llama.cpp/examples/imatrix
Liu Jia 0a4ce78681
common : Changed tuple to struct (TODO fix) (#8823)
* common : Changed tuple to struct (TODO fix)

Use struct `llama_init_result` to replace the previous
std::tuple<struct llama_model *, struct llama_context *>

* delete llama_init_default_params()

* delete the extra whitespace
2024-08-05 18:14:10 +02:00
..
CMakeLists.txt build: rename main → llama-cli, server → llama-server, llava-cli → llama-llava-cli, etc... (#7809) 2024-06-13 00:41:52 +01:00
imatrix.cpp common : Changed tuple to struct (TODO fix) (#8823) 2024-08-05 18:14:10 +02:00
README.md docs : Quantum -> Quantized (#8666) 2024-07-25 11:13:27 +03:00

llama.cpp/examples/imatrix

Compute an importance matrix for a model and given text dataset. Can be used during quantization to enchance the quality of the quantized models. More information is available here: https://github.com/ggerganov/llama.cpp/pull/4861

Usage

./llama-imatrix \
    -m model.gguf -f some-text.txt [-o imatrix.dat] [--process-output] [--verbosity 1] \
    [--no-ppl] [--chunk 123] [--output-frequency 10] [--save-frequency 0] \
    [--in-file imatrix-prev-0.dat --in-file imatrix-prev-1.dat ...]

Here -m with a model name and -f with a file containing training data (such as e.g. wiki.train.raw) are mandatory. The parameters in square brackets are optional and have the following meaning:

  • -o (or --output-file) specifies the name of the file where the computed data will be stored. If missing imatrix.dat is used.
  • --verbosity specifies the verbosity level. If set to 0, no output other than the perplexity of the processed chunks will be generated. If set to 1, each time the results are saved a message is written to stderr. If >=2, a message is output each time data is collected for any tensor. Default verbosity level is 1.
  • --output-frequency specifies how often the so far computed result is saved to disk. Default is 10 (i.e., every 10 chunks)
  • --save-frequency specifies how often to save a copy of the imatrix in a separate file. Default is 0 (i.e., never)
  • --process-output specifies if data will be collected for the output.weight tensor. My experience is that it is better to not utilize the importance matrix when quantizing output.weight, so this is set to false by default.

For faster computation, make sure to use GPU offloading via the -ngl argument

Example

GGML_CUDA=1 make -j

# generate importance matrix (imatrix.dat)
./llama-imatrix -m ggml-model-f16.gguf -f train-data.txt -ngl 99

# use the imatrix to perform a Q4_K_M quantization
./llama-quantize --imatrix imatrix.dat ggml-model-f16.gguf ./ggml-model-q4_k_m.gguf q4_k_m