Fix benchmark generating empty prompts when random_input_len is small (#21492)

2026-03-26 16:24:35 -07:00
parent a93065679b
commit c580ddd19d
1 changed files with 1 additions and 1 deletions
@@ -79,7 +79,7 @@ def sample_random_requests(
        # Need to truncate input_len as server encode will add special token.
        num_special_tokens = int(tokenizer.num_special_tokens_to_add())
        for i in range(num_prompts):
-            input_lens[i] = max(0, input_lens[i] - num_special_tokens)
+            input_lens[i] = max(1, input_lens[i] - num_special_tokens)

    if random_sample:
        # Sample token ids from ShareGPT and repeat/truncate them to satisfy the input_lens