char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Georgi Gerganovtests : add fusion baseline README and broaden fusion CI triggers (#28893)bbdd9f246

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
921 B26 linesraw

Fusion baselines

Per-device baselines for test-fusion, one CSV per backend (e.g. MTL.csv). Rows are arch,moe,mode,label,count. Regenerate a CSV whenever fusion patterns change.

Update a baseline

cmake -B build -DCMAKE_BUILD_TYPE=Release -DGGML_METAL=ON   # enable the target backend
cmake --build build --config Release --target test-llama-archs --target test-fusion -j

rm -rf build-ci-models && mkdir -p build-ci-models
./build/bin/test-llama-archs -o build-ci-models

./build/bin/test-fusion --models build-ci-models --device MTL0 --record MTL.csv

Validate

./build/bin/test-fusion --models build-ci-models --device MTL0 --check MTL.csv

Non-zero exit means a row differs from the baseline. Use --model FILE to run a single architecture. Note --check only sees present rows — a fusion that stops matching is not reported, so diff the recorded CSV to catch removed patterns.

1# Fusion baselines
2
3Per-device baselines for `test-fusion`, one CSV per backend (e.g. `MTL.csv`). Rows are
4`arch,moe,mode,label,count`. Regenerate a CSV whenever fusion patterns change.
5
6## Update a baseline
7
8```sh
9cmake -B build -DCMAKE_BUILD_TYPE=Release -DGGML_METAL=ON   # enable the target backend
10cmake --build build --config Release --target test-llama-archs --target test-fusion -j
11
12rm -rf build-ci-models && mkdir -p build-ci-models
13./build/bin/test-llama-archs -o build-ci-models
14
15./build/bin/test-fusion --models build-ci-models --device MTL0 --record MTL.csv
16```
17
18## Validate
19
20```sh
21./build/bin/test-fusion --models build-ci-models --device MTL0 --check MTL.csv
22```
23
24Non-zero exit means a row differs from the baseline. Use `--model FILE` to run a single
25architecture. Note `--check` only sees present rows — a fusion that stops matching is not
26reported, so diff the recorded CSV to catch removed patterns.