Update grok 1 model (#1095)

2024-08-14 04:40:44 -07:00
parent fe5024325b
commit a59636bb5e
11 changed files with 814 additions and 514 deletions
--- a/python/sglang/bench_latency.py
+++ b/python/sglang/bench_latency.py
@@ -221,6 +221,7 @@ def correctness_test(

    # Prepare inputs
    input_ids, reqs = prepare_inputs_for_correctness_test(bench_args, tokenizer)
+    rank_print(f"{input_ids=}")

    if bench_args.cut_len > 0:
        # Prefill