minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 14 Packages Projects Releases Wiki Activity

[AMD][MoRI] Auto-select dispatch quantization type from MoE weight dtype. (#21040)

Browse Source
This commit is contained in:
Duyi-Wang
2026-03-24 22:53:57 -07:00
committed by GitHub
parent 86e2622097
commit 61a902ce88
5 changed files with 90 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/amd/test_moriep_small.py
+8 -8
View File
@@ -69,7 +69,7 @@ class TestPureDP(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
@@ -113,7 +113,7 @@ class TestMTP(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
@@ -166,7 +166,7 @@ class TestNormal(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
@@ -213,7 +213,7 @@ class TestLowLatency(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
# FIXME(billishyahao): enable p2p due to no rdma devices on CI machine
@@ -263,7 +263,7 @@ class TestTBOwithNormal(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
@@ -311,7 +311,7 @@ class TestTBOwithLowLatency(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
# FIXME(billishyahao): enable p2p due to no rdma devices on CI machine
@@ -365,7 +365,7 @@ class TestMTPwithTBONormal(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
@@ -423,7 +423,7 @@ class TestMTPwithTBOLowLatency(CustomTestCase):
env = dict(os.environ)
env["SGLANG_USE_AITER"] = "1"
env["SGLANG_MORI_FP8_DISP"] = "False"
env["SGLANG_MORI_DISPATCH_DTYPE"] = "bf16"
env["SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK"] = "4096"
env["MORI_SHMEM_MODE"] = "ISOLATION" # avoid out of symmetric heap memory
# FIXME(billishyahao): enable p2p due to no rdma devices on CI machine