[embedding] rollback for better FAW performance (#1625)

pull/1630/head
Jiarui Fang 2 years ago committed by GitHub
parent d925122020
commit 38c68b5b9a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

@ -20,15 +20,15 @@ class CachedParamMgr(torch.nn.Module):
CPU maintains the entire original weight.
CUDA maintains a fraction of the weights used in the upcoming computation. The row number in CUDA is controlled by `cuda_row_num`.
During training, GPU needs to transmit embedding rows between CPU and GPU.
Args:
weight (torch.Tensor): the weight of the Embedding layer.
cuda_row_num (int, optional): the number of rows cached in CUDA memory. Defaults to 0.
buffer_size (int, optional): the number of rows in a data transmitter buffer. Defaults to 50_000.
pin_weight (bool, optional): use pin memory to store the cpu weight. If set `True`, the cpu memory usage will increase largely. Defaults to False.
evict_strategy (EvictionStrategy, optional): the eviction strategy. There are two options. `EvictionStrategy.LFU` uses the least frequently used cache. `EvictionStrategy.DATASET`: use the stats collected from the target dataset. It usually leads to less cpu-gpu communication volume.
Default as EvictionStrategy.DATASET.
use_cpu_caching (bool, optional): use cpu to execute cache indexing. It is slower than use gpu.
pin_weight (bool, optional): use pin memory to store the cpu weight. If set `True`, the cpu memory usage will increase largely. Defaults to False.
evict_strategy (EvictionStrategy, optional): the eviction strategy. There are two options.
`EvictionStrategy.LFU`: use the least frequently used cache.
`EvictionStrategy.DATASET`: use the stats collected from the target dataset. It usually leads to less cpu-gpu communication volume.
Defaults to EvictionStrategy.DATASET.
"""
def __init__(
@ -38,7 +38,6 @@ class CachedParamMgr(torch.nn.Module):
buffer_size: int = 0,
pin_weight: bool = False,
evict_strategy: EvictionStrategy = EvictionStrategy.DATASET,
use_cpu_caching=False,
) -> None:
super(CachedParamMgr, self).__init__()
self.buffer_size = buffer_size
@ -48,13 +47,6 @@ class CachedParamMgr(torch.nn.Module):
self.pin_weight = pin_weight
self.elem_size_in_byte = weight.element_size()
self._cpu_caching = use_cpu_caching
if self._cpu_caching:
self._cache_dev = torch.device('cpu')
else:
self._cache_dev = torch.cuda.current_device()
# weight configure
self._init_weight(weight)
@ -69,24 +61,16 @@ class CachedParamMgr(torch.nn.Module):
if self._evict_strategy == EvictionStrategy.LFU:
# cache_row_idx -> frequency, freq of the cache rows.
# classic lfu cache. evict the minimal freq value row in cuda cache.
if self._cpu_caching:
self.freq_cnter = torch.empty(self.cuda_row_num, device=self._cache_dev,
dtype=torch.long).fill_(sys.maxsize)
else:
self.register_buffer("freq_cnter",
torch.empty(self.cuda_row_num, device=self._cache_dev,
dtype=torch.long).fill_(sys.maxsize),
persistent=False)
self.register_buffer("freq_cnter",
torch.empty(self.cuda_row_num, device=torch.cuda.current_device(),
dtype=torch.long).fill_(sys.maxsize),
persistent=False)
def _find_evict_gpu_idxs(self, evict_num: int) -> torch.Tensor:
"""_find_evict_gpu_idxs
Find the gpu idxs to be evicted, according to their freq.
Args:
evict_num (int): how many rows has to be evicted
Returns:
torch.Tensor: a list tensor (1D), contains the gpu_row_idxs.
"""
@ -117,32 +101,26 @@ class CachedParamMgr(torch.nn.Module):
self.weight = weight.pin_memory() if self.pin_weight else weight
# map original id to new id with respect to frequency
# id -> cpu_row_idx
self.register_buffer(
"idx_map",
torch.arange(self.num_embeddings, dtype=torch.long, device=torch.cuda.current_device()),
persistent=False,
)
if self._cpu_caching:
self.idx_map = torch.arange(self.num_embeddings, dtype=torch.long, device=self._cache_dev)
self.cached_idx_map = torch.empty(self.cuda_row_num, device=self._cache_dev, dtype=torch.long).fill_(-1)
self.inverted_cached_idx = torch.zeros(self.num_embeddings, device=self._cache_dev,
dtype=torch.long).fill_(-1)
else:
self.register_buffer(
"idx_map",
torch.arange(self.num_embeddings, dtype=torch.long, device=self._cache_dev),
persistent=False,
)
# cached_idx_map: gpu_row_idx -> cpu_row_idx
self.register_buffer("cached_idx_map",
torch.empty(self.cuda_row_num, device=self._cache_dev, dtype=torch.long).fill_(-1),
persistent=False)
# cpu_row_id -> gpu_row_idx.
# gpu_row_idx as -1 means cpu_row_id not in CUDA.
self.register_buffer("inverted_cached_idx",
torch.zeros(self.num_embeddings, device=self._cache_dev,
dtype=torch.long).fill_(-1),
persistent=False)
self.evict_backlist = torch.tensor([], device=self._cache_dev)
# cached_idx_map: gpu_row_idx -> cpu_row_idx
self.register_buffer("cached_idx_map",
torch.empty(self.cuda_row_num, device=torch.cuda.current_device(),
dtype=torch.long).fill_(-1),
persistent=False)
# cpu_row_id -> gpu_row_idx.
# gpu_row_idx as -1 means cpu_row_id not in CUDA.
self.register_buffer("inverted_cached_idx",
torch.zeros(self.num_embeddings, device=torch.cuda.current_device(),
dtype=torch.long).fill_(-1),
persistent=False)
self.evict_backlist = torch.tensor([], device=torch.cuda.current_device())
# index copy buffer size should less than 10% of cuda weight.
if self.buffer_size > 0:
@ -157,10 +135,8 @@ class CachedParamMgr(torch.nn.Module):
def cpu_weight_data(self, row_idx: int) -> torch.Tensor:
"""
access a row of CPU weight.
Args:
row_idx (int): the idx of rows
Returns:
torch.Tensor: a piece of memory in CPU weight corresponding to row id's payload. The tensor is 1-D.
"""
@ -178,14 +154,12 @@ class CachedParamMgr(torch.nn.Module):
"""reorder
reorder the weight according to ids' frequency in dataset before training.
Execute only once before training, also known as warmup phase.
Note:
If you would like to use the DATASET as the eviction strategy, you must call this function.
Note:
If you are use the LFU as the eviction strategy, you can skip this function. If you still use this function. It will initialize
The frequency in LFU cache using the dataset statistics.
Args:
ids_freq_mapping (List[int]): a list, whose offset is id number, value is freq. if None then not reorder the cpu weight.
warmup_ratio (float): the amount of chunks preloaded in cuda cache
@ -209,24 +183,24 @@ class CachedParamMgr(torch.nn.Module):
# extract rows from cpu weight
if self._evict_strategy == EvictionStrategy.LFU and ids_freq_mapping is not None:
freq_value, preload_cpu_ids = torch.topk(ids_freq_mapping, preload_row_num, dim=0, largest=True)
preload_cuda_row_idxs = torch.arange(preload_row_num).to(self._cache_dev)
preload_cuda_row_idxs = torch.arange(preload_row_num).cuda()
else:
preload_cpu_ids = torch.arange(preload_row_num, device=self.weight.device)
preload_cuda_row_idxs = preload_cpu_ids.to(self._cache_dev)
preload_cpu_ids = torch.arange(preload_row_num)
preload_cuda_row_idxs = preload_cpu_ids.cuda()
if self.buffer_size > 0:
self.limit_buff_index_copyer.index_copy(0,
src_index=preload_cpu_ids,
tgt_index=preload_cuda_row_idxs.cuda(),
tgt_index=preload_cuda_row_idxs,
src=self.weight.view(self.num_embeddings, -1),
tgt=self.cuda_cached_weight.view(self.cuda_row_num, -1))
else:
preload_rows = self.weight.view(self.num_embeddings, -1).index_select(0, preload_cpu_ids).cuda()
self.cuda_cached_weight.view(self.cuda_row_num, -1).index_copy_(0, preload_cuda_row_idxs.cuda(),
self.cuda_cached_weight.view(self.cuda_row_num, -1).index_copy_(0, preload_cuda_row_idxs,
preload_rows)
# update auxiliary info
self.cached_idx_map[preload_cuda_row_idxs] = preload_cpu_ids.to(self._cache_dev)
self.cached_idx_map[preload_cuda_row_idxs] = preload_cpu_ids.cuda()
self.inverted_cached_idx[preload_cpu_ids] = preload_cuda_row_idxs
self._cuda_available_row_num -= preload_row_num
@ -235,7 +209,7 @@ class CachedParamMgr(torch.nn.Module):
if ids_freq_mapping is None:
self.freq_cnter.index_fill_(0, preload_cuda_row_idxs, 0)
else:
self.freq_cnter[preload_cuda_row_idxs] = freq_value.to(self._cache_dev)
self.freq_cnter[preload_cuda_row_idxs] = freq_value.cuda()
print(f'Cache warmup finished cost {timer.elapsed} sec.')
@ -245,7 +219,7 @@ class CachedParamMgr(torch.nn.Module):
"""
slots = torch.nonzero(self.cached_idx_map > -1).squeeze(1)
row_ids = self.cached_idx_map[slots]
rows = self.cuda_cached_weight.view(self.cuda_row_num, -1).index_select(0, slots.cuda()).cpu()
rows = self.cuda_cached_weight.view(self.cuda_row_num, -1).index_select(0, slots).cpu()
self.weight.view(self.num_embeddings, -1).index_copy_(0, row_ids.cpu(), rows)
self.cached_idx_map.index_fill_(0, slots, -1)
self.inverted_cached_idx.index_fill_(0, row_ids, -1)
@ -272,10 +246,8 @@ class CachedParamMgr(torch.nn.Module):
"""
convert ids to indices in self.cuda_cached_weight.
Implemented with parallel operations on GPU.
Args:
ids (torch.Tensor): ids from the dataset
Returns:
torch.Tensor: contains indices in self.cuda_cached_weight
"""
@ -287,14 +259,12 @@ class CachedParamMgr(torch.nn.Module):
def prepare_ids(self, ids: torch.Tensor) -> torch.Tensor:
"""
move the cpu embedding rows w.r.t. ids into CUDA memory
Args:
ids (torch.Tensor): the ids to be computed
Returns:
torch.Tensor: indices on the cuda_cached_weight.
"""
with record_function(f"(pre-id) get unique indices. cache ratio {self.cuda_row_num / self.num_embeddings}"):
ids = ids.to(self._cache_dev)
with record_function("(zhg) get unique indices"):
cpu_row_idxs, repeat_times = torch.unique(self.idx_map.index_select(0, ids), return_counts=True)
assert len(cpu_row_idxs) <= self.cuda_row_num, \
@ -303,29 +273,26 @@ class CachedParamMgr(torch.nn.Module):
f"Please increase cuda_row_num or decrease the training batch size."
self.evict_backlist = cpu_row_idxs
with record_function("(pre-id) get cpu row idxs"):
comm_cpu_row_idxs = cpu_row_idxs[torch.isin(cpu_row_idxs,
self.cached_idx_map,
assume_unique=True,
invert=True)]
with record_function("(zhg) get cpu row idxs"):
comm_cpu_row_idxs = cpu_row_idxs[torch.isin(cpu_row_idxs, self.cached_idx_map, invert=True)]
self.num_hits_history.append(len(cpu_row_idxs) - len(comm_cpu_row_idxs))
self.num_miss_history.append(len(comm_cpu_row_idxs))
self.num_write_back_history.append(0)
# move sure the cuda rows will not be evicted!
with record_function("(pre-id) cache update"):
with record_function("(zhg) cache update"):
self._prepare_rows_on_cuda(comm_cpu_row_idxs)
self.evict_backlist = torch.tensor([], device=cpu_row_idxs.device, dtype=cpu_row_idxs.dtype)
with record_function("(pre-id) embed cpu rows idx -> cache gpu row idxs"):
self.evict_backlist = torch.tensor([], device=cpu_row_idxs.device, dtype=cpu_row_idxs.dtype)
with record_function("(zhg) embed cpu rows idx -> cache gpu row idxs"):
gpu_row_idxs = self._id_to_cached_cuda_id(ids)
# update for LFU.
if self._evict_strategy == EvictionStrategy.LFU:
with record_function("(pre-id) lfu cnter updates"):
unique_gpu_row_idxs = self.inverted_cached_idx[cpu_row_idxs]
self.freq_cnter.scatter_add_(0, unique_gpu_row_idxs, repeat_times)
unique_gpu_row_idxs = self.inverted_cached_idx[cpu_row_idxs]
self.freq_cnter.scatter_add_(0, unique_gpu_row_idxs, repeat_times)
return gpu_row_idxs
@ -341,14 +308,13 @@ class CachedParamMgr(torch.nn.Module):
@torch.no_grad()
def _prepare_rows_on_cuda(self, cpu_row_idxs: torch.Tensor) -> None:
"""prepare rows in cpu_row_idxs on CUDA memory
Args:
cpu_row_idxs (torch.Tensor): the rows to be placed on CUDA
"""
evict_num = cpu_row_idxs.numel() - self.cuda_available_row_num
if evict_num > 0:
with Timer() as timer:
mask_cpu_row_idx = torch.isin(self.cached_idx_map, self.evict_backlist, assume_unique=True)
mask_cpu_row_idx = torch.isin(self.cached_idx_map, self.evict_backlist)
invalid_idxs = torch.nonzero(mask_cpu_row_idx).squeeze(1)
if self._evict_strategy == EvictionStrategy.DATASET:
# mask method.
@ -375,8 +341,7 @@ class CachedParamMgr(torch.nn.Module):
tgt=self.weight.view(self.num_embeddings, -1))
else:
# allocate tmp memory on CPU and copy rows on CUDA to CPU.
rows = self.cuda_cached_weight.view(self.cuda_row_num,
-1).index_select(0, evict_gpu_row_idxs.cuda()).cpu()
rows = self.cuda_cached_weight.view(self.cuda_row_num, -1).index_select(0, evict_gpu_row_idxs).cpu()
self.weight.view(self.num_embeddings, -1).index_copy_(0, evict_info.cpu(), rows)
self.cached_idx_map.index_fill_(0, evict_gpu_row_idxs, -1)
@ -395,12 +360,12 @@ class CachedParamMgr(torch.nn.Module):
if self.buffer_size > 0:
self.limit_buff_index_copyer.index_copy(0,
src_index=cpu_row_idxs.cpu(),
tgt_index=slots.cuda(),
tgt_index=slots,
src=self.weight.view(self.num_embeddings, -1),
tgt=self.cuda_cached_weight.view(self.cuda_row_num, -1))
else:
rows = self.weight.view(self.num_embeddings, -1).index_select(0, cpu_row_idxs.cpu()).cuda()
self.cuda_cached_weight.view(self.cuda_row_num, -1).index_copy_(0, slots.cuda(), rows)
self.cuda_cached_weight.view(self.cuda_row_num, -1).index_copy_(0, slots, rows)
slot_offsets = slots
self.cached_idx_map[slots] = cpu_row_idxs
self.inverted_cached_idx.index_copy_(0, cpu_row_idxs, slot_offsets)
@ -421,7 +386,6 @@ class CachedParamMgr(torch.nn.Module):
def _evict(self) -> int:
"""
deprecated
evict one row from cuda to cpu.
Returns:
(int) : the slot id be evicted.
@ -463,9 +427,7 @@ class CachedParamMgr(torch.nn.Module):
def _admit(self, row_id: int):
"""
deprecated
move in row_id to CUDA
Args:
row_id (int): the id of row to be moved in
"""
@ -491,4 +453,4 @@ class CachedParamMgr(torch.nn.Module):
self._cuda_available_row_num -= 1
self._cpu_to_cuda_numel += self.embedding_dim
self._cpu_to_cuda_elpase += timer.elapsed
self._cpu_to_cuda_elpase += timer.elapsed

@ -90,7 +90,7 @@ def test_reorder_with_freq():
offset_in_chunk = torch.tensor(offset_in_chunk, dtype=torch.long, device=dev)
weight = torch.rand(num_embed, 2)
mgr = CachedParamMgr(weight, num_chunk, use_cpu_caching=dev.type == 'cpu')
mgr = CachedParamMgr(weight, num_chunk)
mgr.reorder(idx_map)

Loading…
Cancel
Save