Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

llama : remove MPI backend #7395

Merged
merged 1 commit into from
May 19, 2024
Merged

llama : remove MPI backend #7395

merged 1 commit into from
May 19, 2024

Conversation

slaren
Copy link
Collaborator

@slaren slaren commented May 19, 2024

Removes the MPI backend in favor of the RPC backend. The MPI backend has not been functional for a very long time, and its functionality can be obtained with the RPC backend.

For more details about the RPC backend, check #6829 and https://github.com/ggerganov/llama.cpp/tree/master/examples/rpc.

@github-actions github-actions bot added build Compilation issues script Script related nix Issues specific to consuming flake.nix, or generally concerned with ❄ Nix-based llama.cpp deployment devops improvements to build systems and github actions ggml changes relating to the ggml tensor library for machine learning labels May 19, 2024
@slaren slaren merged commit d359f30 into master May 19, 2024
76 checks passed
@slaren slaren deleted the sl/remove-mpi branch May 19, 2024 23:17
Copy link
Contributor

📈 llama.cpp server for bench-server-baseline on Standard_NC4as_T4_v3 for phi-2-q4_0: 548 iterations 🚀

Expand details for performance related PR only
  • Concurrent users: 8, duration: 10m
  • HTTP request : avg=8532.78ms p(95)=20422.01ms fails=, finish reason: stop=496 truncated=52
  • Prompt processing (pp): avg=91.33tk/s p(95)=333.81tk/s
  • Token generation (tg): avg=34.4tk/s p(95)=46.13tk/s
  • ggml-org/models/phi-2/ggml-model-q4_0.gguf parallel=8 ctx-size=16384 ngl=33 batch-size=2048 ubatch-size=256 pp=1024 pp+tg=2048 branch=sl/remove-mpi commit=78cded5394c2ada8f119ce75d674b937372e7dd1

prompt_tokens_seconds

More
---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 548 iterations"
    y-axis "llamacpp:prompt_tokens_seconds"
    x-axis "llamacpp:prompt_tokens_seconds" 1716161804 --> 1716162428
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 542.56, 542.56, 542.56, 542.56, 542.56, 754.17, 754.17, 754.17, 754.17, 754.17, 722.31, 722.31, 722.31, 722.31, 722.31, 751.1, 751.1, 751.1, 751.1, 751.1, 823.46, 823.46, 823.46, 823.46, 823.46, 816.0, 816.0, 816.0, 816.0, 816.0, 828.16, 828.16, 828.16, 828.16, 828.16, 845.95, 845.95, 845.95, 845.95, 845.95, 843.38, 843.38, 843.38, 843.38, 843.38, 830.18, 830.18, 830.18, 830.18, 830.18, 847.61, 847.61, 847.61, 847.61, 847.61, 843.31, 843.31, 843.31, 843.31, 843.31, 839.65, 839.65, 839.65, 839.65, 839.65, 846.18, 846.18, 846.18, 846.18, 846.18, 825.1, 825.1, 825.1, 825.1, 825.1, 838.5, 838.5, 838.5, 838.5, 838.5, 836.69, 836.69, 836.69, 836.69, 836.69, 844.15, 844.15, 844.15, 844.15, 844.15, 856.78, 856.78, 856.78, 856.78, 856.78, 858.95, 858.95, 858.95, 858.95, 858.95, 858.17, 858.17, 858.17, 858.17, 858.17, 865.97, 865.97, 865.97, 865.97, 865.97, 868.19, 868.19, 868.19, 868.19, 868.19, 864.5, 864.5, 864.5, 864.5, 864.5, 863.72, 863.72, 863.72, 863.72, 863.72, 865.37, 865.37, 865.37, 865.37, 865.37, 837.68, 837.68, 837.68, 837.68, 837.68, 837.55, 837.55, 837.55, 837.55, 837.55, 837.98, 837.98, 837.98, 837.98, 837.98, 843.5, 843.5, 843.5, 843.5, 843.5, 841.41, 841.41, 841.41, 841.41, 841.41, 842.51, 842.51, 842.51, 842.51, 842.51, 848.82, 848.82, 848.82, 848.82, 848.82, 836.74, 836.74, 836.74, 836.74, 836.74, 836.6, 836.6, 836.6, 836.6, 836.6, 833.12, 833.12, 833.12, 833.12, 833.12, 832.22, 832.22, 832.22, 832.22, 832.22, 831.27, 831.27, 831.27, 831.27, 831.27, 833.12, 833.12, 833.12, 833.12, 833.12, 835.9, 835.9, 835.9, 835.9, 835.9, 836.63, 836.63, 836.63, 836.63, 836.63, 841.22, 841.22, 841.22, 841.22, 841.22, 846.95, 846.95, 846.95, 846.95, 846.95, 845.48, 845.48, 845.48, 845.48, 845.48, 844.24, 844.24, 844.24, 844.24, 844.24, 844.3, 844.3, 844.3, 844.3, 844.3, 848.61, 848.61, 848.61, 848.61, 848.61, 847.44, 847.44, 847.44, 847.44, 847.44, 850.35, 850.35, 850.35, 850.35, 850.35, 851.69, 851.69, 851.69, 851.69, 851.69, 854.04, 854.04, 854.04, 854.04, 854.04, 857.27, 857.27, 857.27, 857.27, 857.27, 852.77, 852.77, 852.77, 852.77, 852.77, 854.27, 854.27, 854.27, 854.27, 854.27, 854.47, 854.47, 854.47, 854.47, 854.47, 854.09, 854.09, 854.09, 854.09, 854.09, 854.05, 854.05, 854.05, 854.05, 854.05, 853.99, 853.99, 853.99, 853.99, 853.99, 855.98, 855.98, 855.98, 855.98, 855.98, 858.93, 858.93, 858.93, 858.93, 858.93, 858.62, 858.62, 858.62]
                    
Loading
predicted_tokens_seconds
More
---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 548 iterations"
    y-axis "llamacpp:predicted_tokens_seconds"
    x-axis "llamacpp:predicted_tokens_seconds" 1716161804 --> 1716162428
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 40.52, 40.52, 40.52, 40.52, 40.52, 37.27, 37.27, 37.27, 37.27, 37.27, 28.86, 28.86, 28.86, 28.86, 28.86, 31.29, 31.29, 31.29, 31.29, 31.29, 31.76, 31.76, 31.76, 31.76, 31.76, 32.54, 32.54, 32.54, 32.54, 32.54, 34.01, 34.01, 34.01, 34.01, 34.01, 34.56, 34.56, 34.56, 34.56, 34.56, 34.37, 34.37, 34.37, 34.37, 34.37, 34.3, 34.3, 34.3, 34.3, 34.3, 34.37, 34.37, 34.37, 34.37, 34.37, 34.23, 34.23, 34.23, 34.23, 34.23, 32.96, 32.96, 32.96, 32.96, 32.96, 32.59, 32.59, 32.59, 32.59, 32.59, 31.98, 31.98, 31.98, 31.98, 31.98, 31.5, 31.5, 31.5, 31.5, 31.5, 31.04, 31.04, 31.04, 31.04, 31.04, 31.17, 31.17, 31.17, 31.17, 31.17, 30.99, 30.99, 30.99, 30.99, 30.99, 30.86, 30.86, 30.86, 30.86, 30.86, 30.83, 30.83, 30.83, 30.83, 30.83, 30.84, 30.84, 30.84, 30.84, 30.84, 31.06, 31.06, 31.06, 31.06, 31.06, 31.06, 31.06, 31.06, 31.06, 31.06, 30.94, 30.94, 30.94, 30.94, 30.94, 31.01, 31.01, 31.01, 31.01, 31.01, 31.25, 31.25, 31.25, 31.25, 31.25, 31.28, 31.28, 31.28, 31.28, 31.28, 31.54, 31.54, 31.54, 31.54, 31.54, 31.72, 31.72, 31.72, 31.72, 31.72, 31.78, 31.78, 31.78, 31.78, 31.78, 32.01, 32.01, 32.01, 32.01, 32.01, 32.15, 32.15, 32.15, 32.15, 32.15, 31.9, 31.9, 31.9, 31.9, 31.9, 31.78, 31.78, 31.78, 31.78, 31.78, 31.75, 31.75, 31.75, 31.75, 31.75, 31.39, 31.39, 31.39, 31.39, 31.39, 31.26, 31.26, 31.26, 31.26, 31.26, 31.35, 31.35, 31.35, 31.35, 31.35, 31.5, 31.5, 31.5, 31.5, 31.5, 31.62, 31.62, 31.62, 31.62, 31.62, 31.62, 31.62, 31.62, 31.62, 31.62, 31.4, 31.4, 31.4, 31.4, 31.4, 31.05, 31.05, 31.05, 31.05, 31.05, 30.72, 30.72, 30.72, 30.72, 30.72, 29.74, 29.74, 29.74, 29.74, 29.74, 29.64, 29.64, 29.64, 29.64, 29.64, 29.57, 29.57, 29.57, 29.57, 29.57, 29.61, 29.61, 29.61, 29.61, 29.61, 29.68, 29.68, 29.68, 29.68, 29.68, 29.75, 29.75, 29.75, 29.75, 29.75, 29.71, 29.71, 29.71, 29.71, 29.71, 29.73, 29.73, 29.73, 29.73, 29.73, 29.63, 29.63, 29.63, 29.63, 29.63, 29.6, 29.6, 29.6, 29.6, 29.6, 29.61, 29.61, 29.61, 29.61, 29.61, 29.8, 29.8, 29.8, 29.8, 29.8, 29.87, 29.87, 29.87, 29.87, 29.87, 29.98, 29.98, 29.98, 29.98, 29.98, 30.04, 30.04, 30.04, 30.04, 30.04, 30.06, 30.06, 30.06]
                    
Loading

Details

kv_cache_usage_ratio

More
---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 548 iterations"
    y-axis "llamacpp:kv_cache_usage_ratio"
    x-axis "llamacpp:kv_cache_usage_ratio" 1716161804 --> 1716162428
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.13, 0.13, 0.13, 0.13, 0.13, 0.29, 0.29, 0.29, 0.29, 0.29, 0.13, 0.13, 0.13, 0.13, 0.13, 0.15, 0.15, 0.15, 0.15, 0.15, 0.16, 0.16, 0.16, 0.16, 0.16, 0.14, 0.14, 0.14, 0.14, 0.14, 0.13, 0.13, 0.13, 0.13, 0.13, 0.2, 0.2, 0.2, 0.2, 0.2, 0.19, 0.19, 0.19, 0.19, 0.19, 0.16, 0.16, 0.16, 0.16, 0.16, 0.24, 0.24, 0.24, 0.24, 0.24, 0.18, 0.18, 0.18, 0.18, 0.18, 0.24, 0.24, 0.24, 0.24, 0.24, 0.31, 0.31, 0.31, 0.31, 0.31, 0.36, 0.36, 0.36, 0.36, 0.36, 0.3, 0.3, 0.3, 0.3, 0.3, 0.13, 0.13, 0.13, 0.13, 0.13, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.29, 0.29, 0.29, 0.29, 0.29, 0.17, 0.17, 0.17, 0.17, 0.17, 0.16, 0.16, 0.16, 0.16, 0.16, 0.15, 0.15, 0.15, 0.15, 0.15, 0.33, 0.33, 0.33, 0.33, 0.33, 0.09, 0.09, 0.09, 0.09, 0.09, 0.15, 0.15, 0.15, 0.15, 0.15, 0.27, 0.27, 0.27, 0.27, 0.27, 0.12, 0.12, 0.12, 0.12, 0.12, 0.11, 0.11, 0.11, 0.11, 0.11, 0.17, 0.17, 0.17, 0.17, 0.17, 0.14, 0.14, 0.14, 0.14, 0.14, 0.16, 0.16, 0.16, 0.16, 0.16, 0.17, 0.17, 0.17, 0.17, 0.17, 0.25, 0.25, 0.25, 0.25, 0.25, 0.17, 0.17, 0.17, 0.17, 0.17, 0.3, 0.3, 0.3, 0.3, 0.3, 0.22, 0.22, 0.22, 0.22, 0.22, 0.16, 0.16, 0.16, 0.16, 0.16, 0.14, 0.14, 0.14, 0.14, 0.14, 0.12, 0.12, 0.12, 0.12, 0.12, 0.11, 0.11, 0.11, 0.11, 0.11, 0.38, 0.38, 0.38, 0.38, 0.38, 0.58, 0.58, 0.58, 0.58, 0.58, 0.52, 0.52, 0.52, 0.52, 0.52, 0.48, 0.48, 0.48, 0.48, 0.48, 0.17, 0.17, 0.17, 0.17, 0.17, 0.28, 0.28, 0.28, 0.28, 0.28, 0.24, 0.24, 0.24, 0.24, 0.24, 0.14, 0.14, 0.14, 0.14, 0.14, 0.17, 0.17, 0.17, 0.17, 0.17, 0.26, 0.26, 0.26, 0.26, 0.26, 0.21, 0.21, 0.21, 0.21, 0.21, 0.22, 0.22, 0.22, 0.22, 0.22, 0.26, 0.26, 0.26, 0.26, 0.26, 0.15, 0.15, 0.15, 0.15, 0.15, 0.2, 0.2, 0.2, 0.2, 0.2, 0.15, 0.15, 0.15, 0.15, 0.15, 0.1, 0.1, 0.1, 0.1, 0.1, 0.12, 0.12, 0.12, 0.12, 0.12, 0.24, 0.24, 0.24, 0.24, 0.24, 0.23, 0.23, 0.23]
                    
Loading
requests_processing
More
---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 548 iterations"
    y-axis "llamacpp:requests_processing"
    x-axis "llamacpp:requests_processing" 1716161804 --> 1716162428
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 4.0, 4.0, 4.0, 4.0, 4.0, 1.0, 1.0, 1.0, 1.0, 1.0, 3.0, 3.0, 3.0, 3.0, 3.0, 6.0, 6.0, 6.0, 6.0, 6.0, 5.0, 5.0, 5.0, 5.0, 5.0, 7.0, 7.0, 7.0, 7.0, 7.0, 2.0, 2.0, 2.0, 2.0, 2.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 6.0, 6.0, 6.0, 6.0, 6.0, 4.0, 4.0, 4.0, 4.0, 4.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 7.0, 6.0, 6.0, 6.0, 6.0, 6.0, 4.0, 4.0, 4.0, 4.0, 4.0, 5.0, 5.0, 5.0, 5.0, 5.0, 6.0, 6.0, 6.0, 6.0, 6.0, 5.0, 5.0, 5.0, 5.0, 5.0, 3.0, 3.0, 3.0, 3.0, 3.0, 7.0, 7.0, 7.0, 7.0, 7.0, 5.0, 5.0, 5.0, 5.0, 5.0, 7.0, 7.0, 7.0, 7.0, 7.0, 6.0, 6.0, 6.0, 6.0, 6.0, 5.0, 5.0, 5.0, 5.0, 5.0, 4.0, 4.0, 4.0, 4.0, 4.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 7.0, 7.0, 7.0, 7.0, 7.0, 6.0, 6.0, 6.0, 6.0, 6.0, 7.0, 7.0, 7.0, 7.0, 7.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 7.0, 7.0, 7.0, 7.0, 7.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 7.0, 7.0, 7.0, 7.0, 7.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 2.0, 2.0, 2.0, 2.0, 2.0, 5.0, 5.0, 5.0, 5.0, 5.0, 3.0, 3.0, 3.0, 3.0, 3.0, 7.0, 7.0, 7.0, 7.0, 7.0, 6.0, 6.0, 6.0, 6.0, 6.0, 5.0, 5.0, 5.0, 5.0, 5.0, 3.0, 3.0, 3.0, 3.0, 3.0, 5.0, 5.0, 5.0, 5.0, 5.0, 3.0, 3.0, 3.0, 3.0, 3.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 4.0, 4.0, 4.0, 4.0, 4.0, 2.0, 2.0, 2.0]
                    
Loading

@mofosyne
Copy link
Collaborator

mofosyne commented May 20, 2024

CI failed here https://github.com/ggml-org/ci/tree/results/llama.cpp/d3/59f30921a9f62a0fd299c412ff3f270286fea6/ggml-4-x86-cuda-v100

Failed on

 20 - test-backend-ops (Failed)

edit: Ah i see this particular ci fail is due to this new sanatiser run

@mofosyne
Copy link
Collaborator

CI issue fixed by #7409

maxstrid added a commit to maxstrid/nixpkgs that referenced this pull request May 21, 2024
llama-cpp no longer supports mpi and rcp is the recommended alternative.
See: ggerganov/llama.cpp#7395

Signed-off-by: Maxwell Henderson <mxwhenderson@gmail.com>
maxstrid added a commit to maxstrid/nixpkgs that referenced this pull request May 22, 2024
llama-cpp no longer supports mpi and rpc is the recommended alternative.
See: ggerganov/llama.cpp#7395

Signed-off-by: Maxwell Henderson <mxwhenderson@gmail.com>
maxstrid added a commit to maxstrid/nixpkgs that referenced this pull request May 22, 2024
llama-cpp no longer supports mpi and rpc is the recommended alternative.
See: ggerganov/llama.cpp#7395

Signed-off-by: Maxwell Henderson <mxwhenderson@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
build Compilation issues devops improvements to build systems and github actions ggml changes relating to the ggml tensor library for machine learning nix Issues specific to consuming flake.nix, or generally concerned with ❄ Nix-based llama.cpp deployment script Script related
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants