refactor

2025-06-11 21:12:57 +08:00
parent b98e1ca305
commit 386290d69e
8 changed files with 31 additions and 35 deletions
--- a/nanovllm/engine/model_runner.py
+++ b/nanovllm/engine/model_runner.py
@@ -169,7 +169,7 @@ class ModelRunner:
        context_lens = torch.zeros(max_bs, dtype=torch.int32)
        block_tables = torch.zeros(max_bs, max_num_blocks, dtype=torch.int32)
        outputs = torch.zeros(max_bs, hf_config.hidden_size)
-        self.graph_bs = [1, 2, 4, 8, 16] + list(range(32, max_bs + 1, 32))
+        self.graph_bs = [1, 2, 4, 8, 16] + list(range(16, max_bs + 1, 16))
        self.graphs = {}
        self.graph_pool = None