k-quants : support for super-block size of 64 (#2001)

* k_quants: WIP super-blocks with 64 weights

* k_quants: WIP super-blocks with 64 weights

Q6_K scalar and AVX2 works

* k_quants: WIP super-blocks with 64 weights

Q4_K scalar and AVX2 works

* k_quants: WIP super-blocks with 64 weights

Q2_K scalar and AVX2 works. Q2_K is way too slow (it is actually slower
than the scalar implementation)

* k_quants: WIP super-blocks with 64 weights

Q3_K scalar and AVX2 works.

* k_quants: WIP super-blocks with 64 weights

Q5_K scalar and AVX2 works, and with that all
k_quants are done on AVX2 and scalar

* k_quants: WIP super-blocks with 64 weights

Q6_K working on CUDA. Cannot make it run quite as gast as
with super-blocks with 256 weigths: 8% slower on 4080,
20% slower on the 1660 (but there we fit 1 less layer on the
GPU because pf the larger model size), so some fraction of
these 20% is due to that,

* k_quants: WIP super-blocks with 64 weights

Q4_K working on CUDA. ~10% slower on GTX-1660,
16% slower on 4080.

* k_quants: WIP super-blocks with 64 weights

Q2_K working on CUDA. ~3% slower on GTX-1660,
10% slower on 4080.

* k_quants: WIP super-blocks with 64 weights

Q3_K working on CUDA.

* k_quants: WIP super-blocks with 64 weights

Q5_K working on CUDA, and with this CUDA is done.

* k_quants: WIP super-blocks with 64 weights

Q6_K working on ARM_NEON

* k_quants: WIP super-blocks with 64 weights

Q4_K working on ARM_NEON, but quite a bit slower than 256 weights

* k_quants: WIP super-blocks with 64 weights

Q2_K working on ARM_NEON, but quite a bit slower than 256 weights

* k_quants: WIP super-blocks with 64 weights

Q3_K working on ARM_NEON, but quite a bit slower than 256 weights.

* k_quants: WIP super-blocks with 64 weights

Q5_K working on ARM_NEON, but quite a bit slower than 256 weights.

With that, we have full support for ARM_NEON, although
performance is not quite there.

* k_quants: WIP super-blocks with 64 weights

Slightly more efficient Q3_K and Q5_K

* k_quants: WIP super-blocks with 64 weights

Another small improvement for Q3_K and Q5_K on ARM_NEON

* k_quants: WIP super-blocks with 64 weights

Yet another speedup for Q5_K on ARM_NEON.
We are now within 10% of the QK_K = 256 version.

* k_quants: WIP super-blocks with 64 weights

* We are able to pass preprocessor macros to the Metal
  compiler
* Q6_K works and is actually slightly more efficient than
  the QK_K = 256 version (25.2 ms vs 25.8 ms)

* k_quants: WIP super-blocks with 64 weights

Q4_K works on Metal and is actually slightly faster
than QK_K = 256 (21.95 ms vs 24.0 ms).

* k_quants: WIP super-blocks with 64 weights

Q2_K works on Metal and is very slightly faster
than QK_K = 256 (23.8 ms vs 24.2 ms).

* k_quants: WIP super-blocks with 64 weights

Q3_K works on Metal and is slightly faster
than QK_K = 256 (26.6 ms vs 28.3 ms).

* k_quants: WIP super-blocks with 64 weights

Q5_K works on Metal and is slightly faster
than QK_K = 256 (23.7 ms vs 26.3 ms).

* k_quants: call them _K, not _k, also on Metal

* k_quants: correctly define QK_K in llama.cpp

* Fixed bug in q4_K quantization added with the 64-block addition

* Simplify via lambda

* k_quants: swicth Q3_K to 4-bit scales when QK_K = 64

Otherwise there isn't much benefit from this
quantization type. There is some very slight loss
in accuracy, but we reduce size by ~7%.
E.g., for OpenLLaMA-3B, Q3_K_S perplexity is
8.6131 with 8-bit scales and 8.6352 with 4-bit,
while file size decreases from 1.53G to 1.44G.

* k_quants: switch Q4_K to 4-bit scales when QK_K = 64

 Here the loss in accuracy is greater than for Q3_K,
 but the Q4_K points still move further to the left on
 the perplexity vs size curve.

* k_quants: forgot to add the Metal changes in last commit

* k_quants: change Q5_K to be type 0 when QK_K = 64

Still needs AVX2 implementation

* k_quants: AVX2 implementation for new 64-weight Q5_K

* k_quants: 10% faster ARM_NEON Q5_K dot product

* k_quants: fixed issue caused by merging with master

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

1 files changed, 36 insertions, 30 deletions

diff --git a/ggml-metal.m b/ggml-metal.m
index a7e104d..7551231 100644
--- a/ggml-metal.m
+++ b/ggml-metal.m
@@ -51,21 +51,21 @@ struct ggml_metal_context {
     GGML_METAL_DECL_KERNEL(get_rows_f16);
     GGML_METAL_DECL_KERNEL(get_rows_q4_0);
     GGML_METAL_DECL_KERNEL(get_rows_q4_1);
-    GGML_METAL_DECL_KERNEL(get_rows_q2_k);
-    GGML_METAL_DECL_KERNEL(get_rows_q3_k);
-    GGML_METAL_DECL_KERNEL(get_rows_q4_k);
-    GGML_METAL_DECL_KERNEL(get_rows_q5_k);
-    GGML_METAL_DECL_KERNEL(get_rows_q6_k);
+    GGML_METAL_DECL_KERNEL(get_rows_q2_K);
+    GGML_METAL_DECL_KERNEL(get_rows_q3_K);
+    GGML_METAL_DECL_KERNEL(get_rows_q4_K);
+    GGML_METAL_DECL_KERNEL(get_rows_q5_K);
+    GGML_METAL_DECL_KERNEL(get_rows_q6_K);
     GGML_METAL_DECL_KERNEL(rms_norm);
     GGML_METAL_DECL_KERNEL(norm);
     GGML_METAL_DECL_KERNEL(mul_mat_f16_f32);
     GGML_METAL_DECL_KERNEL(mul_mat_q4_0_f32);
     GGML_METAL_DECL_KERNEL(mul_mat_q4_1_f32);
-    GGML_METAL_DECL_KERNEL(mul_mat_q2_k_f32);
-    GGML_METAL_DECL_KERNEL(mul_mat_q3_k_f32);
-    GGML_METAL_DECL_KERNEL(mul_mat_q4_k_f32);
-    GGML_METAL_DECL_KERNEL(mul_mat_q5_k_f32);
-    GGML_METAL_DECL_KERNEL(mul_mat_q6_k_f32);
+    GGML_METAL_DECL_KERNEL(mul_mat_q2_K_f32);
+    GGML_METAL_DECL_KERNEL(mul_mat_q3_K_f32);
+    GGML_METAL_DECL_KERNEL(mul_mat_q4_K_f32);
+    GGML_METAL_DECL_KERNEL(mul_mat_q5_K_f32);
+    GGML_METAL_DECL_KERNEL(mul_mat_q6_K_f32);
     GGML_METAL_DECL_KERNEL(rope);
     GGML_METAL_DECL_KERNEL(alibi_f32);
     GGML_METAL_DECL_KERNEL(cpy_f32_f16);
@@ -132,7 +132,13 @@ struct ggml_metal_context * ggml_metal_init(void) {
             exit(1);
         }
 
+#ifdef GGML_QKK_64
+        MTLCompileOptions* options = [MTLCompileOptions new];
+        options.preprocessorMacros = @{ @"QK_K" : @(64) };
+        ctx->library = [ctx->device newLibraryWithSource:src options:options error:&error];
+#else
         ctx->library = [ctx->device newLibraryWithSource:src options:nil error:&error];
+#endif
         if (error) {
             fprintf(stderr, "%s: error: %s\n", __func__, [[error description] UTF8String]);
             exit(1);
@@ -159,21 +165,21 @@ struct ggml_metal_context * ggml_metal_init(void) {
         GGML_METAL_ADD_KERNEL(get_rows_f16);
         GGML_METAL_ADD_KERNEL(get_rows_q4_0);
         GGML_METAL_ADD_KERNEL(get_rows_q4_1);
-        GGML_METAL_ADD_KERNEL(get_rows_q2_k);
-        GGML_METAL_ADD_KERNEL(get_rows_q3_k);
-        GGML_METAL_ADD_KERNEL(get_rows_q4_k);
-        GGML_METAL_ADD_KERNEL(get_rows_q5_k);
-        GGML_METAL_ADD_KERNEL(get_rows_q6_k);
+        GGML_METAL_ADD_KERNEL(get_rows_q2_K);
+        GGML_METAL_ADD_KERNEL(get_rows_q3_K);
+        GGML_METAL_ADD_KERNEL(get_rows_q4_K);
+        GGML_METAL_ADD_KERNEL(get_rows_q5_K);
+        GGML_METAL_ADD_KERNEL(get_rows_q6_K);
         GGML_METAL_ADD_KERNEL(rms_norm);
         GGML_METAL_ADD_KERNEL(norm);
         GGML_METAL_ADD_KERNEL(mul_mat_f16_f32);
         GGML_METAL_ADD_KERNEL(mul_mat_q4_0_f32);
         GGML_METAL_ADD_KERNEL(mul_mat_q4_1_f32);
-        GGML_METAL_ADD_KERNEL(mul_mat_q2_k_f32);
-        GGML_METAL_ADD_KERNEL(mul_mat_q3_k_f32);
-        GGML_METAL_ADD_KERNEL(mul_mat_q4_k_f32);
-        GGML_METAL_ADD_KERNEL(mul_mat_q5_k_f32);
-        GGML_METAL_ADD_KERNEL(mul_mat_q6_k_f32);
+        GGML_METAL_ADD_KERNEL(mul_mat_q2_K_f32);
+        GGML_METAL_ADD_KERNEL(mul_mat_q3_K_f32);
+        GGML_METAL_ADD_KERNEL(mul_mat_q4_K_f32);
+        GGML_METAL_ADD_KERNEL(mul_mat_q5_K_f32);
+        GGML_METAL_ADD_KERNEL(mul_mat_q6_K_f32);
         GGML_METAL_ADD_KERNEL(rope);
         GGML_METAL_ADD_KERNEL(alibi_f32);
         GGML_METAL_ADD_KERNEL(cpy_f32_f16);
@@ -662,7 +668,7 @@ void ggml_metal_graph_compute(
 
                                             nth0 = 4;
                                             nth1 = 16;
-                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q2_k_f32];
+                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q2_K_f32];
                                         } break;
                                     case GGML_TYPE_Q3_K:
                                         {
@@ -671,7 +677,7 @@ void ggml_metal_graph_compute(
 
                                             nth0 = 4;
                                             nth1 = 16;
-                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q3_k_f32];
+                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q3_K_f32];
                                         } break;
                                     case GGML_TYPE_Q4_K:
                                         {
@@ -680,7 +686,7 @@ void ggml_metal_graph_compute(
 
                                             nth0 = 4;
                                             nth1 = 16;
-                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q4_k_f32];
+                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q4_K_f32];
                                         } break;
                                     case GGML_TYPE_Q5_K:
                                         {
@@ -689,7 +695,7 @@ void ggml_metal_graph_compute(
 
                                             nth0 = 4;
                                             nth1 = 16;
-                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q5_k_f32];
+                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q5_K_f32];
                                         } break;
                                     case GGML_TYPE_Q6_K:
                                         {
@@ -698,7 +704,7 @@ void ggml_metal_graph_compute(
 
                                             nth0 = 4;
                                             nth1 = 16;
-                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q6_k_f32];
+                                            [encoder setComputePipelineState:ctx->pipeline_mul_mat_q6_K_f32];
                                         } break;
                                     default:
                                         {
@@ -750,11 +756,11 @@ void ggml_metal_graph_compute(
                                 case GGML_TYPE_F16:  [encoder setComputePipelineState:ctx->pipeline_get_rows_f16]; break;
                                 case GGML_TYPE_Q4_0: [encoder setComputePipelineState:ctx->pipeline_get_rows_q4_0]; break;
                                 case GGML_TYPE_Q4_1: [encoder setComputePipelineState:ctx->pipeline_get_rows_q4_1]; break;
-                                case GGML_TYPE_Q2_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q2_k]; break;
-                                case GGML_TYPE_Q3_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q3_k]; break;
-                                case GGML_TYPE_Q4_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q4_k]; break;
-                                case GGML_TYPE_Q5_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q5_k]; break;
-                                case GGML_TYPE_Q6_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q6_k]; break;
+                                case GGML_TYPE_Q2_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q2_K]; break;
+                                case GGML_TYPE_Q3_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q3_K]; break;
+                                case GGML_TYPE_Q4_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q4_K]; break;
+                                case GGML_TYPE_Q5_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q5_K]; break;
+                                case GGML_TYPE_Q6_K: [encoder setComputePipelineState:ctx->pipeline_get_rows_q6_K]; break;
                                 default: GGML_ASSERT(false && "not implemented");
                             }