vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

iq3_xxs: fix failing quantization test

Browse source 
Dot product still fails. Is this real?
This commit is contained in:
Iwan Kawrakow 2024-01-29 16:33:03 +02:00
parent 7e4e7488ae
commit 6efbc690dc
1 changed files with 6 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

8
tests/test-quantize-fns.cpp
View file

@ -17,6 +17,7 @@ constexpr float MAX_QUANTIZATION_REFERENCE_ERROR = 0.0001f;
constexpr float MAX_QUANTIZATION_TOTAL_ERROR = 0.002f; constexpr float MAX_QUANTIZATION_TOTAL_ERROR = 0.002f;
constexpr float MAX_QUANTIZATION_TOTAL_ERROR_2BITS = 0.0075f; constexpr float MAX_QUANTIZATION_TOTAL_ERROR_2BITS = 0.0075f;
constexpr float MAX_QUANTIZATION_TOTAL_ERROR_3BITS = 0.0040f; constexpr float MAX_QUANTIZATION_TOTAL_ERROR_3BITS = 0.0040f;
constexpr float MAX_QUANTIZATION_TOTAL_ERROR_3BITS_XXS = 0.0050f;
constexpr float MAX_DOT_PRODUCT_ERROR = 0.02f; constexpr float MAX_DOT_PRODUCT_ERROR = 0.02f;
static const char* RESULT_STR[] = {"ok", "FAILED"}; static const char* RESULT_STR[] = {"ok", "FAILED"};
@ -135,18 +136,21 @@ int main(int argc, char * argv[]) {
} }
const ggml_type ei = (ggml_type)i; const ggml_type ei = (ggml_type)i;
if (ei == GGML_TYPE_IQ2_XXS || ei == GGML_TYPE_IQ2_XS) { if (ei == GGML_TYPE_IQ2_XXS || ei == GGML_TYPE_IQ2_XS) {
printf("Skip %s due to missing quantization functionality\n", ggml_type_name(ei)); printf("Skip %s due to missing quantization functionality\n", ggml_type_name(ei));
continue; continue;
} }
printf("Testing %s\n", ggml_type_name((ggml_type) i)); printf("Testing %s\n", ggml_type_name((ggml_type) i));
ggml_quantize_init(ei);
if (qfns.from_float && qfns.to_float) { if (qfns.from_float && qfns.to_float) {
const float total_error = total_quantization_error(qfns, test_size, test_data.data()); const float total_error = total_quantization_error(qfns, test_size, test_data.data());
const float max_quantization_error = const float max_quantization_error =
type == GGML_TYPE_Q2_K ? MAX_QUANTIZATION_TOTAL_ERROR_2BITS : type == GGML_TYPE_Q2_K ? MAX_QUANTIZATION_TOTAL_ERROR_2BITS :
type == GGML_TYPE_Q3_K ? MAX_QUANTIZATION_TOTAL_ERROR_3BITS : MAX_QUANTIZATION_TOTAL_ERROR; type == GGML_TYPE_Q3_K ? MAX_QUANTIZATION_TOTAL_ERROR_3BITS :
type == GGML_TYPE_IQ3_XXS ? MAX_QUANTIZATION_TOTAL_ERROR_3BITS_XXS : MAX_QUANTIZATION_TOTAL_ERROR;
failed = !(total_error < max_quantization_error); failed = !(total_error < max_quantization_error);
num_failed += failed; num_failed += failed;
if (failed || verbose) { if (failed || verbose) {