A high-throughput and memory-efficient inference and serving engine for LLMs
[Bugfix] [DSV4] [ROCm] Pin apache-tvm-ffi version to `0.1.10` (#45169)
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
TJian committed
Jun 11, 2026 at 01:41 UTC
bfe1001ab6d59d7da1f70a3bc107cebaff91f925
2 files changed
+3
-1
requirements/rocm.txt
+2
-1
@@ -23,6 +23,7 @@ timm>=1.0.17
23
# To be consistent with test_quark.py
24
amd-quark>=0.8.99
25
tilelang==0.1.10
26
-
26
+# Required apache-tvm-ffi matching tilelang version
27
+apache-tvm-ffi==0.1.10
28
# Required for faster safetensors model loading
29
fastsafetensors >= 0.3.2
requirements/test/rocm.txt
+1
@@ -44,6 +44,7 @@ anyio==4.13.0
44
# watchfiles
45
apache-tvm-ffi==0.1.10
46
# via
47
+ # -c requirements/rocm.txt
48
# tilelang
49
# xgrammar
50
arctic-inference==0.1.1