Unverified Commit fd5a55cf authored by Yubo Wang's avatar Yubo Wang Committed by GitHub
Browse files

Use public model for FA3 speculative decode testing (#5152)

parent 804d9f2e
......@@ -7,8 +7,6 @@ import torch
from sglang.srt.utils import get_device_sm, kill_process_tree
from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
from sglang.test.test_utils import (
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_MLA_MODEL_NAME_FOR_TEST,
DEFAULT_MODEL_NAME_FOR_TEST,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
......@@ -125,7 +123,7 @@ class TestFlashAttention3MLA(BaseFlashAttentionTest):
class TestFlashAttention3SpeculativeDecode(BaseFlashAttentionTest):
"""Test FlashAttention3 with speculative decode enabled."""
model = DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST
model = "meta-llama/Llama-3.1-8B-Instruct"
@classmethod
def get_server_args(cls):
......@@ -137,7 +135,7 @@ class TestFlashAttention3SpeculativeDecode(BaseFlashAttentionTest):
"--speculative-algorithm",
"EAGLE3",
"--speculative-draft",
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
"jamesliu1/sglang-EAGLE3-Llama-3.1-Instruct-8B",
"--speculative-num-steps",
"3",
"--speculative-eagle-topk",
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment