Unverified Commit 31484afb authored by Sylvain Gugger's avatar Sylvain Gugger Committed by GitHub
Browse files

Add test for new model parallelism features (#17401)

parent 56b35ce3
...@@ -1734,6 +1734,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix ...@@ -1734,6 +1734,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
same device. same device.
To have Accelerate compute the most optimized `device_map` automatically, set `device_map="auto"`. To have Accelerate compute the most optimized `device_map` automatically, set `device_map="auto"`.
max_memory (`Dict`, *optional*):
A dictionary device identifier to maximum memory. Will default to the maximum memory available for each
GPU and the available CPU RAM if unset.
offload_folder (`str` or `os.PathLike`, *optional*): offload_folder (`str` or `os.PathLike`, *optional*):
If the `device_map` contains any value `"disk"`, the folder where we will offload weights. If the `device_map` contains any value `"disk"`, the folder where we will offload weights.
offload_state_dict (`bool`, *optional*, defaults to `False`): offload_state_dict (`bool`, *optional*, defaults to `False`):
...@@ -1822,6 +1825,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix ...@@ -1822,6 +1825,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
torch_dtype = kwargs.pop("torch_dtype", None) torch_dtype = kwargs.pop("torch_dtype", None)
low_cpu_mem_usage = kwargs.pop("low_cpu_mem_usage", None) low_cpu_mem_usage = kwargs.pop("low_cpu_mem_usage", None)
device_map = kwargs.pop("device_map", None) device_map = kwargs.pop("device_map", None)
max_memory = kwargs.pop("max_memory", None)
offload_folder = kwargs.pop("offload_folder", None) offload_folder = kwargs.pop("offload_folder", None)
offload_state_dict = kwargs.pop("offload_state_dict", False) offload_state_dict = kwargs.pop("offload_state_dict", False)
...@@ -2119,7 +2123,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix ...@@ -2119,7 +2123,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
if model._no_split_modules is None: if model._no_split_modules is None:
raise ValueError(f"{model.__class__.__name__} does not support `device_map='auto'` yet.") raise ValueError(f"{model.__class__.__name__} does not support `device_map='auto'` yet.")
no_split_modules = model._no_split_modules no_split_modules = model._no_split_modules
device_map = infer_auto_device_map(model, no_split_module_classes=no_split_modules, dtype=torch_dtype) device_map = infer_auto_device_map(
model, no_split_module_classes=no_split_modules, dtype=torch_dtype, max_memory=max_memory
)
if from_tf: if from_tf:
if resolved_archive_file.endswith(".index"): if resolved_archive_file.endswith(".index"):
......
...@@ -420,14 +420,12 @@ class T5Attention(nn.Module): ...@@ -420,14 +420,12 @@ class T5Attention(nn.Module):
relative_buckets += torch.where(is_small, relative_position, relative_position_if_large) relative_buckets += torch.where(is_small, relative_position, relative_position_if_large)
return relative_buckets return relative_buckets
def compute_bias(self, query_length, key_length): def compute_bias(self, query_length, key_length, device=None):
"""Compute binned relative position bias""" """Compute binned relative position bias"""
context_position = torch.arange( if device is None:
query_length, dtype=torch.long, device=self.relative_attention_bias.weight.device device = self.relative_attention_bias.weight.device
)[:, None] context_position = torch.arange(query_length, dtype=torch.long, device=device)[:, None]
memory_position = torch.arange( memory_position = torch.arange(key_length, dtype=torch.long, device=device)[None, :]
key_length, dtype=torch.long, device=self.relative_attention_bias.weight.device
)[None, :]
relative_position = memory_position - context_position # shape (query_length, key_length) relative_position = memory_position - context_position # shape (query_length, key_length)
relative_position_bucket = self._relative_position_bucket( relative_position_bucket = self._relative_position_bucket(
relative_position, # shape (query_length, key_length) relative_position, # shape (query_length, key_length)
...@@ -522,7 +520,7 @@ class T5Attention(nn.Module): ...@@ -522,7 +520,7 @@ class T5Attention(nn.Module):
if self.gradient_checkpointing and self.training: if self.gradient_checkpointing and self.training:
position_bias.requires_grad = True position_bias.requires_grad = True
else: else:
position_bias = self.compute_bias(real_seq_length, key_length) position_bias = self.compute_bias(real_seq_length, key_length, device=scores.device)
# if key and values are already calculated # if key and values are already calculated
# we want only the last query position bias # we want only the last query position bias
......
...@@ -51,7 +51,9 @@ from transformers.testing_utils import ( ...@@ -51,7 +51,9 @@ from transformers.testing_utils import (
is_pt_flax_cross_test, is_pt_flax_cross_test,
is_pt_tf_cross_test, is_pt_tf_cross_test,
is_staging_test, is_staging_test,
require_accelerate,
require_torch, require_torch,
require_torch_gpu,
require_torch_multi_gpu, require_torch_multi_gpu,
require_usr_bin_time, require_usr_bin_time,
slow, slow,
...@@ -60,6 +62,7 @@ from transformers.testing_utils import ( ...@@ -60,6 +62,7 @@ from transformers.testing_utils import (
from transformers.utils import ( from transformers.utils import (
WEIGHTS_INDEX_NAME, WEIGHTS_INDEX_NAME,
WEIGHTS_NAME, WEIGHTS_NAME,
is_accelerate_available,
is_flax_available, is_flax_available,
is_tf_available, is_tf_available,
is_torch_fx_available, is_torch_fx_available,
...@@ -72,6 +75,10 @@ sys.path.append(str(Path(__file__).parent.parent / "utils")) ...@@ -72,6 +75,10 @@ sys.path.append(str(Path(__file__).parent.parent / "utils"))
from test_module.custom_configuration import CustomConfig, NoSuperInitConfig # noqa E402 from test_module.custom_configuration import CustomConfig, NoSuperInitConfig # noqa E402
if is_accelerate_available():
from accelerate.utils import compute_module_sizes
if is_torch_available(): if is_torch_available():
import torch import torch
from torch import nn from torch import nn
...@@ -2178,6 +2185,86 @@ class ModelTesterMixin: ...@@ -2178,6 +2185,86 @@ class ModelTesterMixin:
model.parallelize() model.parallelize()
model.generate(**cast_to_device(inputs_dict, "cuda:0"), num_beams=2) model.generate(**cast_to_device(inputs_dict, "cuda:0"), num_beams=2)
def check_device_map_is_respected(self, model, device_map):
for param_name, param in model.named_parameters():
# Find device in device_map
while len(param_name) > 0 and param_name not in device_map:
param_name = ".".join(param_name.split(".")[:-1])
if param_name not in device_map:
raise ValueError("device map is incomplete, it does not contain any device for `param_name`.")
param_device = device_map[param_name]
if param_device in ["cpu", "disk"]:
self.assertEqual(param.device, torch.device("meta"))
else:
self.assertEqual(param.device, torch.device(param_device))
@require_accelerate
@require_torch_gpu
def test_cpu_offload(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
if config.num_hidden_layers < 5:
config.num_hidden_layers = 5
for model_class in self.all_model_classes:
if model_class._no_split_modules is None:
continue
inputs_dict = self._prepare_for_class(inputs_dict, model_class)
model = model_class(config).eval()
model = model.to(torch_device)
base_output = model(**inputs_dict)
model_size = compute_module_sizes(model)[""]
# We test several splits of sizes to make sure it works.
max_gpu_sizes = [int(p * model_size) for p in [0.5, 0.7, 0.9]]
with tempfile.TemporaryDirectory() as tmp_dir:
model.cpu().save_pretrained(tmp_dir)
for max_size in max_gpu_sizes:
max_memory = {0: max_size, "cpu": model_size * 2}
new_model = model_class.from_pretrained(tmp_dir, device_map="auto", max_memory=max_memory)
# Making sure part of the model will actually end up offloaded
self.assertSetEqual(set(new_model.hf_device_map.values()), {0, "cpu"})
self.check_device_map_is_respected(new_model, new_model.hf_device_map)
new_output = new_model(**inputs_dict)
self.assertTrue(torch.allclose(base_output[0], new_output[0]))
@require_accelerate
@require_torch_multi_gpu
def test_model_parallelism(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
if config.num_hidden_layers < 5:
config.num_hidden_layers = 5
for model_class in self.all_model_classes:
if model_class._no_split_modules is None:
continue
inputs_dict = self._prepare_for_class(inputs_dict, model_class)
model = model_class(config).eval()
model = model.to(torch_device)
base_output = model(**inputs_dict)
model_size = compute_module_sizes(model)[""]
# We test several splits of sizes to make sure it works.
max_gpu_sizes = [int(p * model_size) for p in [0.5, 0.7, 0.9]]
with tempfile.TemporaryDirectory() as tmp_dir:
model.cpu().save_pretrained(tmp_dir)
for max_size in max_gpu_sizes:
max_memory = {0: max_size, 1: model_size * 2, "cpu": model_size * 2}
new_model = model_class.from_pretrained(tmp_dir, device_map="auto", max_memory=max_memory)
# Making sure part of the model will actually end up offloaded
self.assertSetEqual(set(new_model.hf_device_map.values()), {0, 1})
self.check_device_map_is_respected(new_model, new_model.hf_device_map)
new_output = new_model(**inputs_dict)
self.assertTrue(torch.allclose(base_output[0], new_output[0]))
def test_problem_types(self): def test_problem_types(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common() config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
...@@ -2547,6 +2634,7 @@ class ModelUtilsTest(TestCasePlus): ...@@ -2547,6 +2634,7 @@ class ModelUtilsTest(TestCasePlus):
for p1, p2 in zip(model.parameters(), ref_model.parameters()): for p1, p2 in zip(model.parameters(), ref_model.parameters()):
self.assertTrue(torch.allclose(p1, p2)) self.assertTrue(torch.allclose(p1, p2))
@require_accelerate
def test_from_pretrained_low_cpu_mem_usage_functional(self): def test_from_pretrained_low_cpu_mem_usage_functional(self):
# test that we can use `from_pretrained(..., low_cpu_mem_usage=True)` with normal and # test that we can use `from_pretrained(..., low_cpu_mem_usage=True)` with normal and
# sharded models # sharded models
...@@ -2559,6 +2647,7 @@ class ModelUtilsTest(TestCasePlus): ...@@ -2559,6 +2647,7 @@ class ModelUtilsTest(TestCasePlus):
_ = BertModel.from_pretrained(mname, low_cpu_mem_usage=True) _ = BertModel.from_pretrained(mname, low_cpu_mem_usage=True)
@require_usr_bin_time @require_usr_bin_time
@require_accelerate
def test_from_pretrained_low_cpu_mem_usage_measured(self): def test_from_pretrained_low_cpu_mem_usage_measured(self):
# test that `from_pretrained(..., low_cpu_mem_usage=True)` uses less cpu memory than default # test that `from_pretrained(..., low_cpu_mem_usage=True)` uses less cpu memory than default
...@@ -2597,6 +2686,7 @@ class ModelUtilsTest(TestCasePlus): ...@@ -2597,6 +2686,7 @@ class ModelUtilsTest(TestCasePlus):
# functionality to load models directly on gpu, this test can be rewritten to use torch's # functionality to load models directly on gpu, this test can be rewritten to use torch's
# cuda memory tracking and then we should be able to do a much more precise test. # cuda memory tracking and then we should be able to do a much more precise test.
@require_accelerate
@require_torch_multi_gpu @require_torch_multi_gpu
@slow @slow
def test_model_parallelism_gpt2(self): def test_model_parallelism_gpt2(self):
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment