Skip to content

Commit 524cf49

Browse files
Henry LinjamäkiRealTimeChris
Henry Linjamäki
authored and
RealTimeChris
committed
opencl: use OpenCL C standard supported by the device (ggml-org#12221)
This patch nudges the llama.cpp a bit to be supported on PoCL which doesn't support OpenCL C CL2.0. The issue is solved by querying the device for the supported OpenCL C versions and using the highest one available.
0 parents  commit 524cf49

File tree

1,224 files changed

+404681
-0
lines changed

Some content is hidden

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

1,224 files changed

+404681
-0
lines changed

‎.clang-format‎

Lines changed: 161 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,161 @@
1+
---
2+
Language: Cpp
3+
AlignAfterOpenBracket: Align
4+
AlignArrayOfStructures: Left
5+
AlignConsecutiveAssignments: AcrossComments
6+
AlignConsecutiveBitFields: AcrossComments
7+
AlignConsecutiveDeclarations: AcrossComments
8+
AlignConsecutiveMacros: AcrossComments
9+
# AlignConsecutiveShortCaseStatements: AcrossComments
10+
AlignEscapedNewlines: Left # LeftWithLastLine
11+
AlignOperands: Align
12+
AlignTrailingComments:
13+
Kind: Always
14+
OverEmptyLines: 1
15+
AllowAllArgumentsOnNextLine: true
16+
AllowAllParametersOfDeclarationOnNextLine: false
17+
# AllowBreakBeforeNoexceptSpecifier: OnlyWithParen
18+
AllowShortBlocksOnASingleLine: Never
19+
AllowShortCaseLabelsOnASingleLine: false
20+
AllowShortFunctionsOnASingleLine: Inline
21+
AllowShortIfStatementsOnASingleLine: Never
22+
AllowShortLambdasOnASingleLine: Inline
23+
AllowShortLoopsOnASingleLine: false
24+
AlwaysBreakBeforeMultilineStrings: true
25+
BinPackArguments: true
26+
BinPackParameters: true # OnePerLine
27+
BitFieldColonSpacing: Both
28+
BreakBeforeBraces: Custom # Attach
29+
BraceWrapping:
30+
AfterCaseLabel: true
31+
AfterClass: false
32+
AfterControlStatement: false
33+
AfterEnum: false
34+
AfterFunction: false
35+
AfterNamespace: false
36+
AfterObjCDeclaration: false
37+
AfterStruct: false
38+
AfterUnion: false
39+
AfterExternBlock: false
40+
BeforeCatch: false
41+
BeforeElse: false
42+
BeforeLambdaBody: false
43+
BeforeWhile: false
44+
IndentBraces: false
45+
SplitEmptyFunction: false
46+
SplitEmptyRecord: false
47+
SplitEmptyNamespace: false
48+
# BreakAdjacentStringLiterals: true
49+
BreakAfterAttributes: Never
50+
BreakBeforeBinaryOperators: None
51+
BreakBeforeInlineASMColon: OnlyMultiline
52+
BreakBeforeTernaryOperators: false
53+
# BreakBinaryOperations: Never
54+
BreakConstructorInitializers: AfterColon
55+
# BreakFunctionDefinitionParameters: false
56+
BreakInheritanceList: AfterComma
57+
BreakStringLiterals: true
58+
# BreakTemplateDeclarations: Yes
59+
ColumnLimit: 120
60+
CommentPragmas: '^ IWYU pragma:'
61+
CompactNamespaces: false
62+
ConstructorInitializerIndentWidth: 4
63+
ContinuationIndentWidth: 4
64+
Cpp11BracedListStyle: false
65+
DerivePointerAlignment: false
66+
DisableFormat: false
67+
EmptyLineBeforeAccessModifier: Leave
68+
EmptyLineAfterAccessModifier: Never
69+
ExperimentalAutoDetectBinPacking: false
70+
FixNamespaceComments: true
71+
IncludeBlocks: Regroup
72+
IncludeCategories:
73+
- Regex: '^<.*\.h>'
74+
Priority: 1
75+
SortPriority: 0
76+
- Regex: '^<.*'
77+
Priority: 2
78+
SortPriority: 0
79+
- Regex: '.*'
80+
Priority: 3
81+
SortPriority: 0
82+
IncludeIsMainRegex: '([-_](test|unittest))?$'
83+
IncludeIsMainSourceRegex: ''
84+
IndentAccessModifiers: false
85+
IndentCaseBlocks: true
86+
IndentCaseLabels: true
87+
IndentExternBlock: NoIndent
88+
IndentGotoLabels: false
89+
IndentPPDirectives: AfterHash
90+
IndentWidth: 4
91+
IndentWrappedFunctionNames: false
92+
InsertBraces: true # NOTE: may lead to incorrect formatting
93+
InsertNewlineAtEOF: true
94+
JavaScriptQuotes: Leave
95+
JavaScriptWrapImports: true
96+
KeepEmptyLinesAtTheStartOfBlocks: false
97+
LambdaBodyIndentation: Signature
98+
LineEnding: LF
99+
MacroBlockBegin: ''
100+
MacroBlockEnd: ''
101+
MaxEmptyLinesToKeep: 1
102+
NamespaceIndentation: None
103+
ObjCBinPackProtocolList: Auto
104+
ObjCBlockIndentWidth: 4
105+
ObjCSpaceAfterProperty: true
106+
ObjCSpaceBeforeProtocolList: true
107+
PPIndentWidth: -1
108+
PackConstructorInitializers: CurrentLine
109+
PenaltyBreakAssignment: 2
110+
PenaltyBreakBeforeFirstCallParameter: 1
111+
PenaltyBreakComment: 300
112+
PenaltyBreakFirstLessLess: 120
113+
PenaltyBreakString: 1000
114+
PenaltyBreakTemplateDeclaration: 10
115+
PenaltyExcessCharacter: 1000000
116+
PenaltyReturnTypeOnItsOwnLine: 200
117+
PointerAlignment: Middle
118+
QualifierAlignment: Left
119+
#QualifierOrder: ['static', 'inline', 'friend', 'constexpr', 'const', 'volatile', 'type', 'restrict']
120+
RawStringFormats:
121+
- Language: Cpp
122+
Delimiters:
123+
- cc
124+
- CC
125+
- cpp
126+
- Cpp
127+
- CPP
128+
- 'c++'
129+
- 'C++'
130+
CanonicalDelimiter: ''
131+
ReferenceAlignment: Middle
132+
ReflowComments: false # IndentOnly
133+
SeparateDefinitionBlocks: Always
134+
SortIncludes: CaseInsensitive
135+
SortUsingDeclarations: LexicographicNumeric
136+
SpaceAfterCStyleCast: true
137+
SpaceAfterLogicalNot: false
138+
SpaceAfterTemplateKeyword: true
139+
SpaceBeforeAssignmentOperators: true
140+
SpaceBeforeCpp11BracedList: false
141+
SpaceBeforeCtorInitializerColon: true
142+
SpaceBeforeInheritanceColon: true
143+
SpaceBeforeParens: ControlStatements
144+
SpaceBeforeRangeBasedForLoopColon: true
145+
SpaceInEmptyBlock: false
146+
SpaceInEmptyParentheses: false
147+
SpacesBeforeTrailingComments: 2
148+
SpacesInAngles: Never
149+
SpacesInContainerLiterals: true
150+
SpacesInLineCommentPrefix:
151+
Minimum: 1
152+
Maximum: -1
153+
SpacesInParentheses: false
154+
SpacesInSquareBrackets: false
155+
SpaceBeforeSquareBrackets: false
156+
Standard: c++17
157+
TabWidth: 4
158+
UseTab: Never
159+
WhitespaceSensitiveMacros: ['STRINGIZE']
160+
...
161+

‎.clang-tidy‎

Lines changed: 26 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,26 @@
1+
---
2+
Checks: >
3+
bugprone-*,
4+
-bugprone-easily-swappable-parameters,
5+
-bugprone-implicit-widening-of-multiplication-result,
6+
-bugprone-misplaced-widening-cast,
7+
-bugprone-narrowing-conversions,
8+
readability-*,
9+
-readability-avoid-unconditional-preprocessor-if,
10+
-readability-function-cognitive-complexity,
11+
-readability-identifier-length,
12+
-readability-implicit-bool-conversion,
13+
-readability-magic-numbers,
14+
-readability-uppercase-literal-suffix,
15+
-readability-simplify-boolean-expr,
16+
clang-analyzer-*,
17+
-clang-analyzer-security.insecureAPI.DeprecatedOrUnsafeBufferHandling,
18+
performance-*,
19+
portability-*,
20+
-portability-simd-intrinsics,
21+
misc-*,
22+
-misc-const-correctness,
23+
-misc-non-private-member-variables-in-classes,
24+
-misc-no-recursion,
25+
-misc-use-anonymous-namespace,
26+
FormatStyle: none

‎.devops/cloud-v-pipeline‎

Lines changed: 22 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,22 @@
1+
node('x86_runner1'){ // Running on x86 runner containing latest vector qemu, latest vector gcc and all the necessary libraries
2+
stage('Cleanup'){
3+
cleanWs() // Cleaning previous CI build in workspace
4+
}
5+
stage('checkout repo'){
6+
retry(5){ // Retry if the cloning fails due to some reason
7+
checkout scm // Clone the repo on Runner
8+
}
9+
}
10+
stage('Compiling llama.cpp'){
11+
sh'''#!/bin/bash
12+
make RISCV=1 RISCV_CROSS_COMPILE=1 # Compiling llama for RISC-V
13+
'''
14+
}
15+
stage('Running llama.cpp'){
16+
sh'''#!/bin/bash
17+
module load gnu-bin2/0.1 # loading latest versions of vector qemu and vector gcc
18+
qemu-riscv64 -L /softwares/gnu-bin2/sysroot -cpu rv64,v=true,vlen=256,elen=64,vext_spec=v1.0 ./llama-cli -m /home/alitariq/codellama-7b.Q4_K_M.gguf -p "Anything" -n 9 > llama_log.txt # Running llama.cpp on vector qemu-riscv64
19+
cat llama_log.txt # Printing results
20+
'''
21+
}
22+
}

‎.devops/cpu.Dockerfile‎

Lines changed: 92 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,92 @@
1+
ARG UBUNTU_VERSION=22.04
2+
3+
FROM ubuntu:$UBUNTU_VERSION AS build
4+
5+
ARG TARGETARCH
6+
7+
ARG GGML_CPU_ARM_ARCH=armv8-a
8+
9+
RUN apt-get update && \
10+
apt-get install -y build-essential git cmake libcurl4-openssl-dev
11+
12+
WORKDIR /app
13+
14+
COPY . .
15+
16+
RUN if [ "$TARGETARCH" = "amd64" ]; then \
17+
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release -DLLAMA_CURL=ON -DGGML_NATIVE=OFF -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON; \
18+
elif [ "$TARGETARCH" = "arm64" ]; then \
19+
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release -DLLAMA_CURL=ON -DGGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=${GGML_CPU_ARM_ARCH}; \
20+
else \
21+
echo "Unsupported architecture"; \
22+
exit 1; \
23+
fi && \
24+
cmake --build build -j $(nproc)
25+
26+
RUN mkdir -p /app/lib && \
27+
find build -name "*.so" -exec cp {} /app/lib \;
28+
29+
RUN mkdir -p /app/full \
30+
&& cp build/bin/* /app/full \
31+
&& cp *.py /app/full \
32+
&& cp -r gguf-py /app/full \
33+
&& cp -r requirements /app/full \
34+
&& cp requirements.txt /app/full \
35+
&& cp .devops/tools.sh /app/full/tools.sh
36+
37+
## Base image
38+
FROM ubuntu:$UBUNTU_VERSION AS base
39+
40+
RUN apt-get update \
41+
&& apt-get install -y libgomp1 curl\
42+
&& apt autoremove -y \
43+
&& apt clean -y \
44+
&& rm -rf /tmp/* /var/tmp/* \
45+
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
46+
&& find /var/cache -type f -delete
47+
48+
COPY --from=build /app/lib/ /app
49+
50+
### Full
51+
FROM base AS full
52+
53+
COPY --from=build /app/full /app
54+
55+
WORKDIR /app
56+
57+
RUN apt-get update \
58+
&& apt-get install -y \
59+
git \
60+
python3 \
61+
python3-pip \
62+
&& pip install --upgrade pip setuptools wheel \
63+
&& pip install -r requirements.txt \
64+
&& apt autoremove -y \
65+
&& apt clean -y \
66+
&& rm -rf /tmp/* /var/tmp/* \
67+
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
68+
&& find /var/cache -type f -delete
69+
70+
ENTRYPOINT ["/app/tools.sh"]
71+
72+
### Light, CLI only
73+
FROM base AS light
74+
75+
COPY --from=build /app/full/llama-cli /app
76+
77+
WORKDIR /app
78+
79+
ENTRYPOINT [ "/app/llama-cli" ]
80+
81+
### Server, Server only
82+
FROM base AS server
83+
84+
ENV LLAMA_ARG_HOST=0.0.0.0
85+
86+
COPY --from=build /app/full/llama-server /app
87+
88+
WORKDIR /app
89+
90+
HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]
91+
92+
ENTRYPOINT [ "/app/llama-server" ]

‎.devops/cuda.Dockerfile‎

Lines changed: 94 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,94 @@
1+
ARG UBUNTU_VERSION=22.04
2+
# This needs to generally match the container host's environment.
3+
ARG CUDA_VERSION=12.4.0
4+
# Target the CUDA build image
5+
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
6+
7+
ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}
8+
9+
FROM ${BASE_CUDA_DEV_CONTAINER} AS build
10+
11+
# CUDA architecture to build for (defaults to all supported archs)
12+
ARG CUDA_DOCKER_ARCH=default
13+
14+
RUN apt-get update && \
15+
apt-get install -y build-essential cmake python3 python3-pip git libcurl4-openssl-dev libgomp1
16+
17+
WORKDIR /app
18+
19+
COPY . .
20+
21+
RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
22+
export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
23+
fi && \
24+
cmake -B build -DGGML_NATIVE=OFF -DGGML_CUDA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
25+
cmake --build build --config Release -j$(nproc)
26+
27+
RUN mkdir -p /app/lib && \
28+
find build -name "*.so" -exec cp {} /app/lib \;
29+
30+
RUN mkdir -p /app/full \
31+
&& cp build/bin/* /app/full \
32+
&& cp *.py /app/full \
33+
&& cp -r gguf-py /app/full \
34+
&& cp -r requirements /app/full \
35+
&& cp requirements.txt /app/full \
36+
&& cp .devops/tools.sh /app/full/tools.sh
37+
38+
## Base image
39+
FROM ${BASE_CUDA_RUN_CONTAINER} AS base
40+
41+
RUN apt-get update \
42+
&& apt-get install -y libgomp1 curl\
43+
&& apt autoremove -y \
44+
&& apt clean -y \
45+
&& rm -rf /tmp/* /var/tmp/* \
46+
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
47+
&& find /var/cache -type f -delete
48+
49+
COPY --from=build /app/lib/ /app
50+
51+
### Full
52+
FROM base AS full
53+
54+
COPY --from=build /app/full /app
55+
56+
WORKDIR /app
57+
58+
RUN apt-get update \
59+
&& apt-get install -y \
60+
git \
61+
python3 \
62+
python3-pip \
63+
&& pip install --upgrade pip setuptools wheel \
64+
&& pip install -r requirements.txt \
65+
&& apt autoremove -y \
66+
&& apt clean -y \
67+
&& rm -rf /tmp/* /var/tmp/* \
68+
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
69+
&& find /var/cache -type f -delete
70+
71+
72+
ENTRYPOINT ["/app/tools.sh"]
73+
74+
### Light, CLI only
75+
FROM base AS light
76+
77+
COPY --from=build /app/full/llama-cli /app
78+
79+
WORKDIR /app
80+
81+
ENTRYPOINT [ "/app/llama-cli" ]
82+
83+
### Server, Server only
84+
FROM base AS server
85+
86+
ENV LLAMA_ARG_HOST=0.0.0.0
87+
88+
COPY --from=build /app/full/llama-server /app
89+
90+
WORKDIR /app
91+
92+
HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]
93+
94+
ENTRYPOINT [ "/app/llama-server" ]

0 commit comments

Comments
 (0)