ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

CUDA: MMQ code deduplication + iquant support (#8495)

Browse source 
* CUDA: MMQ code deduplication + iquant support

* 1 less parallel job for CI build
This commit is contained in:
Johannes Gäßler 2024-07-20 22:25:26 +02:00 • committed by GitHub
parent 07283b1a90
commit 69c487f4ed
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
11 changed files with 800 additions and 639 deletions
Download patch file Download diff file Expand all files Collapse all files

24
ggml/src/ggml-cuda/mmq.cu
View file

@ -59,6 +59,24 @@ void ggml_cuda_op_mul_mat_q(
case GGML_TYPE_Q6_K:
mul_mat_q_case<GGML_TYPE_Q6_K>(ctx, args, stream);
break;
case GGML_TYPE_IQ2_XXS:
mul_mat_q_case<GGML_TYPE_IQ2_XXS>(ctx, args, stream);
break;
case GGML_TYPE_IQ2_XS:
mul_mat_q_case<GGML_TYPE_IQ2_XS>(ctx, args, stream);
break;
case GGML_TYPE_IQ2_S:
mul_mat_q_case<GGML_TYPE_IQ2_S>(ctx, args, stream);
break;
case GGML_TYPE_IQ3_XXS:
mul_mat_q_case<GGML_TYPE_IQ3_XXS>(ctx, args, stream);
break;
case GGML_TYPE_IQ3_S:
mul_mat_q_case<GGML_TYPE_IQ3_S>(ctx, args, stream);
break;
case GGML_TYPE_IQ1_S:
mul_mat_q_case<GGML_TYPE_IQ1_S>(ctx, args, stream);
break;
case GGML_TYPE_IQ4_XS:
mul_mat_q_case<GGML_TYPE_IQ4_XS>(ctx, args, stream);
break;
@ -93,6 +111,12 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11) {
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ2_S:
case GGML_TYPE_IQ3_XXS:
case GGML_TYPE_IQ3_S:
case GGML_TYPE_IQ1_S:
case GGML_TYPE_IQ4_XS:
case GGML_TYPE_IQ4_NL:
mmq_supported = true;