Skip to content

Commit df94522

Browse files
voidstackloopclaude
voidstackloop
and
claude
committed
Merge master into feat/hybrid-moe-expert-cache
Resolve conflicts with master (1b29621): - llama-moe-cache: master turned the LRU into an SLRU (ggml-org#30637). Rename the PR's profile prefill, hit counting and adaptive-swap code to the SLRU struct (moe_cache_slru); plan() keeps n_hit_profile and clears from_profile on a fill. - llama-graph: keep the PR's host_ids parameter and add master's build_hadamard_input call in build_lora_mm_id; apply master's MELLUM precision condition inside the PR's build_experts lambda. - llama-context: keep the hybrid graph-node reservation next to master's new sampler node count. - server-context: keep master's "graph reallocs" log and the PR's MoE cache stats. - test-llama-archs: master refactored tensor filling; port the quantized expert branch to it. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2 parents f6dea51 + 1b29621 commit df94522

881 files changed

Lines changed: 74901 additions & 31349 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.

‎.devops/cann.Dockerfile‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -145,7 +145,7 @@ ENTRYPOINT ["/app/tools.sh"]
145145
# ==============================================================================
146146
FROM base AS light
147147

148-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
148+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
149149

150150
ENTRYPOINT [ "/app/llama-cli" ]
151151

@@ -155,8 +155,9 @@ ENTRYPOINT [ "/app/llama-cli" ]
155155
FROM base AS server
156156

157157
ENV LLAMA_ARG_HOST=0.0.0.0
158+
ENV LLAMA_ARG_PORT=8080
158159

159-
COPY --from=build /app/full/llama /app/full/llama-server /app
160+
COPY --from=build /app/full/llama /app/full/llama-server /app/
160161

161162
HEALTHCHECK --interval=5m CMD [ "curl", "-f", "http://localhost:8080/health" ]
162163

‎.devops/cpu.Dockerfile‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -104,7 +104,7 @@ ENTRYPOINT ["/app/tools.sh"]
104104
### Light, CLI only
105105
FROM base AS light
106106

107-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
107+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
108108

109109
WORKDIR /app
110110

@@ -114,8 +114,9 @@ ENTRYPOINT [ "/app/llama-cli" ]
114114
FROM base AS server
115115

116116
ENV LLAMA_ARG_HOST=0.0.0.0
117+
ENV LLAMA_ARG_PORT=8080
117118

118-
COPY --from=build /app/full/llama /app/full/llama-server /app
119+
COPY --from=build /app/full/llama /app/full/llama-server /app/
119120

120121
WORKDIR /app
121122

‎.devops/cuda.Dockerfile‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -113,7 +113,7 @@ ENTRYPOINT ["/app/tools.sh"]
113113
### Light, CLI only
114114
FROM base AS light
115115

116-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
116+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
117117

118118
WORKDIR /app
119119

@@ -123,8 +123,9 @@ ENTRYPOINT [ "/app/llama-cli" ]
123123
FROM base AS server
124124

125125
ENV LLAMA_ARG_HOST=0.0.0.0
126+
ENV LLAMA_ARG_PORT=8080
126127

127-
COPY --from=build /app/full/llama /app/full/llama-server /app
128+
COPY --from=build /app/full/llama /app/full/llama-server /app/
128129

129130
WORKDIR /app
130131

‎.devops/intel.Dockerfile‎

Lines changed: 11 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,12 @@ RUN npm ci
1919
COPY tools/ui/ ./
2020
RUN LLAMA_BUILD_NUMBER="$APP_VERSION" npm run build
2121

22-
FROM docker.io/intel/oneapi-toolkit:$ONEAPI_VERSION AS build
22+
FROM docker.io/intel/deep-learning-essentials:$ONEAPI_VERSION AS build
23+
24+
RUN echo "Download oneDNN installation package" && \
25+
wget https://registrationcenter-download.intel.com/akdlm/IRC_NAS/94c3dbac-0852-45be-a57d-21c204cada3e/intel-onednn-2026.0.2.46_offline.sh -O intel-onednn_offline.sh && \
26+
echo "Install oneDNN components" && \
27+
bash intel-onednn_offline.sh -s -a --silent --eula accept
2328

2429
ARG GGML_SYCL_F16=ON
2530
ARG LEVEL_ZERO_VERSION=1.28.2
@@ -45,7 +50,7 @@ RUN if [ "${GGML_SYCL_F16}" = "ON" ]; then \
4550
fi && \
4651
echo "Building with dynamic libs" && \
4752
cmake -B build -DGGML_NATIVE=OFF -DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DLLAMA_BUILD_TESTS=OFF ${OPT_SYCL_F16} && \
48-
cmake --build build --config Release -j$(nproc)
53+
cmake --build build --config Release -j$((($(nproc)+1)/2))
4954

5055
RUN mkdir -p /app/lib && \
5156
find build -name "*.so*" -exec cp -P {} /app/lib \;
@@ -102,7 +107,7 @@ RUN mkdir /tmp/neo/ && cd /tmp/neo/ \
102107
&& dpkg --install *.deb
103108

104109
RUN apt-get update \
105-
&& apt-get install -y libgomp1 curl ffmpeg \
110+
&& apt-get install -y libgomp1 curl ffmpeg openssl libssl-dev \
106111
&& apt autoremove -y \
107112
&& apt clean -y \
108113
&& rm -rf /tmp/* /var/tmp/* \
@@ -141,7 +146,7 @@ ENTRYPOINT ["/app/tools.sh"]
141146
FROM base AS light
142147

143148
COPY --from=build /app/lib/ /app
144-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
149+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
145150

146151
WORKDIR /app
147152

@@ -151,9 +156,10 @@ ENTRYPOINT [ "/app/llama-cli" ]
151156
FROM base AS server
152157

153158
ENV LLAMA_ARG_HOST=0.0.0.0
159+
ENV LLAMA_ARG_PORT=8080
154160

155161
COPY --from=build /app/lib/ /app
156-
COPY --from=build /app/full/llama /app/full/llama-server /app
162+
COPY --from=build /app/full/llama /app/full/llama-server /app/
157163

158164
WORKDIR /app
159165

‎.devops/musa.Dockerfile‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -120,7 +120,7 @@ ENTRYPOINT ["/app/tools.sh"]
120120
### Light, CLI only
121121
FROM base AS light
122122

123-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
123+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
124124

125125
WORKDIR /app
126126

@@ -130,8 +130,9 @@ ENTRYPOINT [ "/app/llama-cli" ]
130130
FROM base AS server
131131

132132
ENV LLAMA_ARG_HOST=0.0.0.0
133+
ENV LLAMA_ARG_PORT=8080
133134

134-
COPY --from=build /app/full/llama /app/full/llama-server /app
135+
COPY --from=build /app/full/llama /app/full/llama-server /app/
135136

136137
WORKDIR /app
137138

‎.devops/openvino.Dockerfile‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -227,6 +227,7 @@ ENTRYPOINT [ "/app/llama-cli" ]
227227
FROM base AS server
228228

229229
ENV LLAMA_ARG_HOST=0.0.0.0
230+
ENV LLAMA_ARG_PORT=8080
230231

231232
COPY --from=build /app/full/llama /app/full/llama-server /app/
232233

‎.devops/rocm.Dockerfile‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -126,7 +126,7 @@ ENTRYPOINT ["/app/tools.sh"]
126126
### Light, CLI only
127127
FROM base AS light
128128

129-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
129+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
130130

131131
WORKDIR /app
132132

@@ -136,8 +136,9 @@ ENTRYPOINT [ "/app/llama-cli" ]
136136
FROM base AS server
137137

138138
ENV LLAMA_ARG_HOST=0.0.0.0
139+
ENV LLAMA_ARG_PORT=8080
139140

140-
COPY --from=build /app/full/llama /app/full/llama-server /app
141+
COPY --from=build /app/full/llama /app/full/llama-server /app/
141142

142143
WORKDIR /app
143144

‎.devops/s390x.Dockerfile‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -133,6 +133,7 @@ ENTRYPOINT [ "/llama.cpp/bin/llama-cli" ]
133133
FROM base AS server
134134

135135
ENV LLAMA_ARG_HOST=0.0.0.0
136+
ENV LLAMA_ARG_PORT=8080
136137

137138
WORKDIR /llama.cpp/bin
138139

‎.devops/vulkan.Dockerfile‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -107,7 +107,7 @@ ENTRYPOINT ["/app/tools.sh"]
107107
### Light, CLI only
108108
FROM base AS light
109109

110-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
110+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
111111

112112
WORKDIR /app
113113

@@ -117,8 +117,9 @@ ENTRYPOINT [ "/app/llama-cli" ]
117117
FROM base AS server
118118

119119
ENV LLAMA_ARG_HOST=0.0.0.0
120+
ENV LLAMA_ARG_PORT=8080
120121

121-
COPY --from=build /app/full/llama /app/full/llama-server /app
122+
COPY --from=build /app/full/llama /app/full/llama-server /app/
122123

123124
WORKDIR /app
124125

‎.devops/zendnn.Dockerfile‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -97,7 +97,7 @@ ENTRYPOINT ["/app/tools.sh"]
9797
### Light, CLI only
9898
FROM base AS light
9999

100-
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app
100+
COPY --from=build /app/full/llama /app/full/llama-cli /app/full/llama-completion /app/
101101

102102
WORKDIR /app
103103

@@ -107,8 +107,9 @@ ENTRYPOINT [ "/app/llama-cli" ]
107107
FROM base AS server
108108

109109
ENV LLAMA_ARG_HOST=0.0.0.0
110+
ENV LLAMA_ARG_PORT=8080
110111

111-
COPY --from=build /app/full/llama /app/full/llama-server /app
112+
COPY --from=build /app/full/llama /app/full/llama-server /app/
112113

113114
WORKDIR /app
114115

0 commit comments

Comments
 (0)