Skip to content

Commit

Permalink
sync: add ggml as submodule (#5)
Browse files Browse the repository at this point in the history
  • Loading branch information
@PABannier
PABannier authored Oct 1, 2023
1 parent 23d89e7 commit 94775ca
Show file tree
Hide file tree
Showing 8 changed files with 68 additions and 18,140 deletions.
8 changes: 6 additions & 2 deletions .github/workflows/build.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,9 @@ jobs:
steps:
- name: Clone
id: checkout
uses: actions/checkout@v2
uses: actions/checkout@v4
with:
submodules: true

- name: Dependencies
id: depends
Expand All @@ -44,7 +46,9 @@ jobs:
steps:
- name: Clone
id: checkout
uses: actions/checkout@v2
uses: actions/checkout@v4
with:
submodules: true

- name: Dependencies
id: depends
Expand Down
3 changes: 3 additions & 0 deletions .gitmodules
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,3 @@
[submodule "ggml"]
path = ggml
url = https://github.com/ggerganov/ggml.git
2 changes: 2 additions & 0 deletions CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,8 @@ endif()

set(ENCODEC_LIB encodec.cpp)

add_subdirectory(ggml)

add_library(
${ENCODEC_LIB}
encodec
Expand Down
158 changes: 54 additions & 104 deletions encodec.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -13,33 +13,55 @@ static void read_safe(std::ifstream& infile, T& dest) {
infile.read((char*)& dest, sizeof(T));
}

static void encodec_sigmoid_impl(struct ggml_tensor * dst, const struct ggml_tensor * src, int ith, int nth, void * userdata) {
GGML_ASSERT(userdata == NULL);
GGML_ASSERT(ggml_are_same_shape(dst, src));
GGML_ASSERT(ggml_is_contiguous(dst));
GGML_ASSERT(ggml_is_contiguous(src));

const float * src_data = ggml_get_data_f32(src);
float * dst_data = ggml_get_data_f32(dst);

const int ne = (int)ggml_nelements(dst);
const int dr = (ne + nth - 1) / nth;
const int ie0 = dr * ith;
const int ie1 = std::min(ie0 + dr, ne);

for (int i = ie0; i < ie1; ++i) {
dst_data[i] = 1.0f / (1.0f + expf(-src_data[i]));
}
}

static struct ggml_tensor * encodec_sigmoid(ggml_context * ctx, struct ggml_tensor * x) {
return ggml_map_custom1(ctx, x, encodec_sigmoid_impl, GGML_N_TASKS_MAX, NULL);
}

static int get_extra_padding_for_conv_1d(ggml_tensor * inp, float kernel_size, float stride, float padding_total) {
float length = inp->ne[0];
float n_frames = (length - kernel_size + padding_total) / stride + 1.0f;
int ideal_length = (std::ceilf(n_frames) - 1) * stride + (kernel_size - padding_total);
int ideal_length = (ceilf(n_frames) - 1) * stride + (kernel_size - padding_total);
return ideal_length - length;
}

static struct ggml_tensor * pad_1d(ggml_context * ctx0, ggml_tensor * inp, int padding_left, int padding_right) {
int length = inp->ne[0];
int dim = inp->ne[1];
ENCODEC_ASSERT(padding_left >= 0);
ENCODEC_ASSERT(padding_right >= 0);

const int max_pad = std::max(padding_left, padding_right);
int extra_pad = 0;

if (length <= max_pad) {
extra_pad = max_pad - length + 1;
int padding[2] = {0, extra_pad};
inp = ggml_pad_1d_constant(ctx0, inp, padding, 0);

// constant padding
struct ggml_tensor * out = ggml_new_tensor_2d(ctx0, inp->type, length+extra_pad, dim);
ggml_set_zero(out);
out = ggml_set_2d(ctx0, out, inp, out->nb[1], 0);
}

int padding[2] = {padding_left, padding_right};
struct ggml_tensor * padded = ggml_pad_1d_reflective(ctx0, inp, padding);
struct ggml_tensor * padded = ggml_pad_reflec_1d(ctx0, inp, padding_left, padding_right);

const int end = padded->ne[0] - extra_pad;

struct ggml_tensor *dest = ggml_view_2d(ctx0, padded, end, dim, padded->nb[1], 0);

return dest;
Expand All @@ -61,20 +83,18 @@ static struct ggml_tensor * unpad_1d(ggml_context * ctx0, ggml_tensor * inp, int
return dst;
}

struct ggml_tensor * strided_conv_1d(
static struct ggml_tensor * strided_conv_1d(
ggml_context * ctx0,
ggml_tensor * inp,
ggml_tensor * conv_w,
ggml_tensor * conv_b,
int stride) {
int kernel_size = conv_w->ne[0];
int padding_total = kernel_size - stride;

int extra_padding = get_extra_padding_for_conv_1d(inp, kernel_size, stride, padding_total);

struct ggml_tensor * padded_inp = pad_1d(ctx0, inp, padding_total, extra_padding);

struct ggml_tensor * dst = ggml_conv_1d(ctx0, conv_w, padded_inp, stride);
struct ggml_tensor * dst = ggml_conv_1d(ctx0, conv_w, padded_inp, stride, 0, 1);

// add bias
dst = ggml_transpose(ctx0, dst);
Expand All @@ -84,27 +104,28 @@ struct ggml_tensor * strided_conv_1d(
return dst;
}

struct ggml_tensor * forward_pass_lstm_unilayer(
struct ggml_context * ctx0,
struct ggml_tensor * inp,
struct ggml_tensor * weight_ih,
struct ggml_tensor * weight_hh,
struct ggml_tensor * bias_ih,
struct ggml_tensor * bias_hh) {
static struct ggml_tensor * forward_pass_lstm_unilayer(
struct ggml_context * ctx0,
struct ggml_tensor * inp,
struct ggml_tensor * weight_ih,
struct ggml_tensor * weight_hh,
struct ggml_tensor * bias_ih,
struct ggml_tensor * bias_hh) {

const int input_dim = inp->ne[1];
const int hidden_dim = weight_ih->ne[1]/4;
const int seq_length = inp->ne[0];

struct ggml_tensor * hs = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hidden_dim, seq_length);

struct ggml_tensor * c_t = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, hidden_dim);
struct ggml_tensor * c_t = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, hidden_dim);
struct ggml_tensor * h_t = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, hidden_dim);

ggml_set_zero(h_t);
h_t = ggml_set_zero(h_t);
c_t = ggml_set_zero(c_t);

struct ggml_tensor * current = ggml_cont(ctx0, ggml_transpose(ctx0, inp));

for (int t = 0; t < seq_length; t++) {
struct ggml_tensor * x_t = ggml_view_1d(ctx0, current, input_dim, t*current->nb[1]);

Expand All @@ -116,10 +137,10 @@ struct ggml_tensor * forward_pass_lstm_unilayer(

struct ggml_tensor * out_gates = ggml_add(ctx0, inp_gates, hid_gates);

struct ggml_tensor * i_t = ggml_sigmoid(ctx0, ggml_view_1d(ctx0, out_gates, hidden_dim, 0*sizeof(float)*hidden_dim));
struct ggml_tensor * f_t = ggml_sigmoid(ctx0, ggml_view_1d(ctx0, out_gates, hidden_dim, 1*sizeof(float)*hidden_dim));
struct ggml_tensor * i_t = encodec_sigmoid(ctx0, ggml_view_1d(ctx0, out_gates, hidden_dim, 0*sizeof(float)*hidden_dim));
struct ggml_tensor * f_t = encodec_sigmoid(ctx0, ggml_view_1d(ctx0, out_gates, hidden_dim, 1*sizeof(float)*hidden_dim));
struct ggml_tensor * g_t = ggml_tanh (ctx0, ggml_view_1d(ctx0, out_gates, hidden_dim, 2*sizeof(float)*hidden_dim));
struct ggml_tensor * o_t = ggml_sigmoid(ctx0, ggml_view_1d(ctx0, out_gates, hidden_dim, 3*sizeof(float)*hidden_dim));
struct ggml_tensor * o_t = encodec_sigmoid(ctx0, ggml_view_1d(ctx0, out_gates, hidden_dim, 3*sizeof(float)*hidden_dim));

c_t = ggml_add(ctx0, ggml_mul(ctx0, f_t, c_t), ggml_mul(ctx0, i_t, g_t));
h_t = ggml_mul(ctx0, o_t, ggml_tanh(ctx0, c_t));
Expand All @@ -132,23 +153,23 @@ struct ggml_tensor * forward_pass_lstm_unilayer(
return hs;
}

struct ggml_tensor * strided_conv_transpose_1d(
ggml_context * ctx0,
ggml_tensor * inp,
ggml_tensor * conv_w,
ggml_tensor * conv_b,
int stride) {
static struct ggml_tensor * strided_conv_transpose_1d(
ggml_context * ctx0,
ggml_tensor * inp,
ggml_tensor * conv_w,
ggml_tensor * conv_b,
int stride) {
int kernel_size = conv_w->ne[0];
int padding_total = kernel_size - stride;

struct ggml_tensor * dst = ggml_transpose_conv_1d(ctx0, conv_w, inp, stride);
struct ggml_tensor * dst = ggml_conv_transpose_1d(ctx0, conv_w, inp, stride, 0, 1);

// add bias
dst = ggml_transpose(ctx0, dst);
dst = ggml_add(ctx0, ggml_repeat(ctx0, conv_b, dst), dst);
dst = ggml_cont(ctx0, ggml_transpose(ctx0, dst));

int padding_right = std::ceilf(padding_total);
int padding_right = ceilf(padding_total);
int padding_left = padding_total - padding_right;

struct ggml_tensor * unpadded = unpad_1d(ctx0, dst, padding_left, padding_right);
Expand Down Expand Up @@ -730,76 +751,5 @@ static void encodec_model_eval(
ggml_build_forward_expand(&gf, out);
ggml_graph_compute (ctx0, &gf);

printf("\n");
printf("seq_length = %d\n", out->ne[0]);
printf("n_channels = %d\n", out->ne[1]);
printf("out_channels = %d\n", out->ne[2]);
printf("\n");

out = ggml_view_2d(ctx0, out, 1000, out->ne[1], out->nb[1], 0);

for(int i = 0; i < out->ne[1]; i++) {
for (int j = 0; j < out->ne[0]; j++) {
float val = *(float *) ((char *) out->data + j*out->nb[0] + i*out->nb[1]);
printf("%.4f ", val);
}
printf("\n");
}

// for(int i = 0; i < out->ne[1]; i++) {
// for (int j = 0; j < out->ne[0]; j++) {
// int32_t val = *(int32_t *) ((char *) out->data + j*out->nb[0] + i*out->nb[1]);
// printf("%d ", val);
// }
// printf("\n");
// }

ggml_free(ctx0);
}

int main(int argc, char* argv[]) {
const int64_t t_main_start_us = ggml_time_us();

int64_t t_load_us = 0;
int64_t t_compr_us = 0;

encodec_model model;

// load the model
{
const int64_t t_start_us = ggml_time_us();
std::string model_path = "./ggml_weights/ggml-model.bin";

if (!encodec_model_load(model_path, model)) {
fprintf(stderr, "%s: failed to load model from '%s'\n", __func__, model_path.c_str());
return 1;
}

t_load_us = ggml_time_us() - t_start_us;
}

// generate toy data
std::vector<float> raw_audio(1000, 0.4);

// encode
const int64_t t_compr_us_start = ggml_time_us();

// TODO change n_threads to be passed by params
encodec_model_eval(raw_audio, model, 1);

t_compr_us = ggml_time_us() - t_compr_us_start;

// report timing
{
const int64_t t_main_end_us = ggml_time_us();

printf("\n\n");
printf("%s: load time = %8.2f ms\n", __func__, t_load_us/1000.0f);
printf("%s: compress time = %8.2f ms\n", __func__, t_compr_us/1000.0f);
printf("%s: total time = %8.2f ms\n", __func__, (t_main_end_us - t_main_start_us)/1000.0f);
}

ggml_free(model.ctx);

return 0;
}
26 changes: 2 additions & 24 deletions encodec.h
View file
Original file line number Diff line number Diff line change
@@ -1,7 +1,5 @@
#pragma once

#include "ggml.h"

#include <cmath>
#include <fstream>
#include <iostream>
Expand All @@ -10,6 +8,8 @@
#include <string>
#include <vector>

#include "ggml.h"

#define ENCODEC_FILE_MAGIC 'ggml'
#define ENCODEC_FILE_VERSION 1

Expand Down Expand Up @@ -140,25 +140,3 @@ struct encodec_model {

std::map<std::string, struct ggml_tensor *> tensors;
};

struct ggml_tensor * strided_conv_1d(
ggml_context * ctx0,
ggml_tensor * inp,
ggml_tensor * conv_w,
ggml_tensor * conv_b,
int stride);

struct ggml_tensor * forward_pass_lstm_unilayer(
ggml_context * ctx0,
ggml_tensor * inp,
ggml_tensor * weight_ih,
ggml_tensor * weight_hh,
ggml_tensor * bias_ih,
ggml_tensor * bias_hh);

struct ggml_tensor * strided_conv_transpose_1d(
ggml_context * ctx0,
ggml_tensor * inp,
ggml_tensor * conv_w,
ggml_tensor * conv_b,
int stride);
1 change: 1 addition & 0 deletions ggml
Submodule ggml added at a16b01
Loading

0 comments on commit 94775ca

Please sign in to comment.