Rooba
/
spaCy
mirror of https://github.com/explosion/spaCy.git
1
1
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #6855 from adrianeboyd/docs/trf-sentencepiece [ci skip] 

Update transfomers install docs
This commit is contained in:
Ines Montani 2021-01-29 23:34:01 +11:00 committed by GitHub
parent d600bc6d5d e3e87e7275
commit 095055ac48
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 29 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
website/docs/usage/embeddings-transformers.md
View File

@ -204,14 +204,25 @@ drop-in replacements that let you achieve **higher accuracy** in exchange for
> downloaded: 3GB CUDA runtime, 800MB PyTorch, 400MB CuPy, 500MB weights, 200MB > downloaded: 3GB CUDA runtime, 800MB PyTorch, 400MB CuPy, 500MB weights, 200MB
> spaCy and dependencies. > spaCy and dependencies.
Once you have CUDA installed, you'll need to install two pip packages, Once you have CUDA installed, we recommend installing PyTorch separately
[`cupy`](https://docs.cupy.dev/en/stable/install.html) and following the
[`spacy-transformers`](https://github.com/explosion/spacy-transformers). `cupy` [PyTorch installation guidelines](https://pytorch.org/get-started/locally/) for
is just like `numpy`, but for GPU. The best way to install it is to choose a your package manager and CUDA version. If you skip this step, pip will install
wheel that matches the version of CUDA you're using. You may also need to set PyTorch as a dependency below, but it may not find the best version for your
the `CUDA_PATH` environment variable if your CUDA runtime is installed in a setup.
non-standard location. Putting it all together, if you had installed CUDA 10.2
in `/opt/nvidia/cuda`, you would run: ```bash
### Example: Install PyTorch 1.7.1 for CUDA 10.1 with pip
$ pip install torch==1.7.1+cu101 torchvision==0.8.2+cu101 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html
```
Next, install spaCy with the extras for your CUDA version and transformers. The
CUDA extra (e.g., `cuda92`, `cuda102`, `cuda111`) installs the correct version
of [`cupy`](https://docs.cupy.dev/en/stable/install.html#installing-cupy), which
is just like `numpy`, but for GPU. You may also need to set the `CUDA_PATH`
environment variable if your CUDA runtime is installed in a non-standard
location. Putting it all together, if you had installed CUDA 10.2 in
`/opt/nvidia/cuda`, you would run:
```bash ```bash
### Installation with CUDA ### Installation with CUDA
@ -219,6 +230,16 @@ $ export CUDA_PATH="/opt/nvidia/cuda"
$ pip install -U %%SPACY_PKG_NAME[cuda102,transformers]%%SPACY_PKG_FLAGS $ pip install -U %%SPACY_PKG_NAME[cuda102,transformers]%%SPACY_PKG_FLAGS
``` ```
For [`transformers`](https://huggingface.co/transformers/) v4.0.0+ and models
that require [`SentencePiece`](https://github.com/google/sentencepiece) (e.g.,
ALBERT, CamemBERT, XLNet, Marian, and T5), install the additional dependencies
with:
```bash
### Install sentencepiece
$ pip install transformers[sentencepiece]
```
### Runtime usage {#transformers-runtime} ### Runtime usage {#transformers-runtime}
Transformer models can be used as **drop-in replacements** for other types of Transformer models can be used as **drop-in replacements** for other types of