dnrops
/
llama.swift
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

remove prints to stderr and return NSErrors

This commit is contained in:
Alex Rozanski 2023-03-14 11:30:37 +01:00
parent 2b27f14035
commit 81e97c8585
2 changed files with 51 additions and 39 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
llama/LlamaError.h
View File

@ -13,8 +13,9 @@ extern NSString *const LlamaErrorDomain;
typedef NS_ENUM(NSInteger, LlamaErrorCode) {
LlamaErrorCodeUnknown = -1,
LlamaErrorCodeFailedToLoadModel = -1000,
LlamaErrorCodePredictionFailed = -1001
LlamaErrorCodePredictionFailed = -1001,
};
NS_ASSUME_NONNULL_END

87
llama/bridge/LlamaPredictOperation.mm
View File

@ -87,11 +87,19 @@ struct llama_model {
std::map<std::string, struct ggml_tensor *> tensors;
};
NSError *makeLlamaError(LlamaErrorCode errorCode, NSString *description)
{
return [[NSError alloc] initWithDomain:LlamaErrorDomain code:errorCode userInfo:@{
NSLocalizedDescriptionKey: description
}];
}
// load the model's weights from a file
bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab & vocab, int n_ctx) {
bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab & vocab, int n_ctx, NSError **outError) {
auto fin = std::ifstream(fname, std::ios::binary);
if (!fin) {
fprintf(stderr, "%s: failed to open '%s'\n", __func__, fname.c_str());
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"failed to open '%s'", fname.c_str()]);
return false;
}
@ -100,7 +108,8 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
uint32_t magic;
fin.read((char *) &magic, sizeof(magic));
if (magic != 0x67676d6c) {
fprintf(stderr, "%s: invalid model file '%s' (bad magic)\n", __func__, fname.c_str());
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"invalid model file '%s' (bad magic)", fname.c_str()]);
return false;
}
}
@ -132,8 +141,8 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
const int32_t n_vocab = model.hparams.n_vocab;
if (n_vocab != model.hparams.n_vocab) {
fprintf(stderr, "%s: invalid model file '%s' (bad vocab size %d != %d)\n",
__func__, fname.c_str(), n_vocab, model.hparams.n_vocab);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"invalid model file '%s' (bad vocab size %d != %d)", fname.c_str(), n_vocab, model.hparams.n_vocab]);
return false;
}
@ -164,8 +173,8 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
case 3: wtype = GGML_TYPE_Q4_1; break;
default:
{
fprintf(stderr, "%s: invalid model file '%s' (bad f16 value %d)\n",
__func__, fname.c_str(), model.hparams.f16);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"invalid model file '%s' (bad f16 value %d)", fname.c_str(), model.hparams.f16]);
return false;
}
}
@ -218,7 +227,7 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
model.ctx = ggml_init(params);
if (!model.ctx) {
fprintf(stderr, "%s: ggml_init() failed\n", __func__);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel, [NSString stringWithFormat:@"ggml_init() failed"]);
return false;
}
}
@ -341,7 +350,8 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
fin.read(&name[0], length);
if (model.tensors.find(name.data()) == model.tensors.end()) {
fprintf(stderr, "%s: unknown tensor '%s' in model file\n", __func__, name.data());
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"unknown tensor '%s' in model file", name.data()]);
return false;
}
@ -381,33 +391,35 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
if (n_dims == 1) {
if (ggml_nelements(tensor) != nelements) {
fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data());
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"tensor '%s' has wrong size in model file", name.data()]);
return false;
}
} else {
if (ggml_nelements(tensor)/n_parts != nelements) {
fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data());
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"tensor '%s' has wrong size in model file", name.data()]);
return false;
}
}
if (n_dims == 1) {
if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) {
fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]\n",
__func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]", name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]]);
return false;
}
} else {
if (split_type == 0) {
if (tensor->ne[0]/n_parts != ne[0] || tensor->ne[1] != ne[1]) {
fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]\n",
__func__, name.data(), tensor->ne[0]/n_parts, tensor->ne[1], ne[0], ne[1]);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]", name.data(), tensor->ne[0]/n_parts, tensor->ne[1], ne[0], ne[1]]);
return false;
}
} else {
if (tensor->ne[0] != ne[0] || tensor->ne[1]/n_parts != ne[1]) {
fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]\n",
__func__, name.data(), tensor->ne[0], tensor->ne[1]/n_parts, ne[0], ne[1]);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]", name.data(), tensor->ne[0], tensor->ne[1]/n_parts, ne[0], ne[1]]);
return false;
}
}
@ -426,15 +438,15 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
case 3: bpe = ggml_type_size(GGML_TYPE_Q4_1); assert(ne[0] % 64 == 0); break;
default:
{
fprintf(stderr, "%s: unknown ftype %d in model file\n", __func__, ftype);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel, [NSString stringWithFormat:@"unknown ftype %d in model file", ftype]);
return false;
}
};
if (n_dims == 1 || n_parts == 1) {
if ((nelements*bpe)/ggml_blck_size(tensor->type) != ggml_nbytes(tensor)) {
fprintf(stderr, "%s: tensor '%s' has wrong size in model file: got %zu, expected %zu\n",
__func__, name.data(), ggml_nbytes(tensor), nelements*bpe);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"tensor '%s' has wrong size in model file: got %zu, expected %zu", name.data(), ggml_nbytes(tensor), nelements*bpe]);
return false;
}
@ -447,8 +459,8 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
total_size += ggml_nbytes(tensor);
} else {
if ((nelements*bpe)/ggml_blck_size(tensor->type) != ggml_nbytes(tensor)/n_parts) {
fprintf(stderr, "%s: tensor '%s' has wrong size in model file: got %zu, expected %zu\n",
__func__, name.data(), ggml_nbytes(tensor)/n_parts, nelements*bpe);
*outError = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"tensor '%s' has wrong size in model file: got %zu, expected %zu", name.data(), ggml_nbytes(tensor)/n_parts, nelements*bpe]);
return false;
}
@ -501,7 +513,9 @@ bool llama_eval(
const int n_past,
const std::vector<gpt_vocab::id> & embd_inp,
std::vector<float> & embd_w,
size_t & mem_per_token) {
size_t & mem_per_token,
NSError **outError
) {
const int N = embd_inp.size();
const auto & hparams = model.hparams;
@ -526,7 +540,8 @@ bool llama_eval(
buf_size = buf_size_new;
buf = realloc(buf, buf_size);
if (buf == nullptr) {
fprintf(stderr, "%s: failed to allocate %zu bytes\n", __func__, buf_size);
*outError = makeLlamaError(LlamaErrorCodePredictionFailed,
[NSString stringWithFormat:@"failed to allocate %zu bytes", buf_size]);
return false;
}
}
@ -727,13 +742,6 @@ void sigint_handler(int signo) {
}
#endif
NSError *makeLlamaError(LlamaErrorCode errorCode, NSString *description)
{
return [[NSError alloc] initWithDomain:LlamaErrorDomain code:errorCode userInfo:@{
NSLocalizedDescriptionKey: description
}];
}
@interface LlamaPredictOperation () {
gpt_params _params;
LlamaPredictOperationEventHandler _eventHandler;
@ -778,10 +786,9 @@ NSError *makeLlamaError(LlamaErrorCode errorCode, NSString *description)
const int64_t t_start_us = ggml_time_us();
if (!llama_model_load(_params.model, model, vocab, 512)) { // TODO: set context from user input ??
NSError *error = makeLlamaError(LlamaErrorCodeFailedToLoadModel,
[NSString stringWithFormat:@"Failed to load model from '%s'", _params.model.c_str()]);
[self postEvent:[_LlamaEvent failedWithError:error]];
NSError *loadError = nil;
if (!llama_model_load(_params.model, model, vocab, 512, &loadError)) { // TODO: set context from user input ??
[self postEvent:[_LlamaEvent failedWithError:loadError]];
return;
}
@ -811,7 +818,11 @@ NSError *makeLlamaError(LlamaErrorCode errorCode, NSString *description)
// determine the required inference memory per token:
size_t mem_per_token = 0;
llama_eval(model, _params.n_threads, 0, { 0, 1, 2, 3 }, logits, mem_per_token);
NSError *error = nil;
if (!llama_eval(model, _params.n_threads, 0, { 0, 1, 2, 3 }, logits, mem_per_token, &error)) {
[self postEvent:[_LlamaEvent failedWithError:error]];
return;
}
int last_n_size = _params.repeat_last_n;
std::vector<gpt_vocab::id> last_n_tokens(last_n_size);
@ -825,8 +836,8 @@ NSError *makeLlamaError(LlamaErrorCode errorCode, NSString *description)
if (embd.size() > 0) {
const int64_t t_start_us = ggml_time_us();
if (!llama_eval(model, _params.n_threads, n_past, embd, logits, mem_per_token)) {
NSError *error = makeLlamaError(LlamaErrorCodePredictionFailed, @"Failed to predict");
NSError *error = nil;
if (!llama_eval(model, _params.n_threads, n_past, embd, logits, mem_per_token, &error)) {
[self postEvent:[_LlamaEvent failedWithError:error]];
return;
}