Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Quantization improvements for k_quants #2707

Merged
merged 12 commits into from
Aug 22, 2023
Merged

Quantization improvements for k_quants #2707

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
f26f9ef
Improve LLaMA-2 2-, 3- and 4-bit quantization
Kawrakow Aug 13, 2023
77aea72
Minor 4-bit quantization improvement
Kawrakow Aug 13, 2023
ec9cb75
Some more fine tuning
Kawrakow Aug 13, 2023
4f8dcb1
Adding make_qkx2_quants
Kawrakow Aug 14, 2023
e9f1340
Another minor improvement
Kawrakow Aug 14, 2023
1c1f985
Q2_K improvement
Kawrakow Aug 14, 2023
404e43c
Iterating
Kawrakow Aug 16, 2023
9f78d4c
Revert Q5_K back to make_qkx1_quants
Kawrakow Aug 20, 2023
e2af308
Better Q6_K
Kawrakow Aug 21, 2023
b706339
make_qkx2_quants is better for Q5_K after all
Kawrakow Aug 22, 2023
35a0b97
Fix after rebasing on master
Kawrakow Aug 22, 2023
fdf73db
Fix for changed tensor names
Kawrakow Aug 22, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
164 changes: 110 additions & 54 deletions k_quants.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,6 +77,11 @@ static float make_qx_quants(int n, int nmax, const float * restrict x, int8_t *
}
return 1/iscale;
}
bool return_early = false;
if (rmse_type < 0) {
rmse_type = -rmse_type;
return_early = true;
}
int weight_type = rmse_type%2;
float sumlx = 0;
float suml2 = 0;
Expand All @@ -89,56 +94,9 @@ static float make_qx_quants(int n, int nmax, const float * restrict x, int8_t *
suml2 += w*l*l;
}
float scale = sumlx/suml2;
if (return_early) return suml2 > 0 ? 0.5f*(scale + 1/iscale) : 1/iscale;
float best = scale * sumlx;
for (int itry = 0; itry < 3; ++itry) {
iscale = 1/scale;
float slx = 0;
float sl2 = 0;
bool changed = false;
for (int i = 0; i < n; ++i) {
int l = nearest_int(iscale * x[i]);
l = MAX(-nmax, MIN(nmax-1, l));
if (l + nmax != L[i]) { changed = true; }
float w = weight_type == 1 ? x[i] * x[i] : 1.f;
slx += w*x[i]*l;
sl2 += w*l*l;
}
if (!changed || sl2 == 0 || slx*slx <= best*sl2) { break; }
for (int i = 0; i < n; ++i) {
int l = nearest_int(iscale * x[i]);
L[i] = nmax + MAX(-nmax, MIN(nmax-1, l));
}
sumlx = slx; suml2 = sl2;
scale = sumlx/suml2;
best = scale * sumlx;
}
for (int itry = 0; itry < 5; ++itry) {
int n_changed = 0;
for (int i = 0; i < n; ++i) {
float w = weight_type == 1 ? x[i]*x[i] : 1;
int l = L[i] - nmax;
float slx = sumlx - w*x[i]*l;
if (slx > 0) {
float sl2 = suml2 - w*l*l;
int new_l = nearest_int(x[i] * sl2 / slx);
new_l = MAX(-nmax, MIN(nmax-1, new_l));
if (new_l != l) {
slx += w*x[i]*new_l;
sl2 += w*new_l*new_l;
if (sl2 > 0 && slx*slx*suml2 > sumlx*sumlx*sl2) {
L[i] = nmax + new_l; sumlx = slx; suml2 = sl2;
scale = sumlx / suml2; best = scale * sumlx;
++n_changed;
}
}
}
}
if (!n_changed) { break; }
}
if (rmse_type < 3) {
return scale;
}
for (int is = -4; is <= 4; ++is) {
for (int is = -9; is <= 9; ++is) {
if (is == 0) {
continue;
}
Expand Down Expand Up @@ -221,12 +179,17 @@ static float make_q3_quants(int n, int nmax, const float * restrict x, int8_t *
return 1/iscale;
}

static float make_qkx1_quants(int n, int nmax, const float * restrict x, uint8_t * restrict L, float * restrict the_min, int ntry) {
static float make_qkx1_quants(int n, int nmax, const float * restrict x, uint8_t * restrict L, float * restrict the_min,
int ntry, float alpha) {
float min = x[0];
float max = x[0];
float sum_x = 0;
float sum_x2 = 0;
for (int i = 1; i < n; ++i) {
if (x[i] < min) min = x[i];
if (x[i] > max) max = x[i];
sum_x += x[i];
sum_x2 += x[i]*x[i];
}
if (max == min) {
for (int i = 0; i < n; ++i) L[i] = 0;
Expand Down Expand Up @@ -254,7 +217,7 @@ static float make_qkx1_quants(int n, int nmax, const float * restrict x, uint8_t
for (int i = 0; i < n; ++i) {
sum += x[i] - scale*L[i];
}
min = sum/n;
min = alpha*min + (1 - alpha)*sum/n;
if (min > 0) min = 0;
iscale = 1/scale;
if (!did_change) break;
Expand All @@ -263,6 +226,82 @@ static float make_qkx1_quants(int n, int nmax, const float * restrict x, uint8_t
return scale;
}

static float make_qkx2_quants(int n, int nmax, const float * restrict x, const float * restrict weights,
uint8_t * restrict L, float * restrict the_min, uint8_t * restrict Laux,
float rmin, float rdelta, int nstep, bool use_mad) {
float min = x[0];
float max = x[0];
float sum_w = weights[0];
float sum_x = sum_w * x[0];
for (int i = 1; i < n; ++i) {
if (x[i] < min) min = x[i];
if (x[i] > max) max = x[i];
float w = weights[i];
sum_w += w;
sum_x += w * x[i];
}
if (min > 0) min = 0;
if (max == min) {
for (int i = 0; i < n; ++i) L[i] = 0;
*the_min = -min;
return 0.f;
}
float iscale = nmax/(max - min);
float scale = 1/iscale;
float best_mad = 0;
for (int i = 0; i < n; ++i) {
int l = nearest_int(iscale*(x[i] - min));
L[i] = MAX(0, MIN(nmax, l));
float diff = scale * L[i] + min - x[i];
diff = use_mad ? fabsf(diff) : diff * diff;
float w = weights[i];
best_mad += w * diff;
}
if (nstep < 1) {
*the_min = -min;
return scale;
}
for (int is = 0; is <= nstep; ++is) {
iscale = (rmin + rdelta*is + nmax)/(max - min);
float sum_l = 0, sum_l2 = 0, sum_xl = 0;
for (int i = 0; i < n; ++i) {
int l = nearest_int(iscale*(x[i] - min));
l = MAX(0, MIN(nmax, l));
Laux[i] = l;
float w = weights[i];
sum_l += w*l;
sum_l2 += w*l*l;
sum_xl += w*l*x[i];
}
float D = sum_w * sum_l2 - sum_l * sum_l;
if (D > 0) {
float this_scale = (sum_w * sum_xl - sum_x * sum_l)/D;
float this_min = (sum_l2 * sum_x - sum_l * sum_xl)/D;
if (this_min > 0) {
this_min = 0;
this_scale = sum_xl / sum_l2;
}
float mad = 0;
for (int i = 0; i < n; ++i) {
float diff = this_scale * Laux[i] + this_min - x[i];
diff = use_mad ? fabsf(diff) : diff * diff;
float w = weights[i];
mad += w * diff;
}
if (mad < best_mad) {
for (int i = 0; i < n; ++i) {
L[i] = Laux[i];
}
best_mad = mad;
scale = this_scale;
min = this_min;
}
}
}
*the_min = -min;
return scale;
}

#if QK_K == 256
static inline void get_scale_min_k4(int j, const uint8_t * restrict q, uint8_t * restrict d, uint8_t * restrict m) {
if (j < 4) {
Expand All @@ -281,6 +320,8 @@ void quantize_row_q2_K_reference(const float * restrict x, block_q2_K * restrict
const int nb = k / QK_K;

uint8_t L[QK_K];
uint8_t Laux[16];
float weights[16];
float mins[QK_K/16];
float scales[QK_K/16];

Expand All @@ -291,7 +332,8 @@ void quantize_row_q2_K_reference(const float * restrict x, block_q2_K * restrict
float max_scale = 0; // as we are deducting the min, scales are always positive
float max_min = 0;
for (int j = 0; j < QK_K/16; ++j) {
scales[j] = make_qkx1_quants(16, 3, x + 16*j, L + 16*j, &mins[j], 5);
for (int l = 0; l < 16; ++l) weights[l] = fabsf(x[16*j + l]);
scales[j] = make_qkx2_quants(16, 3, x + 16*j, weights, L + 16*j, &mins[j], Laux, -0.5f, 0.1f, 15, true);
float scale = scales[j];
if (scale > max_scale) {
max_scale = scale;
Expand Down Expand Up @@ -637,6 +679,8 @@ void quantize_row_q4_K_reference(const float * restrict x, block_q4_K * restrict
const int nb = k / QK_K;

uint8_t L[QK_K];
uint8_t Laux[32];
float weights[32];
float mins[QK_K/32];
float scales[QK_K/32];

Expand All @@ -645,7 +689,12 @@ void quantize_row_q4_K_reference(const float * restrict x, block_q4_K * restrict
float max_scale = 0; // as we are deducting the min, scales are always positive
float max_min = 0;
for (int j = 0; j < QK_K/32; ++j) {
scales[j] = make_qkx1_quants(32, 15, x + 32*j, L + 32*j, &mins[j], 5);
//scales[j] = make_qkx1_quants(32, 15, x + 32*j, L + 32*j, &mins[j], 9, 0.5f);
float sum_x2 = 0;
for (int l = 0; l < 32; ++l) sum_x2 += x[32*j + l] * x[32*j + l];
float av_x = sqrtf(sum_x2/32);
for (int l = 0; l < 32; ++l) weights[l] = av_x + fabsf(x[32*j + l]);
scales[j] = make_qkx2_quants(32, 15, x + 32*j, weights, L + 32*j, &mins[j], Laux, -1.f, 0.1f, 20, false);
float scale = scales[j];
if (scale > max_scale) {
max_scale = scale;
Expand Down Expand Up @@ -798,6 +847,8 @@ void quantize_row_q5_K_reference(const float * restrict x, block_q5_K * restrict
uint8_t L[QK_K];
float mins[QK_K/32];
float scales[QK_K/32];
float weights[32];
uint8_t Laux[32];
#else
int8_t L[QK_K];
float scales[QK_K/16];
Expand All @@ -810,7 +861,12 @@ void quantize_row_q5_K_reference(const float * restrict x, block_q5_K * restrict
float max_scale = 0; // as we are deducting the min, scales are always positive
float max_min = 0;
for (int j = 0; j < QK_K/32; ++j) {
scales[j] = make_qkx1_quants(32, 31, x + 32*j, L + 32*j, &mins[j], 5);
//scales[j] = make_qkx1_quants(32, 31, x + 32*j, L + 32*j, &mins[j], 9, 0.5f);
float sum_x2 = 0;
for (int l = 0; l < 32; ++l) sum_x2 += x[32*j + l] * x[32*j + l];
float av_x = sqrtf(sum_x2/32);
for (int l = 0; l < 32; ++l) weights[l] = av_x + fabsf(x[32*j + l]);
scales[j] = make_qkx2_quants(32, 31, x + 32*j, weights, L + 32*j, &mins[j], Laux, -0.5f, 0.1f, 15, false);
float scale = scales[j];
if (scale > max_scale) {
max_scale = scale;
Expand Down
24 changes: 20 additions & 4 deletions llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3717,24 +3717,40 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
new_type = GGML_TYPE_Q6_K;
}
} else if (name.find("attn_v.weight") != std::string::npos) {
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q4_K;
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M) {
new_type = i_attention_wv < 2 ? GGML_TYPE_Q5_K : GGML_TYPE_Q4_K;
}
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q5_K;
else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) &&
use_more_bits(i_attention_wv, n_attention_wv)) new_type = GGML_TYPE_Q6_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S && i_attention_wv < 4) new_type = GGML_TYPE_Q5_K;
else if (QK_K == 64 && (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_S) &&
(i_attention_wv < n_attention_wv/8 || i_attention_wv >= 7*n_attention_wv/8)) new_type = GGML_TYPE_Q6_K;
++i_attention_wv;
} else if (name.find("ffn_down.weight") != std::string::npos) {
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q4_K;
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M) {
new_type = i_feed_forward_w2 < 2 ? GGML_TYPE_Q5_K : GGML_TYPE_Q4_K;
}
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q5_K;
else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) &&
use_more_bits(i_feed_forward_w2, n_feed_forward_w2)) new_type = GGML_TYPE_Q6_K;
//else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S && i_feed_forward_w2 < n_feed_forward_w2/8) new_type = GGML_TYPE_Q6_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S && i_feed_forward_w2 < 4) new_type = GGML_TYPE_Q5_K;
++i_feed_forward_w2;
} else if (name.find("attn_output.weight") != std::string::npos) {
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q4_K;
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K ) new_type = GGML_TYPE_Q3_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M) new_type = GGML_TYPE_Q4_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q5_K;
}
else if (name.find("ffn_gate.weight") != std::string::npos || name.find("ffn_up.weight") != std::string::npos) {
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
}
// This can be used to reduce the size of the Q5_K_S model.
// The associated PPL increase is fully in line with the size reduction
//else {
// if (ftype == LLAMA_FTYPE_MOSTLY_Q5_K_S) new_type = GGML_TYPE_Q4_K;
//}
bool convert_incompatible_tensor = false;
if (new_type == GGML_TYPE_Q2_K || new_type == GGML_TYPE_Q3_K || new_type == GGML_TYPE_Q4_K ||
new_type == GGML_TYPE_Q5_K || new_type == GGML_TYPE_Q6_K) {
Expand Down
Loading