gpt4all/gpt4all-backend/gptj.cpp

820 lines
25 KiB
C++
Raw Normal View History

#define GPTJ_H_I_KNOW_WHAT_I_AM_DOING_WHEN_INCLUDING_THIS_FILE
#include "gptj_impl.h"
2023-04-09 03:28:39 +00:00
#include "utils.h"
#include "llmodel_shared.h"
2023-04-09 03:28:39 +00:00
#include <cassert>
#include <cinttypes>
2023-04-09 03:28:39 +00:00
#include <cmath>
#include <cstdio>
#include <cstring>
#include <map>
#include <string>
#include <vector>
#include <iostream>
#if defined(_WIN32) && defined(_MSC_VER)
#define WIN32_LEAN_AND_MEAN
#ifndef NOMINMAX
#define NOMINMAX
#endif
#include <windows.h>
#include <io.h>
#include <stdio.h>
#else
#include <unistd.h>
#endif
2023-05-05 14:00:05 +00:00
#include <sstream>
#include <unordered_set>
#include <ggml.h>
namespace {
const char *modelType_ = "GPT-J";
}
2023-05-05 14:00:05 +00:00
// default hparams (GPT-J 6B)
2023-04-09 03:28:39 +00:00
struct gptj_hparams {
int32_t n_vocab = 50400;
int32_t n_ctx = 2048;
int32_t n_embd = 4096;
int32_t n_head = 16;
int32_t n_layer = 28;
int32_t n_rot = 64;
2023-09-28 20:17:06 +00:00
float norm_eps = 1e-5;
2023-04-09 03:28:39 +00:00
};
struct gptj_layer {
// normalization
struct ggml_tensor * ln_1_g;
struct ggml_tensor * ln_1_b;
// attention
struct ggml_tensor * c_attn_q_proj_w;
struct ggml_tensor * c_attn_k_proj_w;
struct ggml_tensor * c_attn_v_proj_w;
struct ggml_tensor * c_attn_proj_w;
// ff
struct ggml_tensor * c_mlp_fc_w;
struct ggml_tensor * c_mlp_fc_b;
struct ggml_tensor * c_mlp_proj_w;
struct ggml_tensor * c_mlp_proj_b;
};
struct gptj_model {
gptj_hparams hparams;
// normalization
struct ggml_tensor * ln_f_g;
struct ggml_tensor * ln_f_b;
struct ggml_tensor * wte; // position embedding
struct ggml_tensor * lmh_g; // language model head
struct ggml_tensor * lmh_b; // language model bias
std::vector<gptj_layer> layers;
// key + value memory
struct llm_kv_cache kv_self;
2023-04-09 03:28:39 +00:00
//
struct ggml_context * ctx;
std::map<std::string, struct ggml_tensor *> tensors;
2023-05-05 14:00:05 +00:00
llm_buffer eval_buf;
llm_buffer scr0_buf;
llm_buffer scr1_buf;
2023-05-05 14:00:05 +00:00
~gptj_model() {
if (ctx) {
ggml_free(ctx);
}
}
2023-04-09 03:28:39 +00:00
};
2023-05-05 14:00:05 +00:00
static bool kv_cache_init(
const struct gptj_hparams & hparams,
struct llm_kv_cache & cache,
2023-05-05 14:00:05 +00:00
ggml_type wtype,
int n_ctx) {
const int n_embd = hparams.n_embd;
const int n_layer = hparams.n_layer;
const int64_t n_mem = (int64_t)n_layer*n_ctx;
const int64_t n_elements = n_embd*n_mem;
2023-06-16 19:56:22 +00:00
cache.buf.resize(2u*n_elements*ggml_type_size(wtype) + 2_MiB);
2023-05-05 14:00:05 +00:00
struct ggml_init_params params;
params.mem_size = cache.buf.size;
params.mem_buffer = cache.buf.addr;
params.no_alloc = false;
cache.ctx = ggml_init(params);
if (!cache.ctx) {
fprintf(stderr, "%s: failed to allocate memory for kv cache\n", __func__);
return false;
}
cache.k = ggml_new_tensor_1d(cache.ctx, wtype, n_elements);
cache.v = ggml_new_tensor_1d(cache.ctx, wtype, n_elements);
return true;
}
2023-09-28 20:17:06 +00:00
// load the model's weights from a file path
bool gptj_model_load(const std::string &fname, gptj_model & model, gpt_vocab & vocab, size_t * mem_req = nullptr) {
2023-04-09 03:28:39 +00:00
printf("%s: loading model from '%s' - please wait ...\n", __func__, fname.c_str());
if(mem_req != nullptr) {
*mem_req = 0;
}
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
// create the ggml context
struct gguf_init_params params = {
/*.no_alloc = */ false,
/*.ctx = */ &model.ctx,
};
gguf_context *ggufctx = gguf_init_from_file(fname.c_str(), params);
if (!ggufctx) {
fprintf(stderr, "%s: gguf_init_from_file() failed\n", __func__);
return false;
2023-04-09 03:28:39 +00:00
}
// load hparams
{
auto & hparams = model.hparams;
2023-09-28 20:17:06 +00:00
bool ok = false;
int keyidx;
do {
keyidx = gguf_find_key(ggufctx, "gptj.context_length");
if (keyidx == -1) { break; }
hparams.n_ctx = gguf_get_val_u32(ggufctx, keyidx);
keyidx = gguf_find_key(ggufctx, "gptj.embedding_length");
if (keyidx == -1) { break; }
hparams.n_embd = gguf_get_val_u32(ggufctx, keyidx);
keyidx = gguf_find_key(ggufctx, "gptj.attention.head_count");
if (keyidx == -1) { break; }
hparams.n_head = gguf_get_val_u32(ggufctx, keyidx);
keyidx = gguf_find_key(ggufctx, "gptj.block_count");
if (keyidx == -1) { break; }
hparams.n_layer = gguf_get_val_u32(ggufctx, keyidx);
keyidx = gguf_find_key(ggufctx, "gptj.rope.dimension_count");
if (keyidx == -1) { break; }
hparams.n_rot = gguf_get_val_u32(ggufctx, keyidx);
keyidx = gguf_find_key(ggufctx, "gptj.attention.layer_norm_epsilon");
if (keyidx == -1) { break; }
hparams.norm_eps = gguf_get_val_f32(ggufctx, keyidx);
ok = true;
} while (false);
if (!ok) {
fprintf(stderr, "%s: required hparam missing!\n", __func__);
return false;
}
2023-04-09 03:28:39 +00:00
printf("%s: n_ctx = %d\n", __func__, hparams.n_ctx);
printf("%s: n_embd = %d\n", __func__, hparams.n_embd);
printf("%s: n_head = %d\n", __func__, hparams.n_head);
printf("%s: n_layer = %d\n", __func__, hparams.n_layer);
printf("%s: n_rot = %d\n", __func__, hparams.n_rot);
}
// load vocab
{
2023-09-28 20:17:06 +00:00
auto & hparams = model.hparams;
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
int keyidx = gguf_find_key(ggufctx, "tokenizer.ggml.model");
if (keyidx == -1) {
fprintf(stderr, "%s: tokenizer model not found!\n", __func__);
return false;
}
if (strcmp(gguf_get_val_str(ggufctx, keyidx), "gpt2") != 0) {
fprintf(stderr, "%s: tokenizer model not supported!\n", __func__);
2023-04-09 03:28:39 +00:00
return false;
}
2023-09-28 20:17:06 +00:00
int tokens_keyidx = gguf_find_key(ggufctx, "tokenizer.ggml.tokens");
if (tokens_keyidx == -1) {
fprintf(stderr, "%s: gpt2 tokenizer vocab not found!\n", __func__);
return false;
}
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
hparams.n_vocab = gguf_get_arr_n(ggufctx, tokens_keyidx);
printf("%s: gpt2 tokenizer vocab = %d\n", __func__, int(hparams.n_vocab));
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
for (int i = 0; i < hparams.n_vocab; i++) {
std::string word = gguf_get_arr_str(ggufctx, tokens_keyidx, i);
2023-04-09 03:28:39 +00:00
vocab.token_to_id[word] = i;
vocab.id_to_token[i] = word;
}
}
auto & ctx = model.ctx;
2023-09-28 20:17:06 +00:00
size_t ctx_size = ggml_get_mem_size(ctx);
printf("%s: ggml ctx size = %6.2f MB\n", __func__, ctx_size / (1024.0 * 1024.0));
2023-04-09 03:28:39 +00:00
if (mem_req != nullptr) {
2023-09-28 20:17:06 +00:00
*mem_req = ctx_size;
gguf_free(ggufctx);
return false;
}
2023-04-09 03:28:39 +00:00
// prepare memory for the weights
{
const auto & hparams = model.hparams;
2023-10-02 14:04:38 +00:00
model.layers.resize(hparams.n_layer);
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
model.wte = ggml_get_tensor(ctx, "token_embd.weight");
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
model.ln_f_g = ggml_get_tensor(ctx, "output_norm.weight");
model.ln_f_b = ggml_get_tensor(ctx, "output_norm.bias");
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
model.lmh_g = ggml_get_tensor(ctx, "output.weight");
model.lmh_b = ggml_get_tensor(ctx, "output.bias");
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
auto name = [](int i, std::string n) {
static std::string key;
key = "blk." + std::to_string(i) + "." + n;
return key.c_str();
};
2023-04-09 03:28:39 +00:00
2023-10-02 14:04:38 +00:00
for (int i = 0; i < hparams.n_layer; ++i) {
2023-04-09 03:28:39 +00:00
auto & layer = model.layers[i];
2023-09-28 20:17:06 +00:00
layer.ln_1_g = ggml_get_tensor(ctx, name(i, "attn_norm.weight"));
layer.ln_1_b = ggml_get_tensor(ctx, name(i, "attn_norm.bias"));
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
layer.c_attn_q_proj_w = ggml_get_tensor(ctx, name(i, "attn_q.weight"));
layer.c_attn_k_proj_w = ggml_get_tensor(ctx, name(i, "attn_k.weight"));
layer.c_attn_v_proj_w = ggml_get_tensor(ctx, name(i, "attn_v.weight"));
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
layer.c_attn_proj_w = ggml_get_tensor(ctx, name(i, "attn_output.weight"));
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
layer.c_mlp_fc_w = ggml_get_tensor(ctx, name(i, "ffn_up.weight"));
layer.c_mlp_fc_b = ggml_get_tensor(ctx, name(i, "ffn_up.bias"));
2023-04-09 03:28:39 +00:00
2023-09-28 20:17:06 +00:00
layer.c_mlp_proj_w = ggml_get_tensor(ctx, name(i, "ffn_down.weight"));
layer.c_mlp_proj_b = ggml_get_tensor(ctx, name(i, "ffn_down.bias"));
2023-04-09 03:28:39 +00:00
}
}
// key + value memory
{
const auto & hparams = model.hparams;
if (!kv_cache_init(hparams, model.kv_self, GGML_TYPE_F16, model.hparams.n_ctx)) {
2023-05-05 14:00:05 +00:00
fprintf(stderr, "%s: kv_cache_init() failed for self-attention cache\n", __func__);
ggml_free(ctx);
return false;
}
2023-04-09 03:28:39 +00:00
2023-05-05 14:00:05 +00:00
const size_t memory_size = ggml_nbytes(model.kv_self.k) + ggml_nbytes(model.kv_self.v);
printf("%s: kv self size = %7.2f MB\n", __func__, memory_size / 1024.0 / 1024.0);
2023-04-09 03:28:39 +00:00
}
model.scr0_buf.resize(256u * 1024 * 1024);
model.scr1_buf.resize(256u * 1024 * 1024);
2023-04-09 03:28:39 +00:00
return true;
}
// evaluate the transformer
//
// - model: the model
// - n_threads: number of threads to use
// - n_past: the context size so far
// - embd_inp: the embeddings of the tokens in the context
// - embd_w: the predicted logits for the next token
//
// The GPT-J model requires about 16MB of memory per input token.
//
bool gptj_eval(
2023-05-05 14:00:05 +00:00
gptj_model & model,
2023-04-09 03:28:39 +00:00
const int n_threads,
const int n_past,
const std::vector<gpt_vocab::id> & embd_inp,
std::vector<float> & embd_w,
size_t & mem_per_token) {
const int N = embd_inp.size();
const auto & hparams = model.hparams;
const int n_embd = hparams.n_embd;
const int n_layer = hparams.n_layer;
const int n_ctx = hparams.n_ctx;
const int n_head = hparams.n_head;
const int n_vocab = hparams.n_vocab;
const int n_rot = hparams.n_rot;
2023-06-16 19:56:22 +00:00
const size_t init_buf_size = 1024_MiB;
if (!model.eval_buf.addr || model.eval_buf.size < init_buf_size)
model.eval_buf.resize(init_buf_size);
2023-04-09 03:28:39 +00:00
if (mem_per_token > 0 && mem_per_token*N > model.eval_buf.size) {
2023-04-09 03:28:39 +00:00
const size_t buf_size_new = 1.1*(mem_per_token*N); // add 10% to account for ggml object overhead
printf("\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, model.eval_buf.size, buf_size_new);
2023-04-09 03:28:39 +00:00
// reallocate
model.eval_buf.resize(buf_size_new);
if (model.eval_buf.addr == nullptr) {
fprintf(stderr, "%s: failed to allocate %zu bytes\n", __func__, model.eval_buf.size);
2023-04-09 03:28:39 +00:00
return false;
}
}
struct ggml_init_params params = {
.mem_size = model.eval_buf.size,
.mem_buffer = model.eval_buf.addr,
.no_alloc = false
2023-04-09 03:28:39 +00:00
};
struct ggml_context * ctx0 = ggml_init(params);
struct ggml_cgraph gf = {};
2023-04-09 03:28:39 +00:00
struct ggml_tensor * embd = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, N);
memcpy(embd->data, embd_inp.data(), N*ggml_element_size(embd));
// wte
struct ggml_tensor * inpL = ggml_get_rows(ctx0, model.wte, embd);
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * cur;
ggml_set_scratch(ctx0, {0, model.scr0_buf.size, model.scr0_buf.addr, });
2023-04-09 03:28:39 +00:00
// norm
{
2023-09-28 20:17:06 +00:00
cur = ggml_norm(ctx0, inpL, model.hparams.norm_eps);
2023-04-09 03:28:39 +00:00
// cur = ln_1_g*cur + ln_1_b
cur = ggml_add(ctx0,
ggml_mul(ctx0,
ggml_repeat(ctx0, model.layers[il].ln_1_g, cur),
cur),
ggml_repeat(ctx0, model.layers[il].ln_1_b, cur));
}
struct ggml_tensor * inpSA = cur;
// self-attention
{
struct ggml_tensor * Qcur = ggml_rope(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model.layers[il].c_attn_q_proj_w, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, 0);
struct ggml_tensor * Kcur = ggml_rope(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model.layers[il].c_attn_k_proj_w, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, 0);
2023-04-09 03:28:39 +00:00
// store key and value to memory
2023-05-08 19:08:48 +00:00
{
struct ggml_tensor * Vcur = ggml_transpose(ctx0, ggml_mul_mat(ctx0, model.layers[il].c_attn_v_proj_w, cur));
2023-05-05 14:00:05 +00:00
struct ggml_tensor * k = ggml_view_1d(ctx0, model.kv_self.k, N*n_embd, (ggml_element_size(model.kv_self.k)*n_embd)*(il*n_ctx + n_past));
struct ggml_tensor * v = ggml_view_2d(ctx0, model.kv_self.v, N, n_embd,
( n_ctx)*ggml_element_size(model.kv_self.v),
(il*n_ctx)*ggml_element_size(model.kv_self.v)*n_embd + n_past*ggml_element_size(model.kv_self.v));
2023-04-09 03:28:39 +00:00
ggml_build_forward_expand(&gf, ggml_cpy(ctx0, Kcur, k));
ggml_build_forward_expand(&gf, ggml_cpy(ctx0, Vcur, v));
}
// Q = Qcur.contiguous().view(n_embd/n_head, n_head, N).permute(0, 2, 1, 3)
struct ggml_tensor * Q = ggml_permute(ctx0, Qcur, 0, 2, 1, 3);
2023-04-09 03:28:39 +00:00
// K = Kmem.view(n_embd/n_head, n_head, n_past + N).permute(0, 2, 1, 3)
struct ggml_tensor * K =
ggml_permute(ctx0,
ggml_reshape_3d(ctx0,
ggml_view_1d(ctx0, model.kv_self.k, (n_past + N)*n_embd, il*n_ctx*ggml_element_size(model.kv_self.k)*n_embd),
n_embd/n_head, n_head, n_past + N),
2023-04-09 03:28:39 +00:00
0, 2, 1, 3);
// K * Q
struct ggml_tensor * KQ = ggml_mul_mat(ctx0, K, Q);
// KQ_scaled = KQ / sqrt(n_embd/n_head)
struct ggml_tensor * KQ_scaled =
ggml_scale(ctx0,
KQ,
ggml_new_f32(ctx0, 1.0f/sqrt(float(n_embd)/n_head))
);
// KQ_masked = mask_past(KQ_scaled)
struct ggml_tensor * KQ_masked = ggml_diag_mask_inf(ctx0, KQ_scaled, n_past);
// KQ = soft_max(KQ_masked)
struct ggml_tensor * KQ_soft_max = ggml_soft_max(ctx0, KQ_masked);
// V_trans = Vmem.view(n_embd/n_head, n_head, n_past + N).permute(1, 2, 0, 3).contiguous()
struct ggml_tensor * V =
ggml_view_3d(ctx0, model.kv_self.v,
n_past + N, n_embd/n_head, n_head,
n_ctx*ggml_element_size(model.kv_self.v),
n_ctx*ggml_element_size(model.kv_self.v)*n_embd/n_head,
il*n_ctx*ggml_element_size(model.kv_self.v)*n_embd);
2023-04-09 03:28:39 +00:00
// KQV = transpose(V) * KQ_soft_max
struct ggml_tensor * KQV = ggml_mul_mat(ctx0, V, KQ_soft_max);
2023-04-09 03:28:39 +00:00
// KQV_merged = KQV.permute(0, 2, 1, 3)
struct ggml_tensor * KQV_merged = ggml_permute(ctx0, KQV, 0, 2, 1, 3);
// cur = KQV_merged.contiguous().view(n_embd, N)
cur = ggml_cpy(ctx0,
KQV_merged,
ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, N));
// projection (no bias)
cur = ggml_mul_mat(ctx0,
model.layers[il].c_attn_proj_w,
cur);
}
struct ggml_tensor * inpFF = cur;
ggml_set_scratch(ctx0, {0, model.scr1_buf.size, model.scr1_buf.addr, });
2023-04-09 03:28:39 +00:00
// feed-forward network
// this is independent of the self-attention result, so it could be done in parallel to the self-attention
{
// note here we pass inpSA instead of cur
cur = ggml_mul_mat(ctx0,
model.layers[il].c_mlp_fc_w,
inpSA);
cur = ggml_add(ctx0,
ggml_repeat(ctx0, model.layers[il].c_mlp_fc_b, cur),
cur);
// GELU activation
cur = ggml_gelu(ctx0, cur);
// projection
// cur = proj_w*cur + proj_b
cur = ggml_mul_mat(ctx0,
model.layers[il].c_mlp_proj_w,
cur);
cur = ggml_add(ctx0,
ggml_repeat(ctx0, model.layers[il].c_mlp_proj_b, cur),
cur);
}
// self-attention + FF
cur = ggml_add(ctx0, cur, inpFF);
// input for next layer
inpL = ggml_add(ctx0, cur, inpL);
}
ggml_set_scratch(ctx0, {0, model.scr0_buf.size, model.scr0_buf.addr, });
2023-04-09 03:28:39 +00:00
// norm
{
2023-09-28 20:17:06 +00:00
inpL = ggml_norm(ctx0, inpL, model.hparams.norm_eps);
2023-04-09 03:28:39 +00:00
// inpL = ln_f_g*inpL + ln_f_b
inpL = ggml_add(ctx0,
ggml_mul(ctx0,
ggml_repeat(ctx0, model.ln_f_g, inpL),
inpL),
ggml_repeat(ctx0, model.ln_f_b, inpL));
}
ggml_set_scratch(ctx0, { 0, 0, nullptr, });
2023-04-09 03:28:39 +00:00
// lm_head
{
inpL = ggml_mul_mat(ctx0, model.lmh_g, inpL);
inpL = ggml_add(ctx0,
ggml_repeat(ctx0, model.lmh_b, inpL),
inpL);
}
// logits -> probs
//inpL = ggml_soft_max(ctx0, inpL);
ggml_build_forward_expand(&gf, inpL);
2023-09-28 20:17:06 +00:00
// run the computation
{
std::unique_ptr<uint8_t []> data;
auto plan = ggml_graph_plan(&gf, n_threads);
if (plan.work_size > 0) {
data.reset(new uint8_t[plan.work_size]);
plan.work_data = data.get();
}
ggml_graph_compute(&gf, &plan);
}
2023-04-09 03:28:39 +00:00
//if (n_past%100 == 0) {
// ggml_graph_print (&gf);
// ggml_graph_dump_dot(&gf, NULL, "gpt-2.dot");
//}
//embd_w.resize(n_vocab*N);
//memcpy(embd_w.data(), ggml_get_data(inpL), sizeof(float)*n_vocab*N);
// return result for just the last token
embd_w.resize(n_vocab);
memcpy(embd_w.data(), (float *) ggml_get_data(inpL) + (n_vocab*(N-1)), sizeof(float)*n_vocab);
if (mem_per_token == 0) {
mem_per_token = ggml_used_mem(ctx0)/N;
}
//printf("used_mem = %zu\n", ggml_used_mem(ctx0));
ggml_free(ctx0);
return true;
}
2023-05-05 14:00:05 +00:00
#define GPTJ_MAX_RNG_STATE 64*1024
size_t gptj_get_state_size(const gptj_model &model)
{
// we don't know size of rng until we actually serialize it. so reserve more than enough memory for its serialized state.
// for reference, std::mt19937(1337) serializes to 6701 bytes.
const size_t s_rng_size = sizeof(size_t);
const size_t s_rng = GPTJ_MAX_RNG_STATE;
const size_t s_kv_size = sizeof(size_t);
const size_t s_kv_ntok = sizeof(int);
const size_t s_kv = model.kv_self.buf.size;
const size_t s_total = (
+ s_rng_size
+ s_rng
+ s_kv_size
+ s_kv_ntok
+ s_kv
);
fflush(stdout);
return s_total;
}
size_t gptj_copy_state_data(const gptj_model &model, const std::mt19937 &rng, uint8_t *dest)
{
uint8_t * out = dest;
fflush(stdout);
// copy rng
{
std::stringstream rng_ss;
rng_ss << rng;
const size_t rng_size = rng_ss.str().size();
char rng_buf[GPTJ_MAX_RNG_STATE];
memset(&rng_buf[0], 0, GPTJ_MAX_RNG_STATE);
memcpy(&rng_buf[0], rng_ss.str().data(), rng_ss.str().size());
memcpy(out, &rng_size, sizeof(rng_size)); out += sizeof(rng_size);
memcpy(out, &rng_buf[0], GPTJ_MAX_RNG_STATE); out += GPTJ_MAX_RNG_STATE;
}
// copy kv cache
{
const size_t kv_size = model.kv_self.buf.size;
const int kv_ntok = model.kv_self.n;
memcpy(out, &kv_size, sizeof(kv_size)); out += sizeof(kv_size);
memcpy(out, &kv_ntok, sizeof(kv_ntok)); out += sizeof(kv_ntok);
if (kv_size) {
memcpy(out, model.kv_self.buf.addr, kv_size); out += kv_size;
}
}
const size_t written = out - dest;
assert(written == gptj_get_state_size(model));
2023-05-05 14:00:05 +00:00
fflush(stdout);
return written;
}
size_t gptj_set_state_data(gptj_model *model, std::mt19937 *rng, const uint8_t *src)
{
const uint8_t * in = src;
// set rng
{
size_t rng_size;
char rng_buf[GPTJ_MAX_RNG_STATE];
memcpy(&rng_size, in, sizeof(rng_size)); in += sizeof(rng_size);
memcpy(&rng_buf[0], in, GPTJ_MAX_RNG_STATE); in += GPTJ_MAX_RNG_STATE;
std::stringstream rng_ss;
rng_ss.str(std::string(&rng_buf[0], rng_size));
rng_ss >> *rng;
assert(rng_ss.fail() == false);
}
// set kv cache
{
size_t kv_size;
int kv_ntok;
memcpy(&kv_size, in, sizeof(kv_size)); in += sizeof(kv_size);
memcpy(&kv_ntok, in, sizeof(kv_ntok)); in += sizeof(kv_ntok);
if (kv_size) {
assert(model->kv_self.buf.size == kv_size);
void * k_data = model->kv_self.k->data; // remember data pointers
void * v_data = model->kv_self.v->data; // because their value is stored in buf and overwritten by memcpy
memcpy(model->kv_self.buf.addr, in, kv_size); in += kv_size;
model->kv_self.k->data = k_data; // restore correct data pointers
model->kv_self.v->data = v_data;
}
model->kv_self.n = kv_ntok;
}
const size_t nread = in - src;
assert(nread == gptj_get_state_size(*model));
2023-05-05 14:00:05 +00:00
fflush(stdout);
return nread;
}
2023-04-09 03:28:39 +00:00
struct GPTJPrivate {
const std::string modelPath;
bool modelLoaded;
gpt_vocab vocab;
2023-05-05 14:00:05 +00:00
gptj_model *model = nullptr;
2023-04-09 03:28:39 +00:00
int64_t n_threads = 0;
size_t mem_per_token = 0;
2023-04-09 03:28:39 +00:00
std::mt19937 rng;
};
GPTJ::GPTJ()
: d_ptr(new GPTJPrivate) {
2023-05-05 14:00:05 +00:00
d_ptr->model = new gptj_model;
d_ptr->model->ctx = nullptr;
2023-04-09 03:28:39 +00:00
d_ptr->modelLoaded = false;
}
size_t GPTJ::requiredMem(const std::string &modelPath) {
gptj_model dummy_model;
gpt_vocab dummy_vocab;
size_t mem_req;
2023-09-28 20:17:06 +00:00
gptj_model_load(modelPath, dummy_model, dummy_vocab, &mem_req);
return mem_req;
}
2023-04-29 00:15:10 +00:00
bool GPTJ::loadModel(const std::string &modelPath) {
2023-04-09 03:28:39 +00:00
std::mt19937 rng(time(NULL));
d_ptr->rng = rng;
// load the model
2023-09-28 20:17:06 +00:00
if (!gptj_model_load(modelPath, *d_ptr->model, d_ptr->vocab)) {
2023-04-10 20:33:14 +00:00
std::cerr << "GPT-J ERROR: failed to load model from " << modelPath;
return false;
2023-04-09 03:28:39 +00:00
}
d_ptr->n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
d_ptr->modelLoaded = true;
fflush(stdout);
2023-04-09 03:28:39 +00:00
return true;
}
2023-04-18 13:46:03 +00:00
void GPTJ::setThreadCount(int32_t n_threads) {
d_ptr->n_threads = n_threads;
}
int32_t GPTJ::threadCount() const
{
2023-04-18 13:46:03 +00:00
return d_ptr->n_threads;
}
2023-04-09 03:28:39 +00:00
GPTJ::~GPTJ()
{
2023-05-05 14:00:05 +00:00
delete d_ptr->model;
2023-04-09 03:28:39 +00:00
}
bool GPTJ::isModelLoaded() const
{
return d_ptr->modelLoaded;
}
2023-05-05 14:00:05 +00:00
size_t GPTJ::stateSize() const
{
return gptj_get_state_size(*d_ptr->model);
}
size_t GPTJ::saveState(uint8_t *dest) const
{
return gptj_copy_state_data(*d_ptr->model, d_ptr->rng, dest);
}
size_t GPTJ::restoreState(const uint8_t *src)
{
return gptj_set_state_data(d_ptr->model, &d_ptr->rng, src);
}
std::vector<LLModel::Token> GPTJ::tokenize(PromptContext &, const std::string &str) const
{
return ::gpt_tokenize(d_ptr->vocab, str);
}
LLModel::Token GPTJ::sampleToken(PromptContext &promptCtx) const
{
const size_t n_prev_toks = std::min((size_t) promptCtx.repeat_last_n, promptCtx.tokens.size());
return gpt_sample_top_k_top_p(d_ptr->model->hparams.n_vocab,
promptCtx.tokens.data() + promptCtx.tokens.size() - n_prev_toks,
n_prev_toks,
promptCtx.logits,
promptCtx.top_k, promptCtx.top_p, promptCtx.temp,
promptCtx.repeat_penalty,
d_ptr->rng);
}
2023-04-26 23:08:37 +00:00
std::string GPTJ::tokenToString(Token id) const
{
return d_ptr->vocab.id_to_token[id];
}
2023-04-09 03:28:39 +00:00
bool GPTJ::evalTokens(PromptContext &ctx, const std::vector<int32_t> &tokens) const
{
2023-04-09 03:28:39 +00:00
// determine the required inference memory per token:
2023-04-09 14:27:35 +00:00
static bool initialized = false;
if (!initialized) {
gptj_eval(*d_ptr->model, d_ptr->n_threads, 0, { 0, 1, 2, 3 }, ctx.logits,
d_ptr->mem_per_token);
2023-04-09 14:27:35 +00:00
initialized = true;
}
2023-04-09 03:28:39 +00:00
return gptj_eval(*d_ptr->model, d_ptr->n_threads, ctx.n_past, tokens, ctx.logits, d_ptr->mem_per_token);
}
int32_t GPTJ::contextLength() const
{
return d_ptr->model->hparams.n_ctx;
2023-04-09 03:28:39 +00:00
}
const std::vector<LLModel::Token> &GPTJ::endTokens() const
{
static const std::vector<LLModel::Token> fres = {50256};
return fres;
}
2023-09-28 20:17:06 +00:00
std::string get_arch_name(gguf_context *ctx_gguf) {
std::string arch_name;
const int kid = gguf_find_key(ctx_gguf, "general.architecture");
enum gguf_type ktype = gguf_get_kv_type(ctx_gguf, kid);
if (ktype != GGUF_TYPE_STRING) {
throw std::runtime_error("ERROR: Can't get general architecture from gguf file.");
}
return gguf_get_val_str(ctx_gguf, kid);
}
#if defined(_WIN32)
#define DLL_EXPORT __declspec(dllexport)
#else
#define DLL_EXPORT __attribute__ ((visibility ("default")))
#endif
extern "C" {
DLL_EXPORT bool is_g4a_backend_model_implementation() {
return true;
}
DLL_EXPORT const char *get_model_type() {
return modelType_;
}
DLL_EXPORT const char *get_build_variant() {
return GGML_BUILD_VARIANT;
}
2023-09-28 20:17:06 +00:00
DLL_EXPORT bool magic_match(const char * fname) {
struct ggml_context * ctx_meta = NULL;
struct gguf_init_params params = {
/*.no_alloc = */ true,
/*.ctx = */ &ctx_meta,
};
gguf_context *ctx_gguf = gguf_init_from_file(fname, params);
if (!ctx_gguf)
return false;
bool isValid = gguf_get_version(ctx_gguf) <= 2;
isValid = isValid && get_arch_name(ctx_gguf) == "gptj";
gguf_free(ctx_gguf);
return isValid;
}
DLL_EXPORT LLModel *construct() {
return new GPTJ;
}
}