Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Security] fix security problem for prune_by_memory_estimation #61320

Merged
merged 2 commits into from
Jan 30, 2024
Merged

[Security] fix security problem for prune_by_memory_estimation #61320

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c434000
OS Command Injection prune_by_memory_estimation fix
vn-ncvinh Jan 24, 2024
e70602b
Fix CodeStyle
vn-ncvinh Jan 30, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Next Next commit
OS Command Injection prune_by_memory_estimation fix
  • Loading branch information
@vn-ncvinh
vn-ncvinh authored Jan 24, 2024
commit c4340008880416bb86cff5a35fb10b88e4ac9b99
62 changes: 32 additions & 30 deletions python/paddle/distributed/auto_tuner/prune.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -528,47 +528,49 @@ def prune_by_memory_estimation(tuner_cfg, cur_cfg, history_cfgs=[]):
micro_batch_size = cur_cfg['micro_batch_size']
recompute_granularity = cur_cfg['recompute_granularity']

memory_estimation_cmd = f"python {memory_estimation_tool} --dp_degree {dp_degree} --mp_degree {mp_degree} \
--pp_degree {pp_degree} --vpp_degree {vpp_degree} \
--sharding_degree {sharding_degree} --sharding_stage {sharding_stage} \
--use_recompute {use_recompute} --micro_batch_size {micro_batch_size} \
--recompute_granularity {recompute_granularity}"
memory_estimation_cmd = [
"python",
memory_estimation_tool,
"--dp_degree", str(dp_degree),
"--mp_degree", str(mp_degree),
"--pp_degree", str(pp_degree),
"--vpp_degree", str(vpp_degree),
"--sharding_degree", str(sharding_degree),
"--sharding_stage", str(sharding_stage),
"--use_recompute", str(use_recompute),
"--micro_batch_size", str(micro_batch_size),
"--recompute_granularity", str(recompute_granularity),
]


# get model config
hidden_size = model_cfg.get('hidden_size', None)
memory_estimation_cmd += (
f" --hidden_size {hidden_size}" if hidden_size is not None else ""
)
if hidden_size is not None:
memory_estimation_cmd.extend(["--hidden_size", str(hidden_size)])

num_attention_heads = model_cfg.get('num_attention_heads', None)
memory_estimation_cmd += (
f" --num_attention_heads {num_attention_heads}"
if num_attention_heads is not None
else ""
)
if num_attention_heads is not None:
memory_estimation_cmd.extend(["--num_attention_heads", str(num_attention_heads)])

num_layers = model_cfg.get('num_layers', None)
memory_estimation_cmd += (
f" --num_layers {num_layers}" if num_layers is not None else ""
)
if num_layers is not None:
memory_estimation_cmd.extend(["--num_layers", str(num_layers)])

max_sequence_length = model_cfg.get('max_sequence_length', None)
memory_estimation_cmd += (
f" --max_sequence_length {max_sequence_length}"
if max_sequence_length is not None
else ""
)
if max_sequence_length is not None:
memory_estimation_cmd.extend(["--max_sequence_length", str(max_sequence_length)])

vocab_size = model_cfg.get('vocab_size', None)
memory_estimation_cmd += (
f" --vocab_size {vocab_size}" if vocab_size is not None else ""
)
if vocab_size is not None:
memory_estimation_cmd.extend(["--vocab_size", str(vocab_size)])

intermediate_size = model_cfg.get('intermediate_size', None)
memory_estimation_cmd += (
f" --intermediate_size {intermediate_size}"
if intermediate_size is not None
else ""
)
if intermediate_size is not None:
memory_estimation_cmd.extend(["--intermediate_size", str(intermediate_size)])


result = subprocess.run(
memory_estimation_cmd,
shell=True,
capture_output=True,
text=True,
)
Expand Down