A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 23:45 UTC
[Fix] Update mamba_ssm to 2.2.5 (#21421)
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
elvischenv committed
Jul 24, 2025 at 18:25 UTC
5a19a6c6705fe83db2e3517a2d2f473586901743
4 files changed
+6
-12
docker/Dockerfile
-8
@@ -276,10 +276,6 @@ ARG PYTORCH_CUDA_INDEX_BASE_URL
276
ENV UV_HTTP_TIMEOUT=500
277
ENV UV_INDEX_STRATEGY="unsafe-best-match"
278
279
-# Workaround for #17068
280
-RUN --mount=type=cache,target=/root/.cache/uv \
281
- uv pip install --system --no-build-isolation "git+https://github.com/state-spaces/mamba@v2.2.4"
282
-
279
COPY requirements/lint.txt requirements/lint.txt
280
COPY requirements/test.txt requirements/test.txt
281
COPY requirements/dev.txt requirements/dev.txt
@@ -452,10 +448,6 @@ ARG PIP_EXTRA_INDEX_URL UV_EXTRA_INDEX_URL
448
ENV UV_HTTP_TIMEOUT=500
449
ENV UV_INDEX_STRATEGY="unsafe-best-match"
450
455
-# Workaround for #17068
456
-RUN --mount=type=cache,target=/root/.cache/uv \
457
- uv pip install --system --no-build-isolation "git+https://github.com/state-spaces/mamba@v2.2.4"
458
-
451
# install development dependencies (for testing)
452
RUN --mount=type=cache,target=/root/.cache/uv \
453
CUDA_MAJOR="${CUDA_VERSION%%.*}"; \
docs/contributing/ci/update_pytorch_version.md
+1
-1
@@ -134,7 +134,7 @@ MAX_JOBS=16 uv pip install --system \
134
135
```bash
136
uv pip install --system \
137
- --no-build-isolation "git+https://github.com/state-spaces/mamba@v2.2.4"
137
+ --no-build-isolation "git+https://github.com/state-spaces/mamba@v2.2.5"
138
```
139
140
### causal-conv1d
requirements/test.in
+1
-1
@@ -26,7 +26,7 @@ torch==2.7.1
26
torchaudio==2.7.1
27
torchvision==0.22.1
28
transformers_stream_generator # required for qwen-vl test
29
-mamba_ssm # required for plamo2 test
29
+mamba_ssm==2.2.5 # required for plamo2 test
30
matplotlib # required for qwen-vl test
31
mistral_common[image,audio] >= 1.8.2 # required for voxtral test
32
num2words # required for smolvlm test
requirements/test.txt
+4
-2
@@ -421,7 +421,7 @@ lxml==5.3.0
421
# sacrebleu
422
mako==1.3.10
423
# via alembic
424
-mamba-ssm==2.2.4
424
+mamba-ssm==2.2.5
425
# via -r requirements/test.in
426
markdown==3.8.2
427
# via mlflow
@@ -1152,7 +1152,9 @@ transformers==4.53.2
1152
transformers-stream-generator==0.0.5
1153
# via -r requirements/test.in
1154
triton==3.3.1
1155
- # via torch
1155
+ # via
1156
+ # mamba-ssm
1157
+ # torch
1158
tritonclient==2.51.0
1159
# via
1160
# -r requirements/test.in