vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

added ability to fast forward in time through partially duplicated prompts

Browse source
This commit is contained in:
Concedo 2023-03-24 18:50:16 +08:00
parent 1166fda943
commit 706e19e9b4
3 changed files with 53 additions and 53 deletions
Download patch file Download diff file Expand all files Collapse all files

70
expose.cpp
View file

@ -10,6 +10,21 @@
#include "main.cpp"
#include "extra.h"
void print_tok_vec(std::vector<llama_token> & embd)
{
std::cout << "[";
bool first = true;
for (auto i: embd) {
if(!first)
{
std::cout << ',';
}
first = false;
std::cout << i;
}
std::cout << "]";
}
extern "C" {
struct load_model_inputs
@ -31,7 +46,6 @@ extern "C" {
const float top_p;
const float rep_pen;
const int rep_pen_range;
const bool reset_state = true; //determines if we can continue off the previous prompt state
};
struct generation_outputs
{
@ -43,12 +57,12 @@ extern "C" {
llama_context_params ctx_params;
gpt_params params;
int n_past = 0;
llama_token old_embd_id = -1;
int n_threads = 4;
int n_batch = 8;
std::string model;
llama_context * ctx;
std::vector<llama_token> last_n_tokens;
std::vector<llama_token> current_context_tokens;
bool load_model(const load_model_inputs inputs)
{
@ -80,6 +94,10 @@ extern "C" {
printf("\n---\nWarning: Your model is using an OUTDATED format. Please reconvert it for better results!\n");
}
//determine mem per token
const std::vector<llama_token> tmp = { 0, 1, 2, 3 };
llama_eval(ctx, tmp.data(), tmp.size(), 0, params.n_threads);
return true;
}
@ -97,12 +115,6 @@ extern "C" {
params.n_batch = n_batch;
params.n_threads = n_threads;
bool reset_state = inputs.reset_state;
if(n_past==0)
{
reset_state = true;
}
if(params.repeat_last_n<1)
{
params.repeat_last_n = 1;
@ -116,10 +128,7 @@ extern "C" {
params.seed = time(NULL);
}
if(reset_state)
{
params.prompt.insert(0, 1, ' ');
}
// tokenize the prompt
std::vector<llama_token> embd_inp;
@ -136,6 +145,9 @@ extern "C" {
int offset = embd_inp.size() - params.n_ctx + params.n_predict;
embd_inp = std::vector<llama_token>(embd_inp.begin() + offset, embd_inp.end());
}
//determine how much npast we have to rewind from the current state
std::vector<llama_token> embd;
int last_n_size = params.repeat_last_n;
@ -146,25 +158,29 @@ extern "C" {
// char * tst2 = (char*)tst.c_str();
// gpt_print_usage(1,&tst2,params);
if(reset_state)
{
const std::vector<llama_token> tmp = { 0, 1, 2, 3 };
llama_eval(ctx, tmp.data(), tmp.size(), 0, params.n_threads);
std::fill(last_n_tokens.begin(), last_n_tokens.end(), 0);
n_past = 0;
//fast forward the past based on identical tokens, stop once a divergence is noted
for(int i=0;i<current_context_tokens.size();++i)
{
if(current_context_tokens[i]==embd_inp[0])
{
n_past += 1;
embd_inp.erase(embd_inp.begin());
last_n_tokens.erase(last_n_tokens.begin());
last_n_tokens.push_back(current_context_tokens[i]);
}
else
{
//strip out the reset token (1) at the start of the embedding
if(embd_inp.size()>0)
break;
}
if(embd_inp.size()<=1)
{
embd_inp.erase(embd_inp.begin());
}
if(old_embd_id!=-1)
{
embd.push_back(old_embd_id);
break;
}
}
current_context_tokens.resize(n_past);
int remaining_tokens = params.n_predict;
int input_consumed = 0;
@ -181,10 +197,7 @@ extern "C" {
if (embd.size() > 0)
{
printf("|");
// for (auto i: embd) {
// std::cout << i << ',';
// }
// printf("\nnp:%d embd:%d",n_past,embd.size());
//printf("\nnp:%d embd:%d txt:%s",n_past,embd.size(),llama_token_to_str(ctx, embd[0]));
if (llama_eval(ctx, embd.data(), embd.size(), n_past, params.n_threads))
{
fprintf(stderr, "Failed to predict\n");
@ -222,13 +235,12 @@ extern "C" {
last_n_tokens.erase(last_n_tokens.begin());
last_n_tokens.push_back(id);
current_context_tokens.push_back(id);
}
// add it to the context
old_embd_id = id;
embd.push_back(id);
// decrement remaining sampling budget
--remaining_tokens;
//printf("\nid:%d word:%s\n",id,llama_token_to_str(ctx, id));
@ -239,10 +251,10 @@ extern "C" {
// some user input remains from prompt or interaction, forward it to processing
while ((int) embd_inp.size() > input_consumed)
{
old_embd_id = embd_inp[input_consumed];
embd.push_back(embd_inp[input_consumed]);
last_n_tokens.erase(last_n_tokens.begin());
last_n_tokens.push_back(embd_inp[input_consumed]);
current_context_tokens.push_back(embd_inp[input_consumed]);
++input_consumed;
if ((int) embd.size() >= params.n_batch)
{

22
llama_for_kobold.py
View file

@ -23,8 +23,7 @@ class generation_inputs(ctypes.Structure):
("top_k", ctypes.c_int),
("top_p", ctypes.c_float),
("rep_pen", ctypes.c_float),
("rep_pen_range", ctypes.c_int),
("reset_state", ctypes.c_bool)]
("rep_pen_range", ctypes.c_int)]
class generation_outputs(ctypes.Structure):
_fields_ = [("status", ctypes.c_int),
@ -48,7 +47,7 @@ def load_model(model_filename,batch_size=8,max_context_length=512,n_parts_overwr
ret = handle.load_model(inputs)
return ret
def generate(prompt,max_length=20, max_context_length=512,temperature=0.8,top_k=100,top_p=0.85,rep_pen=1.1,rep_pen_range=128,seed=-1,reset_state=True):
def generate(prompt,max_length=20, max_context_length=512,temperature=0.8,top_k=100,top_p=0.85,rep_pen=1.1,rep_pen_range=128,seed=-1):
inputs = generation_inputs()
outputs = ctypes.create_unicode_buffer(ctypes.sizeof(generation_outputs))
inputs.prompt = prompt.encode("UTF-8")
@ -60,7 +59,6 @@ def generate(prompt,max_length=20, max_context_length=512,temperature=0.8,top_k=
inputs.rep_pen = rep_pen
inputs.rep_pen_range = rep_pen_range
inputs.seed = seed
inputs.reset_state = reset_state
ret = handle.generate(inputs,outputs)
if(ret.status==1):
return ret.text.decode("UTF-8")
@ -80,7 +78,6 @@ maxctx = 2048
maxlen = 128
modelbusy = False
port = 5001
last_context = ""
embedded_kailite = None
class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
@ -130,7 +127,6 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
def do_POST(self):
global modelbusy
global last_context
content_length = int(self.headers['Content-Length'])
body = self.rfile.read(content_length)
@ -159,17 +155,13 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
self.end_headers()
return
print("\nInput: " + json.dumps(genparams))
fresh_state = True
modelbusy = True
if kai_api_flag:
fullprompt = genparams.get('prompt', "")
else:
fullprompt = genparams.get('text', "")
newprompt = fullprompt
if last_context!="" and newprompt.startswith(last_context):
fresh_state = False
newprompt = newprompt[len(last_context):]
print("Resuming state, new input len: " + str(len(newprompt)))
recvtxt = ""
@ -183,11 +175,9 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
top_p=genparams.get('top_p', 0.85),
rep_pen=genparams.get('rep_pen', 1.1),
rep_pen_range=genparams.get('rep_pen_range', 128),
seed=-1,
reset_state=fresh_state
seed=-1
)
print("\nOutput: " + recvtxt)
last_context = fullprompt + recvtxt
res = {"results": [{"text": recvtxt}]}
self.send_response(200)
self.end_headers()
@ -201,11 +191,9 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
top_p=genparams.get('top_p', 0.85),
rep_pen=genparams.get('rep_pen', 1.1),
rep_pen_range=genparams.get('rep_pen_range', 128),
seed=-1,
reset_state=fresh_state
seed=-1
)
print("\nOutput: " + recvtxt)
last_context = fullprompt + recvtxt
res = {"data": {"seqs":[recvtxt]}}
self.send_response(200)
self.end_headers()

BIN
llamacpp.dll
View file

Binary file not shown.