apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[CI] Change 21b ep4 to tp1_dp4 in 4_cards_tests (#6745)

Browse Source 
* [CI] Change 21b ep4 to tp1_dp4 in 4_cards_tests
This commit is contained in:
YuBaoku
2026-03-20 20:42:23 +08:00
committed by GitHub
parent 030820db4c
commit 0b4c1cba9b
2 changed files with 71 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/e2e/4cards_cases/test_ernie_21b_ep4.py → tests/e2e/4cards_cases/test_ernie_21b_tp1_dp4.py
+37 -14
View File
@@ -1,4 +1,4 @@
# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved. # Copyright (c) 2026 PaddlePaddle Authors. All Rights Reserved.
# #
# Licensed under the Apache License, Version 2.0 (the "License"); # Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License. # you may not use this file except in compliance with the License.
@@ -50,9 +50,17 @@ def setup_and_run_server(api_url):
ports_to_add = [ ports_to_add = [
FD_API_PORT + 1, FD_API_PORT + 1,
FD_API_PORT + 2,
FD_API_PORT + 3,
FD_METRICS_PORT + 1, FD_METRICS_PORT + 1,
FD_METRICS_PORT + 2,
FD_METRICS_PORT + 3,
FD_CACHE_QUEUE_PORT + 1, FD_CACHE_QUEUE_PORT + 1,
FD_CACHE_QUEUE_PORT + 2,
FD_CACHE_QUEUE_PORT + 3,
FD_ENGINE_QUEUE_PORT + 1, FD_ENGINE_QUEUE_PORT + 1,
FD_ENGINE_QUEUE_PORT + 2,
FD_ENGINE_QUEUE_PORT + 3,
] ]
for port in ports_to_add: for port in ports_to_add:
@@ -77,22 +85,22 @@ def setup_and_run_server(api_url):
"-m", "-m",
"fastdeploy.entrypoints.openai.multi_api_server", "fastdeploy.entrypoints.openai.multi_api_server",
"--num-servers", "--num-servers",
"2", "4",
"--ports", "--ports",
f"{FD_API_PORT},{FD_API_PORT + 1}", f"{FD_API_PORT},{FD_API_PORT + 1},{FD_API_PORT + 2},{FD_API_PORT + 3}",
"--metrics-ports", "--metrics-ports",
f"{FD_METRICS_PORT},{FD_METRICS_PORT + 1}", f"{FD_METRICS_PORT},{FD_METRICS_PORT + 1},{FD_METRICS_PORT + 2},{FD_METRICS_PORT + 3}",
"--args", "--args",
"--model", "--model",
model_path, model_path,
"--engine-worker-queue-port", "--engine-worker-queue-port",
f"{FD_ENGINE_QUEUE_PORT},{FD_ENGINE_QUEUE_PORT + 1}", f"{FD_ENGINE_QUEUE_PORT},{FD_ENGINE_QUEUE_PORT + 1},{FD_ENGINE_QUEUE_PORT + 2},{FD_ENGINE_QUEUE_PORT + 3}",
"--cache-queue-port", "--cache-queue-port",
f"{FD_CACHE_QUEUE_PORT},{FD_CACHE_QUEUE_PORT + 1}", f"{FD_CACHE_QUEUE_PORT},{FD_CACHE_QUEUE_PORT + 1},{FD_CACHE_QUEUE_PORT + 2},{FD_CACHE_QUEUE_PORT + 3}",
"--tensor-parallel-size", "--tensor-parallel-size",
"2", "1",
"--data-parallel-size", "--data-parallel-size",
"2", "4",
"--max-model-len", "--max-model-len",
"65536", "65536",
"--max-num-seqs", "--max-num-seqs",
@@ -328,12 +336,17 @@ def test_text_diff(api_url):
base_path = os.getenv("MODEL_PATH") base_path = os.getenv("MODEL_PATH")
if base_path: if base_path:
base_file = os.path.join(base_path, "21b_ep4_text_baseline_dev_0311.txt") base_file = os.path.join(base_path, "21b_tp1_dp4_text_baseline.txt")
else: else:
base_file = "21b_ep4_text_baseline_dev_0311.txt" base_file = "21b_tp1_dp4_text_baseline.txt"
with open(base_file, "r", encoding="utf-8") as f: with open(base_file, "r", encoding="utf-8") as f:
baseline = f.read() baseline = f.read()
if result != baseline:
with open("21b_tp1_dp4_text_tmp.txt", "w", encoding="utf-8") as f:
f.write(result)
assert result == baseline, f"Text mismatch with baseline\nresult: {result}\nbaseline: {baseline}" assert result == baseline, f"Text mismatch with baseline\nresult: {result}\nbaseline: {baseline}"
@@ -596,12 +609,17 @@ def test_non_stream_with_logprobs(api_url):
base_path = os.getenv("MODEL_PATH") base_path = os.getenv("MODEL_PATH")
if base_path: if base_path:
base_file = os.path.join(base_path, "21b_ep4_logprobs_non_stream_static_baseline_dev_0311.txt") base_file = os.path.join(base_path, "21b_tp1_dp4_logprobs_non_stream_static_baseline.txt")
else: else:
base_file = "21b_ep4_logprobs_non_stream_static_baseline_dev_0311.txt" base_file = "21b_tp1_dp4_logprobs_non_stream_static_baseline.txt"
with open(base_file, "r", encoding="utf-8") as f: with open(base_file, "r", encoding="utf-8") as f:
baseline = json.load(f) baseline = json.load(f)
if logprobs != baseline:
with open("21b_tp1_dp4_logprobs_non_stream_static_tmp.txt", "w", encoding="utf-8") as f:
f.write(json.dumps(logprobs, ensure_ascii=False, indent=2))
assert logprobs == baseline assert logprobs == baseline
@@ -629,10 +647,15 @@ def test_stream_with_logprobs(api_url):
base_path = os.getenv("MODEL_PATH") base_path = os.getenv("MODEL_PATH")
if base_path: if base_path:
base_file = os.path.join(base_path, "21b_ep4_logprobs_stream_static_baseline_dev_0311.txt") base_file = os.path.join(base_path, "21b_tp1_dp4_logprobs_stream_static_baseline.txt")
else: else:
base_file = "21b_ep4_logprobs_stream_static_baseline_dev_0311.txt" base_file = "21b_tp1_dp4_logprobs_stream_static_baseline.txt"
with open(base_file, "r", encoding="utf-8") as f: with open(base_file, "r", encoding="utf-8") as f:
baseline = json.load(f) baseline = json.load(f)
if logprobs != baseline:
with open("21b_tp1_dp4_logprobs_stream_static_tmp.txt", "w", encoding="utf-8") as f:
f.write(json.dumps(logprobs, ensure_ascii=False, indent=2))
assert logprobs == baseline assert logprobs == baseline
tests/e2e/4cards_cases/test_ernie_21b_ep4_mtp.py → tests/e2e/4cards_cases/test_ernie_21b_tp1_dp4_mtp.py
+34 -14
View File
@@ -1,4 +1,4 @@
# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved. # Copyright (c) 2026 PaddlePaddle Authors. All Rights Reserved.
# #
# Licensed under the Apache License, Version 2.0 (the "License"); # Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License. # you may not use this file except in compliance with the License.
@@ -50,9 +50,17 @@ def setup_and_run_server(api_url):
ports_to_add = [ ports_to_add = [
FD_API_PORT + 1, FD_API_PORT + 1,
FD_API_PORT + 2,
FD_API_PORT + 3,
FD_METRICS_PORT + 1, FD_METRICS_PORT + 1,
FD_METRICS_PORT + 2,
FD_METRICS_PORT + 3,
FD_CACHE_QUEUE_PORT + 1, FD_CACHE_QUEUE_PORT + 1,
FD_CACHE_QUEUE_PORT + 2,
FD_CACHE_QUEUE_PORT + 3,
FD_ENGINE_QUEUE_PORT + 1, FD_ENGINE_QUEUE_PORT + 1,
FD_ENGINE_QUEUE_PORT + 2,
FD_ENGINE_QUEUE_PORT + 3,
] ]
for port in ports_to_add: for port in ports_to_add:
@@ -79,22 +87,22 @@ def setup_and_run_server(api_url):
"-m", "-m",
"fastdeploy.entrypoints.openai.multi_api_server", "fastdeploy.entrypoints.openai.multi_api_server",
"--num-servers", "--num-servers",
"2", "4",
"--ports", "--ports",
f"{FD_API_PORT},{FD_API_PORT + 1}", f"{FD_API_PORT},{FD_API_PORT + 1},{FD_API_PORT + 2},{FD_API_PORT + 3}",
"--metrics-ports", "--metrics-ports",
f"{FD_METRICS_PORT},{FD_METRICS_PORT + 1}", f"{FD_METRICS_PORT},{FD_METRICS_PORT + 1},{FD_METRICS_PORT + 2},{FD_METRICS_PORT + 3}",
"--args", "--args",
"--model", "--model",
model_path, model_path,
"--engine-worker-queue-port", "--engine-worker-queue-port",
f"{FD_ENGINE_QUEUE_PORT},{FD_ENGINE_QUEUE_PORT + 1}", f"{FD_ENGINE_QUEUE_PORT},{FD_ENGINE_QUEUE_PORT + 1},{FD_ENGINE_QUEUE_PORT + 2},{FD_ENGINE_QUEUE_PORT + 3}",
"--cache-queue-port", "--cache-queue-port",
f"{FD_CACHE_QUEUE_PORT},{FD_CACHE_QUEUE_PORT + 1}", f"{FD_CACHE_QUEUE_PORT},{FD_CACHE_QUEUE_PORT + 1},{FD_CACHE_QUEUE_PORT + 2},{FD_CACHE_QUEUE_PORT + 3}",
"--tensor-parallel-size", "--tensor-parallel-size",
"2", "1",
"--data-parallel-size", "--data-parallel-size",
"2", "4",
"--max-model-len", "--max-model-len",
"65536", "65536",
"--max-num-seqs", "--max-num-seqs",
@@ -337,13 +345,17 @@ def test_text_diff(api_url):
base_path = os.getenv("MODEL_PATH") base_path = os.getenv("MODEL_PATH")
if base_path: if base_path:
base_file = os.path.join(base_path, "21b_ep4_mtp_text_baseline_dev_0311.txt") base_file = os.path.join(base_path, "21b_tp1_dp4_mtp_text_baseline.txt")
else: else:
base_file = "21b_ep4_mtp_text_baseline_dev_0311.txt" base_file = "21b_tp1_dp4_mtp_text_baseline.txt"
with open(base_file, "r", encoding="utf-8") as f: with open(base_file, "r", encoding="utf-8") as f:
baseline = f.read() baseline = f.read()
if result != baseline:
with open("21b_tp1_dp4_mtp_text_tmp.txt", "w", encoding="utf-8") as f:
f.write(result)
assert result == baseline, f"Text mismatch with baseline\nresult: {result}\nbaseline: {baseline}" assert result == baseline, f"Text mismatch with baseline\nresult: {result}\nbaseline: {baseline}"
@@ -504,13 +516,17 @@ def test_non_stream_with_logprobs(api_url):
base_path = os.getenv("MODEL_PATH") base_path = os.getenv("MODEL_PATH")
if base_path: if base_path:
base_file = os.path.join(base_path, "21b_ep4_mtp_logprobs_non_stream_static_baseline_dev_0311.txt") base_file = os.path.join(base_path, "21b_tp1_dp4_mtp_logprobs_non_stream_static_baseline.txt")
else: else:
base_file = "21b_ep4_mtp_logprobs_non_stream_static_baseline_dev_0311.txt" base_file = "21b_tp1_dp4_mtp_logprobs_non_stream_static_baseline.txt"
with open(base_file, "r", encoding="utf-8") as f: with open(base_file, "r", encoding="utf-8") as f:
baseline = json.load(f) baseline = json.load(f)
if logprobs != baseline:
with open("21b_tp1_dp4_mtp_logprobs_non_stream_static_tmp.txt", "w", encoding="utf-8") as f:
f.write(json.dumps(logprobs, ensure_ascii=False, indent=2))
assert logprobs == baseline assert logprobs == baseline
@@ -539,11 +555,15 @@ def test_stream_with_logprobs(api_url):
base_path = os.getenv("MODEL_PATH") base_path = os.getenv("MODEL_PATH")
if base_path: if base_path:
base_file = os.path.join(base_path, "21b_ep4_mtp_logprobs_stream_static_baseline_dev_0311.txt") base_file = os.path.join(base_path, "21b_tp1_dp4_mtp_logprobs_stream_static_baseline.txt")
else: else:
base_file = "21b_ep4_mtp_logprobs_stream_static_baseline_dev_0311.txt" base_file = "21b_tp1_dp4_mtp_logprobs_stream_static_baseline.txt"
with open(base_file, "r", encoding="utf-8") as f: with open(base_file, "r", encoding="utf-8") as f:
baseline = json.load(f) baseline = json.load(f)
if logprobs != baseline:
with open("21b_tp1_dp4_mtp_logprobs_stream_static_tmp.txt", "w", encoding="utf-8") as f:
f.write(json.dumps(logprobs, ensure_ascii=False, indent=2))
assert logprobs == baseline assert logprobs == baseline