happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

llama_fit_params: return enum for fail vs. error (#18374)

This commit is contained in:
Johannes Gäßler 2025-12-27 09:59:19 +01:00 committed by GitHub
parent 9045c9afe5
commit a52dc60ba3
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 28 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
include/llama.h
View File

@ -467,10 +467,16 @@ extern "C" {
// Frees all allocated memory
LLAMA_API void llama_free(struct llama_context * ctx);
enum llama_params_fit_status {
LLAMA_PARAMS_FIT_STATUS_SUCCESS = 0, // found allocations that are projected to fit
LLAMA_PARAMS_FIT_STATUS_FAILURE = 1, // could not find allocations that are projected to fit
LLAMA_PARAMS_FIT_STATUS_ERROR = 2, // a hard error occured, e.g. because no model could be found at the specified path
};
// fits mparams and cparams to free device memory (assumes system memory is unlimited)
// returns true if the parameters could be successfully modified to fit device memory
// this function is NOT thread safe because it modifies the global llama logger state
LLAMA_API bool llama_params_fit(
LLAMA_API enum llama_params_fit_status llama_params_fit(
const char * path_model,
struct llama_model_params * mparams,
struct llama_context_params * cparams,

31
src/llama.cpp
View File

@ -140,6 +140,10 @@ enum layer_fraction_t {
};
// this enum is only used in llama_params_fit_impl but needs to be defined outside of it to fix a Windows compilation issue
class llama_params_fit_exception : public std::runtime_error {
using std::runtime_error::runtime_error;
};
static void llama_params_fit_impl(
const char * path_model, struct llama_model_params * mparams, struct llama_context_params * cparams,
float * tensor_split, struct llama_model_tensor_buft_override * tensor_buft_overrides,
@ -281,28 +285,28 @@ static void llama_params_fit_impl(
}
if (mparams->n_gpu_layers != default_mparams.n_gpu_layers) {
throw std::runtime_error("n_gpu_layers already set by user to " + std::to_string(mparams->n_gpu_layers) + ", abort");
throw llama_params_fit_exception("n_gpu_layers already set by user to " + std::to_string(mparams->n_gpu_layers) + ", abort");
}
if (nd > 1) {
if (!tensor_split) {
throw std::runtime_error("did not provide a buffer to write the tensor_split to, abort");
throw llama_params_fit_exception("did not provide a buffer to write the tensor_split to, abort");
}
if (mparams->tensor_split) {
for (size_t id = 0; id < nd; id++) {
if (mparams->tensor_split[id] != 0.0f) {
throw std::runtime_error("model_params::tensor_split already set by user, abort");
throw llama_params_fit_exception("model_params::tensor_split already set by user, abort");
}
}
}
if (mparams->split_mode == LLAMA_SPLIT_MODE_ROW) {
throw std::runtime_error("changing weight allocation for LLAMA_SPLIT_MODE_ROW not implemented, abort");
throw llama_params_fit_exception("changing weight allocation for LLAMA_SPLIT_MODE_ROW not implemented, abort");
}
}
if (!tensor_buft_overrides) {
throw std::runtime_error("did not provide buffer to set tensor_buft_overrides, abort");
throw llama_params_fit_exception("did not provide buffer to set tensor_buft_overrides, abort");
}
if (mparams->tensor_buft_overrides && (mparams->tensor_buft_overrides->pattern || mparams->tensor_buft_overrides->buft)) {
throw std::runtime_error("model_params::tensor_buft_overrides already set by user, abort");
throw llama_params_fit_exception("model_params::tensor_buft_overrides already set by user, abort");
}
// step 3: iteratively fill the back to front with "dense" layers
@ -385,7 +389,7 @@ static void llama_params_fit_impl(
tensor_buft_overrides[itbo].buft = nullptr;
itbo++;
mparams.tensor_buft_overrides = tensor_buft_overrides;
throw std::runtime_error("llama_params_fit_n_tensor_buft_overrides() == "
throw llama_params_fit_exception("llama_params_fit_n_tensor_buft_overrides() == "
+ std::to_string(ntbo) + " is insufficient for model\n");
}
tensor_buft_overrides[itbo].pattern = get_overflow_pattern(il, il == il0 ? ngl_per_device[id].overflow_type : LAYER_FRACTION_MOE);
@ -683,22 +687,25 @@ static void llama_params_fit_impl(
set_ngl_tensor_split_tbo(ngl_per_device, overflow_bufts, *mparams);
}
bool llama_params_fit(
enum llama_params_fit_status llama_params_fit(
const char * path_model, struct llama_model_params * mparams, struct llama_context_params * cparams,
float * tensor_split, struct llama_model_tensor_buft_override * tensor_buft_overrides,
size_t margin_s, uint32_t n_ctx_min, enum ggml_log_level log_level) {
const int64_t t0_us = llama_time_us();
bool ok = true;
llama_params_fit_status status = LLAMA_PARAMS_FIT_STATUS_SUCCESS;
try {
llama_params_fit_impl(path_model, mparams, cparams, tensor_split, tensor_buft_overrides, margin_s, n_ctx_min, log_level);
LLAMA_LOG_INFO("%s: successfully fit params to free device memory\n", __func__);
} catch (const std::runtime_error & e) {
} catch (const llama_params_fit_exception & e) {
LLAMA_LOG_WARN("%s: failed to fit params to free device memory: %s\n", __func__, e.what());
ok = false;
status = LLAMA_PARAMS_FIT_STATUS_FAILURE;
} catch (const std::runtime_error & e) {
LLAMA_LOG_ERROR("%s: encountered an error while trying to fit params to free device memory: %s\n", __func__, e.what());
status = LLAMA_PARAMS_FIT_STATUS_ERROR;
}
const int64_t t1_us = llama_time_us();
LLAMA_LOG_INFO("%s: fitting params to free memory took %.2f seconds\n", __func__, (t1_us - t0_us) * 1e-6);
return ok;
return status;
}
struct llama_sampler_chain_params llama_sampler_chain_default_params() {

4
tools/fit-params/fit-params.cpp
View File

@ -26,10 +26,10 @@ int main(int argc, char ** argv) {
llama_numa_init(params.numa);
auto mparams = common_model_params_to_llama(params);
auto cparams = common_context_params_to_llama(params);
const bool success = llama_params_fit(params.model.path.c_str(), &mparams, &cparams,
const llama_params_fit_status status = llama_params_fit(params.model.path.c_str(), &mparams, &cparams,
params.tensor_split, params.tensor_buft_overrides.data(), params.fit_params_target, params.fit_params_min_ctx,
params.verbosity >= 4 ? GGML_LOG_LEVEL_DEBUG : GGML_LOG_LEVEL_ERROR);
if (!success) {
if (status != LLAMA_PARAMS_FIT_STATUS_SUCCESS) {
LOG_ERR("%s: failed to fit CLI arguments to free memory, exiting...\n", __func__);
exit(1);
}