mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-09-27 04:46:16 +08:00
2e9e53ff7e
* remove max_num_batched_tokens in parallel config * remove max_num_seqs * update test case * fix test * fix --------- Co-authored-by: Jiang-Jia-Jun <163579578+Jiang-Jia-Jun@users.noreply.github.com>
95 lines
3.2 KiB
Python
95 lines
3.2 KiB
Python
import unittest
|
|
|
|
from fastdeploy import envs
|
|
from fastdeploy.config import (
|
|
CacheConfig,
|
|
FDConfig,
|
|
GraphOptimizationConfig,
|
|
ParallelConfig,
|
|
SchedulerConfig,
|
|
)
|
|
|
|
|
|
class TestConfig(unittest.TestCase):
|
|
def test_fdconfig_nnode(self):
|
|
parallel_config = ParallelConfig({"tensor_parallel_size": 16, "expert_parallel_size": 1})
|
|
graph_opt_config = GraphOptimizationConfig({})
|
|
cache_config = CacheConfig({})
|
|
scheduler_config = SchedulerConfig({})
|
|
fd_config = FDConfig(
|
|
parallel_config=parallel_config,
|
|
graph_opt_config=graph_opt_config,
|
|
cache_config=cache_config,
|
|
scheduler_config=scheduler_config,
|
|
ips=["1.1.1.1", "0.0.0.0"],
|
|
test_mode=True,
|
|
)
|
|
assert fd_config.nnode == 2
|
|
assert fd_config.is_master is False
|
|
|
|
def test_fdconfig_ips(self):
|
|
parallel_config = ParallelConfig({})
|
|
graph_opt_config = GraphOptimizationConfig({})
|
|
cache_config = CacheConfig({})
|
|
scheduler_config = SchedulerConfig({})
|
|
fd_config = FDConfig(
|
|
parallel_config=parallel_config,
|
|
graph_opt_config=graph_opt_config,
|
|
cache_config=cache_config,
|
|
scheduler_config=scheduler_config,
|
|
ips="0.0.0.0",
|
|
test_mode=True,
|
|
)
|
|
assert fd_config.master_ip == "0.0.0.0"
|
|
|
|
def test_fdconfig_max_num_tokens(self):
|
|
parallel_config = ParallelConfig({})
|
|
graph_opt_config = GraphOptimizationConfig({})
|
|
cache_config = CacheConfig({})
|
|
cache_config.enable_chunked_prefill = True
|
|
scheduler_config = SchedulerConfig({})
|
|
fd_config = FDConfig(
|
|
parallel_config=parallel_config,
|
|
graph_opt_config=graph_opt_config,
|
|
cache_config=cache_config,
|
|
scheduler_config=scheduler_config,
|
|
ips="0.0.0.0",
|
|
test_mode=True,
|
|
)
|
|
if not envs.ENABLE_V1_KVCACHE_SCHEDULER:
|
|
assert fd_config.scheduler_config.max_num_batched_tokens == 2048
|
|
|
|
cache_config.enable_chunked_prefill = False
|
|
fd_config = FDConfig(
|
|
parallel_config=parallel_config,
|
|
graph_opt_config=graph_opt_config,
|
|
cache_config=cache_config,
|
|
scheduler_config=scheduler_config,
|
|
ips="0.0.0.0",
|
|
test_mode=True,
|
|
)
|
|
if not envs.ENABLE_V1_KVCACHE_SCHEDULER:
|
|
assert fd_config.scheduler_config.max_num_batched_tokens == 8192
|
|
|
|
def test_fdconfig_init_cache(self):
|
|
parallel_config = ParallelConfig({})
|
|
graph_opt_config = GraphOptimizationConfig({})
|
|
cache_config = CacheConfig({})
|
|
cache_config.cache_transfer_protocol = "rdma,ipc"
|
|
cache_config.pd_comm_port = "2334"
|
|
scheduler_config = SchedulerConfig({})
|
|
fd_config = FDConfig(
|
|
parallel_config=parallel_config,
|
|
graph_opt_config=graph_opt_config,
|
|
cache_config=cache_config,
|
|
scheduler_config=scheduler_config,
|
|
splitwise_role="prefill",
|
|
test_mode=True,
|
|
)
|
|
fd_config.init_cache_info()
|
|
assert fd_config.disaggregate_info["role"] == "prefill"
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|