happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update usage()

This commit is contained in:
Ed Addario 2026-01-07 18:32:01 +00:00
parent d18ddbac9c
commit 26213bc805
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
tools/quantize/quantize.cpp
View File

@ -119,7 +119,7 @@ static bool try_parse_ftype(const std::string & ftype_str_in, llama_ftype & ftyp
[[noreturn]]
static void usage(const char * executable) {
printf("usage: %s [--help] [--allow-requantize] [--leave-output-tensor] [--pure] [--imatrix] [--include-weights] [--exclude-weights]\n", executable);
printf(" [--target-bpw n] [--target-size n] [--no-importance] [--keep-bpw-state] [--bpw-state filename] [--output-tensor-type] [--token-embedding-type]\n");
printf(" [--target-bpw n] [--target-size n] [--no-importance] [--save-state] [--state-file filename] [--output-tensor-type] [--token-embedding-type]\n");
printf(" [--tensor-type] [--prune-layers] [--keep-split] [--override-kv] model-f32.gguf [model-quant.gguf] type [nthreads]\n\n");
printf(" --allow-requantize: allows requantizing tensors that have already been quantized. Warning: This can severely reduce quality compared to quantizing from 16bit or 32bit\n");
printf(" --leave-output-tensor: will leave output.weight un(re)quantized. Increases model size but may also increase quality, especially when requantizing\n");
@ -133,14 +133,14 @@ static void usage(const char * executable) {
printf(" Advanced option to selectively quantize tensors. May be specified multiple times.\n");
printf(" --prune-layers L0,L1,L2...comma-separated list of layer numbers to prune from the model\n");
printf(" Advanced option to remove all tensors from the given layers\n");
printf(" --target-bpw: target a total bits per weight (bpw). Must be a positive number between 0.0 and 16.0\n");
printf(" --target-bpw N: target a total bits per weight (bpw). N must be a positive number between 0.0 and 16.0\n");
printf(" Advanced option to automatically select quantization types to achieve a total bits per weight (bpw) target\n");
printf(" --target-size: target a file size. Must be a positive number\n");
printf(" --target-size N[unit]: target a file size. N must be a positive number with an optional unit (b, kb, mb, gb, tb)\n");
printf(" Advanced option to automatically select quantization types to achieve a target file size\n");
printf(" --no-importance: distribute bpw budget equitably across all tensors\n");
printf(" --ignore-tensor-importance: distribute bpw budget equitably across all tensors\n");
printf(" Advanced option to disable assigning more bpw budget to important tensors. It may increase quality for some models\n");
printf(" --keep-bpw-state: save the bpw computations to <model name>-<model hash>.bpw_state\n");
printf(" --bpw-state: file name to use instead of default\n");
printf(" --save-state: save the bpw / file size computations to <model name>-<model hash>-mse.bpw_state\n");
printf(" --state-file file_name: file name to use instead of default\n");
printf(" --keep-split: will generate quantized model in the same shards as input\n");
printf(" --override-kv KEY=TYPE:VALUE\n");
printf(" Advanced option to override model metadata by key in the quantized model. May be specified multiple times.\n");