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

Add experiments data location to run_experiment #46

Merged
merged 5 commits into from
Oct 20, 2023
Merged

Add experiments data location to run_experiment #46

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
86550bb
Use the jagged subclass the whole way through
jbschlosser Oct 12, 2023
bd7340d
Add experiments data location to run_experiment
cpuhrsch Oct 20, 2023
19010dd
Update nightly requirements
cpuhrsch Oct 20, 2023
8083852
Merge remote-tracking branch 'origin/main' into rexp1
cpuhrsch Oct 20, 2023
5a26a06
Comment out baseline runs
cpuhrsch Oct 20, 2023
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
85 changes: 46 additions & 39 deletions experiments/run_experiments.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
import subprocess
import os
import math
import fire
import itertools
import functools

home = "/home/cpuhrsch"

Expand Down Expand Up @@ -30,22 +30,8 @@ def change_sam_commit(commit_name):
assert result.returncode == 0


root_cmd = ["python", "eval_combo.py",
"--coco_root_dir",
"experiments_data/datasets/coco2017",
"--coco_slice_name",
"val2017",
"--sam_checkpoint_base_path",
"experiments_data/checkpoints",
"--sam_model_type",
"vit_b",
"--point_sampling_cache_dir",
"experiments_data/tmp/sam_coco_mask_center_cache",
"--mask_debug_out_dir",
"experiments_data/tmp/sam_eval_masks_out"]


def run_experiment(idx,
def run_experiment(experiments_data,
idx,
sam_commit_name,
model_type,
batch_size,
Expand All @@ -61,6 +47,19 @@ def run_experiment(idx,
profile_path=None,
profile_top=False,
memory_path=None):
root_cmd = ["python", "eval_combo.py",
"--coco_root_dir",
f"{experiments_data}/datasets/coco2017",
"--coco_slice_name",
"val2017",
"--sam_checkpoint_base_path",
f"{experiments_data}/checkpoints",
"--sam_model_type",
"vit_b",
"--point_sampling_cache_dir",
f"{experiments_data}/tmp/sam_coco_mask_center_cache",
"--mask_debug_out_dir",
f"{experiments_data}/tmp/sam_eval_masks_out"]
args = root_cmd
args = args + ["--sam_model_type", model_type]
args = args + ["--batch_size", str(batch_size)]
Expand Down Expand Up @@ -139,24 +138,32 @@ def run_traces(*args, **kwargs):
result = subprocess.run(conversion_cmd, capture_output=True)
assert result.returncode == 0

# run_traces("fp32", "default", "vit_b", 16, 32, print_header=True)
# run_traces("fp16", "codesign", "vit_b", 16, 32, use_half=True)
# run_traces("compile", "codesign", "vit_b", 16, 32, use_half=True, use_compile="max-autotune")
# run_traces("SDPA", "sdpa-decoder", "vit_b", 16, 32, use_half=True, use_compile="max-autotune")
# run_traces("Triton", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune")
# run_traces("NT", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune", use_nested_tensor=True)
# run_traces("int8", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune", use_nested_tensor=True, compress="dynamic_quant")
# run_traces("sparse", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune", use_nested_tensor=True, compress="sparse")

print_header = True
for bs, model in itertools.product([1, 32], ["vit_b", "vit_h"]):
# run_experiment("fp32", "default", model, bs, 32, print_header=print_header)
# print_header = False
# run_experiment("bf16", "codesign", model, bs, 32, use_half="bfloat16")
# run_experiment("compile", "codesign", model, bs, 32, use_half="bfloat16", use_compile="max-autotune")
# run_experiment("SDPA", "sdpa-decoder", model, bs, 32, use_half="bfloat16", use_compile="max-autotune")
run_experiment("Triton", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune")
if bs > 1:
run_experiment("NT", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune", use_nested_tensor=(bs > 1))
run_experiment("int8", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune", use_nested_tensor=(bs > 1), compress="dynamic_quant")
run_experiment("sparse", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune", use_nested_tensor=(bs > 1), compress="sparse")
def run(experiments_data=None):
if experiments_data is None:
experiments_data = "experiments_data"

# run_traces("fp32", "default", "vit_b", 16, 32, print_header=True)
# run_traces("fp16", "codesign", "vit_b", 16, 32, use_half=True)
# run_traces("compile", "codesign", "vit_b", 16, 32, use_half=True, use_compile="max-autotune")
# run_traces("SDPA", "sdpa-decoder", "vit_b", 16, 32, use_half=True, use_compile="max-autotune")
# run_traces("Triton", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune")
# run_traces("NT", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune", use_nested_tensor=True)
# run_traces("int8", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune", use_nested_tensor=True, compress="dynamic_quant")
# run_traces("sparse", "local-fork", "vit_b", 16, 32, use_half=True, use_compile="max-autotune", use_nested_tensor=True, compress="sparse")

rexp = functools.partial(run_experiment, experiments_data)
print_header = True
for bs, model in itertools.product([1, 32], ["vit_b", "vit_h"]):
# rexp("fp32", "default", model, bs, 32, print_header=print_header)
print_header = False
# rexp("bf16", "codesign", model, bs, 32, use_half="bfloat16")
# rexp("compile", "codesign", model, bs, 32, use_half="bfloat16", use_compile="max-autotune")
# rexp("SDPA", "sdpa-decoder", model, bs, 32, use_half="bfloat16", use_compile="max-autotune")
rexp("Triton", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune", capture_output=False)
if bs > 1:
rexp("NT", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune", use_nested_tensor=(bs > 1))
rexp("int8", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune", use_nested_tensor=(bs > 1), compress="dynamic_quant")
rexp("sparse", "local-fork", model, bs, 32, use_half="bfloat16", use_compile="max-autotune", use_nested_tensor=(bs > 1), compress="sparse")

if __name__ == '__main__':
fire.Fire(run)
4 changes: 2 additions & 2 deletions setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,8 @@
version='0.2',
packages=packages,
install_requires=[
'torch>=2.2.0.dev20231015',
'torchvision>=0.17.0.dev20231015',
'torch>=2.2.0.dev20231019',
'torchvision>=0.17.0.dev20231019',
'diskcache',
'pycocotools',
'scipy',
Expand Down
Loading