llama.cpp/examples/control-vector-generator/control-vector-generator.cpp

792 lines
29 KiB
C++

#include "common.h"
#include "llama.h"
#include "ggml.h"
#ifdef GGML_USE_CUDA
#include "ggml-cuda.h"
#endif
#ifdef GGML_USE_METAL
#include "ggml-metal.h"
#endif
#include <cstdio>
#include <string>
#include <tuple>
#include <vector>
#include <algorithm>
#include <iostream>
#include <fstream>
// TODO read everything over and make sure it makes sense because I'm dropping logic errors left and right - Christian
struct callback_data {
std::vector<uint8_t> data;
int n_tokens = 0;
int n_embd = 0;
bool is_eval_pos = true;
// each element of the vector correspond to one layer
std::vector<struct ggml_tensor *> v_pos; // vector of matrices of size [n_embd, n_tokens]
std::vector<struct ggml_tensor *> v_neg; // vector of matrices of size [n_embd, n_tokens]
std::vector<struct ggml_tensor *> v_final; // vector of finished vectors of size [n_embd]
std::vector<struct ggml_tensor *> v_diff; // vector of matrices of size [n_embd, m] where m ~ n_tokens * n_completions
// each element of the outer vector correspond to one layer, each element of the inner vector correspond to one prompt pass
std::vector<std::vector<struct ggml_tensor *>> v_diffs_wrapped; // vector of compiled diff matrices to be concatenated
// TODO ggml destructor?
~callback_data() {
for (auto ptr : v_pos) free(ptr);
for (auto ptr : v_neg) free(ptr);
for (auto ptr : v_diff) free(ptr);
for (auto ptr : v_final) free(ptr);
for (auto & vec : v_diffs_wrapped) for (auto ptr : vec) free(ptr);
}
};
struct ctrl_params {
/* default meta parameters */
bool always_reload = false;
int n_completions = 64;
int n_threads = 8;
/* default filepaths */
std::string outfile = "control_vector.gguf";
std::string completions_file = "examples/control-vector-generator/completions.txt";
std::string positive_prompts_file = "examples/control-vector-generator/positive.txt";
std::string negative_prompts_file = "examples/control-vector-generator/negative.txt";
/* pair of prompts to be used for generating the vectors */
std::vector<std::string> positive_prompts;
std::vector<std::string> negative_prompts;
/* pair of prompts to be used for testing */
std::vector<std::string> positive_entries;
std::vector<std::string> negative_entries;
};
template <typename T>
static std::string to_string(const T & val) {
std::stringstream ss;
ss << val;
return ss.str();
}
static void print_usage(const char * executable) {
printf("\n");
printf("usage: %s [options] -m <model> [gpt-opts]", executable);
printf("\n");
printf("Creates a GGUF control vector for a given model.");
printf("\n");
printf("options:\n");
printf(" -h, --help show this help message and exit\n");
printf(" -o, --outfile output file\n");
printf(" default: 'control_vector.gguf'\n");
printf(" -pf, --positive-file positive prompts file, one prompt per line\n");
printf(" default: 'examples/control-vector-generator/positive.txt'\n");
printf(" -nf, --negative-file negative prompts file, one prompt per line\n");
printf(" default: 'examples/control-vector-generator/negative.txt'\n");
printf(" -cf, --completions-file completions file\n");
printf(" default: 'examples/control-vector-generator/completions.txt'\n");
printf(" -nc, --num-completions N number of lines of completions file to use\n");
printf(" default: 64\n");
printf(" -t, --num-threads N number of threads to use (do not confuse with gpt-opts -t)\n");
printf(" default: 8\n");
printf(" --always-reload reload the model for every new template to parse (not recommended)\n");
printf("\n");
printf("gpt-opts:\n");
printf(" other options from main\n");
printf("\n");
}
static int ctrlvec_params_parse_ex(int argc, char ** argv, ctrl_params & params) {
std::string arg;
const std::string arg_prefix = "-";
// hack to skip ctrlvec args in gpt_parse_params but we'll leave it as is
int skipme = 0;
for(int arg_idx = 1; arg_idx < argc; ++arg_idx) {
arg = argv[arg_idx];
if (arg.compare(0, arg_prefix.size(), arg_prefix) == 0) {
std::replace(arg.begin(), arg.end(), '_', '-');
}
if (arg == "-h" || arg == "--help") {
print_usage(argv[0]);
exit(0);
}
if (arg == "--version") {
fprintf(stderr, "version: %d (%s)\n", LLAMA_BUILD_NUMBER, LLAMA_COMMIT);
fprintf(stderr, "built with %s for %s\n", LLAMA_COMPILER, LLAMA_BUILD_TARGET);
exit(0);
}
if (arg == "--outfile" || arg == "-o") {
if (++arg_idx < argc && strncmp(argv[arg_idx], arg_prefix.c_str(), 2) != 0) {
params.outfile = argv[arg_idx];
skipme += 2;
} else {
throw std::invalid_argument("error: missing argument for " + arg);
}
}
if (arg == "--completions-file" || arg == "-cf") {
if (++arg_idx < argc && strncmp(argv[arg_idx], arg_prefix.c_str(), 2) != 0) {
params.completions_file = argv[arg_idx];
skipme += 2;
} else {
throw std::invalid_argument("error: missing argument for " + arg);
}
}
if (arg == "--positive-file" || arg == "-pf") {
if (++arg_idx < argc && strncmp(argv[arg_idx], arg_prefix.c_str(), 2) != 0) {
params.positive_prompts_file = argv[arg_idx];
skipme += 2;
} else {
throw std::invalid_argument("error: missing argument for " + arg);
}
}
if (arg == "--negative-file" || arg == "-nf") {
if (++arg_idx < argc && strncmp(argv[arg_idx], arg_prefix.c_str(), 2) != 0) {
params.negative_prompts_file = argv[arg_idx];
skipme += 2;
} else {
throw std::invalid_argument("error: missing argument for " + arg);
}
}
if (arg == "--num-completions" || arg == "-nc") {
if (++arg_idx < argc && strncmp(argv[arg_idx], arg_prefix.c_str(), 2) != 0) {
try {
params.n_completions = std::stoi(argv[arg_idx]);
}
catch (const std::invalid_argument & ex) {
throw std::invalid_argument("error: invalid argument for " + arg);
}
skipme += 2;
} else {
throw std::invalid_argument("error: missing argument for " + arg);
}
}
if (arg == "--num-threads" || arg == "-t") {
if (++arg_idx < argc && strncmp(argv[arg_idx], arg_prefix.c_str(), 2) != 0) {
try {
params.n_threads = std::stoi(argv[arg_idx]);
}
catch (const std::invalid_argument & ex) {
throw std::invalid_argument("error: invalid argument for " + arg);
}
skipme += 2;
} else {
throw std::invalid_argument("error: missing argument for " + arg);
}
}
if (arg == "--always-reload") {
params.always_reload = true;
skipme += 1;
}
// TODO it might be nice QoL to have single positive/negative args
// we do not handle any other unknown arguments here because they will be handled by gpt_parse_params
}
return skipme;
}
static int ctrlvec_params_parse(int argc, char ** argv, ctrl_params & params) {
int skipme = 0;
try {
skipme = ctrlvec_params_parse_ex(argc, argv, params);
}
catch (const std::invalid_argument & ex) {
fprintf(stderr, "%s\n", ex.what());
print_usage(argv[0]);
exit(EXIT_FAILURE);
}
return skipme;
}
static std::vector<std::string> ctrlvec_load_prompt_file(std::string path) {
std::vector<std::string> output;
std::ifstream file(path);
if (!file.is_open()) {
throw std::runtime_error("Unable to open file " + path);
}
std::string line;
while (std::getline(file, line)) {
if (!line.empty()) { // skip empty lines
output.push_back(line);
}
}
file.close();
return output;
}
static std::string format_template(std::string persona, std::string suffix) {
//const std::string user_tag = "[INST]";
//const std::string asst_tag = "[/INST]";
//return user_tag + " Act as if you're extremely " + persona + ". " + asst_tag + " " + suffix;
// TODO make this dynamic - allow the user to change it somehow - and adapt based on model
return persona + " " + suffix; // entry in positive/negative.txt must already be formatted i.e. "[INST] Act as if you're extremely happy. [/INST]"
}
static void populate_entries(ctrl_params & cparams, std::string positive, std::string negative) {
std::string line;
std::ifstream completions_file(cparams.completions_file);
int i = 0;
if (completions_file.is_open()) {
while (std::getline(completions_file, line) && i < cparams.n_completions) {
// TODO replicate the truncations done by the python implementation
cparams.positive_entries.push_back(format_template(positive, line));
cparams.negative_entries.push_back(format_template(negative, line));
i++;
}
completions_file.close();
} else {
throw std::invalid_argument("error: invalid completions file or file could not be opened");
}
}
static std::string ggml_ne_string(const ggml_tensor * t) {
std::string str;
for (int i = 0; i < GGML_MAX_DIMS; ++i) {
str += std::to_string(t->ne[i]);
if (i + 1 < GGML_MAX_DIMS) {
str += ", ";
}
}
return str;
}
static bool cb_eval(struct ggml_tensor * t, bool ask, void * user_data) {
auto * cb_data = (callback_data *) user_data;
static const char * l_out_name = "l_out";
const bool is_l_out = strncmp(t->name, l_out_name, strlen(l_out_name)) == 0;
const struct ggml_tensor * src0 = t->src[0];
const struct ggml_tensor * src1 = t->src[1];
if (ask) {
return is_l_out;
}
if (!is_l_out || t->ne[1] != cb_data->n_tokens) {
return true;
}
char src1_str[128] = {0};
if (src1) {
sprintf(src1_str, "%s{%s}", src1->name, ggml_ne_string(src1).c_str());
}
printf("%s: %24s = (%s) %10s(%s{%s}, %s}) = {%s}\n", __func__,
t->name, ggml_type_name(t->type), ggml_op_desc(t),
src0->name, ggml_ne_string(src0).c_str(),
src1 ? src1_str : "",
ggml_ne_string(t).c_str());
// copy the data from the GPU memory if needed
const bool is_host = ggml_backend_buffer_is_host(t->buffer);
// TODO does this look right?
struct ggml_tensor * t_host;
if (!is_host) {
auto n_bytes = ggml_nbytes(t);
struct ggml_init_params params = {
/*.mem_size =*/ n_bytes,
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ false,
};
struct ggml_context * ctx_data = ggml_init(params);
t_host = ggml_new_tensor_2d(ctx_data, t->type, t->ne[0], t->ne[1]);
ggml_backend_tensor_get(t, t_host->data, 0, n_bytes);
}
else t_host = t;
if (t_host->type == GGML_TYPE_F32) {
if (cb_data->is_eval_pos) {
cb_data->v_pos.push_back(t_host);
} else {
cb_data->v_neg.push_back(t_host);
}
}
return true;
}
static bool get_hidden_layers(llama_context * ctx, std::vector<llama_token> & tokens) {
if (llama_decode(ctx, llama_batch_get_one(tokens.data(), tokens.size(), 0, 0))) {
fprintf(stderr, "%s : failed to eval\n", __func__);
return false;
}
return true;
}
static void padding_seq(llama_context * ctx, std::vector<llama_token> & tokens, size_t len) {
// TODO: customize padding token
std::vector<llama_token> pad_tokens = ::llama_tokenize(ctx, " ", false);
llama_token pad_tok = pad_tokens.back();
while (tokens.size() < len) {
tokens.push_back(pad_tok);
}
}
static bool is_row_all_zeros(struct ggml_tensor * diff, size_t row, size_t cols, float eps = 1e-6) {
for (size_t i = 0; i < cols; ++i) if (ggml_get_f32_nd(diff, row, i, 0, 0) > eps) return false;
return true;
}
static void calc_diff(callback_data & cb_data) {
// TODO: assert cb_data.v_pos.size() == cb_data.v_neg.size()
cb_data.v_diffs_wrapped.resize(cb_data.v_pos.size());
for (size_t il = 0; il < cb_data.v_pos.size(); il++) {
auto & inp_pos = cb_data.v_pos[il];
auto & inp_neg = cb_data.v_neg[il];
auto n_bytes = ggml_nbytes(inp_pos);
struct ggml_init_params params = {
/*.mem_size =*/ n_bytes,
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ false,
};
struct ggml_context * ctx_data = ggml_init(params);
// TODO is this the best way to get dimension? i don't know which way n_embd/n_tokens go
// for that matter can we get rid of n_embd/n_tokens fields in favor of ne[0]/ne[1]?
struct ggml_tensor * dest = ggml_new_tensor_2d(ctx_data, GGML_TYPE_F32, inp_pos->ne[0], inp_pos->ne[1]);
for (size_t i = 0; i < cb_data.n_embd; i++) {
for (size_t j = 0; j < cb_data.n_tokens; j++) {
ggml_set_f32_nd(dest, i, j, 0, 0, ggml_get_f32_nd(inp_pos, i, j, 0, 0) - ggml_get_f32_nd(inp_neg, i, j, 0, 0));
}
}
// TODO can we make this faster? like check during the above operation rather than on a second pass?
// strip zero rows
std::vector<size_t> nonzero_rows;
for (int i = 0; i < cb_data.n_tokens; ++i) {
if (!is_row_all_zeros(dest, i, cb_data.n_embd)) {
nonzero_rows.push_back(i);
}
}
/* debug
if(cb_data.n_tokens != nonzero_rows.size()) {
std::cout << "original n_tokens: " << cb_data.n_tokens << std::endl;
std::cout << "zero rows in layer " << il << ": " << cb_data.n_tokens - nonzero_rows.size() << std::endl;
} */
// TODO I don't know if this is the right dimension but presumably it is
struct ggml_tensor * diff = ggml_new_tensor_2d(ctx_data, GGML_TYPE_F32, nonzero_rows.size(), inp_pos->ne[1]);
size_t offset = 0;
for (size_t i = 0; i < nonzero_rows.size(); ++i) {
float * origin = (float *)(dest->data) + nonzero_rows[i] * cb_data.n_embd;
memcpy((float *)(diff->data) + offset, origin, cb_data.n_embd * sizeof(float));
offset += cb_data.n_embd;
}
cb_data.v_diffs_wrapped[il].push_back(diff);
free(dest);
}
}
static void concatenate_diffs(callback_data & cb_data) {
for (size_t i = 0; i < cb_data.v_diffs_wrapped.size(); ++i) {
std::vector<struct ggml_tensor *> & vec = cb_data.v_diffs_wrapped[i];
size_t n_rows_total = 0;
for (size_t j = 0; j < vec.size(); ++j) {
// TODO likewise no clue if this is right
n_rows_total += vec[j]->ne[0];
}
struct ggml_init_params params = {
/*.mem_size =*/ cb_data.n_embd * n_rows_total * sizeof(float),
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ false,
};
struct ggml_context * ctx_data = ggml_init(params);
// std::cout << "n_rows_total: " << n_rows_total << std::endl;
struct ggml_tensor * diff = ggml_new_tensor_2d(ctx_data, GGML_TYPE_F32, cb_data.n_embd, n_rows_total);
size_t offset = 0;
for (size_t j = 0; j < vec.size(); ++j) {
float * origin = (float *)(vec[j]->data);
// TODO again not sure about dimension
memcpy((float *)(diff->data) + offset, origin, vec[j]->ne[0] * cb_data.n_embd * sizeof(float));
offset += vec[j]->ne[0] * cb_data.n_embd;
}
cb_data.v_diff.push_back(diff);
}
}
struct pca_model {
struct ggml_tensor * v_diff_original;
struct ggml_tensor * square;
struct ggml_tensor * eigenvector;
ggml_backend_t backend = NULL;
ggml_backend_buffer_t buffer;
struct ggml_context * ctx;
};
void load_pca_model(pca_model & model, struct ggml_tensor * v_diff_original, const int n_embd) {
#ifdef GGML_USE_CUDA
fprintf(stderr, "%s: using CUDA backend\n", __func__);
model.backend = ggml_backend_cuda_init(0); // init device 0
if (!model.backend) {
fprintf(stderr, "%s: ggml_backend_cuda_init() failed\n", __func__);
}
#endif
#ifdef GGML_USE_METAL
fprintf(stderr, "%s: using Metal backend\n", __func__);
ggml_backend_metal_log_set_callback(ggml_log_callback_default, nullptr);
model.backend = ggml_backend_metal_init();
if (!model.backend) {
fprintf(stderr, "%s: ggml_backend_metal_init() failed\n", __func__);
}
#endif
// if there aren't GPU Backends fallback to CPU backend
if (!model.backend) {
model.backend = ggml_backend_cpu_init();
}
const int num_tensors = 3;
struct ggml_init_params params {
/*.mem_size =*/ ggml_tensor_overhead() * num_tensors,
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ true,
};
model.ctx = ggml_init(params);
model.v_diff_original = ggml_new_tensor_2d(model.ctx, GGML_TYPE_F32, v_diff_original->ne[0], v_diff_original->ne[1]);
model.square = ggml_new_tensor_2d(model.ctx, GGML_TYPE_F32, n_embd, n_embd);
model.eigenvector = ggml_new_tensor_1d(model.ctx, GGML_TYPE_F32, n_embd);
model.buffer = ggml_backend_alloc_ctx_tensors(model.ctx, model.backend);
ggml_backend_tensor_set(model.v_diff_original, v_diff_original->data, 0, ggml_nbytes(v_diff_original));
// no need to load anything into square yet
// initialize model.eigenvector to random vector
std::vector<float> random_vec = std::vector<float>();
std::default_random_engine generator(static_cast<unsigned int>(std::time(0)));
std::uniform_real_distribution<float> distribution(0.0, 1.0);
for (int i = 0; i < n_embd; ++i) {
random_vec.push_back(distribution(generator));
}
// we don't normalize it at first but that shouldn't be a problem
ggml_backend_tensor_set(model.eigenvector, random_vec.data(), 0, ggml_nbytes(model.eigenvector));
}
struct ggml_cgraph * square_diff_graph(const pca_model & model) {
static size_t buf_size = ggml_tensor_overhead()*GGML_DEFAULT_GRAPH_SIZE + ggml_graph_overhead();
static std::vector<uint8_t> buf(buf_size);
struct ggml_init_params params0 = {
/*.mem_size =*/ buf_size,
/*.mem_buffer =*/ buf.data(),
/*.no_alloc =*/ true, // the tensors will be allocated later by ggml_allocr_alloc_graph()
};
struct ggml_context * ctx0 = ggml_init(params0);
struct ggml_cgraph * gf = ggml_new_graph(ctx0);
struct ggml_tensor * square = ggml_mul_mat(ctx0, model.v_diff_original, model.v_diff_original);
ggml_build_forward_expand(gf, square);
ggml_free(ctx0);
return gf;
}
struct ggml_tensor * compute_square(const pca_model & model, ggml_gallocr_t allocr, int n_threads) {
struct ggml_cgraph * gf = square_diff_graph(model);
ggml_gallocr_alloc_graph(allocr, gf);
if (ggml_backend_is_cpu(model.backend)) {
ggml_backend_cpu_set_n_threads(model.backend, n_threads);
}
#ifdef GGML_USE_METAL
if (ggml_backend_is_metal(model.backend)) {
ggml_backend_metal_set_n_cb(model.backend, n_threads);
}
#endif
ggml_backend_graph_compute(model.backend, gf);
return gf->nodes[gf->n_nodes - 1];
}
struct ggml_cgraph * power_iteration_graph(const pca_model & model, float tolerance) {
static size_t buf_size = ggml_tensor_overhead()*GGML_DEFAULT_GRAPH_SIZE + ggml_graph_overhead();
static std::vector<uint8_t> buf(buf_size);
struct ggml_init_params params0 = {
/*.mem_size =*/ buf_size,
/*.mem_buffer =*/ buf.data(),
/*.no_alloc =*/ true, // the tensors will be allocated later by ggml_allocr_alloc_graph()
};
struct ggml_context * ctx0 = ggml_init(params0);
struct ggml_cgraph * gf = ggml_new_graph(ctx0);
struct ggml_tensor * b_tensor = ggml_mul_mat(ctx0, model.square, model.eigenvector);
// TODO difference between ggml_norm and ggml_norm_inplace?
// also is this the right way to do multi-step graphs?
b_tensor = ggml_norm_inplace(ctx0, b_tensor, tolerance);
ggml_build_forward_expand(gf, b_tensor);
ggml_free(ctx0);
return gf;
}
struct ggml_tensor * compute_piter(const pca_model & model, ggml_gallocr_t allocr, int n_threads, float tolerance) {
struct ggml_cgraph * gf = power_iteration_graph(model, tolerance);
ggml_gallocr_alloc_graph(allocr, gf);
if (ggml_backend_is_cpu(model.backend)) {
ggml_backend_cpu_set_n_threads(model.backend, n_threads);
}
#ifdef GGML_USE_METAL
if (ggml_backend_is_metal(model.backend)) {
ggml_backend_metal_set_n_cb(model.backend, n_threads);
}
#endif
ggml_backend_graph_compute(model.backend, gf);
return gf->nodes[gf->n_nodes - 1];
}
static void power_iteration(callback_data & cb_data, int idx, int n_threads, int maxIterations = 1000, float tolerance = 1e-8) {
pca_model model;
load_pca_model(model, cb_data.v_diff[idx], cb_data.n_embd);
ggml_gallocr_t allocr = NULL;
{
allocr = ggml_gallocr_new(ggml_backend_get_default_buffer_type(model.backend));
// create the worst case graph for memory usage estimation
struct ggml_cgraph * gf = square_diff_graph(model);
ggml_gallocr_reserve(allocr, gf);
size_t mem_size = ggml_gallocr_get_buffer_size(allocr, 0);
fprintf(stderr, "%s: square diff, compute buffer size: %.4f KB\n", __func__, mem_size/1024.0);
}
struct ggml_tensor * square = compute_square(model, allocr, n_threads);
ggml_backend_tensor_get(square, model.square->data, 0, ggml_nbytes(square));
// yes?
ggml_gallocr_free(allocr);
for (int iter = 0; iter < maxIterations; ++iter) {
// TODO do I need to reset it like this every time?
allocr = ggml_gallocr_new(ggml_backend_get_default_buffer_type(model.backend));
struct ggml_tensor * host_new_eigenvector = ggml_new_tensor_1d(model.ctx, GGML_TYPE_F32, cb_data.n_embd);
struct ggml_tensor * b_tensor = compute_piter(model, allocr, n_threads, tolerance);
ggml_backend_tensor_get(b_tensor, host_new_eigenvector->data, 0, ggml_nbytes(b_tensor));
// convergence check
float diff = 0.0;
for (int i = 0; i < cb_data.n_embd; ++i) {
diff += std::pow(((float *)(host_new_eigenvector->data))[i] - ((float *)(model.eigenvector->data))[i], 2);
}
// update eigenvector
ggml_backend_tensor_set(model.eigenvector, host_new_eigenvector->data, 0, ggml_nbytes(model.eigenvector));
try {
if (std::sqrt(diff) < tolerance) {
break;
}
}
catch (std::exception & e) {
// catch division by zero I guess
break;
}
}
// push back v_final with eigenvector
ggml_backend_tensor_get(model.eigenvector, cb_data.v_final[idx]->data, 0, ggml_nbytes(model.eigenvector));
}
static void pca(callback_data & cb_data, int n_threads) {
printf("Running PCA...\n");
for (int il = 0; il < cb_data.v_diff.size(); ++il) {
struct ggml_init_params params = {
/*.mem_size =*/ cb_data.n_embd * sizeof(float),
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ false,
};
struct ggml_context * ctx_data = ggml_init(params);
cb_data.v_final.push_back(ggml_new_tensor_1d(ctx_data, GGML_TYPE_F32, cb_data.n_embd));
power_iteration(cb_data, il, n_threads);
printf("Done with layer %d\n", il);
printf("il = %d | %f %f \n", il, ggml_get_f32_1d(cb_data.v_final[il], 0), ggml_get_f32_1d(cb_data.v_final[il], 1));
}
printf("Done with PCA.\n");
printf("Done with PCA.\n");
}
static void export_gguf(callback_data & cb_data, int n_layers, const std::string fname, const std::string model_hint) {
struct gguf_context * ctx = gguf_init_empty();
size_t v_final_size_eff = n_layers - 1;
const std::string arch = "controlvector";
gguf_set_val_str(ctx, "general.architecture", arch.c_str());
gguf_set_val_str(ctx, (arch + ".model_hint").c_str(), model_hint.c_str());
gguf_set_val_i32(ctx, (arch + ".layer_count").c_str(), v_final_size_eff);
for (size_t i = 0; i < v_final_size_eff; ++i) {
// TODO this number is probably not right - figure out which layer is which
// i'm pretty sure it's right now
const std::string name = "direction." + to_string(i+1);
ggml_set_name(cb_data.v_final[i], name.c_str());
gguf_add_tensor(ctx, cb_data.v_final[i]);
printf("Added tensor %zu\n", i);
}
printf("Writing file...\n");
gguf_write_to_file(ctx, fname.c_str(), false);
printf("%s: wrote file '%s'\n", __func__, fname.c_str());
gguf_free(ctx);
}
// END NON-GGML IMPLEMENTATION
int main(int argc, char ** argv) {
ctrl_params cparams;
int skipme = ctrlvec_params_parse(argc, argv, cparams);
argc -= skipme;
argv += skipme;
gpt_params params;
if (!gpt_params_parse(argc, argv, params)) {
return 1;
}
// load prompts
cparams.positive_prompts = ctrlvec_load_prompt_file(cparams.positive_prompts_file);
cparams.negative_prompts = ctrlvec_load_prompt_file(cparams.negative_prompts_file);
if (cparams.positive_prompts.size() != cparams.negative_prompts.size()) {
fprintf(stderr, "number of positive and negative prompts must be equal");
return 1;
}
if (cparams.positive_prompts.empty()) {
fprintf(stderr, "must provide at least one prompt pair");
return 1;
}
callback_data cb_data;
// pass the callback to the backend scheduler
// it will be executed for each node during the graph computation
params.cb_eval = cb_eval;
params.cb_eval_user_data = &cb_data;
params.warmup = false;
print_build_info();
llama_backend_init();
llama_numa_init(params.numa);
// load the model to get hparams
llama_model * model;
llama_context * ctx;
std::tie(model, ctx) = llama_init_from_gpt_params(params);
int n_ctx = llama_n_ctx(ctx);
int n_layers = llama_n_layer(model);
int n_embd = llama_n_embd(model);
cb_data.n_embd = n_embd;
int n_prompts = cparams.positive_prompts.size();
// create templated prompts
for (int i = 0; i < n_prompts; ++i) {
populate_entries(cparams, cparams.positive_prompts[i], cparams.negative_prompts[i]);
}
const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
int token_ct = 0;
for(size_t i = 0; i < cparams.positive_entries.size(); ++i) {
std::string positive_prompt = cparams.positive_entries[i];
std::string negative_prompt = cparams.negative_entries[i];
std::vector<llama_token> tokens_pos = ::llama_tokenize(ctx, positive_prompt, add_bos);
std::vector<llama_token> tokens_neg = ::llama_tokenize(ctx, negative_prompt, add_bos);
size_t max_seq_len = std::max(tokens_pos.size(), tokens_neg.size());
padding_seq(ctx, tokens_pos, max_seq_len);
padding_seq(ctx, tokens_neg, max_seq_len);
cb_data.n_tokens = max_seq_len;
// need to reload the model so it doesn't run out of context
// this should scale with -c option passed by main
token_ct += 2 * max_seq_len;
if (token_ct > n_ctx || cparams.always_reload) {
//break;
llama_free(ctx);
llama_free_model(model);
std::tie(model, ctx) = llama_init_from_gpt_params(params);
token_ct = 2 * max_seq_len;
}
if (token_ct > n_ctx) {
fprintf(stderr, "context size exceeded on iteration %zu\n", i);
break;
}
printf("Evaluating prompt: \"%s\" - \"%s\" (%ld tokens)\n", positive_prompt.c_str(), negative_prompt.c_str(), max_seq_len);
cb_data.is_eval_pos = true;
get_hidden_layers(ctx, tokens_pos);
cb_data.is_eval_pos = false;
get_hidden_layers(ctx, tokens_neg);
// TODO check whether the same tokens correspond to zero rows because we don't seem to be getting many zero rows anymore
// we get a lot of zero rows for the first few prompts and then they drop off
// likewise most of the zero rows are in the first few layers for each prompt
calc_diff(cb_data);
// reset for next iteration
for (auto ptr : cb_data.v_pos) free(ptr);
for (auto ptr : cb_data.v_neg) free(ptr);
cb_data.v_pos.clear();
cb_data.v_neg.clear();
}
concatenate_diffs(cb_data);
pca(cb_data, cparams.n_threads);
//printf("v_final %f %f \n", cb_data.v_final[0][0], cb_data.v_final[0][1]);
llama_free(ctx);
llama_free_model(model);
// TODO figure out how to extract this from model - there's no API exposed to get model arch string
// we need get_arch_name() from llama.cpp
// TODO also has support been implemeneted for arches other than llama yet? see #5970
std::string model_hint = "llama";
export_gguf(cb_data, n_layers, cparams.outfile, model_hint);
llama_backend_free();
return 0;
}