mirror of
https://github.com/jart/cosmopolitan.git
synced 2025-01-31 11:37:35 +00:00
8fdb31681a
llama.com can now load weights that use the new file format which was introduced a few weeks ago. Note that, unlike llama.cpp, we will keep support for old file formats in our tool so you don't need to convert your weights when the upstream project makes breaking changes. Please note that using ggjt v3 does make avx2 inference go 5% faster for me.
221 lines
9.8 KiB
C
221 lines
9.8 KiB
C
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:4;tab-width:8;coding:utf-8 -*-│
|
|
│vi: set net ft=c ts=4 sts=4 sw=4 fenc=utf-8 :vi│
|
|
╚──────────────────────────────────────────────────────────────────────────────╝
|
|
│ │
|
|
│ GGML │
|
|
│ Copyright (c) 2023 Georgi Gerganov │
|
|
│ │
|
|
│ Permission is hereby granted, free of charge, to any person obtaining │
|
|
│ a copy of this software and associated documentation files (the │
|
|
│ "Software"), to deal in the Software without restriction, including │
|
|
│ without limitation the rights to use, copy, modify, merge, publish, │
|
|
│ distribute, sublicense, and/or sell copies of the Software, and to │
|
|
│ permit persons to whom the Software is furnished to do so, subject to │
|
|
│ the following conditions: │
|
|
│ │
|
|
│ The above copyright notice and this permission notice shall be │
|
|
│ included in all copies or substantial portions of the Software. │
|
|
│ │
|
|
│ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, │
|
|
│ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF │
|
|
│ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. │
|
|
│ IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY │
|
|
│ CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, │
|
|
│ TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE │
|
|
│ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. │
|
|
│ │
|
|
╚─────────────────────────────────────────────────────────────────────────────*/
|
|
#include "third_party/ggml/ggjt.v2.q8_1.h"
|
|
#include "libc/assert.h"
|
|
#include "libc/macros.internal.h"
|
|
#include "libc/math.h"
|
|
#include "third_party/ggml/ggjt.v2.internal.h"
|
|
// clang-format off
|
|
|
|
static_assert(sizeof(block_v2_q8_1) == 2*sizeof(float) + V2_QK8_1,
|
|
"wrong q8_1 block size/padding");
|
|
|
|
#if __AVX__ || __AVX2__ || __AVX512F__
|
|
static inline int hsum_i32_8(const __m256i a) {
|
|
const __m128i sum128 = _mm_add_epi32(_mm256_castsi256_si128(a), _mm256_extractf128_si256(a, 1));
|
|
const __m128i hi64 = _mm_unpackhi_epi64(sum128, sum128);
|
|
const __m128i sum64 = _mm_add_epi32(hi64, sum128);
|
|
const __m128i hi32 = _mm_shuffle_epi32(sum64, _MM_SHUFFLE(2, 3, 0, 1));
|
|
return _mm_cvtsi128_si32(_mm_add_epi32(sum64, hi32));
|
|
}
|
|
#endif
|
|
|
|
void quantize_row_v2_q8_1(const float * restrict x, void * restrict vy, int k) {
|
|
assert(k % V2_QK8_1 == 0);
|
|
const int nb = k / V2_QK8_1;
|
|
|
|
block_v2_q8_1 * restrict y = vy;
|
|
|
|
#if defined(__ARM_NEON)
|
|
for (int i = 0; i < nb; i++) {
|
|
float32x4_t srcv [8];
|
|
float32x4_t asrcv[8];
|
|
float32x4_t amaxv[8];
|
|
|
|
for (int j = 0; j < 8; j++) srcv[j] = vld1q_f32(x + i*32 + 4*j);
|
|
for (int j = 0; j < 8; j++) asrcv[j] = vabsq_f32(srcv[j]);
|
|
|
|
for (int j = 0; j < 4; j++) amaxv[2*j] = vmaxq_f32(asrcv[2*j], asrcv[2*j+1]);
|
|
for (int j = 0; j < 2; j++) amaxv[4*j] = vmaxq_f32(amaxv[4*j], amaxv[4*j+2]);
|
|
for (int j = 0; j < 1; j++) amaxv[8*j] = vmaxq_f32(amaxv[8*j], amaxv[8*j+4]);
|
|
|
|
const float amax = vmaxvq_f32(amaxv[0]);
|
|
|
|
const float d = amax / ((1 << 7) - 1);
|
|
const float id = d ? 1.0f/d : 0.0f;
|
|
|
|
y[i].d = d;
|
|
|
|
int32x4_t accv = vdupq_n_s32(0);
|
|
|
|
for (int j = 0; j < 8; j++) {
|
|
const float32x4_t v = vmulq_n_f32(srcv[j], id);
|
|
const int32x4_t vi = vcvtnq_s32_f32(v);
|
|
|
|
y[i].qs[4*j + 0] = vgetq_lane_s32(vi, 0);
|
|
y[i].qs[4*j + 1] = vgetq_lane_s32(vi, 1);
|
|
y[i].qs[4*j + 2] = vgetq_lane_s32(vi, 2);
|
|
y[i].qs[4*j + 3] = vgetq_lane_s32(vi, 3);
|
|
|
|
accv = vaddq_s32(accv, vi);
|
|
}
|
|
|
|
y[i].s = d * vaddvq_s32(accv);
|
|
}
|
|
#elif defined(__AVX2__) || defined(__AVX__)
|
|
for (int i = 0; i < nb; i++) {
|
|
// Load elements into 4 AVX vectors
|
|
__m256 v0 = _mm256_loadu_ps( x );
|
|
__m256 v1 = _mm256_loadu_ps( x + 8 );
|
|
__m256 v2 = _mm256_loadu_ps( x + 16 );
|
|
__m256 v3 = _mm256_loadu_ps( x + 24 );
|
|
x += 32;
|
|
|
|
// Compute max(abs(e)) for the block
|
|
const __m256 signBit = _mm256_set1_ps( -0.0f );
|
|
__m256 maxAbs = _mm256_andnot_ps( signBit, v0 );
|
|
maxAbs = _mm256_max_ps( maxAbs, _mm256_andnot_ps( signBit, v1 ) );
|
|
maxAbs = _mm256_max_ps( maxAbs, _mm256_andnot_ps( signBit, v2 ) );
|
|
maxAbs = _mm256_max_ps( maxAbs, _mm256_andnot_ps( signBit, v3 ) );
|
|
|
|
__m128 max4 = _mm_max_ps( _mm256_extractf128_ps( maxAbs, 1 ), _mm256_castps256_ps128( maxAbs ) );
|
|
max4 = _mm_max_ps( max4, _mm_movehl_ps( max4, max4 ) );
|
|
max4 = _mm_max_ss( max4, _mm_movehdup_ps( max4 ) );
|
|
const float maxScalar = _mm_cvtss_f32( max4 );
|
|
|
|
// Quantize these floats
|
|
const float d = maxScalar / 127.f;
|
|
y[i].d = d;
|
|
const float id = ( maxScalar != 0.0f ) ? 127.f / maxScalar : 0.0f;
|
|
const __m256 mul = _mm256_set1_ps( id );
|
|
|
|
// Apply the multiplier
|
|
v0 = _mm256_mul_ps( v0, mul );
|
|
v1 = _mm256_mul_ps( v1, mul );
|
|
v2 = _mm256_mul_ps( v2, mul );
|
|
v3 = _mm256_mul_ps( v3, mul );
|
|
|
|
// Round to nearest integer
|
|
v0 = _mm256_round_ps( v0, _MM_ROUND_NEAREST );
|
|
v1 = _mm256_round_ps( v1, _MM_ROUND_NEAREST );
|
|
v2 = _mm256_round_ps( v2, _MM_ROUND_NEAREST );
|
|
v3 = _mm256_round_ps( v3, _MM_ROUND_NEAREST );
|
|
|
|
// Convert floats to integers
|
|
__m256i i0 = _mm256_cvtps_epi32( v0 );
|
|
__m256i i1 = _mm256_cvtps_epi32( v1 );
|
|
__m256i i2 = _mm256_cvtps_epi32( v2 );
|
|
__m256i i3 = _mm256_cvtps_epi32( v3 );
|
|
|
|
#if defined(__AVX2__)
|
|
// Compute the sum of the quants and set y[i].s
|
|
y[i].s = d * hsum_i32_8(_mm256_add_epi32(_mm256_add_epi32(i0, i1), _mm256_add_epi32(i2, i3)));
|
|
|
|
// Convert int32 to int16
|
|
i0 = _mm256_packs_epi32( i0, i1 ); // 0, 1, 2, 3, 8, 9, 10, 11, 4, 5, 6, 7, 12, 13, 14, 15
|
|
i2 = _mm256_packs_epi32( i2, i3 ); // 16, 17, 18, 19, 24, 25, 26, 27, 20, 21, 22, 23, 28, 29, 30, 31
|
|
// Convert int16 to int8
|
|
i0 = _mm256_packs_epi16( i0, i2 ); // 0, 1, 2, 3, 8, 9, 10, 11, 16, 17, 18, 19, 24, 25, 26, 27, 4, 5, 6, 7, 12, 13, 14, 15, 20, 21, 22, 23, 28, 29, 30, 31
|
|
|
|
// We got our precious signed bytes, but the order is now wrong
|
|
// These AVX2 pack instructions process 16-byte pieces independently
|
|
// The following instruction is fixing the order
|
|
const __m256i perm = _mm256_setr_epi32( 0, 4, 1, 5, 2, 6, 3, 7 );
|
|
i0 = _mm256_permutevar8x32_epi32( i0, perm );
|
|
|
|
_mm256_storeu_si256((__m256i *)y[i].qs, i0);
|
|
#else
|
|
// Since we don't have in AVX some necessary functions,
|
|
// we split the registers in half and call AVX2 analogs from SSE
|
|
__m128i ni0 = _mm256_castsi256_si128( i0 );
|
|
__m128i ni1 = _mm256_extractf128_si256( i0, 1);
|
|
__m128i ni2 = _mm256_castsi256_si128( i1 );
|
|
__m128i ni3 = _mm256_extractf128_si256( i1, 1);
|
|
__m128i ni4 = _mm256_castsi256_si128( i2 );
|
|
__m128i ni5 = _mm256_extractf128_si256( i2, 1);
|
|
__m128i ni6 = _mm256_castsi256_si128( i3 );
|
|
__m128i ni7 = _mm256_extractf128_si256( i3, 1);
|
|
|
|
// Compute the sum of the quants and set y[i].s
|
|
const __m128i s0 = _mm_add_epi32(_mm_add_epi32(ni0, ni1), _mm_add_epi32(ni2, ni3));
|
|
const __m128i s1 = _mm_add_epi32(_mm_add_epi32(ni4, ni5), _mm_add_epi32(ni6, ni7));
|
|
y[i].s = d * hsum_i32_4(_mm_add_epi32(s0, s1));
|
|
|
|
// Convert int32 to int16
|
|
ni0 = _mm_packs_epi32( ni0, ni1 );
|
|
ni2 = _mm_packs_epi32( ni2, ni3 );
|
|
ni4 = _mm_packs_epi32( ni4, ni5 );
|
|
ni6 = _mm_packs_epi32( ni6, ni7 );
|
|
// Convert int16 to int8
|
|
ni0 = _mm_packs_epi16( ni0, ni2 );
|
|
ni4 = _mm_packs_epi16( ni4, ni6 );
|
|
|
|
_mm_storeu_si128((__m128i *)(y[i].qs + 0), ni0);
|
|
_mm_storeu_si128((__m128i *)(y[i].qs + 16), ni4);
|
|
#endif
|
|
}
|
|
#else
|
|
// scalar
|
|
quantize_row_q8_1_reference(x, y, k);
|
|
#endif
|
|
}
|
|
|
|
void quantize_row_v2_q8_1_reference(const float * restrict x, block_v2_q8_1 * restrict y, int k) {
|
|
assert(V2_QK8_1 == 32);
|
|
assert(k % V2_QK8_1 == 0);
|
|
const int nb = k / V2_QK8_1;
|
|
|
|
for (int i = 0; i < nb; i++) {
|
|
float amax = 0.0f; // absolute max
|
|
|
|
for (int j = 0; j < V2_QK8_1; j++) {
|
|
const float v = x[i*V2_QK8_1 + j];
|
|
amax = MAX(amax, fabsf(v));
|
|
}
|
|
|
|
const float d = amax / ((1 << 7) - 1);
|
|
const float id = d ? 1.0f/d : 0.0f;
|
|
|
|
y[i].d = d;
|
|
|
|
int sum = 0;
|
|
|
|
for (int j = 0; j < V2_QK8_1/2; ++j) {
|
|
const float v0 = x[i*V2_QK8_1 + j]*id;
|
|
const float v1 = x[i*V2_QK8_1 + V2_QK8_1/2 + j]*id;
|
|
|
|
y[i].qs[ j] = roundf(v0);
|
|
y[i].qs[V2_QK8_1/2 + j] = roundf(v1);
|
|
|
|
sum += y[i].qs[ j];
|
|
sum += y[i].qs[V2_QK8_1/2 + j];
|
|
}
|
|
|
|
y[i].s = d * sum;
|
|
}
|
|
}
|