mirror of
https://github.com/ggerganov/whisper.cpp.git
synced 2024-12-18 20:27:53 +00:00
ggml : disable iq4_nl interleave size 8 (llama/10709)
ggml-ci
This commit is contained in:
parent
e990d1b791
commit
d0a050b51f
@ -3748,16 +3748,18 @@ static block_iq4_nlx4 make_block_iq4_nlx4(block_iq4_nl * in, unsigned int blck_s
|
||||
|
||||
const int end = QK4_NL * 2 / blck_size_interleave;
|
||||
|
||||
if (blck_size_interleave == 8) {
|
||||
for (int i = 0; i < end; ++i) {
|
||||
int src_id = i % 4;
|
||||
int src_offset = (i / 4) * blck_size_interleave;
|
||||
int dst_offset = i * blck_size_interleave;
|
||||
// TODO: this branch seems wrong
|
||||
//if (blck_size_interleave == 8) {
|
||||
// for (int i = 0; i < end; ++i) {
|
||||
// int src_id = i % 4;
|
||||
// int src_offset = (i / 4) * blck_size_interleave;
|
||||
// int dst_offset = i * blck_size_interleave;
|
||||
|
||||
// Using memcpy to avoid unaligned memory accesses
|
||||
memcpy(&out.qs[dst_offset], &in[src_id].qs[src_offset], sizeof(uint64_t));
|
||||
}
|
||||
} else if (blck_size_interleave == 4) {
|
||||
// // Using memcpy to avoid unaligned memory accesses
|
||||
// memcpy(&out.qs[dst_offset], &in[src_id].qs[src_offset], sizeof(uint64_t));
|
||||
// }
|
||||
//} else
|
||||
if (blck_size_interleave == 4) {
|
||||
for (int i = 0; i < end; ++i) {
|
||||
int src_id = i % 4;
|
||||
int src_offset = (i / 4) * blck_size_interleave;
|
||||
@ -3774,7 +3776,8 @@ static block_iq4_nlx4 make_block_iq4_nlx4(block_iq4_nl * in, unsigned int blck_s
|
||||
|
||||
static int repack_iq4_nl_to_iq4_nl_4_bl(struct ggml_tensor * t, int interleave_block, const void * GGML_RESTRICT data, size_t data_size) {
|
||||
GGML_ASSERT(t->type == GGML_TYPE_IQ4_NL);
|
||||
GGML_ASSERT(interleave_block == 4 || interleave_block == 8);
|
||||
//GGML_ASSERT(interleave_block == 4 || interleave_block == 8);
|
||||
GGML_ASSERT(interleave_block == 4);
|
||||
|
||||
block_iq4_nlx4 * dst = (block_iq4_nlx4 *)t->data;
|
||||
const block_iq4_nl * src = (const block_iq4_nl *)data;
|
||||
@ -3825,9 +3828,10 @@ template <> int repack<block_iq4_nl, 4, 4>(struct ggml_tensor * t, const void *
|
||||
return repack_iq4_nl_to_iq4_nl_4_bl(t, 4, data, data_size);
|
||||
}
|
||||
|
||||
template <> int repack<block_iq4_nl, 8, 4>(struct ggml_tensor * t, const void * data, size_t data_size) {
|
||||
return repack_iq4_nl_to_iq4_nl_4_bl(t, 8, data, data_size);
|
||||
}
|
||||
// TODO: needs to be revisited
|
||||
//template <> int repack<block_iq4_nl, 8, 4>(struct ggml_tensor * t, const void * data, size_t data_size) {
|
||||
// return repack_iq4_nl_to_iq4_nl_4_bl(t, 8, data, data_size);
|
||||
//}
|
||||
|
||||
// gemv
|
||||
template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS>
|
||||
|
Loading…
Reference in New Issue
Block a user