2021-10-28 16:21:23 +00:00
|
|
|
#!/usr/bin/env python
|
|
|
|
# -*- encoding: utf-8 -*-
|
|
|
|
|
|
|
|
import torch
|
|
|
|
import torch.nn.functional as F
|
2023-04-06 06:51:35 +00:00
|
|
|
|
2023-09-18 08:31:06 +00:00
|
|
|
from colossalai.legacy.context.parallel_mode import ParallelMode
|
|
|
|
from colossalai.legacy.context.random import add_seed, reset_seeds, seed, set_mode
|
|
|
|
from colossalai.legacy.utils.activation_checkpoint import checkpoint
|
2023-04-06 06:51:35 +00:00
|
|
|
from colossalai.testing import clear_cache_before_run, parameterize
|
2021-10-28 16:21:23 +00:00
|
|
|
|
|
|
|
|
|
|
|
def forward(x, weight):
|
|
|
|
out = torch.matmul(x, weight)
|
|
|
|
with seed(ParallelMode.DATA):
|
|
|
|
out_ = F.dropout(out, p=0.4, training=True)
|
|
|
|
return out_
|
|
|
|
|
2022-03-11 06:48:11 +00:00
|
|
|
|
2022-08-16 07:39:20 +00:00
|
|
|
def forward_inplace_ckpt(x, weight, cpu_offload=False):
|
|
|
|
out = torch.matmul(x, weight)
|
|
|
|
bn = torch.nn.BatchNorm1d(4, affine=False)
|
|
|
|
bn = bn.to(device="cuda")
|
|
|
|
out = bn(out)
|
|
|
|
|
|
|
|
def ckpt0(x):
|
|
|
|
return F.relu(x, inplace=True)
|
|
|
|
|
|
|
|
out = checkpoint(ckpt0, cpu_offload, out, use_reentrant=False)
|
|
|
|
return out
|
|
|
|
|
|
|
|
|
|
|
|
def forward_inplace(x, weight):
|
|
|
|
out = torch.matmul(x, weight)
|
|
|
|
bn = torch.nn.BatchNorm1d(4, affine=False)
|
|
|
|
bn = bn.to(device="cuda")
|
|
|
|
out = bn(out)
|
|
|
|
out = F.relu(out, inplace=True)
|
|
|
|
return out
|
|
|
|
|
|
|
|
|
2023-04-06 06:51:35 +00:00
|
|
|
@clear_cache_before_run()
|
|
|
|
@parameterize("use_reentrant", [True, False])
|
|
|
|
@parameterize("cpu_offload", [True, False])
|
2022-08-16 07:39:20 +00:00
|
|
|
def test_activation_checkpointing(cpu_offload, use_reentrant):
|
|
|
|
# as seed manager is singleton
|
|
|
|
# if we don't reset seeds here,
|
|
|
|
# other tests might affect this test
|
|
|
|
reset_seeds()
|
2022-04-02 13:58:47 +00:00
|
|
|
|
2023-05-10 09:12:03 +00:00
|
|
|
# We put initialization here to avoid change cuda rng state below
|
2023-09-19 06:20:26 +00:00
|
|
|
inputs = torch.rand(2, 2, requires_grad=True, device="cuda")
|
|
|
|
weight = torch.rand(2, 4, requires_grad=True, device="cuda")
|
2022-04-02 13:58:47 +00:00
|
|
|
|
|
|
|
# Get a copy of input tensors
|
2023-09-19 06:20:26 +00:00
|
|
|
inputs_ = torch.empty(2, 2, requires_grad=True, device="cuda")
|
2022-04-02 13:58:47 +00:00
|
|
|
inputs_.data.copy_(inputs.data)
|
2023-09-19 06:20:26 +00:00
|
|
|
weight_ = torch.empty(2, 4, requires_grad=True, device="cuda")
|
2022-04-02 13:58:47 +00:00
|
|
|
weight_.data.copy_(weight.data)
|
|
|
|
|
2022-03-11 06:48:11 +00:00
|
|
|
add_seed(ParallelMode.GLOBAL, 1024)
|
|
|
|
add_seed(ParallelMode.DATA, 1026)
|
2021-10-28 16:21:23 +00:00
|
|
|
set_mode(ParallelMode.GLOBAL)
|
|
|
|
global_cuda_rng_state = torch.cuda.get_rng_state()
|
|
|
|
set_mode(ParallelMode.DATA)
|
|
|
|
data_parallel_cuda_rng_state = torch.cuda.get_rng_state()
|
|
|
|
set_mode(ParallelMode.GLOBAL)
|
|
|
|
|
2022-04-02 13:58:47 +00:00
|
|
|
out = forward(inputs, weight)
|
2021-10-28 16:21:23 +00:00
|
|
|
loss = out.sum()
|
|
|
|
loss.backward()
|
|
|
|
|
2022-04-02 13:58:47 +00:00
|
|
|
# Recover cuda rng states
|
2021-10-28 16:21:23 +00:00
|
|
|
set_mode(ParallelMode.GLOBAL)
|
|
|
|
torch.cuda.set_rng_state(global_cuda_rng_state)
|
|
|
|
set_mode(ParallelMode.DATA)
|
|
|
|
torch.cuda.set_rng_state(data_parallel_cuda_rng_state)
|
|
|
|
set_mode(ParallelMode.GLOBAL)
|
2022-04-02 13:58:47 +00:00
|
|
|
|
2022-08-16 07:39:20 +00:00
|
|
|
out = checkpoint(forward, cpu_offload, inputs_, weight_, use_reentrant=use_reentrant)
|
2021-10-28 16:21:23 +00:00
|
|
|
loss = out.sum()
|
|
|
|
loss.backward()
|
|
|
|
|
2023-09-19 06:20:26 +00:00
|
|
|
assert torch.all(inputs.grad == inputs_.grad), "Gradient of the input does not match"
|
2021-12-16 02:32:08 +00:00
|
|
|
torch.cuda.empty_cache()
|
2022-03-14 14:05:30 +00:00
|
|
|
|
2022-08-16 07:39:20 +00:00
|
|
|
# Extra test for use_reentrant=False
|
|
|
|
if use_reentrant == False:
|
|
|
|
# Recover cuda rng states
|
|
|
|
set_mode(ParallelMode.GLOBAL)
|
|
|
|
torch.cuda.set_rng_state(global_cuda_rng_state)
|
|
|
|
set_mode(ParallelMode.DATA)
|
|
|
|
torch.cuda.set_rng_state(data_parallel_cuda_rng_state)
|
|
|
|
set_mode(ParallelMode.GLOBAL)
|
|
|
|
|
|
|
|
out = forward_inplace(inputs, weight)
|
|
|
|
loss = out.sum()
|
|
|
|
loss.backward()
|
|
|
|
|
|
|
|
# Recover cuda rng states
|
|
|
|
set_mode(ParallelMode.GLOBAL)
|
|
|
|
torch.cuda.set_rng_state(global_cuda_rng_state)
|
|
|
|
set_mode(ParallelMode.DATA)
|
|
|
|
torch.cuda.set_rng_state(data_parallel_cuda_rng_state)
|
|
|
|
set_mode(ParallelMode.GLOBAL)
|
|
|
|
|
|
|
|
out = forward_inplace_ckpt(inputs_, weight_, cpu_offload=cpu_offload)
|
|
|
|
loss = out.sum()
|
|
|
|
loss.backward()
|
|
|
|
|
2023-09-19 06:20:26 +00:00
|
|
|
assert torch.all(inputs.grad == inputs_.grad), "Gradient of the input does not match"
|
2022-08-16 07:39:20 +00:00
|
|
|
torch.cuda.empty_cache()
|
|
|
|
|
2022-03-11 06:48:11 +00:00
|
|
|
# as seed manager is singleton
|
|
|
|
# if we don't reset seeds here,
|
|
|
|
# other tests will fail if running together with this test
|
|
|
|
# as other tests can't overwrite the seed set by this test
|
|
|
|
reset_seeds()
|
2022-08-16 07:39:20 +00:00
|
|
|
|
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
test_activation_checkpointing(False, False)
|