Sync changes on io_struct.py and deterministic ops (#11498)
This commit is contained in:
@@ -455,7 +455,6 @@ class ServerArgs:
|
||||
debug_tensor_dump_output_folder: Optional[str] = None
|
||||
debug_tensor_dump_input_file: Optional[str] = None
|
||||
debug_tensor_dump_inject: bool = False
|
||||
debug_tensor_dump_prefill_only: bool = False
|
||||
|
||||
# PD disaggregation: can be "null" (not disaggregated), "prefill" (prefill-only), or "decode" (decode-only)
|
||||
disaggregation_mode: Literal["null", "prefill", "decode"] = "null"
|
||||
@@ -2831,11 +2830,6 @@ class ServerArgs:
|
||||
default=ServerArgs.debug_tensor_dump_inject,
|
||||
help="Inject the outputs from jax as the input of every layer.",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--debug-tensor-dump-prefill-only",
|
||||
action="store_true",
|
||||
help="Only dump the tensors for prefill requests (i.e. batch size > 1).",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--enable-dynamic-batch-tokenizer",
|
||||
action="store_true",
|
||||
|
||||
Reference in New Issue
Block a user