Skip to content

[fp8] fix cb scheduler step tests #491

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

Merged
merged 5 commits into from
Oct 1, 2025
Merged

[fp8] fix cb scheduler step tests #491

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
46552d0
min compiled batch size of 4 for cb tests
yannicks1 Sep 29, 2025
343a99a
compiling for the next power of two for batch size
yannicks1 Sep 29, 2025
bf2638a
fmt
yannicks1 Sep 29, 2025
6e3598a
fix fmt
yannicks1 Sep 29, 2025
e737961
add some comments
yannicks1 Sep 30, 2025
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
9 changes: 8 additions & 1 deletion tests/llm_cache.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,11 +187,15 @@ def get_engine(
# then adjust these limits in the engine's scheduler for tests.

# Setup the engine
# Round max_num_seqs (batch size) to the next power of two for
# Spyre compilation. This seems more robust and helps that all tests in
# tests/e2e/test_spyre_cb_inference_steps.py pass on Spyre.
max_num_seqs_compiled = 1 << (max_num_seqs - 1).bit_length()
engine_args = EngineArgs(
model=model_name,
tokenizer=model_name,
max_model_len=max(max_model_len, 256),
max_num_seqs=max_num_seqs,
max_num_seqs=max_num_seqs_compiled,
num_gpu_blocks_override=None,
revision=revision,
)
Expand All @@ -202,7 +206,10 @@ def get_engine(
executor_class=executor_class,
log_stats=False)

# Set scheduler configs for max_model_len and max_num_seqs to the
# original values. They were changed for more robust compilation only.
engine_core.scheduler.scheduler_config.max_model_len = max_model_len
engine_core.scheduler.scheduler_config.max_num_seqs = max_num_seqs

if available_blocks is not None:
worker = engine_core.model_executor.driver_worker.worker
Expand Down