github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-08-14 22:23:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

[FCE] update interface for frequency statistics in FreqCacheEmbedding (#1462)

Browse Source
This commit is contained in:
Geng Zhang 2022-08-23 17:38:24 +08:00 committed by GitHub
parent ede326298b
commit 0aad53c62b
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 30 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
colossalai/nn/parallel/layers/cache_embedding/cache_mgr.py
View File

@ -14,12 +14,17 @@ class CachedParamMgr(torch.nn.Module):
During training, GPU needs to transmit rows between CPU and GPU. During training, GPU needs to transmit rows between CPU and GPU.
""" """
def __init__(self, weight: torch.Tensor, cuda_row_num: int = 0, buffer_size: int = 50_000) -> None: def __init__(self,
weight: torch.Tensor,
cuda_row_num: int = 0,
buffer_size: int = 50_000,
pin_weight=False) -> None:
super(CachedParamMgr, self).__init__() super(CachedParamMgr, self).__init__()
self.buffer_size = buffer_size self.buffer_size = buffer_size
self.num_embeddings, self.embedding_dim = weight.shape self.num_embeddings, self.embedding_dim = weight.shape
self.cuda_row_num = cuda_row_num self.cuda_row_num = cuda_row_num
self._cuda_available_row_num = self.cuda_row_num self._cuda_available_row_num = self.cuda_row_num
self.pin_weight = pin_weight
self.elem_size_in_byte = weight.element_size() self.elem_size_in_byte = weight.element_size()
@ -43,8 +48,7 @@ class CachedParamMgr(torch.nn.Module):
dtype=weight.dtype)) dtype=weight.dtype))
# pin memory cpu for higher CPU-GPU copy bandwidth # pin memory cpu for higher CPU-GPU copy bandwidth
self.weight = weight.contiguous().cpu().pin_memory() self.weight = weight.pin_memory() if self.pin_weight else weight
# map original id to new id with respect to frequency # map original id to new id with respect to frequency
# id -> cpu_row_idx # id -> cpu_row_idx
self.register_buffer( self.register_buffer(
@ -109,7 +113,7 @@ class CachedParamMgr(torch.nn.Module):
warmup_ratio (float): the amount of chunks preloaded in cuda cache warmup_ratio (float): the amount of chunks preloaded in cuda cache
""" """
if ids_freq_mapping is not None: if ids_freq_mapping is not None:
tmp_idx = torch.argsort(torch.from_numpy(ids_freq_mapping).cuda(), descending=True) tmp_idx = torch.argsort(ids_freq_mapping, descending=True)
sorted_idx = torch.argsort(tmp_idx) sorted_idx = torch.argsort(tmp_idx)
self.idx_map.data.copy_(sorted_idx) self.idx_map.data.copy_(sorted_idx)

17
colossalai/nn/parallel/layers/cache_embedding/freq_aware_embedding.py
View File

@ -27,20 +27,19 @@ class FreqAwareEmbeddingBag(BaseEmbeddingBag):
ids_freq_mapping=None, ids_freq_mapping=None,
warmup_ratio=0.7, warmup_ratio=0.7,
buffer_size=50_000, buffer_size=50_000,
pin_weight=False,
): ):
super(FreqAwareEmbeddingBag, self).__init__(num_embeddings, embedding_dim, padding_idx, max_norm, norm_type, super(FreqAwareEmbeddingBag, self).__init__(num_embeddings, embedding_dim, padding_idx, max_norm, norm_type,
scale_grad_by_freq, sparse, mode, include_last_offset) scale_grad_by_freq, sparse, mode, include_last_offset)
if _weight is None: if _weight is None:
_weight = self._weight_alloc(dtype, device) _weight = self._weight_alloc(dtype, device)
else:
_weight = _weight
# configure weight & cache # configure weight & cache
self._preprocess(_weight, cuda_row_num, ids_freq_mapping, warmup_ratio, buffer_size) self._preprocess(_weight, cuda_row_num, ids_freq_mapping, warmup_ratio, buffer_size, pin_weight)
def _weight_alloc(self, dtype, device): def _weight_alloc(self, dtype, device):
weight = torch.empty(self.num_embeddings, self.embedding_dim, dtype=dtype, device=device, pin_memory=True) weight = torch.empty(self.num_embeddings, self.embedding_dim, dtype=dtype, device=device)
with torch.no_grad(): with torch.no_grad():
weight.data.uniform_(-1 / self.num_embeddings, 1 / self.num_embeddings) weight.data.uniform_(-1 / self.num_embeddings, 1 / self.num_embeddings)
if self.padding_idx is not None: if self.padding_idx is not None:
@ -52,7 +51,8 @@ class FreqAwareEmbeddingBag(BaseEmbeddingBag):
cuda_row_num: int, cuda_row_num: int,
ids_freq_mapping: Optional[List[int]] = None, ids_freq_mapping: Optional[List[int]] = None,
warmup_ratio=0.7, warmup_ratio=0.7,
buffer_size=50_000): buffer_size=50_000,
pin_weight=False):
""" """
Called after initialized. Called after initialized.
Reorder the weight rows according to the ids_freq_mapping. Reorder the weight rows according to the ids_freq_mapping.
@ -63,17 +63,18 @@ class FreqAwareEmbeddingBag(BaseEmbeddingBag):
ids_freq_mapping (List[int]): a list, idx is id number, value is freq ids_freq_mapping (List[int]): a list, idx is id number, value is freq
warmup_ratio (float): the amount of rows preloaded in cuda cache warmup_ratio (float): the amount of rows preloaded in cuda cache
""" """
self.cache_weight_mgr = CachedParamMgr(weight, cuda_row_num, buffer_size) self.cache_weight_mgr = CachedParamMgr(weight, cuda_row_num, buffer_size, pin_weight)
self.cache_weight_mgr.reorder(ids_freq_mapping, warmup_ratio) self.cache_weight_mgr.reorder(ids_freq_mapping, warmup_ratio)
def forward(self, indices, offsets=None, per_sample_weights=None): def forward(self, indices, offsets=None, per_sample_weights=None, shape_hook=None):
with torch.no_grad(): with torch.no_grad():
reorder_ids = self.cache_weight_mgr.prepare_ids(indices) reorder_ids = self.cache_weight_mgr.prepare_ids(indices)
embeddings = F.embedding_bag(reorder_ids, self.cache_weight_mgr.cuda_cached_weight, offsets, self.max_norm, embeddings = F.embedding_bag(reorder_ids, self.cache_weight_mgr.cuda_cached_weight, offsets, self.max_norm,
self.norm_type, self.scale_grad_by_freq, self.mode, self.sparse, self.norm_type, self.scale_grad_by_freq, self.mode, self.sparse,
per_sample_weights, self.include_last_offset, self.padding_idx) per_sample_weights, self.include_last_offset, self.padding_idx)
if shape_hook is not None:
embeddings = shape_hook(embeddings)
return embeddings return embeddings
@property @property

16
colossalai/nn/parallel/layers/cache_embedding/parallel_freq_aware_embedding.py
View File

@ -3,8 +3,6 @@ import torch.nn.functional as F
from typing import List, Optional, Iterator, Tuple from typing import List, Optional, Iterator, Tuple
from .freq_aware_embedding import FreqAwareEmbeddingBag from .freq_aware_embedding import FreqAwareEmbeddingBag
from .cache_mgr import CachedParamMgr
from torch.nn.parameter import Parameter
from colossalai.nn._ops._utils import dual_all_to_all from colossalai.nn._ops._utils import dual_all_to_all
from colossalai.tensor import ColoParameter, ShardSpec, ComputePattern, ProcessGroup, ColoTensorSpec, ColoTensor from colossalai.tensor import ColoParameter, ShardSpec, ComputePattern, ProcessGroup, ColoTensorSpec, ColoTensor
@ -49,6 +47,7 @@ class ParallelFreqAwareEmbeddingBag(FreqAwareEmbeddingBag):
ids_freq_mapping=None, ids_freq_mapping=None,
warmup_ratio=0.7, warmup_ratio=0.7,
buffer_size=50_000, buffer_size=50_000,
pin_weight=False,
): ):
self.rank = torch.distributed.get_rank() self.rank = torch.distributed.get_rank()
self.world_size = torch.distributed.get_world_size() self.world_size = torch.distributed.get_world_size()
@ -60,17 +59,18 @@ class ParallelFreqAwareEmbeddingBag(FreqAwareEmbeddingBag):
super(ParallelFreqAwareEmbeddingBag, super(ParallelFreqAwareEmbeddingBag,
self).__init__(num_embeddings, embedding_dim, padding_idx, max_norm, norm_type, scale_grad_by_freq, self).__init__(num_embeddings, embedding_dim, padding_idx, max_norm, norm_type, scale_grad_by_freq,
sparse, _weight, mode, include_last_offset, dtype, device, cuda_row_num, ids_freq_mapping, sparse, _weight, mode, include_last_offset, dtype, device, cuda_row_num, ids_freq_mapping,
warmup_ratio, buffer_size) warmup_ratio, buffer_size, pin_weight)
def _weight_alloc(self, dtype, device): def _weight_alloc(self, dtype, device):
weight = torch.empty(self.num_embeddings, self.embedding_dim_per_partition, device=device, dtype=dtype)
with torch.no_grad():
weight.data.uniform_(-1 / self.num_embeddings, 1 / self.num_embeddings)
if self.padding_idx is not None:
weight[self.padding_idx].fill_(0)
colo_tensor_spec = ColoTensorSpec(pg=ProcessGroup(tp_degree=self.world_size), colo_tensor_spec = ColoTensorSpec(pg=ProcessGroup(tp_degree=self.world_size),
dist_attr=ShardSpec(dims=[-1], num_partitions=[self.world_size]), dist_attr=ShardSpec(dims=[-1], num_partitions=[self.world_size]),
compute_attr=ComputePattern.TP1D) compute_attr=ComputePattern.TP1D)
return ColoTensor.from_torch_tensor(torch.empty(self.num_embeddings, return ColoTensor.from_torch_tensor(weight, spec=colo_tensor_spec)
self.embedding_dim_per_partition,
device=device,
dtype=dtype),
spec=colo_tensor_spec)
def forward(self, indices, offsets=None, per_sample_weights=None, shape_hook=None, scatter_dim=0, gather_dim=-1): def forward(self, indices, offsets=None, per_sample_weights=None, shape_hook=None, scatter_dim=0, gather_dim=-1):
with torch.no_grad(): with torch.no_grad():

10
tests/test_layers/test_cache_embedding.py
View File

@ -44,7 +44,7 @@ def synthesize_1d_sparse_feature(
def test_cachemgr(): def test_cachemgr():
model = torch.nn.EmbeddingBag(10000, 128) model = torch.nn.EmbeddingBag(10000, 128)
# 10 chunks, 5 in cuda # 10 chunks, 5 in cuda
mgr = CachedParamMgr(model.weight, 5) mgr = CachedParamMgr(model.weight.detach(), 5)
assert mgr.cuda_row_num == 5 assert mgr.cuda_row_num == 5
mgr._admit(1) mgr._admit(1)
@ -74,8 +74,8 @@ def test_reorder_with_freq():
chunk_size = 1 chunk_size = 1
num_chunk = 5 num_chunk = 5
idx_map = np.random.randint(10000, size=(num_embed,)) idx_map = torch.randint(10000, size=(num_embed,))
sorted_idx = np.flipud(np.argsort(idx_map)).tolist() sorted_idx = torch.argsort(idx_map, descending=True).tolist()
chunkid, offset_in_chunk = [], [] chunkid, offset_in_chunk = [], []
for i in range(num_embed): for i in range(num_embed):
idx = sorted_idx.index(i) idx = sorted_idx.index(i)
@ -231,6 +231,6 @@ def test_parallel_freq_aware_embed(world_size):
if __name__ == '__main__': if __name__ == '__main__':
# test_cachemgr() test_cachemgr()
# test_freq_aware_embed() # test_freq_aware_embed()
test_parallel_freq_aware_embed(2) # test_parallel_freq_aware_embed(2)