Compare commits

...
Sign in to create a new pull request.

3 commits

Author SHA1 Message Date
Georgi Gerganov
4fbb801a9d
ggml : update ggml_backend_cpu_device_supports_op
ggml-ci
2024-12-17 18:09:02 +02:00
Georgi Gerganov
8cc7145cc7
ggml : disable tests involving i-matrix quantization 2024-12-17 18:03:47 +02:00
Georgi Gerganov
b0597b1493
ggml : fix cpy op for IQ-quants to use reference impl
ggml-ci
2024-12-17 17:54:04 +02:00
2 changed files with 5 additions and 2 deletions

View file

@ -394,8 +394,11 @@ static bool ggml_backend_cpu_device_supports_op(ggml_backend_dev_t dev, const st
switch (op->op) {
case GGML_OP_CPY:
return
op->type != GGML_TYPE_IQ3_XXS &&
op->type != GGML_TYPE_IQ3_S &&
op->type != GGML_TYPE_IQ2_XXS &&
op->type != GGML_TYPE_IQ2_XS &&
op->type != GGML_TYPE_IQ2_S &&
op->type != GGML_TYPE_IQ1_S &&
op->type != GGML_TYPE_IQ1_M; // missing type_traits.from_float
case GGML_OP_MUL_MAT:

View file

@ -3549,8 +3549,8 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
for (ggml_type type_src : {GGML_TYPE_F16, GGML_TYPE_F32}) {
for (ggml_type type_dst : all_types) {
test_cases.emplace_back(new test_cpy(type_src, type_dst, {256, 4, 4, 4}));
test_cases.emplace_back(new test_cpy(type_src, type_dst, {256, 2, 3, 4}, {0, 2, 1, 3})); // cpy by rows
test_cases.emplace_back(new test_cpy(type_src, type_dst, {256, 4, 4, 4}));
test_cases.emplace_back(new test_cpy(type_src, type_dst, {256, 2, 3, 4}, {0, 2, 1, 3})); // cpy by rows
}
}
for (ggml_type type_src : {GGML_TYPE_F16, GGML_TYPE_F32}) {