From 851d96a2784055d3ed65cf10f3a6900cb4662a4e Mon Sep 17 00:00:00 2001 From: LRL-ModelCloud Date: Fri, 26 Sep 2025 16:12:41 +0800 Subject: [PATCH] fix test_ppl --- tests/test_perplexity.py | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/tests/test_perplexity.py b/tests/test_perplexity.py index cc984b473..11ff0f7b9 100644 --- a/tests/test_perplexity.py +++ b/tests/test_perplexity.py @@ -118,13 +118,13 @@ def calculate_native_ppl(self, format): @parameterized.expand( [ - # (QUANT_METHOD.GPTQ, FORMAT.GPTQ, 8, 32, True), # A100, 4889 max ram - (METHOD.GPTQ, FORMAT.GPTQ, 8, 32, False), # A100, 6571 max ram - # (QUANT_METHOD.GPTQ, FORMAT.GPTQ_V2, 8, 32, False), - # (QUANT_METHOD.GPTQ, FORMAT.GPTQ_V2, 4, 32, False), - # (QUANT_METHOD.GPTQ, FORMAT.GPTQ, 4, 32, False), - # (QUANT_METHOD.GPTQ, FORMAT.BITBLAS, 4, 32, False), - # (QUANT_METHOD.AUTO_ROUND, FORMAT.GPTQ, 4, 32, False), + # (QUANT_METHOD.GPTQ, FORMAT.GPTQ, 8, 32), # A100, 4889 max ram + (METHOD.GPTQ, FORMAT.GPTQ, 8, 32), # A100, 6571 max ram + # (QUANT_METHOD.GPTQ, FORMAT.GPTQ_V2, 8, 32), + # (QUANT_METHOD.GPTQ, FORMAT.GPTQ_V2, 4, 32), + # (QUANT_METHOD.GPTQ, FORMAT.GPTQ, 4, 32), + # (QUANT_METHOD.GPTQ, FORMAT.BITBLAS, 4, 32), + # (QUANT_METHOD.AUTO_ROUND, FORMAT.GPTQ, 4, 32), ] ) def test_quantized_perplexity(self, method: METHOD, format: FORMAT, bits: int, group_size: int):