biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[ci][test] adjust max wait time for cpu offloading test (#7709)

Browse Source
This commit is contained in:
youkaichao
2024-08-20 17:12:44 -07:00
committed by GitHub
parent 6e4658c7aa
commit 9e51b6a626
2 changed files with 33 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
tests/quantization/test_cpu_offload.py
View File

@@ -14,10 +14,12 @@ def test_cpu_offload_fp8():
# Test quantization of an unquantized checkpoint
compare_two_settings("meta-llama/Meta-Llama-3-8B-Instruct",
["--quantization", "fp8"],
["--quantization", "fp8", "--cpu-offload-gb", "2"])
["--quantization", "fp8", "--cpu-offload-gb", "2"],
max_wait_seconds=480)
# Test loading a quantized checkpoint
compare_two_settings("neuralmagic/Meta-Llama-3-8B-Instruct-FP8", [],
["--cpu-offload-gb", "2"])
["--cpu-offload-gb", "2"],
max_wait_seconds=480)
@pytest.mark.skipif(not is_quant_method_supported("gptq_marlin"),
@@ -25,11 +27,13 @@ def test_cpu_offload_fp8():
def test_cpu_offload_gptq():
# Test GPTQ Marlin
compare_two_settings("Qwen/Qwen2-1.5B-Instruct-GPTQ-Int4", [],
["--cpu-offload-gb", "1"])
["--cpu-offload-gb", "1"],
max_wait_seconds=480)
# Test GPTQ
compare_two_settings("Qwen/Qwen2-1.5B-Instruct-GPTQ-Int4",
["--quantization", "gptq"],
["--quantization", "gptq", "--cpu-offload-gb", "1"])
["--quantization", "gptq", "--cpu-offload-gb", "1"],
max_wait_seconds=480)
@pytest.mark.skipif(not is_quant_method_supported("awq_marlin"),
@@ -37,11 +41,13 @@ def test_cpu_offload_gptq():
def test_cpu_offload_awq():
# Test AWQ Marlin
compare_two_settings("Qwen/Qwen2-1.5B-Instruct-AWQ", [],
["--cpu-offload-gb", "1"])
["--cpu-offload-gb", "1"],
max_wait_seconds=480)
# Test AWQ
compare_two_settings("Qwen/Qwen2-1.5B-Instruct-AWQ",
["--quantization", "awq"],
["--quantization", "awq", "--cpu-offload-gb", "1"])
["--quantization", "awq", "--cpu-offload-gb", "1"],
max_wait_seconds=480)
@pytest.mark.skipif(not is_quant_method_supported("gptq_marlin"),
@@ -49,11 +55,14 @@ def test_cpu_offload_awq():
def test_cpu_offload_compressed_tensors():
# Test wNa16
compare_two_settings("nm-testing/tinyllama-oneshot-w4a16-channel-v2", [],
["--cpu-offload-gb", "1"])
["--cpu-offload-gb", "1"],
max_wait_seconds=480)
# Test w4a16_marlin24
compare_two_settings("nm-testing/llama7b-one-shot-2_4-w4a16-marlin24-t",
[], ["--cpu-offload-gb", "1"])
[], ["--cpu-offload-gb", "1"],
max_wait_seconds=480)
# Test w8a8
compare_two_settings(
"nm-testing/tinyllama-oneshot-w8w8-test-static-shape-change", [],
["--cpu-offload-gb", "1"])
["--cpu-offload-gb", "1"],
max_wait_seconds=480)