Skip to content

training_server: Fix UnboundLocalError in TTFT model training #1843

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

Open
wants to merge 1 commit into
base: slo-prediction-experimental
Choose a base branch
from
+15 −17
Open

training_server: Fix UnboundLocalError in TTFT model training #1843

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
32 changes: 15 additions & 17 deletions latencypredictor-v1/training_server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -628,22 +628,22 @@ def train(self):
if len(df_ttft) >= settings.MIN_SAMPLES_FOR_RETRAIN:
# Updated TTFT features to include prefix_cache_score
ttft_feature_cols_tree = [
'kv_cache_percentage','input_token_length','num_request_waiting',
'num_request_running','prefix_cache_score','effective_input_tokens','prefill_score_bucket'
]
ttft_feature_cols_br = [
'kv_cache_percentage','input_token_length','num_request_waiting',
'num_request_running','prefix_cache_score','effective_input_tokens'
]

# Build X_ttft for all model types, then trim for BR
X_ttft = df_ttft[ttft_feature_cols_tree]
if self.model_type == ModelType.BAYESIAN_RIDGE:
X_ttft = X_ttft[ttft_feature_cols_br]
'kv_cache_percentage','input_token_length','num_request_waiting',
'num_request_running','prefix_cache_score','effective_input_tokens','prefill_score_bucket'
]
ttft_feature_cols_br = [
'kv_cache_percentage','input_token_length','num_request_waiting',
'num_request_running','prefix_cache_score','effective_input_tokens'
]

# Build X_ttft for all model types, then trim for BR
X_ttft = df_ttft[ttft_feature_cols_tree]
if self.model_type == ModelType.BAYESIAN_RIDGE:
X_ttft = X_ttft[ttft_feature_cols_br]

y_ttft = raw_ttft['actual_ttft_ms']
y_ttft = raw_ttft['actual_ttft_ms']

try:
try:
# raw_ttft still has the original columns including 'prefix_cache_score'
raw_ttft['_prefix_bucket'] = raw_ttft['prefix_cache_score'].clip(0, 1).apply(
lambda s: min(int(s * self.prefix_buckets), self.prefix_buckets - 1)
Expand Down Expand Up @@ -677,8 +677,6 @@ def train(self):
new_ttft_model, new_ttft_scaler, test_records, cols, 'actual_ttft_ms'
)



if ql is not None:
self.ttft_quantile_loss_scores.append(ql)
self.ttft_coverage_scores.append(coverage)
Expand All @@ -690,7 +688,7 @@ def train(self):
else:
logging.info(f"TTFT model trained on {len(df_ttft)} samples. Quantile metrics = N/A (insufficient test data)")

except Exception:
except Exception:
logging.error("Error training TTFT model", exc_info=True)


Expand Down