Skip to content

Commit

Permalink
TF: purge TFTrainer (huggingface#28483)
Browse files Browse the repository at this point in the history
  • Loading branch information
gante authored and MadElf1337 committed Jan 15, 2024
1 parent deacbf9 commit 3a0702b
Show file tree
Hide file tree
Showing 15 changed files with 240 additions and 1,689 deletions.
1 change: 0 additions & 1 deletion docs/source/en/main_classes/deepspeed.md
Original file line number Diff line number Diff line change
Expand Up @@ -2049,7 +2049,6 @@ In this case you usually need to raise the value of `initial_scale_power`. Setti
### Notes
- DeepSpeed works with the PyTorch [`Trainer`] but not TF [`TFTrainer`].
- While DeepSpeed has a pip installable PyPI package, it is highly recommended that it gets installed from [source](https://github.com/microsoft/deepspeed#installation) to best match your hardware and also if you need to enable
certain features, like 1-bit Adam, which aren't available in the pypi distribution.
- You don't have to use the [`Trainer`] to use DeepSpeed with 🤗 Transformers - you can use any model
Expand Down
469 changes: 231 additions & 238 deletions docs/source/it/migration.md

Large diffs are not rendered by default.

3 changes: 1 addition & 2 deletions docs/source/ja/main_classes/deepspeed.md
Original file line number Diff line number Diff line change
Expand Up @@ -1910,7 +1910,7 @@ SW: Model with 2783M total params, 65M largest layer params.

3. 次の出力:

```bash
```bash
python -c 'import torch; print(f"torch: {torch.__version__}")'
python -c 'import transformers; print(f"transformers: {transformers.__version__}")'
python -c 'import deepspeed; print(f"deepspeed: {deepspeed.__version__}")'
Expand Down Expand Up @@ -1994,7 +1994,6 @@ SW: Model with 2783M total params, 65M largest layer params.
### Notes
- DeepSpeed は PyTorch [`Trainer`] では動作しますが、TF [`TFTrainer`] では動作しません。
- DeepSpeed には pip でインストール可能な PyPI パッケージがありますが、ハードウェアに最も適合するように、また有効にする必要がある場合は、[ソース](https://github.com/microsoft/deepspeed#installation) からインストールすることを強くお勧めします。
1 ビット Adam などの特定の機能は、pypi ディストリビューションでは利用できません。
- 🤗 Transformers で DeepSpeed を使用するために [`Trainer`] を使用する必要はありません - 任意のモデルを使用できます
Expand Down
3 changes: 1 addition & 2 deletions docs/source/zh/main_classes/deepspeed.md
Original file line number Diff line number Diff line change
Expand Up @@ -249,7 +249,7 @@ recommend ZeRO-3 config as starting one. -->
注意:

- 如果您需要在特定的 GPU 上运行,而不是 GPU 0,则无法使用 `CUDA_VISIBLE_DEVICES` 来限制可用 GPU 的可见范围。相反,您必须使用以下语法:

```bash
deepspeed --include localhost:1 examples/pytorch/translation/run_translation.py ...
```
Expand Down Expand Up @@ -1845,7 +1845,6 @@ SW: Model with 2783M total params, 65M largest layer params.
### 注意事项
- DeepSpeed 与 PyTorch [`Trainer`] 一起工作,但不与 TF [`TFTrainer`] 一起工作。
- 尽管 DeepSpeed 有一个可安装的 PyPI 包,但强烈建议从源代码安装它,以最好地匹配您的硬件,如果您需要启用某些功能,如 1-bit Adam,这些功能在 pypi 发行版中不可用。
- 您不必使用🤗 Transformers的 [`Trainer`] 来使用 DeepSpeed - 您可以使用任何模型与自己的训练器,您还需要根据 [DeepSpeed 集成说明](https://www.deepspeed.ai/getting-started/#writing-deepspeed-models) 调整后者。
Expand Down
313 changes: 0 additions & 313 deletions examples/legacy/text-classification/run_tf_text_classification.py

This file was deleted.

Loading

0 comments on commit 3a0702b

Please sign in to comment.