Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-19 17:24:57 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix CI

Browse Source
This commit is contained in:
Johannes Gäßler
2026-03-05 15:26:38 +01:00
parent 4b7f407ae8
commit 7e466072f3
4 changed files with 35 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-model.cpp
+2 -1
View File
@@ -2690,7 +2690,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
// helper: try merged gate_up_exps first, fall back to separate gate and up
auto create_tensor_gate_up_exps = [&](llama_layer & layer, int bid, int64_t n_embd_, int64_t n_ff_, int64_t n_expert_, int flags) {
layer.ffn_gate_up_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_UP_EXPS, "weight", bid), {n_embd_, n_ff_ * 2, n_expert_}, TENSOR_NOT_REQUIRED | TENSOR_SKIP_IF_VIRTUAL);
layer.ffn_gate_up_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_UP_EXPS, "weight", bid), {n_embd_, n_ff_ * 2, n_expert_}, TENSOR_NOT_REQUIRED);
if (layer.ffn_gate_up_exps == nullptr) {
layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", bid), {n_embd_, n_ff_, n_expert_}, flags);
layer.ffn_up_exps = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS, "weight", bid), {n_embd_, n_ff_, n_expert_}, flags);
@@ -3452,6 +3452,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0);
layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
// FIXME test-llama-archs crashes if q_norm is created
layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED | TENSOR_SKIP_IF_VIRTUAL);
layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED | TENSOR_SKIP_IF_VIRTUAL);
src/models/baichuan.cpp
-1
View File
@@ -1,4 +1,3 @@
#include "llama-model.h"
#include "models.h"
tests/test-llama-archs.cpp
+32 -12
View File
@@ -60,7 +60,7 @@ static void set_tensor_data(struct ggml_tensor * tensor, void * userdata) {
}
static void usage(char ** argv) {
printf("Usage: %s [-s/--seed seed] [-v/--verbose]\n", argv[0]);
printf("Usage: %s [-a/--arch arch] [-s/--seed seed] [-v/--verbose]\n", argv[0]);
}
static std::vector<llama_token> get_tokens(const uint32_t n_tokens, const uint32_t n_vocab, const size_t seed){
@@ -234,8 +234,8 @@ static std::pair<llama_model_ptr, llama_context_ptr> get_model_and_ctx(
llama_context_params ctx_params = llama_context_default_params();
ctx_params.n_ctx = 0;
ctx_params.n_threads = 4;
ctx_params.n_threads_batch = 4;
ctx_params.n_threads = 1; // FIXME LLAMA_SANITIZE_THREAD workaround
ctx_params.n_threads_batch = 1;
size_t tmp = seed;
llama_model_ptr model(llama_model_init_from_user(gguf_ctx, set_tensor_data, &tmp, model_params));
@@ -343,7 +343,7 @@ static bool moe_implemented(const llm_arch arch) {
}
}
static int save_models(const size_t seed, const ggml_log_level log_level, const std::string & dir) {
static int save_models(const llm_arch target_arch, const size_t seed, const ggml_log_level log_level, const std::string & dir) {
GGML_ABORT("llama_model_save_to_file is broken");
struct user_data_t {
struct {
@@ -363,6 +363,9 @@ static int save_models(const size_t seed, const ggml_log_level log_level, const
}, &ud);
for (const llm_arch & arch : llm_arch_all()) {
if (target_arch != LLM_ARCH_UNKNOWN && arch != target_arch) {
continue;
}
if (arch == LLM_ARCH_CLIP || arch == LLM_ARCH_GPTJ || arch == LLM_ARCH_UNKNOWN) {
continue; // These models don't have usable implementations.
}
@@ -384,7 +387,7 @@ static int save_models(const size_t seed, const ggml_log_level log_level, const
return 0;
}
static int test_backends(const size_t seed, const ggml_log_level log_level) {
static int test_backends(const llm_arch target_arch, const size_t seed, const ggml_log_level log_level) {
struct user_data_t {
struct {
ggml_log_callback callback;
@@ -409,6 +412,9 @@ static int test_backends(const size_t seed, const ggml_log_level log_level) {
printf("|%15s|%30s|%6s|%8s|%6s|\n", "Model arch.", "Device", "Config", "NMSE", "Status");
printf("|---------------|------------------------------|------|--------|------|\n");
for (const llm_arch & arch : llm_arch_all()) {
if (target_arch != LLM_ARCH_UNKNOWN && arch != target_arch) {
continue;
}
if (arch == LLM_ARCH_CLIP || arch == LLM_ARCH_GPTJ || arch == LLM_ARCH_UNKNOWN) {
continue; // These models don't have usable implementations.
}
@@ -448,7 +454,7 @@ static int test_backends(const size_t seed, const ggml_log_level log_level) {
auto model_and_ctx_dev = get_model_and_ctx(gguf_ctx.get(), seed, {dev});
const std::vector<float> logits_dev = get_logits(model_and_ctx_dev.first.get(), model_and_ctx_dev.second.get(), tokens, encode);
const double nmse_val = nmse(logits_cpu, logits_dev);
const bool ok = nmse_val <= 1e-6;
const bool ok = nmse_val <= 1e-5;
all_ok = all_ok && ok;
char nmse_str[10];
snprintf(nmse_str, sizeof(nmse_str), "%.2e", nmse_val);
@@ -465,22 +471,28 @@ int main(int argc, char ** argv) {
common_init();
std::random_device rd;
llm_arch arch = LLM_ARCH_UNKNOWN;
size_t seed = rd();
ggml_log_level log_level = GGML_LOG_LEVEL_ERROR;
std::string out;
for (int i = 1; i < argc; i++) {
if (strcmp(argv[i], "-s") == 0 || strcmp(argv[i], "--seed") == 0) {
if (strcmp(argv[i], "-a") == 0 || strcmp(argv[i], "--arch") == 0) {
if (i + 1 < argc) {
seed = std::stoull(argv[++i]);
const std::string arch_name = argv[++i];
arch = llm_arch_from_string(arch_name);
if (arch == LLM_ARCH_UNKNOWN) {
LOG_ERR("%s: unkown LLM architecture: %s\n", __func__, arch_name.c_str());
return 1;
}
} else {
usage(argv);
return 1;
}
}
if (strcmp(argv[i], "-o") == 0 || strcmp(argv[i], "--out") == 0) {
if (strcmp(argv[i], "-s") == 0 || strcmp(argv[i], "--seed") == 0) {
if (i + 1 < argc) {
out = argv[++i];
seed = std::stoull(argv[++i]);
} else {
usage(argv);
return 1;
@@ -490,13 +502,21 @@ int main(int argc, char ** argv) {
log_level = GGML_LOG_LEVEL_INFO;
continue;
}
if (strcmp(argv[i], "-o") == 0 || strcmp(argv[i], "--out") == 0) {
if (i + 1 < argc) {
out = argv[++i];
} else {
usage(argv);
return 1;
}
}
}
try {
if (!out.empty()) {
return save_models(seed, log_level, out);
return save_models(arch, seed, log_level, out);
}
return test_backends(seed, log_level);
return test_backends(arch, seed, log_level);
} catch (const std::exception & err) {
fprintf(stderr, "encountered runtime error: %s\n", err.what());
return -1;
tools/results/results.cpp
+1 -1
View File
@@ -130,7 +130,7 @@ int main(int argc, char ** argv) {
const double nmse_val = nmse(logits_disk, logits_calc);
LOG_INF("%s: NMSE=%.3e\n", __func__, nmse_val);
if (nmse_val > 1e-6) {
if (nmse_val > 1e-5) {
printf("\033[1;31mFAIL\033[0m\n");
return 1;
}