Answer 1

在我的 Windows 机器上，我可以简单地这样做：

conda create -n tf tensorflow-gpu
conda activate tf
conda install -c huggingface

这安装了 tensorflow-gpu 版本 2.3.0 和转换器 4.2.2

然后我可以运行来自their website 的示例代码之一：

(tf) C:\> python
Python 3.8.5 (default, Sep  3 2020, 21:29:08) [MSC v.1916 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> from transformers import pipeline
>>> nlp = pipeline("sentiment-analysis")
Downloading: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 629/629 [00:00<00:00, 180kB/s]
Downloading: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 268M/268M [00:21<00:00, 12.5MB/s]
2021-02-03 08:59:08.884730: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN)to use the following CPU instructions in performance-critical operations:  AVX AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-02-03 08:59:09.109504: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.
Some layers from the model checkpoint at distilbert-base-uncased-finetuned-sst-2-english were not used when initializing TFDistilBertModel: ['pre_classifier', 'dropout_19', 'classifier']
- This IS expected if you are initializing TFDistilBertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing TFDistilBertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
All the layers of TFDistilBertModel were initialized from the model checkpoint at distilbert-base-uncased-finetuned-sst-2-english.
If your task is similar to the task the model of the checkpoint was trained on, you can already use TFDistilBertModel for predictions without further training.
Downloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 232k/232k [00:00<00:00, 528kB/s]
Some layers from the model checkpoint at distilbert-base-uncased-finetuned-sst-2-english were not used when initializing TFDistilBertForSequenceClassification: ['dropout_19']
- This IS expected if you are initializing TFDistilBertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing TFDistilBertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some layers of TFDistilBertForSequenceClassification were not initialized from the model checkpoint at distilbert-base-uncased-finetuned-sst-2-english and are newly initialized: ['dropout_38']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
>>> result = nlp("I hate you")[0]
>>> print(f"label: {result['label']}, with score: {round(result['score'], 4)}")
label: NEGATIVE, with score: 0.9991
>>> result = nlp("I love you")[0]
>>> print(f"label: {result['label']}, with score: {round(result['score'], 4)}")
label: POSITIVE, with score: 0.9999

以及您提到的导入：

>>> from transformers import TFAutoModel
>>>