[CI] Disable non-lazy string operation on logging (#4326)
Co-authored-by: Danny Guinther <dguinther@neuralmagic.com>
This commit is contained in:
co-authored by
Danny Guinther
parent
2f30e7c72f
commit
a88081bf76
+14
-13
@@ -170,8 +170,8 @@ class ModelRunner:
|
||||
)
|
||||
|
||||
self.model_memory_usage = m.consumed_memory
|
||||
logger.info(f"Loading model weights took "
|
||||
f"{self.model_memory_usage / float(2**30):.4f} GB")
|
||||
logger.info("Loading model weights took %.4f GB",
|
||||
self.model_memory_usage / float(2**30))
|
||||
|
||||
if self.lora_config:
|
||||
assert hasattr(self.model, "supported_lora_modules"
|
||||
@@ -196,18 +196,19 @@ class ModelRunner:
|
||||
self.model.load_kv_cache_scales(
|
||||
self.model_config.quantization_param_path)
|
||||
else:
|
||||
raise RuntimeError("Using FP8 KV cache and scaling "
|
||||
"factors provided but model "
|
||||
f"{self.model.__class__} does not "
|
||||
"support loading scaling factors.")
|
||||
raise RuntimeError(
|
||||
"Using FP8 KV cache and scaling factors provided but "
|
||||
"model %s does not support loading scaling factors.",
|
||||
self.model.__class__)
|
||||
else:
|
||||
logger.warn("Using FP8 KV cache but no scaling factors "
|
||||
"provided. Defaulting to scaling factors of 1.0. "
|
||||
"This may lead to less accurate results!")
|
||||
logger.warning(
|
||||
"Using FP8 KV cache but no scaling factors "
|
||||
"provided. Defaulting to scaling factors of 1.0. "
|
||||
"This may lead to less accurate results!")
|
||||
elif self.model_config.quantization_param_path is not None:
|
||||
logger.warn("KV cache scaling factors provided, "
|
||||
"but the KV cache data type is not FP8. "
|
||||
"KV cache scaling factors will not be used.")
|
||||
logger.warning("KV cache scaling factors provided, "
|
||||
"but the KV cache data type is not FP8. "
|
||||
"KV cache scaling factors will not be used.")
|
||||
|
||||
def set_block_size(self, block_size: int) -> None:
|
||||
self.block_size = block_size
|
||||
@@ -1054,7 +1055,7 @@ class ModelRunner:
|
||||
end_time = time.perf_counter()
|
||||
elapsed_time = end_time - start_time
|
||||
# This usually takes < 10 seconds.
|
||||
logger.info(f"Graph capturing finished in {elapsed_time:.0f} secs.")
|
||||
logger.info("Graph capturing finished in %.0f secs.", elapsed_time)
|
||||
|
||||
def __del__(self) -> None:
|
||||
# Delete the CUDA graphs before deleting the pynccl communicator.
|
||||
|
||||
Reference in New Issue
Block a user