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

Introducing the gpu_allocator (#6091) 

* rename 'use_pytorch_for_gpu_memory' to 'gpu_allocator'

* --code instead of --code-path

* update documentation

* avoid querying the "system" section directly

* add explanation of gpu_allocator to TF/PyTorch section in docs

* fix typo

* fix typo 2

* use set_gpu_allocator from thinc 8.0.0a34

* default null instead of empty string
This commit is contained in:
Sofie Van Landeghem 2020-09-19 01:17:02 +02:00 committed by GitHub
parent 0406200a1e
commit 39872de1f6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
13 changed files with 54 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
pyproject.toml
View File

@ -6,7 +6,7 @@ requires = [
"cymem>=2.0.2,<2.1.0", "cymem>=2.0.2,<2.1.0",
"preshed>=3.0.2,<3.1.0", "preshed>=3.0.2,<3.1.0",
"murmurhash>=0.28.0,<1.1.0", "murmurhash>=0.28.0,<1.1.0",
"thinc>=8.0.0a33,<8.0.0a40", "thinc>=8.0.0a34,<8.0.0a40",
"blis>=0.4.0,<0.5.0", "blis>=0.4.0,<0.5.0",
"pytokenizations", "pytokenizations",
"pathy" "pathy"

2
requirements.txt
View File

@ -1,7 +1,7 @@
# Our libraries # Our libraries
cymem>=2.0.2,<2.1.0 cymem>=2.0.2,<2.1.0
preshed>=3.0.2,<3.1.0 preshed>=3.0.2,<3.1.0
thinc>=8.0.0a33,<8.0.0a40 thinc>=8.0.0a34,<8.0.0a40
blis>=0.4.0,<0.5.0 blis>=0.4.0,<0.5.0
ml_datasets==0.2.0a0 ml_datasets==0.2.0a0
murmurhash>=0.28.0,<1.1.0 murmurhash>=0.28.0,<1.1.0

4
setup.cfg
View File

@ -34,13 +34,13 @@ setup_requires =
cymem>=2.0.2,<2.1.0 cymem>=2.0.2,<2.1.0
preshed>=3.0.2,<3.1.0 preshed>=3.0.2,<3.1.0
murmurhash>=0.28.0,<1.1.0 murmurhash>=0.28.0,<1.1.0
thinc>=8.0.0a33,<8.0.0a40 thinc>=8.0.0a34,<8.0.0a40
install_requires = install_requires =
# Our libraries # Our libraries
murmurhash>=0.28.0,<1.1.0 murmurhash>=0.28.0,<1.1.0
cymem>=2.0.2,<2.1.0 cymem>=2.0.2,<2.1.0
preshed>=3.0.2,<3.1.0 preshed>=3.0.2,<3.1.0
thinc>=8.0.0a33,<8.0.0a40 thinc>=8.0.0a34,<8.0.0a40
blis>=0.4.0,<0.5.0 blis>=0.4.0,<0.5.0
wasabi>=0.8.0,<1.1.0 wasabi>=0.8.0,<1.1.0
srsly>=2.1.0,<3.0.0 srsly>=2.1.0,<3.0.0

9
spacy/cli/debug_model.py
View File

@ -2,7 +2,7 @@ from typing import Dict, Any, Optional
from pathlib import Path from pathlib import Path
from wasabi import msg from wasabi import msg
from thinc.api import require_gpu, fix_random_seed, set_dropout_rate, Adam from thinc.api import require_gpu, fix_random_seed, set_dropout_rate, Adam
from thinc.api import Model, data_validation from thinc.api import Model, data_validation, set_gpu_allocator
import typer import typer
from ._util import Arg, Opt, debug_cli, show_validation_error from ._util import Arg, Opt, debug_cli, show_validation_error
@ -53,7 +53,12 @@ def debug_model_cli(
} }
config_overrides = parse_config_overrides(ctx.args) config_overrides = parse_config_overrides(ctx.args)
with show_validation_error(config_path): with show_validation_error(config_path):
config = util.load_config(config_path, overrides=config_overrides) config = util.load_config(
config_path, overrides=config_overrides, interpolate=True
)
allocator = config["training"]["gpu_allocator"]
if use_gpu >= 0 and allocator:
set_gpu_allocator(allocator)
nlp, config = util.load_model_from_config(config_path) nlp, config = util.load_model_from_config(config_path)
seed = config["training"]["seed"] seed = config["training"]["seed"]
if seed is not None: if seed is not None:

17
spacy/cli/pretrain.py
View File

@ -4,10 +4,9 @@ import time
import re import re
from collections import Counter from collections import Counter
from pathlib import Path from pathlib import Path
from thinc.api import Config from thinc.api import require_gpu, set_gpu_allocator
from thinc.api import use_pytorch_for_gpu_memory, require_gpu
from thinc.api import set_dropout_rate, to_categorical, fix_random_seed from thinc.api import set_dropout_rate, to_categorical, fix_random_seed
from thinc.api import CosineDistance, L2Distance from thinc.api import Config, CosineDistance, L2Distance
from wasabi import msg from wasabi import msg
import srsly import srsly
from functools import partial from functools import partial
@ -32,7 +31,7 @@ def pretrain_cli(
ctx: typer.Context, # This is only used to read additional arguments ctx: typer.Context, # This is only used to read additional arguments
config_path: Path = Arg(..., help="Path to config file", exists=True, dir_okay=False), config_path: Path = Arg(..., help="Path to config file", exists=True, dir_okay=False),
output_dir: Path = Arg(..., help="Directory to write weights to on each epoch"), output_dir: Path = Arg(..., help="Directory to write weights to on each epoch"),
code_path: Optional[Path] = Opt(None, "--code-path", "-c", help="Path to Python file with additional code (registered functions) to be imported"), code_path: Optional[Path] = Opt(None, "--code", "-c", help="Path to Python file with additional code (registered functions) to be imported"),
resume_path: Optional[Path] = Opt(None, "--resume-path", "-r", help="Path to pretrained weights from which to resume pretraining"), resume_path: Optional[Path] = Opt(None, "--resume-path", "-r", help="Path to pretrained weights from which to resume pretraining"),
epoch_resume: Optional[int] = Opt(None, "--epoch-resume", "-er", help="The epoch to resume counting from when using --resume-path. Prevents unintended overwriting of existing weight files."), epoch_resume: Optional[int] = Opt(None, "--epoch-resume", "-er", help="The epoch to resume counting from when using --resume-path. Prevents unintended overwriting of existing weight files."),
use_gpu: int = Opt(-1, "--gpu-id", "-g", help="GPU ID or -1 for CPU"), use_gpu: int = Opt(-1, "--gpu-id", "-g", help="GPU ID or -1 for CPU"),
@ -99,10 +98,12 @@ def pretrain(
epoch_resume: Optional[int] = None, epoch_resume: Optional[int] = None,
use_gpu: int = -1, use_gpu: int = -1,
): ):
if config["system"].get("seed") is not None: if config["training"]["seed"] is not None:
fix_random_seed(config["system"]["seed"]) fix_random_seed(config["training"]["seed"])
if use_gpu >= 0 and config["system"].get("use_pytorch_for_gpu_memory"): allocator = config["training"]["gpu_allocator"]
use_pytorch_for_gpu_memory() if use_gpu >= 0 and allocator:
set_gpu_allocator(allocator)
nlp, config = util.load_model_from_config(config) nlp, config = util.load_model_from_config(config)
P_cfg = config["pretraining"] P_cfg = config["pretraining"]
corpus = dot_to_object(config, P_cfg["corpus"]) corpus = dot_to_object(config, P_cfg["corpus"])

2
spacy/cli/templates/quickstart_training.jinja
View File

@ -8,7 +8,7 @@ train = ""
dev = "" dev = ""
[system] [system]
use_pytorch_for_gpu_memory = {{ "true" if use_transformer else "false" }} gpu_allocator = {{ "pytorch" if use_transformer else "" }}
[nlp] [nlp]
lang = "{{ lang }}" lang = "{{ lang }}"

13
spacy/cli/train.py
View File

@ -6,8 +6,7 @@ from pathlib import Path
from wasabi import msg from wasabi import msg
import thinc import thinc
import thinc.schedules import thinc.schedules
from thinc.api import use_pytorch_for_gpu_memory, require_gpu, fix_random_seed from thinc.api import Config, Optimizer, require_gpu, fix_random_seed, set_gpu_allocator
from thinc.api import Config, Optimizer
import random import random
import typer import typer
import logging import logging
@ -29,7 +28,7 @@ def train_cli(
ctx: typer.Context, # This is only used to read additional arguments ctx: typer.Context, # This is only used to read additional arguments
config_path: Path = Arg(..., help="Path to config file", exists=True), config_path: Path = Arg(..., help="Path to config file", exists=True),
output_path: Optional[Path] = Opt(None, "--output", "--output-path", "-o", help="Output directory to store trained pipeline in"), output_path: Optional[Path] = Opt(None, "--output", "--output-path", "-o", help="Output directory to store trained pipeline in"),
code_path: Optional[Path] = Opt(None, "--code-path", "-c", help="Path to Python file with additional code (registered functions) to be imported"), code_path: Optional[Path] = Opt(None, "--code", "-c", help="Path to Python file with additional code (registered functions) to be imported"),
verbose: bool = Opt(False, "--verbose", "-V", "-VV", help="Display more information for debugging purposes"), verbose: bool = Opt(False, "--verbose", "-V", "-VV", help="Display more information for debugging purposes"),
use_gpu: int = Opt(-1, "--gpu-id", "-g", help="GPU ID or -1 for CPU"), use_gpu: int = Opt(-1, "--gpu-id", "-g", help="GPU ID or -1 for CPU"),
resume: bool = Opt(False, "--resume", "-R", help="Resume training"), resume: bool = Opt(False, "--resume", "-R", help="Resume training"),
@ -79,11 +78,11 @@ def train(
config = util.load_config( config = util.load_config(
config_path, overrides=config_overrides, interpolate=True config_path, overrides=config_overrides, interpolate=True
) )
if config.get("training", {}).get("seed") is not None: if config["training"]["seed"] is not None:
fix_random_seed(config["training"]["seed"]) fix_random_seed(config["training"]["seed"])
if config.get("system", {}).get("use_pytorch_for_gpu_memory"): allocator = config["training"]["gpu_allocator"]
# It feels kind of weird to not have a default for this. if use_gpu >= 0 and allocator:
use_pytorch_for_gpu_memory() set_gpu_allocator(allocator)
# Use original config here before it's resolved to functions # Use original config here before it's resolved to functions
sourced_components = get_sourced_components(config) sourced_components = get_sourced_components(config)
with show_validation_error(config_path): with show_validation_error(config_path):

4
spacy/default_config.cfg
View File

@ -6,7 +6,7 @@ init_tok2vec = null
[system] [system]
seed = 0 seed = 0
use_pytorch_for_gpu_memory = false gpu_allocator = null
[nlp] [nlp]
lang = null lang = null
@ -52,6 +52,7 @@ limit = 0
# Training hyper-parameters and additional features. # Training hyper-parameters and additional features.
[training] [training]
seed = ${system.seed} seed = ${system.seed}
gpu_allocator = ${system.gpu_allocator}
dropout = 0.1 dropout = 0.1
accumulate_gradient = 1 accumulate_gradient = 1
# Extra resources for transfer-learning or pseudo-rehearsal # Extra resources for transfer-learning or pseudo-rehearsal
@ -75,7 +76,6 @@ train_corpus = "corpora.train"
[training.logger] [training.logger]
@loggers = "spacy.ConsoleLogger.v1" @loggers = "spacy.ConsoleLogger.v1"
[training.batcher] [training.batcher]
@batchers = "spacy.batch_by_words.v1" @batchers = "spacy.batch_by_words.v1"
discard_oversize = false discard_oversize = false

1
spacy/schemas.py
View File

@ -207,6 +207,7 @@ class ConfigSchemaTraining(BaseModel):
max_steps: StrictInt = Field(..., title="Maximum number of update steps to train for") max_steps: StrictInt = Field(..., title="Maximum number of update steps to train for")
eval_frequency: StrictInt = Field(..., title="How often to evaluate during training (steps)") eval_frequency: StrictInt = Field(..., title="How often to evaluate during training (steps)")
seed: Optional[StrictInt] = Field(..., title="Random seed") seed: Optional[StrictInt] = Field(..., title="Random seed")
gpu_allocator: Optional[StrictStr] = Field(..., title="Memory allocator when running on GPU")
accumulate_gradient: StrictInt = Field(..., title="Whether to divide the batch up into substeps") accumulate_gradient: StrictInt = Field(..., title="Whether to divide the batch up into substeps")
score_weights: Dict[StrictStr, Union[StrictFloat, StrictInt]] = Field(..., title="Scores to report and their weights for selecting final model") score_weights: Dict[StrictStr, Union[StrictFloat, StrictInt]] = Field(..., title="Scores to report and their weights for selecting final model")
init_tok2vec: Optional[StrictStr] = Field(..., title="Path to pretrained tok2vec weights") init_tok2vec: Optional[StrictStr] = Field(..., title="Path to pretrained tok2vec weights")

4
website/docs/api/cli.md
View File

@ -763,6 +763,7 @@ $ python -m spacy train [config_path] [--output] [--code] [--verbose] [overrides
| `--output`, `-o` | Directory to store trained pipeline in. Will be created if it doesn't exist. ~~Optional[Path] \(positional)~~ | | `--output`, `-o` | Directory to store trained pipeline in. Will be created if it doesn't exist. ~~Optional[Path] \(positional)~~ |
| `--code`, `-c` | Path to Python file with additional code to be imported. Allows [registering custom functions](/usage/training#custom-functions) for new architectures. ~~Optional[Path] \(option)~~ | | `--code`, `-c` | Path to Python file with additional code to be imported. Allows [registering custom functions](/usage/training#custom-functions) for new architectures. ~~Optional[Path] \(option)~~ |
| `--verbose`, `-V` | Show more detailed messages during training. ~~bool (flag)~~ | | `--verbose`, `-V` | Show more detailed messages during training. ~~bool (flag)~~ |
| `--gpu-id`, `-g` | GPU ID or `-1` for CPU. Defaults to `-1`. ~~int (option)~~ |
| `--help`, `-h` | Show help message and available arguments. ~~bool (flag)~~ | | `--help`, `-h` | Show help message and available arguments. ~~bool (flag)~~ |
| overrides | Config parameters to override. Should be options starting with `--` that correspond to the config section and value to override, e.g. `--paths.train ./train.spacy`. ~~Any (option/flag)~~ | | overrides | Config parameters to override. Should be options starting with `--` that correspond to the config section and value to override, e.g. `--paths.train ./train.spacy`. ~~Any (option/flag)~~ |
| **CREATES** | The final trained pipeline and the best trained pipeline. | | **CREATES** | The final trained pipeline and the best trained pipeline. |
@ -798,11 +799,12 @@ $ python -m spacy pretrain [config_path] [output_dir] [--code] [--resume-path] [
| Name | Description | | Name | Description |
| ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `output_dir` | Directory to save binary weights to on each epoch. ~~Path (positional)~~ |
| `config_path` | Path to [training config](/api/data-formats#config) file containing all settings and hyperparameters. ~~Path (positional)~~ | | `config_path` | Path to [training config](/api/data-formats#config) file containing all settings and hyperparameters. ~~Path (positional)~~ |
| `output_dir` | Directory to save binary weights to on each epoch. ~~Path (positional)~~ |
| `--code`, `-c` | Path to Python file with additional code to be imported. Allows [registering custom functions](/usage/training#custom-functions) for new architectures. ~~Optional[Path] \(option)~~ | | `--code`, `-c` | Path to Python file with additional code to be imported. Allows [registering custom functions](/usage/training#custom-functions) for new architectures. ~~Optional[Path] \(option)~~ |
| `--resume-path`, `-r` | Path to pretrained weights from which to resume pretraining. ~~Optional[Path] \(option)~~ | | `--resume-path`, `-r` | Path to pretrained weights from which to resume pretraining. ~~Optional[Path] \(option)~~ |
| `--epoch-resume`, `-er` | The epoch to resume counting from when using `--resume-path`. Prevents unintended overwriting of existing weight files. ~~Optional[int] \(option)~~ | | `--epoch-resume`, `-er` | The epoch to resume counting from when using `--resume-path`. Prevents unintended overwriting of existing weight files. ~~Optional[int] \(option)~~ |
| `--gpu-id`, `-g` | GPU ID or `-1` for CPU. Defaults to `-1`. ~~int (option)~~ |
| `--help`, `-h` | Show help message and available arguments. ~~bool (flag)~~ | | `--help`, `-h` | Show help message and available arguments. ~~bool (flag)~~ |
| overrides | Config parameters to override. Should be options starting with `--` that correspond to the config section and value to override, e.g. `--training.dropout 0.2`. ~~Any (option/flag)~~ | | overrides | Config parameters to override. Should be options starting with `--` that correspond to the config section and value to override, e.g. `--training.dropout 0.2`. ~~Any (option/flag)~~ |
| **CREATES** | The pretrained weights that can be used to initialize `spacy train`. | | **CREATES** | The pretrained weights that can be used to initialize `spacy train`. |

1
website/docs/api/data-formats.md
View File

@ -189,6 +189,7 @@ process that are used when you run [`spacy train`](/api/cli#train).
| `dev_corpus` | Dot notation of the config location defining the dev corpus. Defaults to `corpora.dev`. ~~str~~ | | `dev_corpus` | Dot notation of the config location defining the dev corpus. Defaults to `corpora.dev`. ~~str~~ |
| `dropout` | The dropout rate. Defaults to `0.1`. ~~float~~ | | `dropout` | The dropout rate. Defaults to `0.1`. ~~float~~ |
| `eval_frequency` | How often to evaluate during training (steps). Defaults to `200`. ~~int~~ | | `eval_frequency` | How often to evaluate during training (steps). Defaults to `200`. ~~int~~ |
| `gpu_allocator` | Library for cupy to route GPU memory allocation to. Can be "pytorch" or "tensorflow". Defaults to variable `${system.gpu_allocator}`. ~~str~~ |
| `frozen_components` | Pipeline component names that are "frozen" and shouldn't be updated during training. See [here](/usage/training#config-components) for details. Defaults to `[]`. ~~List[str]~~ | | `frozen_components` | Pipeline component names that are "frozen" and shouldn't be updated during training. See [here](/usage/training#config-components) for details. Defaults to `[]`. ~~List[str]~~ |
| `init_tok2vec` | Optional path to pretrained tok2vec weights created with [`spacy pretrain`](/api/cli#pretrain). Defaults to variable `${paths.init_tok2vec}`. ~~Optional[str]~~ | | `init_tok2vec` | Optional path to pretrained tok2vec weights created with [`spacy pretrain`](/api/cli#pretrain). Defaults to variable `${paths.init_tok2vec}`. ~~Optional[str]~~ |
| `max_epochs` | Maximum number of epochs to train for. Defaults to `0`. ~~int~~ | | `max_epochs` | Maximum number of epochs to train for. Defaults to `0`. ~~int~~ |

14
website/docs/api/top-level.md
View File

@ -145,9 +145,10 @@ pipelines.
> nlp = spacy.load("en_core_web_sm") > nlp = spacy.load("en_core_web_sm")
> ``` > ```
| Name | Description | | Name | Description |
| ----------- | --------------------------------------- | | ----------- | ------------------------------------------------ |
| **RETURNS** | Whether the GPU was activated. ~~bool~~ | | `gpu_id` | Device index to select. Defaults to `0`. ~~int~~ |
| **RETURNS** | Whether the GPU was activated. ~~bool~~ |
### spacy.require_gpu {#spacy.require_gpu tag="function" new="2.0.14"} ### spacy.require_gpu {#spacy.require_gpu tag="function" new="2.0.14"}
@ -164,9 +165,10 @@ and _before_ loading any pipelines.
> nlp = spacy.load("en_core_web_sm") > nlp = spacy.load("en_core_web_sm")
> ``` > ```
| Name | Description | | Name | Description |
| ----------- | --------------- | | ----------- | ------------------------------------------------ |
| **RETURNS** | `True` ~~bool~~ | | `gpu_id` | Device index to select. Defaults to `0`. ~~int~~ |
| **RETURNS** | `True` ~~bool~~ |
## displaCy {#displacy source="spacy/displacy"} ## displaCy {#displacy source="spacy/displacy"}

12
website/docs/usage/layers-architectures.md
View File

@ -356,6 +356,18 @@ that training configs are complete and experiments fully reproducible.
</Infobox> </Infobox>
Note that when using a PyTorch or Tensorflow model, it is recommended to set the GPU
memory allocator accordingly. When `gpu_allocator` is set to "pytorch" or
"tensorflow" in the training config, cupy will allocate memory via those respective libraries,
preventing OOM errors when there's available memory sitting in the other
library's pool.
```ini
### config.cfg (excerpt)
[training]
gpu_allocator = "pytorch"
```
## Custom models with Thinc {#thinc} ## Custom models with Thinc {#thinc}
Of course it's also possible to define the `Model` from the previous section Of course it's also possible to define the `Model` from the previous section