[V0 Deprecation] Remove V0 executors (#27142)
Signed-off-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
@@ -128,28 +128,6 @@ class WorkerBase:
|
||||
def execute_model(self, scheduler_output: SchedulerOutput) -> ModelRunnerOutput:
|
||||
raise NotImplementedError
|
||||
|
||||
def start_worker_execution_loop(self) -> None:
|
||||
"""Execute model loop in parallel worker.
|
||||
|
||||
You can stop the loop by executing a driver worker with an empty output.
|
||||
See `stop_remote_worker_execution_loop` for more details.
|
||||
"""
|
||||
raise NotImplementedError("Dead V0 code")
|
||||
|
||||
def determine_num_available_blocks(self) -> tuple[int, int]:
|
||||
"""Determine the number of available blocks for the GPU KV cache and
|
||||
swappable CPU KV cache.
|
||||
|
||||
The implementation may run profiling or other heuristics to determine
|
||||
the size of caches.
|
||||
|
||||
Returns a tuple[num_gpu_blocks, num_cpu_blocks], where num_gpu_blocks
|
||||
are blocks that are "active" on the device and can be appended to.
|
||||
num_cpu_blocks refers to "swapped" blocks in CPU memory and cannot be
|
||||
appended to.
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def get_cache_block_size_bytes(self) -> int:
|
||||
"""Return the size of a single cache block, in bytes. Used in
|
||||
speculative decoding.
|
||||
|
||||
Reference in New Issue
Block a user