root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-30 21:34:36 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

sync op_rope_f16 with recent op_rope_f32 changes

Browse Source
This commit is contained in:
Jared Van Bortel 2024-01-18 11:56:00 -05:00
parent 0f1a958a51
commit 16bc3c3be8
2 changed files with 26 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
kompute-shaders/op_rope_f16.comp
View File

@ -40,16 +40,15 @@ void main() {
} }
} else { } else {
const float inv_ndims = -1.f/pcs.n_dims; const float inv_ndims = -1.f/pcs.n_dims;
for (uint ib = 0; ib < pcs.ne0/pcs.n_dims; ++ib) {
for (uint ic = 0; ic < pcs.n_dims; ic += 2) { for (uint ic = 0; ic < pcs.n_dims; ic += 2) {
const uint cur_rot = ib * pcs.n_dims + ic; const uint cur_rot = ic;
float cos_theta, sin_theta; float cos_theta, sin_theta;
rope_yarn(theta, pcs.freq_scale, corr_dims, cur_rot, pcs.ext_factor, pcs.attn_factor, cos_theta, sin_theta); rope_yarn(theta, pcs.freq_scale, corr_dims, cur_rot, pcs.ext_factor, pcs.attn_factor, cos_theta, sin_theta);
theta *= theta_scale; theta *= theta_scale;
const uint i0 = ib*pcs.n_dims + ic/2; const uint i0 = ic/2;
const uint src = uint((i3*pcs.nb03 + i2*pcs.nb02 + i1*pcs.nb01 + i0*pcs.nb00) / 2) + pcs.inAOff; // Based from in const uint src = uint((i3*pcs.nb03 + i2*pcs.nb02 + i1*pcs.nb01 + i0*pcs.nb00) / 2) + pcs.inAOff; // Based from in
const uint dst_data = uint((i3*pcs.nb3 + i2*pcs.nb2 + i1*pcs.nb1 + i0*pcs.nb0) / 2) + pcs.outOff; // Based from out_ const uint dst_data = uint((i3*pcs.nb3 + i2*pcs.nb2 + i1*pcs.nb1 + i0*pcs.nb0) / 2) + pcs.outOff; // Based from out_
@ -60,6 +59,15 @@ void main() {
out_[dst_data] = float16_t(x0*cos_theta - x1*sin_theta); out_[dst_data] = float16_t(x0*cos_theta - x1*sin_theta);
out_[dst_data+pcs.n_dims/2] = float16_t(x0*sin_theta + x1*cos_theta); out_[dst_data+pcs.n_dims/2] = float16_t(x0*sin_theta + x1*cos_theta);
} }
for (uint ic = pcs.n_dims; ic < pcs.ne0; ic += 2) {
const uint i0 = ic;
const uint src = uint((i3*pcs.nb03 + i2*pcs.nb02 + i1*pcs.nb01 + i0*pcs.nb00) / 2) + pcs.inAOff; // Based from in
const uint dst_data = uint((i3*pcs.nb3 + i2*pcs.nb2 + i1*pcs.nb1 + i0*pcs.nb0) / 2) + pcs.outOff; // Based from out_
out_[dst_data + 0] = inA[src + 0];
out_[dst_data + 1] = inA[src + 1];
} }
} }
} }