[torch.compile] Reorganize vllm/compilation and tests/compile (0/N for vLLM IR) (#33731)

Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Signed-off-by: ProExpertProg <luka.govedic@gmail.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
This commit is contained in:
Luka Govedič
2026-02-06 07:19:49 -05:00
committed by GitHub
parent f79d9dce16
commit ac32e66cf9
47 changed files with 717 additions and 651 deletions

View File

@@ -82,19 +82,17 @@ INDUCTOR_GRAPH_PARTITION = [
]
FUSION_LOG_PATTERNS: dict[str, re.Pattern] = {
"rms_quant_fusion": re.compile(
r"\[(?:compilation/)?fusion.py:\d+] Replaced (\d+) patterns"
),
"act_quant_fusion": re.compile(
r"activation_quant_fusion.py:\d+] Replaced (\d+) patterns"
),
"rms_quant_fusion": re.compile(r"rms_quant_fusion.py:\d+] Replaced (\d+) patterns"),
"act_quant_fusion": re.compile(r"act_quant_fusion.py:\d+] Replaced (\d+) patterns"),
"norm_rope_fusion": re.compile(
r"qk_norm_rope_fusion.py:\d+] Fused QK Norm\+RoPE on (\d+) sites"
),
"attn_quant_fusion": re.compile(
r"fusion_attn.py:\d+] Fused quant onto (\d+) attention nodes"
r"attn_quant_fusion.py:\d+] Fused quant onto (\d+) attention nodes"
),
"ar_rms_fusion": re.compile(
r"allreduce_rms_fusion.py:\d+] Replaced (\d+) patterns"
),
"ar_rms_fusion": re.compile(r"collective_fusion.py:\d+] Replaced (\d+) patterns"),
"sequence_parallel": re.compile(
r"sequence_parallelism.py:\d+] Replaced (\d+) patterns"
),