Skip to content

Commit 5dcbeab

Browse files
ruirui6946ruirui6946
authored
Merge branch 'main' into agent/51912-log-output-token-ids
2 parents 27957b9 + 3d204df commit 5dcbeab

144 files changed

Lines changed: 9478 additions & 443 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

‎.buildkite/test-amd.yaml‎

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1988,8 +1988,8 @@ steps:
19881988
- vllm/
19891989
- tests/models/language/generation
19901990
commands:
1991-
- uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@fix-rocm-7.0-warp-size-constexpr'
1992-
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
1991+
- MAMBA_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@fix-rocm-7.0-warp-size-constexpr'
1992+
- CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
19931993
- pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
19941994

19951995
- label: Language Models Tests (Hybrid) %N # TBD
@@ -2004,8 +2004,8 @@ steps:
20042004
- vllm/
20052005
- tests/models/language/generation
20062006
commands:
2007-
- uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@fix-rocm-7.0-warp-size-constexpr'
2008-
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
2007+
- MAMBA_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@fix-rocm-7.0-warp-size-constexpr'
2008+
- CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
20092009
- pytest -v -s models/language/generation -m hybrid_model --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
20102010

20112011
#---------------------------------------------------- mi300 · models / multimodal ----------------------------------------------------#
@@ -3544,8 +3544,8 @@ steps:
35443544
- vllm/
35453545
- tests/models/language/generation
35463546
commands:
3547-
- uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@rocm-7.0-v2.3.0'
3548-
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
3547+
- MAMBA_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@rocm-7.0-v2.3.0'
3548+
- CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
35493549
- pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
35503550

35513551
- label: Language Models Test (Extended Pooling) # TBD

‎.buildkite/test_areas/lm_eval.yaml‎

Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -101,6 +101,26 @@ steps:
101101
commands:
102102
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-pcp.txt
103103

104+
- label: LM Eval Spec Decode (4xB200)
105+
key: lm-eval-spec-decode-4xb200
106+
timeout_in_minutes: 120
107+
device: b200-k8s
108+
num_devices: 4
109+
optional: true
110+
source_file_dependencies:
111+
- tests/evals/gsm8k/configs/DeepSeek-V4-Flash-DSpark-confidence-TP4.yaml
112+
- tests/evals/gsm8k/configs/models-spec-decode.txt
113+
- vllm/models/deepseek_v4/
114+
- vllm/model_executor/models/qwen3_dspark.py
115+
- vllm/v1/spec_decode/metrics.py
116+
- vllm/v1/worker/gpu/spec_decode/adaptive_verification.py
117+
- vllm/v1/worker/gpu/spec_decode/dflash/
118+
- vllm/v1/worker/gpu/spec_decode/dspark/
119+
- vllm/v1/worker/gpu/spec_decode/rejection_sampler.py
120+
- vllm/v1/worker/gpu/spec_decode/speculator.py
121+
commands:
122+
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-spec-decode.txt
123+
104124
- label: LM Eval Large Models EP (2xB200)
105125
key: lm-eval-large-models-ep-2xb200
106126
timeout_in_minutes: 60

‎.buildkite/test_areas/models_language.yaml‎

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -65,8 +65,8 @@ steps:
6565
- tests/models/language/generation
6666
commands:
6767
# Install fast path packages for testing against transformers
68-
- uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.3.0'
69-
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
68+
- MAMBA_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.3.0'
69+
- CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
7070
# Shard the hybrid language model tests that are numerically stable on Hopper.
7171
- pytest -v -s models/language/generation -m hybrid_model -k 'not granite-4.0-tiny-preview' --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
7272
parallelism: 2
@@ -78,8 +78,8 @@ steps:
7878
depends_on:
7979
- image-build-amd
8080
commands:
81-
- uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@fix-rocm-7.0-warp-size-constexpr'
82-
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
81+
- MAMBA_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/AndreasKaratzas/mamba@fix-rocm-7.0-warp-size-constexpr'
82+
- CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
8383
- pytest -v -s models/language/generation -m hybrid_model --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
8484

8585
# Granite 4 hybrid generation is sensitive to hardware-specific Triton SSD
@@ -93,8 +93,8 @@ steps:
9393
- "!vllm/distributed/kv_transfer/"
9494
- tests/models/language/generation
9595
commands:
96-
- uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.3.0'
97-
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
96+
- MAMBA_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.3.0'
97+
- CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
9898
- pytest -v -s models/language/generation -m hybrid_model -k 'granite-4.0-tiny-preview'
9999

100100
- label: Language Models Test (Extended Generation) # 80min
@@ -108,8 +108,8 @@ steps:
108108
- tests/models/language/generation
109109
commands:
110110
# Install fast path packages for testing against transformers
111-
- uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.3.0'
112-
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
111+
- MAMBA_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.3.0'
112+
- CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
113113
- pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
114114

115115
- label: Language Models Test (PPL)

‎cmake/external_projects/deepgemm.cmake‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -28,9 +28,9 @@ if(DEEPGEMM_SRC_DIR)
2828
message(STATUS "DeepGEMM using local DEEPGEMM_SRC_DIR: ${deepgemm_SOURCE_DIR}")
2929
else()
3030
# Keep in sync with tools/install_deepgemm.sh
31-
set(_DEEPGEMM_UPSTREAM_REPO "https://github.com/vllm-project/DeepGEMM.git")
32-
# TODO: switch to nv_dev branch after it support situ
33-
set(_DEEPGEMM_UPSTREAM_TAG "e21c821f39a2056d68067a466c64ddc942200106")
31+
set(_DEEPGEMM_UPSTREAM_REPO "https://github.com/deepseek-ai/DeepGEMM.git")
32+
# Pinned to the tip of the nv_dev branch (SM120 support).
33+
set(_DEEPGEMM_UPSTREAM_TAG "8b1392b978f5a03c828dd1711090d7fb50958b8a")
3434

3535
set(_deepgemm_fc_root "${FETCHCONTENT_BASE_DIR}")
3636
if(NOT _deepgemm_fc_root)

‎csrc/cpu/cpu_fused_moe.cpp‎

Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -27,10 +27,22 @@
2727
#define NEON_DISPATCH(...) case cpu_utils::ISA::NEON:
2828
#endif
2929

30+
#ifdef __powerpc64__
31+
#include "cpu/micro_gemm/cpu_micro_gemm_vsx.hpp"
32+
#define VSX_DISPATCH(...) \
33+
case cpu_utils::ISA::VSX: { \
34+
using gemm_t = cpu_micro_gemm::MicroGemm<cpu_utils::ISA::VSX, scalar_t>; \
35+
return __VA_ARGS__(); \
36+
}
37+
#else
38+
#define VSX_DISPATCH(...) case cpu_utils::ISA::VSX:
39+
#endif
40+
3041
#define CPU_ISA_DISPATCH_IMPL(ISA_TYPE, ...) \
3142
[&] { \
3243
switch (ISA_TYPE) { \
3344
AMX_DISPATCH(__VA_ARGS__) \
45+
VSX_DISPATCH(__VA_ARGS__) \
3446
case cpu_utils::ISA::VEC: { \
3547
using gemm_t = \
3648
cpu_micro_gemm::MicroGemm<cpu_utils::ISA::VEC, scalar_t>; \

0 commit comments

Comments
 (0)