SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Indent code block in the documentation (#11233)

Browse Source 
* Indent code block

* Indent code blocks version 2

* Quality
This commit is contained in:
Sylvain Gugger
2021-04-13 15:36:36 -04:00
committed by GitHub
parent 9d8e8a8703
commit f38cd4373f
16 changed files with 382 additions and 357 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
docs/source/add_new_model.rst
View File

@@ -388,7 +388,7 @@ Next, you can finally start adding new code to 🤗 Transformers. Go into the cl
:: ::
cd transformers cd transformers
In the special case that you are adding a model whose architecture exactly matches the model architecture of an In the special case that you are adding a model whose architecture exactly matches the model architecture of an
existing model you only have to add a conversion script as described in `this section <#write-a-conversion-script>`__. existing model you only have to add a conversion script as described in `this section <#write-a-conversion-script>`__.
@@ -417,27 +417,27 @@ You should do the following:
:: ::
git checkout -b add_brand_new_bert git checkout -b add_brand_new_bert
2. Commit the automatically generated code: 2. Commit the automatically generated code:
:: ::
git add . git add .
git commit git commit
3. Fetch and rebase to current master 3. Fetch and rebase to current master
:: ::
git fetch upstream git fetch upstream
git rebase upstream/master git rebase upstream/master
4. Push the changes to your account using: 4. Push the changes to your account using:
:: ::
git push -u origin a-descriptive-name-for-my-changes git push -u origin a-descriptive-name-for-my-changes
5. Once you are satisfied, go to the webpage of your fork on GitHub. Click on “Pull request”. Make sure to add the 5. Once you are satisfied, go to the webpage of your fork on GitHub. Click on “Pull request”. Make sure to add the
GitHub handle of some members of the Hugging Face team as reviewers, so that the Hugging Face team gets notified for GitHub handle of some members of the Hugging Face team as reviewers, so that the Hugging Face team gets notified for
@@ -451,8 +451,8 @@ time to time by doing:
:: ::
git fetch upstream git fetch upstream
git merge upstream/master git merge upstream/master
In general, all questions you might have regarding the model or your implementation should be asked in your PR and In general, all questions you might have regarding the model or your implementation should be asked in your PR and
discussed/solved in the PR. This way, the Hugging Face team will always be notified when you are committing new code or discussed/solved in the PR. This way, the Hugging Face team will always be notified when you are committing new code or

92
docs/source/converting_tensorflow_models.rst
View File

@@ -47,12 +47,12 @@ Here is an example of the conversion process for a pre-trained ``BERT-Base Uncas
.. code-block:: shell .. code-block:: shell
export BERT_BASE_DIR=/path/to/bert/uncased_L-12_H-768_A-12 export BERT_BASE_DIR=/path/to/bert/uncased_L-12_H-768_A-12
transformers-cli convert --model_type bert \ transformers-cli convert --model_type bert \
--tf_checkpoint $BERT_BASE_DIR/bert_model.ckpt \ --tf_checkpoint $BERT_BASE_DIR/bert_model.ckpt \
--config $BERT_BASE_DIR/bert_config.json \ --config $BERT_BASE_DIR/bert_config.json \
--pytorch_dump_output $BERT_BASE_DIR/pytorch_model.bin --pytorch_dump_output $BERT_BASE_DIR/pytorch_model.bin
You can download Google's pre-trained models for the conversion `here You can download Google's pre-trained models for the conversion `here
<https://github.com/google-research/bert#pre-trained-models>`__. <https://github.com/google-research/bert#pre-trained-models>`__.
@@ -72,12 +72,12 @@ Here is an example of the conversion process for the pre-trained ``ALBERT Base``
.. code-block:: shell .. code-block:: shell
export ALBERT_BASE_DIR=/path/to/albert/albert_base export ALBERT_BASE_DIR=/path/to/albert/albert_base
transformers-cli convert --model_type albert \ transformers-cli convert --model_type albert \
--tf_checkpoint $ALBERT_BASE_DIR/model.ckpt-best \ --tf_checkpoint $ALBERT_BASE_DIR/model.ckpt-best \
--config $ALBERT_BASE_DIR/albert_config.json \ --config $ALBERT_BASE_DIR/albert_config.json \
--pytorch_dump_output $ALBERT_BASE_DIR/pytorch_model.bin --pytorch_dump_output $ALBERT_BASE_DIR/pytorch_model.bin
You can download Google's pre-trained models for the conversion `here You can download Google's pre-trained models for the conversion `here
<https://github.com/google-research/albert#pre-trained-models>`__. <https://github.com/google-research/albert#pre-trained-models>`__.
@@ -91,13 +91,13 @@ save as the same format than OpenAI pretrained model (see `here <https://github.
.. code-block:: shell .. code-block:: shell
export OPENAI_GPT_CHECKPOINT_FOLDER_PATH=/path/to/openai/pretrained/numpy/weights export OPENAI_GPT_CHECKPOINT_FOLDER_PATH=/path/to/openai/pretrained/numpy/weights
transformers-cli convert --model_type gpt \ transformers-cli convert --model_type gpt \
--tf_checkpoint $OPENAI_GPT_CHECKPOINT_FOLDER_PATH \ --tf_checkpoint $OPENAI_GPT_CHECKPOINT_FOLDER_PATH \
--pytorch_dump_output $PYTORCH_DUMP_OUTPUT \ --pytorch_dump_output $PYTORCH_DUMP_OUTPUT \
[--config OPENAI_GPT_CONFIG] \ [--config OPENAI_GPT_CONFIG] \
[--finetuning_task_name OPENAI_GPT_FINETUNED_TASK] \ [--finetuning_task_name OPENAI_GPT_FINETUNED_TASK] \
OpenAI GPT-2 OpenAI GPT-2
@@ -108,13 +108,13 @@ Here is an example of the conversion process for a pre-trained OpenAI GPT-2 mode
.. code-block:: shell .. code-block:: shell
export OPENAI_GPT2_CHECKPOINT_PATH=/path/to/gpt2/pretrained/weights export OPENAI_GPT2_CHECKPOINT_PATH=/path/to/gpt2/pretrained/weights
transformers-cli convert --model_type gpt2 \ transformers-cli convert --model_type gpt2 \
--tf_checkpoint $OPENAI_GPT2_CHECKPOINT_PATH \ --tf_checkpoint $OPENAI_GPT2_CHECKPOINT_PATH \
--pytorch_dump_output $PYTORCH_DUMP_OUTPUT \ --pytorch_dump_output $PYTORCH_DUMP_OUTPUT \
[--config OPENAI_GPT2_CONFIG] \ [--config OPENAI_GPT2_CONFIG] \
[--finetuning_task_name OPENAI_GPT2_FINETUNED_TASK] [--finetuning_task_name OPENAI_GPT2_FINETUNED_TASK]
Transformer-XL Transformer-XL
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -124,13 +124,13 @@ Here is an example of the conversion process for a pre-trained Transformer-XL mo
.. code-block:: shell .. code-block:: shell
export TRANSFO_XL_CHECKPOINT_FOLDER_PATH=/path/to/transfo/xl/checkpoint export TRANSFO_XL_CHECKPOINT_FOLDER_PATH=/path/to/transfo/xl/checkpoint
transformers-cli convert --model_type transfo_xl \ transformers-cli convert --model_type transfo_xl \
--tf_checkpoint $TRANSFO_XL_CHECKPOINT_FOLDER_PATH \ --tf_checkpoint $TRANSFO_XL_CHECKPOINT_FOLDER_PATH \
--pytorch_dump_output $PYTORCH_DUMP_OUTPUT \ --pytorch_dump_output $PYTORCH_DUMP_OUTPUT \
[--config TRANSFO_XL_CONFIG] \ [--config TRANSFO_XL_CONFIG] \
[--finetuning_task_name TRANSFO_XL_FINETUNED_TASK] [--finetuning_task_name TRANSFO_XL_FINETUNED_TASK]
XLNet XLNet
@@ -140,14 +140,14 @@ Here is an example of the conversion process for a pre-trained XLNet model:
.. code-block:: shell .. code-block:: shell
export TRANSFO_XL_CHECKPOINT_PATH=/path/to/xlnet/checkpoint export TRANSFO_XL_CHECKPOINT_PATH=/path/to/xlnet/checkpoint
export TRANSFO_XL_CONFIG_PATH=/path/to/xlnet/config export TRANSFO_XL_CONFIG_PATH=/path/to/xlnet/config
transformers-cli convert --model_type xlnet \ transformers-cli convert --model_type xlnet \
--tf_checkpoint $TRANSFO_XL_CHECKPOINT_PATH \ --tf_checkpoint $TRANSFO_XL_CHECKPOINT_PATH \
--config $TRANSFO_XL_CONFIG_PATH \ --config $TRANSFO_XL_CONFIG_PATH \
--pytorch_dump_output $PYTORCH_DUMP_OUTPUT \ --pytorch_dump_output $PYTORCH_DUMP_OUTPUT \
[--finetuning_task_name XLNET_FINETUNED_TASK] \ [--finetuning_task_name XLNET_FINETUNED_TASK] \
XLM XLM
@@ -157,13 +157,13 @@ Here is an example of the conversion process for a pre-trained XLM model:
.. code-block:: shell .. code-block:: shell
export XLM_CHECKPOINT_PATH=/path/to/xlm/checkpoint export XLM_CHECKPOINT_PATH=/path/to/xlm/checkpoint
transformers-cli convert --model_type xlm \ transformers-cli convert --model_type xlm \
--tf_checkpoint $XLM_CHECKPOINT_PATH \ --tf_checkpoint $XLM_CHECKPOINT_PATH \
--pytorch_dump_output $PYTORCH_DUMP_OUTPUT --pytorch_dump_output $PYTORCH_DUMP_OUTPUT
[--config XML_CONFIG] \ [--config XML_CONFIG] \
[--finetuning_task_name XML_FINETUNED_TASK] [--finetuning_task_name XML_FINETUNED_TASK]
T5 T5
@@ -173,9 +173,9 @@ Here is an example of the conversion process for a pre-trained T5 model:
.. code-block:: shell .. code-block:: shell
export T5=/path/to/t5/uncased_L-12_H-768_A-12 export T5=/path/to/t5/uncased_L-12_H-768_A-12
transformers-cli convert --model_type t5 \ transformers-cli convert --model_type t5 \
--tf_checkpoint $T5/t5_model.ckpt \ --tf_checkpoint $T5/t5_model.ckpt \
--config $T5/t5_config.json \ --config $T5/t5_config.json \
--pytorch_dump_output $T5/pytorch_model.bin --pytorch_dump_output $T5/pytorch_model.bin

2
docs/source/glossary.rst
View File

@@ -182,7 +182,7 @@ such:
.. code-block:: .. code-block::
>>> # [CLS] SEQUENCE_A [SEP] SEQUENCE_B [SEP] >>> # [CLS] SEQUENCE_A [SEP] SEQUENCE_B [SEP]
We can use our tokenizer to automatically generate such a sentence by passing the two sequences to ``tokenizer`` as two We can use our tokenizer to automatically generate such a sentence by passing the two sequences to ``tokenizer`` as two
arguments (and not a list, like before) like this: arguments (and not a list, like before) like this:

316
docs/source/main_classes/trainer.rst
View File

@@ -293,33 +293,33 @@ with it, you may want to try one of:
.. code-block:: bash .. code-block:: bash
pip install fairscale --no-build-isolation . pip install fairscale --no-build-isolation .
or: or:
.. code-block:: bash .. code-block:: bash
git clone https://github.com/facebookresearch/fairscale/ git clone https://github.com/facebookresearch/fairscale/
cd fairscale cd fairscale
rm -r dist build rm -r dist build
python setup.py bdist_wheel python setup.py bdist_wheel
pip uninstall -y fairscale pip uninstall -y fairscale
pip install dist/fairscale-*.whl pip install dist/fairscale-*.whl
``fairscale`` also has issues with building against pytorch-nightly, so if you use it you may have to try one of: ``fairscale`` also has issues with building against pytorch-nightly, so if you use it you may have to try one of:
.. code-block:: bash .. code-block:: bash
pip uninstall -y fairscale; pip install fairscale --pre \ pip uninstall -y fairscale; pip install fairscale --pre \
-f https://download.pytorch.org/whl/nightly/cu110/torch_nightly.html \ -f https://download.pytorch.org/whl/nightly/cu110/torch_nightly.html \
--no-cache --no-build-isolation --no-cache --no-build-isolation
or: or:
.. code-block:: bash .. code-block:: bash
pip install -v --disable-pip-version-check . \ pip install -v --disable-pip-version-check . \
-f https://download.pytorch.org/whl/nightly/cu110/torch_nightly.html --pre -f https://download.pytorch.org/whl/nightly/cu110/torch_nightly.html --pre
Of course, adjust the urls to match the cuda version you use. Of course, adjust the urls to match the cuda version you use.
@@ -447,12 +447,12 @@ To make a local build for DeepSpeed:
.. code-block:: bash .. code-block:: bash
git clone https://github.com/microsoft/DeepSpeed/ git clone https://github.com/microsoft/DeepSpeed/
cd DeepSpeed cd DeepSpeed
rm -rf build rm -rf build
TORCH_CUDA_ARCH_LIST="6.1;8.6" DS_BUILD_OPS=1 pip install . \ TORCH_CUDA_ARCH_LIST="6.1;8.6" DS_BUILD_OPS=1 pip install . \
--global-option="build_ext" --global-option="-j8" --no-cache -v \ --global-option="build_ext" --global-option="-j8" --no-cache -v \
--disable-pip-version-check 2>&1 | tee build.log --disable-pip-version-check 2>&1 | tee build.log
Edit ``TORCH_CUDA_ARCH_LIST`` to insert the code for the architectures of the GPU cards you intend to use. Edit ``TORCH_CUDA_ARCH_LIST`` to insert the code for the architectures of the GPU cards you intend to use.
@@ -460,11 +460,11 @@ Or if you need to use the same setup on multiple machines, make a binary wheel:
.. code-block:: bash .. code-block:: bash
git clone https://github.com/microsoft/DeepSpeed/ git clone https://github.com/microsoft/DeepSpeed/
cd DeepSpeed cd DeepSpeed
rm -rf build rm -rf build
TORCH_CUDA_ARCH_LIST="6.1;8.6" DS_BUILD_OPS=1 \ TORCH_CUDA_ARCH_LIST="6.1;8.6" DS_BUILD_OPS=1 \
python setup.py build_ext -j8 bdist_wheel python setup.py build_ext -j8 bdist_wheel
it will generate something like ``dist/deepspeed-0.3.13+8cd046f-cp38-cp38-linux_x86_64.whl`` which now you can install it will generate something like ``dist/deepspeed-0.3.13+8cd046f-cp38-cp38-linux_x86_64.whl`` which now you can install
as ``pip install deepspeed-0.3.13+8cd046f-cp38-cp38-linux_x86_64.whl`` locally or on any other machine. as ``pip install deepspeed-0.3.13+8cd046f-cp38-cp38-linux_x86_64.whl`` locally or on any other machine.
@@ -478,20 +478,20 @@ You can check the archs pytorch was built with using:
.. code-block:: bash .. code-block:: bash
python -c "import torch; print(torch.cuda.get_arch_list())" python -c "import torch; print(torch.cuda.get_arch_list())"
Here is how to find out the arch for one of the installed GPU. For example, for GPU 0: Here is how to find out the arch for one of the installed GPU. For example, for GPU 0:
.. code-block:: bash .. code-block:: bash
CUDA_VISIBLE_DEVICES=0 python -c "import torch; \ CUDA_VISIBLE_DEVICES=0 python -c "import torch; \
print(torch.cuda.get_device_properties(torch.device('cuda')))" print(torch.cuda.get_device_properties(torch.device('cuda')))"
If the output is: If the output is:
.. code-block:: bash .. code-block:: bash
_CudaDeviceProperties(name='GeForce RTX 3090', major=8, minor=6, total_memory=24268MB, multi_processor_count=82) _CudaDeviceProperties(name='GeForce RTX 3090', major=8, minor=6, total_memory=24268MB, multi_processor_count=82)
then you know that this card's arch is ``8.6``. then you know that this card's arch is ``8.6``.
@@ -591,18 +591,18 @@ with DeepSpeed is to have at least the following configuration in the configurat
.. code-block:: json .. code-block:: json
{ {
"zero_optimization": { "zero_optimization": {
"stage": 2, "stage": 2,
"allgather_partitions": true, "allgather_partitions": true,
"allgather_bucket_size": 2e8, "allgather_bucket_size": 2e8,
"reduce_scatter": true, "reduce_scatter": true,
"reduce_bucket_size": 2e8, "reduce_bucket_size": 2e8,
"overlap_comm": true, "overlap_comm": true,
"contiguous_gradients": true, "contiguous_gradients": true,
"cpu_offload": true "cpu_offload": true
}, },
} }
which enables ``cpu_offload`` and some other important features. You may experiment with the buffer sizes, you will which enables ``cpu_offload`` and some other important features. You may experiment with the buffer sizes, you will
find more details in the discussion below. find more details in the discussion below.
@@ -710,18 +710,18 @@ shell from a cell. For example, to use ``run_translation.py`` you would launch i
.. code-block:: .. code-block::
!git clone https://github.com/huggingface/transformers !git clone https://github.com/huggingface/transformers
!cd transformers; deepspeed examples/seq2seq/run_translation.py ... !cd transformers; deepspeed examples/seq2seq/run_translation.py ...
or with ``%%bash`` magic, where you can write a multi-line code for the shell program to run: or with ``%%bash`` magic, where you can write a multi-line code for the shell program to run:
.. code-block:: .. code-block::
%%bash %%bash
git clone https://github.com/huggingface/transformers git clone https://github.com/huggingface/transformers
cd transformers cd transformers
deepspeed examples/seq2seq/run_translation.py ... deepspeed examples/seq2seq/run_translation.py ...
In such case you don't need any of the code presented at the beginning of this section. In such case you don't need any of the code presented at the beginning of this section.
@@ -743,16 +743,16 @@ repo <https://github.com/microsoft/DeepSpeedExamples>`__:
.. code-block:: bash .. code-block:: bash
git clone https://github.com/microsoft/DeepSpeedExamples git clone https://github.com/microsoft/DeepSpeedExamples
cd DeepSpeedExamples cd DeepSpeedExamples
find . -name '*json' find . -name '*json'
Continuing the code from above, let's say you're looking to configure the Lamb optimizer. So you can search through the Continuing the code from above, let's say you're looking to configure the Lamb optimizer. So you can search through the
example ``.json`` files with: example ``.json`` files with:
.. code-block:: bash .. code-block:: bash
grep -i Lamb $(find . -name '*json') grep -i Lamb $(find . -name '*json')
Some more examples are to be found in the `main repo <https://github.com/microsoft/DeepSpeed>`__ as well. Some more examples are to be found in the `main repo <https://github.com/microsoft/DeepSpeed>`__ as well.
@@ -1020,49 +1020,49 @@ Here is a full ZeRO-2 all-enabled configuration file ``ds_config_zero2.json``:
.. code-block:: json .. code-block:: json
{ {
"fp16": { "fp16": {
"enabled": true, "enabled": true,
"loss_scale": 0, "loss_scale": 0,
"loss_scale_window": 1000, "loss_scale_window": 1000,
"initial_scale_power": 16, "initial_scale_power": 16,
"hysteresis": 2, "hysteresis": 2,
"min_loss_scale": 1 "min_loss_scale": 1
}, },
"zero_optimization": { "zero_optimization": {
"stage": 2, "stage": 2,
"allgather_partitions": true, "allgather_partitions": true,
"allgather_bucket_size": 2e8, "allgather_bucket_size": 2e8,
"overlap_comm": true, "overlap_comm": true,
"reduce_scatter": true, "reduce_scatter": true,
"reduce_bucket_size": 2e8, "reduce_bucket_size": 2e8,
"contiguous_gradients": true, "contiguous_gradients": true,
"cpu_offload": true "cpu_offload": true
}, },
"optimizer": { "optimizer": {
"type": "AdamW", "type": "AdamW",
"params": { "params": {
"lr": 3e-5, "lr": 3e-5,
"betas": [0.8, 0.999], "betas": [0.8, 0.999],
"eps": 1e-8, "eps": 1e-8,
"weight_decay": 3e-7 "weight_decay": 3e-7
} }
}, },
"scheduler": { "scheduler": {
"type": "WarmupLR", "type": "WarmupLR",
"params": { "params": {
"warmup_min_lr": 0, "warmup_min_lr": 0,
"warmup_max_lr": 3e-5, "warmup_max_lr": 3e-5,
"warmup_num_steps": 500 "warmup_num_steps": 500
} }
}, },
"steps_per_print": 2000, "steps_per_print": 2000,
"wall_clock_breakdown": false "wall_clock_breakdown": false
} }
@@ -1073,54 +1073,54 @@ Here is a full ZeRO-3 all-enabled configuration file ``ds_config_zero3.json``:
.. code-block:: json .. code-block:: json
{ {
"fp16": { "fp16": {
"enabled": true, "enabled": true,
"loss_scale": 0, "loss_scale": 0,
"loss_scale_window": 1000, "loss_scale_window": 1000,
"initial_scale_power": 16, "initial_scale_power": 16,
"hysteresis": 2, "hysteresis": 2,
"min_loss_scale": 1 "min_loss_scale": 1
}, },
"zero_optimization": { "zero_optimization": {
"stage": 3, "stage": 3,
"cpu_offload": true, "cpu_offload": true,
"cpu_offload_params": true, "cpu_offload_params": true,
"cpu_offload_use_pin_memory" : true, "cpu_offload_use_pin_memory" : true,
"overlap_comm": true, "overlap_comm": true,
"contiguous_gradients": true, "contiguous_gradients": true,
"sub_group_size": 1e14, "sub_group_size": 1e14,
"reduce_bucket_size": 1e6, "reduce_bucket_size": 1e6,
"stage3_prefetch_bucket_size": 0.94e6, "stage3_prefetch_bucket_size": 0.94e6,
"stage3_param_persistence_threshold": 1e4, "stage3_param_persistence_threshold": 1e4,
"stage3_max_live_parameters": 1e9, "stage3_max_live_parameters": 1e9,
"stage3_max_reuse_distance": 1e9, "stage3_max_reuse_distance": 1e9,
"stage3_gather_fp16_weights_on_model_save": true "stage3_gather_fp16_weights_on_model_save": true
}, },
"optimizer": { "optimizer": {
"type": "AdamW", "type": "AdamW",
"params": { "params": {
"lr": 3e-5, "lr": 3e-5,
"betas": [0.8, 0.999], "betas": [0.8, 0.999],
"eps": 1e-8, "eps": 1e-8,
"weight_decay": 3e-7 "weight_decay": 3e-7
} }
}, },
"scheduler": { "scheduler": {
"type": "WarmupLR", "type": "WarmupLR",
"params": { "params": {
"warmup_min_lr": 0, "warmup_min_lr": 0,
"warmup_max_lr": 3e-5, "warmup_max_lr": 3e-5,
"warmup_num_steps": 500 "warmup_num_steps": 500
} }
}, },
"steps_per_print": 2000, "steps_per_print": 2000,
"wall_clock_breakdown": false "wall_clock_breakdown": false
} }
Optimizer and Scheduler Optimizer and Scheduler
@@ -1367,26 +1367,26 @@ Let's say your checkpoint folder looks like this:
.. code-block:: bash .. code-block:: bash
$ ls -l output_dir/checkpoint-1/ $ ls -l output_dir/checkpoint-1/
-rw-rw-r-- 1 stas stas 1.4K Mar 27 20:42 config.json -rw-rw-r-- 1 stas stas 1.4K Mar 27 20:42 config.json
drwxrwxr-x 2 stas stas 4.0K Mar 25 19:52 global_step1/ drwxrwxr-x 2 stas stas 4.0K Mar 25 19:52 global_step1/
-rw-rw-r-- 1 stas stas 12 Mar 27 13:16 latest -rw-rw-r-- 1 stas stas 12 Mar 27 13:16 latest
-rw-rw-r-- 1 stas stas 827K Mar 27 20:42 optimizer.pt -rw-rw-r-- 1 stas stas 827K Mar 27 20:42 optimizer.pt
-rw-rw-r-- 1 stas stas 231M Mar 27 20:42 pytorch_model.bin -rw-rw-r-- 1 stas stas 231M Mar 27 20:42 pytorch_model.bin
-rw-rw-r-- 1 stas stas 623 Mar 27 20:42 scheduler.pt -rw-rw-r-- 1 stas stas 623 Mar 27 20:42 scheduler.pt
-rw-rw-r-- 1 stas stas 1.8K Mar 27 20:42 special_tokens_map.json -rw-rw-r-- 1 stas stas 1.8K Mar 27 20:42 special_tokens_map.json
-rw-rw-r-- 1 stas stas 774K Mar 27 20:42 spiece.model -rw-rw-r-- 1 stas stas 774K Mar 27 20:42 spiece.model
-rw-rw-r-- 1 stas stas 1.9K Mar 27 20:42 tokenizer_config.json -rw-rw-r-- 1 stas stas 1.9K Mar 27 20:42 tokenizer_config.json
-rw-rw-r-- 1 stas stas 339 Mar 27 20:42 trainer_state.json -rw-rw-r-- 1 stas stas 339 Mar 27 20:42 trainer_state.json
-rw-rw-r-- 1 stas stas 2.3K Mar 27 20:42 training_args.bin -rw-rw-r-- 1 stas stas 2.3K Mar 27 20:42 training_args.bin
-rwxrw-r-- 1 stas stas 5.5K Mar 27 13:16 zero_to_fp32.py* -rwxrw-r-- 1 stas stas 5.5K Mar 27 13:16 zero_to_fp32.py*
In this example there is just one DeepSpeed checkpoint sub-folder `global_step1`. Therefore to reconstruct the fp32 In this example there is just one DeepSpeed checkpoint sub-folder `global_step1`. Therefore to reconstruct the fp32
weights just run: weights just run:
.. code-block:: bash .. code-block:: bash
python zero_to_fp32.py global_step1 pytorch_model.bin python zero_to_fp32.py global_step1 pytorch_model.bin
The script will automatically handle either ZeRO-2 or ZeRO-3 checkpoint. The script will automatically handle either ZeRO-2 or ZeRO-3 checkpoint.
@@ -1416,18 +1416,18 @@ be seen in the following example:
.. code-block:: python .. code-block:: python
class ModuleZ3(torch.nn.Module): class ModuleZ3(torch.nn.Module):
def __init__(self, *args): def __init__(self, *args):
super().__init__(self, *args) super().__init__(self, *args)
self.layer1 = SomeLayer() self.layer1 = SomeLayer()
self.layer2 = OtherLayer() self.layer2 = OtherLayer()
deepspeed.zero.register_external_parameter(self, self.layer1.weight) deepspeed.zero.register_external_parameter(self, self.layer1.weight)
def forward(self, input): def forward(self, input):
x = self.layer1(input) x = self.layer1(input)
# self.layer1.weight is needed in ModuleZ3.forward # self.layer1.weight is needed in ModuleZ3.forward
y = self.layer2(x, self.layer1.weight) y = self.layer2(x, self.layer1.weight)
return y return y
In general ``transformers`` models don't use this style of referring to other layer's weights so most likely you won't In general ``transformers`` models don't use this style of referring to other layer's weights so most likely you won't
need to use it. need to use it.
@@ -1494,7 +1494,7 @@ Also under ZeRO-3, if you write your own code and run into a model parameter wei
.. code-block:: python .. code-block:: python
tensor([1.], device='cuda:0', dtype=torch.float16, requires_grad=True) tensor([1.], device='cuda:0', dtype=torch.float16, requires_grad=True)
stress on ``tensor([1.])``, or if you get an error where it says the parameter is of size ``1``, instead of some much stress on ``tensor([1.])``, or if you get an error where it says the parameter is of size ``1``, instead of some much
larger multi-dimensional shape, this means that the parameter is partitioned and what you see is a ZeRO-3 placeholder. larger multi-dimensional shape, this means that the parameter is partitioned and what you see is a ZeRO-3 placeholder.

36
docs/source/model_doc/bert_japanese.rst
View File

@@ -33,38 +33,38 @@ Example of using a model with MeCab and WordPiece tokenization:
.. code-block:: .. code-block::
>>> import torch >>> import torch
>>> from transformers import AutoModel, AutoTokenizer >>> from transformers import AutoModel, AutoTokenizer
>>> bertjapanese = AutoModel.from_pretrained("cl-tohoku/bert-base-japanese") >>> bertjapanese = AutoModel.from_pretrained("cl-tohoku/bert-base-japanese")
>>> tokenizer = AutoTokenizer.from_pretrained("cl-tohoku/bert-base-japanese") >>> tokenizer = AutoTokenizer.from_pretrained("cl-tohoku/bert-base-japanese")
>>> ## Input Japanese Text >>> ## Input Japanese Text
>>> line = "吾輩は猫である。" >>> line = "吾輩は猫である。"
>>> inputs = tokenizer(line, return_tensors="pt") >>> inputs = tokenizer(line, return_tensors="pt")
>>> print(tokenizer.decode(inputs['input_ids'][0])) >>> print(tokenizer.decode(inputs['input_ids'][0]))
[CLS] 吾輩 は 猫 で ある 。 [SEP] [CLS] 吾輩 は 猫 で ある 。 [SEP]
>>> outputs = bertjapanese(**inputs) >>> outputs = bertjapanese(**inputs)
Example of using a model with Character tokenization: Example of using a model with Character tokenization:
.. code-block:: .. code-block::
>>> bertjapanese = AutoModel.from_pretrained("cl-tohoku/bert-base-japanese-char") >>> bertjapanese = AutoModel.from_pretrained("cl-tohoku/bert-base-japanese-char")
>>> tokenizer = AutoTokenizer.from_pretrained("cl-tohoku/bert-base-japanese-char") >>> tokenizer = AutoTokenizer.from_pretrained("cl-tohoku/bert-base-japanese-char")
>>> ## Input Japanese Text >>> ## Input Japanese Text
>>> line = "吾輩は猫である。" >>> line = "吾輩は猫である。"
>>> inputs = tokenizer(line, return_tensors="pt") >>> inputs = tokenizer(line, return_tensors="pt")
>>> print(tokenizer.decode(inputs['input_ids'][0])) >>> print(tokenizer.decode(inputs['input_ids'][0]))
[CLS] 吾 輩 は 猫 で あ る 。 [SEP] [CLS] 吾 輩 は 猫 で あ る 。 [SEP]
>>> outputs = bertjapanese(**inputs) >>> outputs = bertjapanese(**inputs)
Tips: Tips:

38
docs/source/model_doc/bertgeneration.rst
View File

@@ -38,22 +38,22 @@ Usage:
.. code-block:: .. code-block::
# leverage checkpoints for Bert2Bert model... # leverage checkpoints for Bert2Bert model...
# use BERT's cls token as BOS token and sep token as EOS token # use BERT's cls token as BOS token and sep token as EOS token
encoder = BertGenerationEncoder.from_pretrained("bert-large-uncased", bos_token_id=101, eos_token_id=102) encoder = BertGenerationEncoder.from_pretrained("bert-large-uncased", bos_token_id=101, eos_token_id=102)
# add cross attention layers and use BERT's cls token as BOS token and sep token as EOS token # add cross attention layers and use BERT's cls token as BOS token and sep token as EOS token
decoder = BertGenerationDecoder.from_pretrained("bert-large-uncased", add_cross_attention=True, is_decoder=True, bos_token_id=101, eos_token_id=102) decoder = BertGenerationDecoder.from_pretrained("bert-large-uncased", add_cross_attention=True, is_decoder=True, bos_token_id=101, eos_token_id=102)
bert2bert = EncoderDecoderModel(encoder=encoder, decoder=decoder) bert2bert = EncoderDecoderModel(encoder=encoder, decoder=decoder)
# create tokenizer... # create tokenizer...
tokenizer = BertTokenizer.from_pretrained("bert-large-uncased") tokenizer = BertTokenizer.from_pretrained("bert-large-uncased")
input_ids = tokenizer('This is a long article to summarize', add_special_tokens=False, return_tensors="pt").input_ids input_ids = tokenizer('This is a long article to summarize', add_special_tokens=False, return_tensors="pt").input_ids
labels = tokenizer('This is a short summary', return_tensors="pt").input_ids labels = tokenizer('This is a short summary', return_tensors="pt").input_ids
# train... # train...
loss = bert2bert(input_ids=input_ids, decoder_input_ids=labels, labels=labels).loss loss = bert2bert(input_ids=input_ids, decoder_input_ids=labels, labels=labels).loss
loss.backward() loss.backward()
- Pretrained :class:`~transformers.EncoderDecoderModel` are also directly available in the model hub, e.g., - Pretrained :class:`~transformers.EncoderDecoderModel` are also directly available in the model hub, e.g.,
@@ -61,15 +61,15 @@ Usage:
.. code-block:: .. code-block::
# instantiate sentence fusion model # instantiate sentence fusion model
sentence_fuser = EncoderDecoderModel.from_pretrained("google/roberta2roberta_L-24_discofuse") sentence_fuser = EncoderDecoderModel.from_pretrained("google/roberta2roberta_L-24_discofuse")
tokenizer = AutoTokenizer.from_pretrained("google/roberta2roberta_L-24_discofuse") tokenizer = AutoTokenizer.from_pretrained("google/roberta2roberta_L-24_discofuse")
input_ids = tokenizer('This is the first sentence. This is the second sentence.', add_special_tokens=False, return_tensors="pt").input_ids input_ids = tokenizer('This is the first sentence. This is the second sentence.', add_special_tokens=False, return_tensors="pt").input_ids
outputs = sentence_fuser.generate(input_ids) outputs = sentence_fuser.generate(input_ids)
print(tokenizer.decode(outputs[0])) print(tokenizer.decode(outputs[0]))
Tips: Tips:

30
docs/source/model_doc/bertweet.rst
View File

@@ -31,28 +31,28 @@ Example of use:
.. code-block:: .. code-block::
import torch import torch
from transformers import AutoModel, AutoTokenizer from transformers import AutoModel, AutoTokenizer
bertweet = AutoModel.from_pretrained("vinai/bertweet-base") bertweet = AutoModel.from_pretrained("vinai/bertweet-base")
# For transformers v4.x+: # For transformers v4.x+:
tokenizer = AutoTokenizer.from_pretrained("vinai/bertweet-base", use_fast=False) tokenizer = AutoTokenizer.from_pretrained("vinai/bertweet-base", use_fast=False)
# For transformers v3.x: # For transformers v3.x:
# tokenizer = AutoTokenizer.from_pretrained("vinai/bertweet-base") # tokenizer = AutoTokenizer.from_pretrained("vinai/bertweet-base")
# INPUT TWEET IS ALREADY NORMALIZED! # INPUT TWEET IS ALREADY NORMALIZED!
line = "SC has first two presumptive cases of coronavirus , DHEC confirms HTTPURL via @USER :cry:" line = "SC has first two presumptive cases of coronavirus , DHEC confirms HTTPURL via @USER :cry:"
input_ids = torch.tensor([tokenizer.encode(line)]) input_ids = torch.tensor([tokenizer.encode(line)])
with torch.no_grad(): with torch.no_grad():
features = bertweet(input_ids) # Models outputs are now tuples features = bertweet(input_ids) # Models outputs are now tuples
## With TensorFlow 2.0+: ## With TensorFlow 2.0+:
# from transformers import TFAutoModel # from transformers import TFAutoModel
# bertweet = TFAutoModel.from_pretrained("vinai/bertweet-base") # bertweet = TFAutoModel.from_pretrained("vinai/bertweet-base")
The original code can be found `here <https://github.com/VinAIResearch/BERTweet>`__. The original code can be found `here <https://github.com/VinAIResearch/BERTweet>`__.

20
docs/source/model_doc/herbert.rst
View File

@@ -40,20 +40,20 @@ Examples of use:
.. code-block:: .. code-block::
from transformers import HerbertTokenizer, RobertaModel from transformers import HerbertTokenizer, RobertaModel
tokenizer = HerbertTokenizer.from_pretrained("allegro/herbert-klej-cased-tokenizer-v1") tokenizer = HerbertTokenizer.from_pretrained("allegro/herbert-klej-cased-tokenizer-v1")
model = RobertaModel.from_pretrained("allegro/herbert-klej-cased-v1") model = RobertaModel.from_pretrained("allegro/herbert-klej-cased-v1")
encoded_input = tokenizer.encode("Kto ma lepszą sztukę, ma lepszy rząd – to jasne.", return_tensors='pt') encoded_input = tokenizer.encode("Kto ma lepszą sztukę, ma lepszy rząd – to jasne.", return_tensors='pt')
outputs = model(encoded_input) outputs = model(encoded_input)
# HerBERT can also be loaded using AutoTokenizer and AutoModel: # HerBERT can also be loaded using AutoTokenizer and AutoModel:
import torch import torch
from transformers import AutoModel, AutoTokenizer from transformers import AutoModel, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("allegro/herbert-klej-cased-tokenizer-v1") tokenizer = AutoTokenizer.from_pretrained("allegro/herbert-klej-cased-tokenizer-v1")
model = AutoModel.from_pretrained("allegro/herbert-klej-cased-v1") model = AutoModel.from_pretrained("allegro/herbert-klej-cased-v1")
The original code can be found `here <https://github.com/allegro/HerBERT>`__. The original code can be found `here <https://github.com/allegro/HerBERT>`__.

20
docs/source/model_doc/layoutlm.rst
View File

@@ -56,24 +56,24 @@ Tips:
.. code-block:: .. code-block::
def normalize_bbox(bbox, width, height): def normalize_bbox(bbox, width, height):
return [ return [
int(1000 * (bbox[0] / width)), int(1000 * (bbox[0] / width)),
int(1000 * (bbox[1] / height)), int(1000 * (bbox[1] / height)),
int(1000 * (bbox[2] / width)), int(1000 * (bbox[2] / width)),
int(1000 * (bbox[3] / height)), int(1000 * (bbox[3] / height)),
] ]
Here, :obj:`width` and :obj:`height` correspond to the width and height of the original document in which the token Here, :obj:`width` and :obj:`height` correspond to the width and height of the original document in which the token
occurs. Those can be obtained using the Python Image Library (PIL) library for example, as follows: occurs. Those can be obtained using the Python Image Library (PIL) library for example, as follows:
.. code-block:: .. code-block::
from PIL import Image from PIL import Image
image = Image.open("name_of_your_document - can be a png file, pdf, etc.") image = Image.open("name_of_your_document - can be a png file, pdf, etc.")
width, height = image.size width, height = image.size
- For a demo which shows how to fine-tune :class:`LayoutLMForTokenClassification` on the `FUNSD dataset - For a demo which shows how to fine-tune :class:`LayoutLMForTokenClassification` on the `FUNSD dataset
<https://guillaumejaume.github.io/FUNSD/>`__ (a collection of annotated forms), see `this notebook <https://guillaumejaume.github.io/FUNSD/>`__ (a collection of annotated forms), see `this notebook

12
docs/source/model_doc/megatron_bert.rst
View File

@@ -53,15 +53,15 @@ BERT-345M-uncased::
.. code-block:: bash .. code-block:: bash
wget --content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_bert_345m/versions/v0.1_uncased/zip wget --content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_bert_345m/versions/v0.1_uncased/zip
-O megatron_bert_345m_v0_1_uncased.zip -O megatron_bert_345m_v0_1_uncased.zip
BERT-345M-cased:: BERT-345M-cased::
.. code-block:: bash .. code-block:: bash
wget --content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_bert_345m/versions/v0.1_cased/zip -O wget --content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_bert_345m/versions/v0.1_cased/zip -O
megatron_bert_345m_v0_1_cased.zip megatron_bert_345m_v0_1_cased.zip
Once you have obtained the checkpoints from NVIDIA GPU Cloud (NGC), you have to convert them to a format that will Once you have obtained the checkpoints from NVIDIA GPU Cloud (NGC), you have to convert them to a format that will
easily be loaded by Hugging Face Transformers and our port of the BERT code. easily be loaded by Hugging Face Transformers and our port of the BERT code.
@@ -71,11 +71,11 @@ The following commands allow you to do the conversion. We assume that the folder
.. code-block:: bash .. code-block:: bash
python3 $PATH_TO_TRANSFORMERS/models/megatron_bert/convert_megatron_bert_checkpoint.py megatron_bert_345m_v0_1_uncased.zip python3 $PATH_TO_TRANSFORMERS/models/megatron_bert/convert_megatron_bert_checkpoint.py megatron_bert_345m_v0_1_uncased.zip
.. code-block:: bash .. code-block:: bash
python3 $PATH_TO_TRANSFORMERS/models/megatron_bert/convert_megatron_bert_checkpoint.py megatron_bert_345m_v0_1_cased.zip python3 $PATH_TO_TRANSFORMERS/models/megatron_bert/convert_megatron_bert_checkpoint.py megatron_bert_345m_v0_1_cased.zip
The original code can be found `here <https://github.com/NVIDIA/Megatron-LM>`__. That repository contains a multi-GPU The original code can be found `here <https://github.com/NVIDIA/Megatron-LM>`__. That repository contains a multi-GPU
and multi-node implementation of the Megatron Language models. In particular, it contains a hybrid model parallel and multi-node implementation of the Megatron Language models. In particular, it contains a hybrid model parallel

6
docs/source/model_doc/megatron_gpt2.rst
View File

@@ -51,8 +51,8 @@ Alternatively, you can directly download the checkpoints using::
.. code-block:: bash .. code-block:: bash
wget --content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_lm_345m/versions/v0.0/zip -O wget --content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_lm_345m/versions/v0.0/zip -O
megatron_gpt2_345m_v0_0.zip megatron_gpt2_345m_v0_0.zip
Once you have obtained the checkpoint from NVIDIA GPU Cloud (NGC), you have to convert it to a format that will easily Once you have obtained the checkpoint from NVIDIA GPU Cloud (NGC), you have to convert it to a format that will easily
be loaded by Hugging Face Transformers GPT2 implementation. be loaded by Hugging Face Transformers GPT2 implementation.
@@ -62,7 +62,7 @@ The following command allows you to do the conversion. We assume that the folder
.. code-block:: bash .. code-block:: bash
python3 $PATH_TO_TRANSFORMERS/models/megatron_gpt2/convert_megatron_gpt2_checkpoint.py megatron_gpt2_345m_v0_0.zip python3 $PATH_TO_TRANSFORMERS/models/megatron_gpt2/convert_megatron_gpt2_checkpoint.py megatron_gpt2_345m_v0_0.zip
The original code can be found `here <https://github.com/NVIDIA/Megatron-LM>`__. That repository contains a multi-GPU The original code can be found `here <https://github.com/NVIDIA/Megatron-LM>`__. That repository contains a multi-GPU
and multi-node implementation of the Megatron Language models. In particular, it contains a hybrid model parallel and multi-node implementation of the Megatron Language models. In particular, it contains a hybrid model parallel

24
docs/source/model_doc/phobert.rst
View File

@@ -31,23 +31,23 @@ Example of use:
.. code-block:: .. code-block::
import torch import torch
from transformers import AutoModel, AutoTokenizer from transformers import AutoModel, AutoTokenizer
phobert = AutoModel.from_pretrained("vinai/phobert-base") phobert = AutoModel.from_pretrained("vinai/phobert-base")
tokenizer = AutoTokenizer.from_pretrained("vinai/phobert-base") tokenizer = AutoTokenizer.from_pretrained("vinai/phobert-base")
# INPUT TEXT MUST BE ALREADY WORD-SEGMENTED! # INPUT TEXT MUST BE ALREADY WORD-SEGMENTED!
line = "Tôi là sinh_viên trường đại_học Công_nghệ ." line = "Tôi là sinh_viên trường đại_học Công_nghệ ."
input_ids = torch.tensor([tokenizer.encode(line)]) input_ids = torch.tensor([tokenizer.encode(line)])
with torch.no_grad(): with torch.no_grad():
features = phobert(input_ids) # Models outputs are now tuples features = phobert(input_ids) # Models outputs are now tuples
## With TensorFlow 2.0+: ## With TensorFlow 2.0+:
# from transformers import TFAutoModel # from transformers import TFAutoModel
# phobert = TFAutoModel.from_pretrained("vinai/phobert-base") # phobert = TFAutoModel.from_pretrained("vinai/phobert-base")
The original code can be found `here <https://github.com/VinAIResearch/PhoBERT>`__. The original code can be found `here <https://github.com/VinAIResearch/PhoBERT>`__.

4
docs/source/model_doc/reformer.rst
View File

@@ -145,8 +145,8 @@ For training, the :class:`~transformers.ReformerModelWithLMHead` should be used
.. code-block:: .. code-block::
input_ids = tokenizer.encode('This is a sentence from the training data', return_tensors='pt') input_ids = tokenizer.encode('This is a sentence from the training data', return_tensors='pt')
loss = model(input_ids, labels=input_ids)[0] loss = model(input_ids, labels=input_ids)[0]
ReformerConfig ReformerConfig

16
docs/source/model_doc/t5.rst
View File

@@ -73,10 +73,10 @@ token. T5 can be trained / fine-tuned both in a supervised and unsupervised fash
.. code-block:: .. code-block::
input_ids = tokenizer('The <extra_id_0> walks in <extra_id_1> park', return_tensors='pt').input_ids input_ids = tokenizer('The <extra_id_0> walks in <extra_id_1> park', return_tensors='pt').input_ids
labels = tokenizer('<extra_id_0> cute dog <extra_id_1> the <extra_id_2>', return_tensors='pt').input_ids labels = tokenizer('<extra_id_0> cute dog <extra_id_1> the <extra_id_2>', return_tensors='pt').input_ids
# the forward function automatically creates the correct decoder_input_ids # the forward function automatically creates the correct decoder_input_ids
loss = model(input_ids=input_ids, labels=labels).loss loss = model(input_ids=input_ids, labels=labels).loss
- Supervised training - Supervised training
@@ -86,10 +86,10 @@ token. T5 can be trained / fine-tuned both in a supervised and unsupervised fash
.. code-block:: .. code-block::
input_ids = tokenizer('translate English to German: The house is wonderful.', return_tensors='pt').input_ids input_ids = tokenizer('translate English to German: The house is wonderful.', return_tensors='pt').input_ids
labels = tokenizer('Das Haus ist wunderbar.', return_tensors='pt').input_ids labels = tokenizer('Das Haus ist wunderbar.', return_tensors='pt').input_ids
# the forward function automatically creates the correct decoder_input_ids # the forward function automatically creates the correct decoder_input_ids
loss = model(input_ids=input_ids, labels=labels).loss loss = model(input_ids=input_ids, labels=labels).loss
T5Config T5Config

78
docs/source/testing.rst
View File

@@ -70,19 +70,19 @@ Run all:
.. code-block:: console .. code-block:: console
pytest pytest
or: or:
.. code-block:: bash .. code-block:: bash
make test make test
Note that the latter is defined as: Note that the latter is defined as:
.. code-block:: bash .. code-block:: bash
python -m pytest -n auto --dist=loadfile -s -v ./tests/ python -m pytest -n auto --dist=loadfile -s -v ./tests/
which tells pytest to: which tells pytest to:
@@ -100,13 +100,13 @@ All tests of the test suite:
.. code-block:: bash .. code-block:: bash
pytest --collect-only -q pytest --collect-only -q
All tests of a given test file: All tests of a given test file:
.. code-block:: bash .. code-block:: bash
pytest tests/test_optimization.py --collect-only -q pytest tests/test_optimization.py --collect-only -q
@@ -117,7 +117,7 @@ To run an individual test module:
.. code-block:: bash .. code-block:: bash
pytest tests/test_logging.py pytest tests/test_logging.py
Run specific tests Run specific tests
@@ -128,7 +128,7 @@ class containing those tests. For example, it could be:
.. code-block:: bash .. code-block:: bash
pytest tests/test_optimization.py::OptimizationTest::test_adam_w pytest tests/test_optimization.py::OptimizationTest::test_adam_w
Here: Here:
@@ -140,7 +140,7 @@ If the file contains multiple classes, you can choose to run only tests of a giv
.. code-block:: bash .. code-block:: bash
pytest tests/test_optimization.py::OptimizationTest pytest tests/test_optimization.py::OptimizationTest
will run all the tests inside that class. will run all the tests inside that class.
@@ -149,7 +149,7 @@ As mentioned earlier you can see what tests are contained inside the ``Optimizat
.. code-block:: bash .. code-block:: bash
pytest tests/test_optimization.py::OptimizationTest --collect-only -q pytest tests/test_optimization.py::OptimizationTest --collect-only -q
You can run tests by keyword expressions. You can run tests by keyword expressions.
@@ -157,7 +157,7 @@ To run only tests whose name contains ``adam``:
.. code-block:: bash .. code-block:: bash
pytest -k adam tests/test_optimization.py pytest -k adam tests/test_optimization.py
Logical ``and`` and ``or`` can be used to indicate whether all keywords should match or either. ``not`` can be used to Logical ``and`` and ``or`` can be used to indicate whether all keywords should match or either. ``not`` can be used to
negate. negate.
@@ -166,19 +166,19 @@ To run all tests except those whose name contains ``adam``:
.. code-block:: bash .. code-block:: bash
pytest -k "not adam" tests/test_optimization.py pytest -k "not adam" tests/test_optimization.py
And you can combine the two patterns in one: And you can combine the two patterns in one:
.. code-block:: bash .. code-block:: bash
pytest -k "ada and not adam" tests/test_optimization.py pytest -k "ada and not adam" tests/test_optimization.py
For example to run both ``test_adafactor`` and ``test_adam_w`` you can use: For example to run both ``test_adafactor`` and ``test_adam_w`` you can use:
.. code-block:: bash .. code-block:: bash
pytest -k "test_adam_w or test_adam_w" tests/test_optimization.py pytest -k "test_adam_w or test_adam_w" tests/test_optimization.py
Note that we use ``or`` here, since we want either of the keywords to match to include both. Note that we use ``or`` here, since we want either of the keywords to match to include both.
@@ -186,7 +186,7 @@ If you want to include only tests that include both patterns, ``and`` is to be u
.. code-block:: bash .. code-block:: bash
pytest -k "test and ada" tests/test_optimization.py pytest -k "test and ada" tests/test_optimization.py
@@ -251,7 +251,7 @@ example, to run all except ``test_modeling_*.py`` tests:
.. code-block:: bash .. code-block:: bash
pytest `ls -1 tests/*py | grep -v test_modeling` pytest `ls -1 tests/*py | grep -v test_modeling`
Clearing state Clearing state
@@ -292,13 +292,13 @@ Repeat tests
.. code-block:: bash .. code-block:: bash
pip install pytest-flakefinder pip install pytest-flakefinder
And then run every test multiple times (50 by default): And then run every test multiple times (50 by default):
.. code-block:: bash .. code-block:: bash
pytest --flake-finder --flake-runs=5 tests/test_failing_test.py pytest --flake-finder --flake-runs=5 tests/test_failing_test.py
.. note:: .. note::
This plugin doesn't work with ``-n`` flag from ``pytest-xdist``. This plugin doesn't work with ``-n`` flag from ``pytest-xdist``.
@@ -322,19 +322,19 @@ As explained earlier this allows detection of coupled tests - where one test's s
.. code-block:: bash .. code-block:: bash
pytest tests pytest tests
[...] [...]
Using --random-order-bucket=module Using --random-order-bucket=module
Using --random-order-seed=573663 Using --random-order-seed=573663
So that if the given particular sequence fails, you can reproduce it by adding that exact seed, e.g.: So that if the given particular sequence fails, you can reproduce it by adding that exact seed, e.g.:
.. code-block:: bash .. code-block:: bash
pytest --random-order-seed=573663 pytest --random-order-seed=573663
[...] [...]
Using --random-order-bucket=module Using --random-order-bucket=module
Using --random-order-seed=573663 Using --random-order-seed=573663
It will only reproduce the exact order if you use the exact same list of tests (or no list at all). Once you start to It will only reproduce the exact order if you use the exact same list of tests (or no list at all). Once you start to
manually narrowing down the list you can no longer rely on the seed, but have to list them manually in the exact order manually narrowing down the list you can no longer rely on the seed, but have to list them manually in the exact order
@@ -342,7 +342,7 @@ they failed and tell pytest to not randomize them instead using ``--random-order
.. code-block:: bash .. code-block:: bash
pytest --random-order-bucket=none tests/test_a.py tests/test_c.py tests/test_b.py pytest --random-order-bucket=none tests/test_a.py tests/test_c.py tests/test_b.py
To disable the shuffling for all tests: To disable the shuffling for all tests:
@@ -369,7 +369,7 @@ progressbar, and show tests that fail and the assert instantly. It gets activate
.. code-block:: bash .. code-block:: bash
pip install pytest-sugar pip install pytest-sugar
To run tests without it, run: To run tests without it, run:
@@ -388,7 +388,7 @@ For a single or a group of tests via ``pytest`` (after ``pip install pytest-pspe
.. code-block:: bash .. code-block:: bash
pytest --pspec tests/test_optimization.py pytest --pspec tests/test_optimization.py
@@ -490,8 +490,8 @@ Inside tests:
.. code-block:: bash .. code-block:: bash
from transformers.testing_utils import get_gpu_count from transformers.testing_utils import get_gpu_count
n_gpu = get_gpu_count() # works with torch and tf n_gpu = get_gpu_count() # works with torch and tf
@@ -514,8 +514,8 @@ You will need at least 2 GPUs to see these tests in action:
.. code-block:: bash .. code-block:: bash
CUDA_VISIBLE_DEVICES="0,1" RUN_SLOW=1 pytest -sv examples/seq2seq/test_finetune_trainer.py \ CUDA_VISIBLE_DEVICES="0,1" RUN_SLOW=1 pytest -sv examples/seq2seq/test_finetune_trainer.py \
examples/seq2seq/test_seq2seq_examples_multi_gpu.py examples/seq2seq/test_seq2seq_examples_multi_gpu.py
Output capture Output capture
@@ -528,13 +528,13 @@ To disable output capturing and to get the ``stdout`` and ``stderr`` normally, u
.. code-block:: bash .. code-block:: bash
pytest -s tests/test_logging.py pytest -s tests/test_logging.py
To send test results to JUnit format output: To send test results to JUnit format output:
.. code-block:: bash .. code-block:: bash
py.test tests --junitxml=result.xml py.test tests --junitxml=result.xml
Color control Color control
@@ -544,7 +544,7 @@ To have no color (e.g., yellow on white background is not readable):
.. code-block:: bash .. code-block:: bash
pytest --color=no tests/test_logging.py pytest --color=no tests/test_logging.py
@@ -555,7 +555,7 @@ Creating a URL for each test failure:
.. code-block:: bash .. code-block:: bash
pytest --pastebin=failed tests/test_logging.py pytest --pastebin=failed tests/test_logging.py
This will submit test run information to a remote Paste service and provide a URL for each failure. You may select This will submit test run information to a remote Paste service and provide a URL for each failure. You may select
tests as usual or add for example -x if you only want to send one particular failure. tests as usual or add for example -x if you only want to send one particular failure.
@@ -564,7 +564,7 @@ Creating a URL for a whole test session log:
.. code-block:: bash .. code-block:: bash
pytest --pastebin=all tests/test_logging.py pytest --pastebin=all tests/test_logging.py
@@ -606,13 +606,13 @@ and you could run just the ``negative`` and ``integer`` sets of params with:
.. code-block:: bash .. code-block:: bash
pytest -k "negative and integer" tests/test_mytest.py pytest -k "negative and integer" tests/test_mytest.py
or all but ``negative`` sub-tests, with: or all but ``negative`` sub-tests, with:
.. code-block:: bash .. code-block:: bash
pytest -k "not negative" tests/test_mytest.py pytest -k "not negative" tests/test_mytest.py
Besides using the ``-k`` filter that was just mentioned, you can find out the exact name of each sub-test and run any Besides using the ``-k`` filter that was just mentioned, you can find out the exact name of each sub-test and run any
or all of them using their exact names. or all of them using their exact names.

27
utils/style_doc.py
View File

@@ -49,6 +49,7 @@ _re_indent = re.compile(r"^(\s*)\S")
_re_table = re.compile(r"(\+-+)+\+\s*$") _re_table = re.compile(r"(\+-+)+\+\s*$")
# Matches a code block in rst `:: `. # Matches a code block in rst `:: `.
_re_code_block = re.compile(r"^\s*::\s*$") _re_code_block = re.compile(r"^\s*::\s*$")
_re_code_block_explicit = re.compile(r"^\.\.\s+code\-block::")
# Matches any block of the form `.. something::` or `.. something:: bla`. # Matches any block of the form `.. something::` or `.. something:: bla`.
_re_ignore = re.compile(r"^\s*\.\.\s+(.*?)\s*::\s*\S*\s*$") _re_ignore = re.compile(r"^\s*\.\.\s+(.*?)\s*::\s*\S*\s*$")
# Matches comment introduction in rst. # Matches comment introduction in rst.
@@ -374,6 +375,28 @@ rst_styler = CodeStyler()
doc_styler = DocstringStyler() doc_styler = DocstringStyler()
def _reindent_code_blocks(text):
"""Checks indent in code blocks is of four"""
lines = text.split("\n")
idx = 0
while idx < len(lines):
# Detect if the line is the start of a new code-block.
if _re_code_block.search(lines[idx]) is not None or _re_code_block_explicit.search(lines[idx]) is not None:
while len(get_indent(lines[idx])) == 0:
idx += 1
indent = len(get_indent(lines[idx]))
should_continue = True
while should_continue:
if len(lines[idx]) > 0 and indent < 4:
lines[idx] = " " * 4 + lines[idx][indent:]
idx += 1
should_continue = (idx < len(lines)) and (len(lines[idx]) == 0 or len(get_indent(lines[idx])) > 0)
else:
idx += 1
return "\n".join(lines)
def _add_new_lines_before_list(text): def _add_new_lines_before_list(text):
"""Add a new empty line before a list begins.""" """Add a new empty line before a list begins."""
lines = text.split("\n") lines = text.split("\n")
@@ -412,8 +435,10 @@ def style_rst_file(doc_file, max_len=119, check_only=False):
with open(doc_file, "r", encoding="utf-8", newline="\n") as f: with open(doc_file, "r", encoding="utf-8", newline="\n") as f:
doc = f.read() doc = f.read()
# Make sure code blocks are indented at 4
clean_doc = _reindent_code_blocks(doc)
# Add missing new lines before lists # Add missing new lines before lists
clean_doc = _add_new_lines_before_list(doc) clean_doc = _add_new_lines_before_list(clean_doc)
# Style # Style
clean_doc = rst_styler.style(clean_doc, max_len=max_len) clean_doc = rst_styler.style(clean_doc, max_len=max_len)