From 4ce82a22851d36ed96ba8ac5006040bea4cd3129 Mon Sep 17 00:00:00 2001
From: Cameron Quilici <cjquilici@gmail.com>
Date: Fri, 18 Sep 2026 15:55:33 -0500
Subject: [PATCH] Delete runners/test_dsv41flash_b300.py

---
 runners/test_dsv41flash_b300.py | 82 ---------------------------------
 1 file changed, 82 deletions(-)
 delete mode 100644 runners/test_dsv41flash_b300.py

diff --git a/runners/test_dsv41flash_b300.py b/runners/test_dsv41flash_b300.py
deleted file mode 100644
index c9dcc3b071..0000000000
--- a/runners/test_dsv41flash_b300.py
+++ /dev/null
@@ -1,82 +0,0 @@
-import json
-import os
-import subprocess
-from pathlib import Path
-
-import pytest
-
-ROOT = Path(__file__).resolve().parents[1]
-
-
-@pytest.mark.parametrize(
-    "tp,concurrency,expected_capture,expected_batched_tokens,expected_memory_utilization",
-    [
-        (4, 1, 2046, 2048, None),
-        (4, 8, 8190, 8192, None),
-        (2, 128, 2046, 2048, "0.97"),
-    ],
-)
-def test_b300_capture_tiers_build_expected_vllm_command(
-    tmp_path: Path,
-    tp: int,
-    concurrency: int,
-    expected_capture: int,
-    expected_batched_tokens: int,
-    expected_memory_utilization: str | None,
-) -> None:
-    env = {
-        **os.environ,
-        "MODEL": "fixture",
-        "TP": str(tp),
-        "CONC": str(concurrency),
-        "KV_OFFLOADING": "none",
-        "TOTAL_CPU_DRAM_GB": "0",
-        "DURATION": "3600",
-        "RESULT_DIR": str(tmp_path),
-        "EVAL_ONLY": "false",
-        "PORT": "18888",
-    }
-    result = subprocess.run(
-        [
-            "bash",
-            "-c",
-            r'''
-source() { :; }
-check_env_vars() { :; }
-require_agentic_kv_offload_none() { :; }
-hf() { :; }
-nvidia-smi() { :; }
-resolve_trace_source() { :; }
-install_agentic_deps() { :; }
-select_available_server_port() { :; }
-wait_for_server_ready() { wait "$SERVER_PID"; }
-build_replay_cmd() { :; }
-run_agentic_replay_and_write_outputs() { :; }
-run_eval() { :; }
-vllm() { command python3 -c 'import json,os,sys; json.dump(sys.argv[1:],open(os.environ["RESULT_DIR"]+"/args.json","w"))' "$@"; }
-builtin source "$1/benchmarks/single_node/agentic/dsv41flash_fp4_b300_vllm_mtp.sh"
-''',
-            "bash",
-            str(ROOT),
-        ],
-        env=env,
-        capture_output=True,
-        check=False,
-        text=True,
-        timeout=15,
-    )
-    assert result.returncode == 0, result.stderr
-
-    args = json.loads((tmp_path / "args.json").read_text())
-    assert int(args[args.index("--max-cudagraph-capture-size") + 1]) == expected_capture
-    assert int(args[args.index("--max-num-batched-tokens") + 1]) == expected_batched_tokens
-    assert args[args.index("--max-num-seqs") + 1] == "256"
-    compilation = json.loads(args[args.index("--compilation-config") + 1])
-    assert compilation["mode"] == "VLLM_COMPILE"
-    assert compilation["cudagraph_mode"] == "FULL_AND_PIECEWISE"
-    assert compilation["cudagraph_capture_sizes"][-1] == expected_capture
-
-    if expected_memory_utilization is None:
-        assert "--gpu-memory-utilization" not in args
-    else:
-        assert args[args.index("--gpu-memory-utilization") + 1] == expected_memory_utilization