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

[Bugfix] Fix AWQ models batch invariance issues (#38670)

Browse Source 
Signed-off-by: yusuf <yusuf@deeplearningmachine.mynet>
Signed-off-by: <>
Co-authored-by: yusuf <yusuf@deeplearningmachine.mynet>
This commit is contained in:
Yusuf Mohammad
2026-04-03 15:54:15 +01:00
committed by GitHub
parent 6b4872240f
commit 46f02e00f2
4 changed files with 27 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/v1/determinism/test_batch_invariance.py
View File

@@ -187,7 +187,7 @@ def test_logprobs_bitwise_batch_invariance_bs1_vs_bsN(
tensor_parallel_size=tp_size,
max_num_seqs=128,
max_model_len=8192,
dtype="bfloat16", # not everything is supported
dtype="auto", # not everything is supported
gpu_memory_utilization=0.9,
enforce_eager=IS_DEVICE_CAPABILITY_BELOW_90,
attention_config={"backend": backend},
@@ -400,7 +400,7 @@ def test_simple_generation(backend):
tensor_parallel_size=int(os.getenv("VLLM_TP_SIZE", "1")),
gpu_memory_utilization=0.9,
max_model_len=2048,
dtype="bfloat16",
dtype="auto",
enable_prefix_caching=False,
enforce_eager=IS_DEVICE_CAPABILITY_BELOW_90,
attention_config={"backend": backend},
@@ -466,7 +466,7 @@ def test_logprobs_without_batch_invariance_should_fail(
tensor_parallel_size=tp_size,
max_num_seqs=32,
max_model_len=8192,
dtype="bfloat16",
dtype="auto",
enforce_eager=IS_DEVICE_CAPABILITY_BELOW_90,
attention_config={"backend": backend},
)
@@ -686,7 +686,7 @@ def test_decode_logprobs_match_prefill_logprobs(
tensor_parallel_size=tp_size,
max_num_seqs=32,
max_model_len=8192,
dtype="bfloat16",
dtype="auto",
enforce_eager=IS_DEVICE_CAPABILITY_BELOW_90,
attention_config={"backend": backend},
)
@@ -931,7 +931,7 @@ def LLM_with_max_seqs(
max_num_seqs=max_num_seqs,
gpu_memory_utilization=gpu_memory_utilization,
max_model_len=max_model_len,
dtype="bfloat16",
dtype="auto",
tensor_parallel_size=int(os.getenv("VLLM_TP_SIZE", "1")),
enable_prefix_caching=False,
enforce_eager=IS_DEVICE_CAPABILITY_BELOW_90,