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

Jump to bottom

Improve log messages around the max sequence length #103

Merged
merged 4 commits into from
Jun 28, 2024
Merged

Improve log messages around the max sequence length #103

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
a98f04c
Improve log messages around the max sequence length
maxdebayser Jun 20, 2024
01c6ec6
Simplify error message
maxdebayser Jun 27, 2024
cdbcfa7
Remove second print of max_batch_weight
maxdebayser Jun 27, 2024
48ed1c9
Merge branch 'main' into improve_seq_len_messages
maxdebayser Jun 28, 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
8 changes: 4 additions & 4 deletions router/src/server.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -222,8 +222,8 @@ impl<'a, B: BatchType> BatchConfigValidator<'a, B> {
self.batch_type.prefill_weight(&single_request_stats, 1);
if max_batch_weight < single_request_prefill_weight {
panic!(
"max_batch_weight ({}) not large enough for (prefill) max_sequence_length ({})",
max_batch_weight, max_sequence_length
"The provided max_sequence length ({}) results in a prefill batch weight that exceeds the estimated capacity ({})",
max_sequence_length, max_batch_weight
)
}

Expand All @@ -232,8 +232,8 @@ impl<'a, B: BatchType> BatchConfigValidator<'a, B> {
.batch_initial_weight(&single_request_stats, 1);
if max_batch_weight < single_request_nexttoken_weight {
panic!(
"max_batch_weight ({}) not large enough for (next-token) max_sequence_length ({})",
max_batch_weight, max_sequence_length
"The provided max_sequence length ({}) results in a next-token batch weight that exceeds the estimated capacity ({})",
max_sequence_length, max_batch_weight
)
}
}
Expand Down
10 changes: 6 additions & 4 deletions server/text_generation_server/server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -409,13 +409,15 @@ def estimate_memory():
memory_scaling_model = estimate_memory()
compile()

max_input = memory_scaling_model.max_input_len_for_nt(1, max_sequence_length-1, sys.maxsize)
max_output = memory_scaling_model.max_output_len_for_nt(1, max_sequence_length-1, sys.maxsize)

if local_rank == 0:
# For a batch of size 1 and an output of 1, get max input limited by max_sequence_length
max_input = memory_scaling_model.max_input_len_for_nt(1, 1, max_sequence_length)
# For a batch of size 1 and an input of 1, get max output limited by max_sequence_length
max_output = memory_scaling_model.max_output_len_for_nt(1, 1, max_sequence_length)
max_theoretical_len = min(max_input, max_output) + 1
print(
"Maximum possible sequence length given available memory (for batch size 1): "
f"{min(max_input, max_output)}"
f"{max_theoretical_len}"
)

elif ESTIMATE_MEMORY == "manual":
Expand Down