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

Reduce warnings. (#439)

Browse source
This commit is contained in:
Ycros 2023-09-16 20:52:09 +10:00 • committed by GitHub
parent 8d90072a2a
commit f6ba36dff6
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
10 changed files with 28 additions and 36 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml-opencl.cpp
View file

@ -1529,7 +1529,7 @@ static void ggml_cl_mul_mat_f32(const ggml_tensor * src0, const ggml_tensor * sr
&queue, &ev_sgemm); &queue, &ev_sgemm);
if (status != clblast::StatusCode::kSuccess) { if (status != clblast::StatusCode::kSuccess) {
printf("\nF32 Matmul Failed (%d): [dims: %lld,%lld,%lld,%lld] You may be out of VRAM. Please check if you have enough.\n",status,ne00,ne01,ne10,ne11); printf("\nF32 Matmul Failed (%d): [dims: %ld,%ld,%ld,%ld] You may be out of VRAM. Please check if you have enough.\n",static_cast<int>(status),ne00,ne01,ne10,ne11);
GGML_ASSERT(false); GGML_ASSERT(false);
} }
@ -1634,7 +1634,7 @@ static void ggml_cl_mul_mat_f16(const ggml_tensor * src0, const ggml_tensor * sr
&queue, &ev_sgemm); &queue, &ev_sgemm);
if (status != clblast::StatusCode::kSuccess) { if (status != clblast::StatusCode::kSuccess) {
printf("\nF16 Matmul Failed (%d): [dims: %lld,%lld,%lld,%lld] You may be out of VRAM. Please check if you have enough.\n",status,ne00,ne01,ne10,ne11); printf("\nF16 Matmul Failed (%d): [dims: %ld,%ld,%ld,%ld] You may be out of VRAM. Please check if you have enough.\n",static_cast<int>(status),ne00,ne01,ne10,ne11);
GGML_ASSERT(false); GGML_ASSERT(false);
} }
@ -1754,7 +1754,7 @@ static void ggml_cl_mul_mat_q_f32(const ggml_tensor * src0, const ggml_tensor *
&queue, events.data() + ev_idx++); &queue, events.data() + ev_idx++);
if (status != clblast::StatusCode::kSuccess) { if (status != clblast::StatusCode::kSuccess) {
printf("\nQF32 Matmul Failed (%d): [dims: %lld,%lld,%lld,%lld] You may be out of VRAM. Please check if you have enough.\n",status,ne00,ne01,ne10,ne11); printf("\nQF32 Matmul Failed (%d): [dims: %ld,%ld,%ld,%ld] You may be out of VRAM. Please check if you have enough.\n",static_cast<int>(status),ne00,ne01,ne10,ne11);
GGML_ASSERT(false); GGML_ASSERT(false);
} }
} }

12
gpttype_adapter.cpp
View file

@ -691,7 +691,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
const struct rwkv_file_header & header = rwkv_ctx_v3->instance->model.header; const struct rwkv_file_header & header = rwkv_ctx_v3->instance->model.header;
const size_t n_vocab = header.n_vocab; const size_t n_vocab = header.n_vocab;
printf("\nDetected Vocab: %d",n_vocab); printf("\nDetected Vocab: %zu",n_vocab);
if(n_vocab>60000) if(n_vocab>60000)
{ {
printf("\nUsing WORLD TOKENIZER"); printf("\nUsing WORLD TOKENIZER");
@ -729,7 +729,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
auto statebufsiz = rwkv_v2_get_state_buffer_element_count(rwkv_ctx_v2) * sizeof(float) + padding; auto statebufsiz = rwkv_v2_get_state_buffer_element_count(rwkv_ctx_v2) * sizeof(float) + padding;
auto logitbufsiz = rwkv_v2_get_logits_buffer_element_count(rwkv_ctx_v2) * sizeof(float) + padding; auto logitbufsiz = rwkv_v2_get_logits_buffer_element_count(rwkv_ctx_v2) * sizeof(float) + padding;
printf("\nRWKV old Init: State Buffer:%u, Logit Buffer:%u\n", statebufsiz, logitbufsiz); printf("\nRWKV old Init: State Buffer:%lu, Logit Buffer:%lu\n", statebufsiz, logitbufsiz);
rwkv_ctx_v2->state_out = (float *)malloc(statebufsiz); rwkv_ctx_v2->state_out = (float *)malloc(statebufsiz);
rwkv_ctx_v2->logits_out = (float *)malloc(logitbufsiz); rwkv_ctx_v2->logits_out = (float *)malloc(logitbufsiz);
rwkv_ctx_v2->state_in = nullptr; rwkv_ctx_v2->state_in = nullptr;
@ -757,7 +757,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
auto statebufsiz = rwkv_get_state_buffer_element_count(rwkv_ctx_v3) * sizeof(float) + padding; auto statebufsiz = rwkv_get_state_buffer_element_count(rwkv_ctx_v3) * sizeof(float) + padding;
auto logitbufsiz = rwkv_get_logits_buffer_element_count(rwkv_ctx_v3) * sizeof(float) + padding; auto logitbufsiz = rwkv_get_logits_buffer_element_count(rwkv_ctx_v3) * sizeof(float) + padding;
printf("\nRWKV Init: State Buffer:%u, Logit Buffer:%u\n", statebufsiz, logitbufsiz); printf("\nRWKV Init: State Buffer:%lu, Logit Buffer:%lu\n", statebufsiz, logitbufsiz);
rwkv_ctx_v3->state_out = (float *)malloc(statebufsiz); rwkv_ctx_v3->state_out = (float *)malloc(statebufsiz);
rwkv_ctx_v3->logits_out = (float *)malloc(logitbufsiz); rwkv_ctx_v3->logits_out = (float *)malloc(logitbufsiz);
rwkv_ctx_v3->state_in = nullptr; rwkv_ctx_v3->state_in = nullptr;
@ -1284,7 +1284,7 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
//prepare banned tokens //prepare banned tokens
if(banned_token_ids.size()==0 && banned_tokens.size()>0) if(banned_token_ids.size()==0 && banned_tokens.size()>0)
{ {
printf("\n[First Run] Banning %d token sequences...",banned_tokens.size()); printf("\n[First Run] Banning %zu token sequences...",banned_tokens.size());
for(int v=0;v<n_vocab;++v) for(int v=0;v<n_vocab;++v)
{ {
std::string word = FileFormatTokenizeID(v,file_format); std::string word = FileFormatTokenizeID(v,file_format);
@ -1297,7 +1297,7 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
} }
} }
} }
printf("\nBanned a total of %d tokens.\n",banned_token_ids.size()); printf("\nBanned a total of %zu tokens.\n",banned_token_ids.size());
} }
if(debugmode!=-1) if(debugmode!=-1)
@ -1337,7 +1337,7 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
//print progress //print progress
if (!startedsampling && debugmode!=-1) if (!startedsampling && debugmode!=-1)
{ {
printf("\rProcessing Prompt%s (%d / %d tokens)", (blasmode ? " [BLAS]" : ""), input_consumed, embd_inp.size()); printf("\rProcessing Prompt%s (%d / %zu tokens)", (blasmode ? " [BLAS]" : ""), input_consumed, embd_inp.size());
} }
fflush(stdout); fflush(stdout);

3
otherarch/ggml_v1.c
View file

@ -1,6 +1,3 @@
// Defines CLOCK_MONOTONIC and asprintf on Linux
#define _GNU_SOURCE
#include "ggml_v1.h" #include "ggml_v1.h"
#if defined(_MSC_VER) || defined(__MINGW32__) #if defined(_MSC_VER) || defined(__MINGW32__)

6
otherarch/ggml_v2-opencl.cpp
View file

@ -573,7 +573,7 @@ static void ggml_v2_cl_mul_mat_f32(const ggml_v2_tensor * src0, const ggml_v2_te
&queue, &ev_sgemm); &queue, &ev_sgemm);
if (status != clblast::StatusCode::kSuccess) { if (status != clblast::StatusCode::kSuccess) {
printf("\nF32 Matmul Failed (%d): [dims: %lld,%lld,%lld,%lld] You may be out of VRAM. Please check if you have enough.\n",status,ne00,ne01,ne10,ne11); printf("\nF32 Matmul Failed (%d): [dims: %ld,%ld,%ld,%ld] You may be out of VRAM. Please check if you have enough.\n",static_cast<int>(status),ne00,ne01,ne10,ne11);
GGML_V2_ASSERT(false); GGML_V2_ASSERT(false);
} }
@ -672,7 +672,7 @@ static void ggml_v2_cl_mul_mat_f16(const ggml_v2_tensor * src0, const ggml_v2_te
&queue, &ev_sgemm); &queue, &ev_sgemm);
if (status != clblast::StatusCode::kSuccess) { if (status != clblast::StatusCode::kSuccess) {
printf("\nF16 Matmul Failed (%d): [dims: %lld,%lld,%lld,%lld] You may be out of VRAM. Please check if you have enough.\n",status,ne00,ne01,ne10,ne11); printf("\nF16 Matmul Failed (%d): [dims: %ld,%ld,%ld,%ld] You may be out of VRAM. Please check if you have enough.\n",static_cast<int>(status),ne00,ne01,ne10,ne11);
GGML_V2_ASSERT(false); GGML_V2_ASSERT(false);
} }
@ -780,7 +780,7 @@ static void ggml_v2_cl_mul_mat_q_f32(const ggml_v2_tensor * src0, const ggml_v2_
&queue, &ev_sgemm); &queue, &ev_sgemm);
if (status != clblast::StatusCode::kSuccess) { if (status != clblast::StatusCode::kSuccess) {
printf("\nQF32 Matmul Failed (%d): [dims: %lld,%lld,%lld,%lld] You may be out of VRAM. Please check if you have enough.\n",status,ne00,ne01,ne10,ne11); printf("\nQF32 Matmul Failed (%d): [dims: %ld,%ld,%ld,%ld] You may be out of VRAM. Please check if you have enough.\n",static_cast<int>(status),ne00,ne01,ne10,ne11);
GGML_V2_ASSERT(false); GGML_V2_ASSERT(false);
} }
} }

3
otherarch/ggml_v2.c
View file

@ -1,6 +1,3 @@
// Defines CLOCK_MONOTONIC on Linux
#define _GNU_SOURCE
#include "ggml_v2.h" #include "ggml_v2.h"
#if defined(_MSC_VER) || defined(__MINGW32__) #if defined(_MSC_VER) || defined(__MINGW32__)

8
otherarch/gpt2_v2.cpp
View file

@ -150,7 +150,7 @@ ModelLoadResult gpt2_v2_model_load(const std::string & fname, gpt2_v2_model & mo
params.mem_size = ctx_size; params.mem_size = ctx_size;
params.mem_buffer = NULL; params.mem_buffer = NULL;
params.no_alloc = false; params.no_alloc = false;
model.ctx = ggml_v2_init(params); model.ctx = ggml_v2_init(params);
if (!model.ctx) { if (!model.ctx) {
@ -237,7 +237,7 @@ ModelLoadResult gpt2_v2_model_load(const std::string & fname, gpt2_v2_model & mo
const int n_mem = n_layer*n_ctx; const int n_mem = n_layer*n_ctx;
const int n_elements = n_embd*n_mem; const int n_elements = n_embd*n_mem;
model.memory_k = ggml_v2_new_tensor_1d(ctx, memory_type, n_elements*1.5); model.memory_k = ggml_v2_new_tensor_1d(ctx, memory_type, n_elements*1.5);
model.memory_v = ggml_v2_new_tensor_1d(ctx, memory_type, n_elements*1.5); model.memory_v = ggml_v2_new_tensor_1d(ctx, memory_type, n_elements*1.5);
@ -287,7 +287,7 @@ ModelLoadResult gpt2_v2_model_load(const std::string & fname, gpt2_v2_model & mo
} }
if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) { if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) {
fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%lld, %lld], expected [%lld, %lld]\n", fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%ld, %ld], expected [%d, %d]\n",
__func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]); __func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]);
return ModelLoadResult::FAIL; return ModelLoadResult::FAIL;
} }
@ -379,7 +379,7 @@ bool gpt2_v2_eval(
params.mem_size = buf_size; params.mem_size = buf_size;
params.mem_buffer = buf; params.mem_buffer = buf;
params.no_alloc = false; params.no_alloc = false;
struct ggml_v2_context * ctx0 = ggml_v2_init(params); struct ggml_v2_context * ctx0 = ggml_v2_init(params);
struct ggml_v2_cgraph gf = {}; struct ggml_v2_cgraph gf = {};

10
otherarch/gptj_v2.cpp
View file

@ -150,7 +150,7 @@ ModelLoadResult gptj_v2_model_load(const std::string & fname, gptj_v2_model & mo
params.mem_size = ctx_size; params.mem_size = ctx_size;
params.mem_buffer = NULL; params.mem_buffer = NULL;
params.no_alloc = false; params.no_alloc = false;
model.ctx = ggml_v2_init(params); model.ctx = ggml_v2_init(params);
if (!model.ctx) { if (!model.ctx) {
@ -281,7 +281,7 @@ ModelLoadResult gptj_v2_model_load(const std::string & fname, gptj_v2_model & mo
fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data()); fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data());
return ModelLoadResult::FAIL; return ModelLoadResult::FAIL;
} }
if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) { if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) {
@ -294,11 +294,11 @@ ModelLoadResult gptj_v2_model_load(const std::string & fname, gptj_v2_model & mo
} }
else else
{ {
fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]\n", fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%ld, %ld], expected [%d, %d]\n",
__func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]); __func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]);
return ModelLoadResult::FAIL; return ModelLoadResult::FAIL;
} }
} }
// for debugging // for debugging
@ -387,7 +387,7 @@ bool gptj_v2_eval(
params.mem_size = buf_size; params.mem_size = buf_size;
params.mem_buffer = buf; params.mem_buffer = buf;
params.no_alloc = false; params.no_alloc = false;
struct ggml_v2_context * ctx0 = ggml_v2_init(params); struct ggml_v2_context * ctx0 = ggml_v2_init(params);
struct ggml_v2_cgraph gf = {}; struct ggml_v2_cgraph gf = {};

2
otherarch/gptj_v3.cpp
View file

@ -304,7 +304,7 @@ ModelLoadResult gptj_model_load(const std::string & fname, gptj_model & model, g
} }
else else
{ {
fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]\n", fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%ld, %ld], expected [%d, %d]\n",
__func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]); __func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]);
return ModelLoadResult::FAIL; return ModelLoadResult::FAIL;
} }

10
otherarch/llama_v3.h
View file

@ -243,10 +243,9 @@ extern "C" {
// Various functions for loading a ggml llama model. // Various functions for loading a ggml llama model.
// Allocate (almost) all memory needed for the model. // Allocate (almost) all memory needed for the model.
// Return NULL on failure // Return NULL on failure
LLAMA_V3_API DEPRECATED(struct llama_v3_context * llama_v3_init_from_file( LLAMA_V3_API struct llama_v3_context * llama_v3_init_from_file(
const char * path_model, const char * path_model,
struct llama_v3_context_params params), struct llama_v3_context_params params);
"please use llama_v3_load_model_from_file combined with llama_v3_new_context_with_model instead");
// Frees all allocated memory // Frees all allocated memory
LLAMA_V3_API void llama_v3_free(struct llama_v3_context * ctx); LLAMA_V3_API void llama_v3_free(struct llama_v3_context * ctx);
@ -263,12 +262,11 @@ extern "C" {
// The model needs to be reloaded before applying a new adapter, otherwise the adapter // The model needs to be reloaded before applying a new adapter, otherwise the adapter
// will be applied on top of the previous one // will be applied on top of the previous one
// Returns 0 on success // Returns 0 on success
LLAMA_V3_API DEPRECATED(int llama_v3_apply_lora_from_file( LLAMA_V3_API int llama_v3_apply_lora_from_file(
struct llama_v3_context * ctx, struct llama_v3_context * ctx,
const char * path_lora, const char * path_lora,
const char * path_base_model, const char * path_base_model,
int n_threads), int n_threads);
"please use llama_v3_model_apply_lora_from_file instead");
LLAMA_V3_API int llama_v3_model_apply_lora_from_file( LLAMA_V3_API int llama_v3_model_apply_lora_from_file(
const struct llama_v3_model * model, const struct llama_v3_model * model,

4
otherarch/rwkv_v2.cpp
View file

@ -367,8 +367,8 @@ struct rwkv_v2_context * rwkv_v2_init_from_file(const char * file_path, uint32_t
// Verify order of dimensions // Verify order of dimensions
struct ggml_v2_tensor * emb = model->emb; struct ggml_v2_tensor * emb = model->emb;
RWKV_V2_ASSERT_NULL(emb->n_dims == 2, "Unexpected dimension count of embedding matrix %d", emb->n_dims); RWKV_V2_ASSERT_NULL(emb->n_dims == 2, "Unexpected dimension count of embedding matrix %d", emb->n_dims);
RWKV_V2_ASSERT_NULL(emb->ne[0] == model->n_embed, "Unexpected dimension of embedding matrix %lld", emb->ne[0]); RWKV_V2_ASSERT_NULL(emb->ne[0] == model->n_embed, "Unexpected dimension of embedding matrix %ld", emb->ne[0]);
RWKV_V2_ASSERT_NULL(emb->ne[1] == model->n_vocab, "Unexpected dimension of embedding matrix %lld", emb->ne[1]); RWKV_V2_ASSERT_NULL(emb->ne[1] == model->n_vocab, "Unexpected dimension of embedding matrix %ld", emb->ne[1]);
int32_t n_embed = model->n_embed; int32_t n_embed = model->n_embed;
int32_t n_layer = model->n_layer; int32_t n_layer = model->n_layer;