[FDConfig] disable chunked_mm_input in ernie5 (#5774)

* disable chunked_mm_input in ernie5

* update code

* update code

* update test case

* update testcase

* upate case
This commit is contained in:
kevin
2025-12-26 15:31:27 +08:00
committed by GitHub
parent 03363cab4c
commit 894f4e312b
17 changed files with 40 additions and 16 deletions
@@ -35,6 +35,7 @@ def make_prefix_cache_manager(max_num_seqs, enable_mm=False, num_gpu_blocks_over
model_cfg = SimpleNamespace(enable_mm=enable_mm, max_model_len=4196)
speculative_cfg = SimpleNamespace(method=None)
model_cfg.print = print
model_cfg.architectures = ["test_model"]
cache_cfg.bytes_per_layer_per_block = 1
parallel_cfg = ParallelConfig(args)
scheduler_cfg = SchedulerConfig(args)