Skip to content

llama : bump max layers from 256 to 512 #8530

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jul 19, 2024
Merged

llama : bump max layers from 256 to 512 #8530

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b268edf
llama : bump max layers from 256 to 512
ggerganov Jul 17, 2024
7d87a0d
llama : replace asserts with exceptions
ggerganov Jul 17, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion include/llama.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,7 +40,7 @@
#define LLAMA_FILE_MAGIC_GGSQ 0x67677371u // 'ggsq'

#define LLAMA_SESSION_MAGIC LLAMA_FILE_MAGIC_GGSN
#define LLAMA_SESSION_VERSION 6
#define LLAMA_SESSION_VERSION 7

#define LLAMA_STATE_SEQ_MAGIC LLAMA_FILE_MAGIC_GGSQ
#define LLAMA_STATE_SEQ_VERSION 1
Expand Down
14 changes: 9 additions & 5 deletions src/llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,7 +112,7 @@

// bump if necessary
#define LLAMA_MAX_NODES 8192
#define LLAMA_MAX_LAYERS 256
#define LLAMA_MAX_LAYERS 512
#define LLAMA_MAX_EXPERTS 160 // DeepSeekV2

//
Expand Down Expand Up @@ -3994,7 +3994,9 @@ struct llama_model_loader {
throw std::runtime_error(format("%s is not a float32, int32 array", key.c_str()));
}

GGML_ASSERT(arr_info.length <= N_MAX);
if (arr_info.length > N_MAX) {
throw std::runtime_error(format("array length %u for key %s exceeds max %u", (uint32_t) arr_info.length, key.c_str(), (uint32_t) N_MAX));
}

std::copy((const T*)arr_info.data, (const T *)arr_info.data + arr_info.length, result.begin());

Expand Down Expand Up @@ -4030,8 +4032,6 @@ struct llama_model_loader {
// get array of n <= N_MAX elements, or a single element repeated n times
template<typename T, size_t N_MAX>
bool get_key_or_arr(const std::string & key, std::array<T, N_MAX> & result, uint32_t n, const bool required = true) {
GGML_ASSERT(n <= N_MAX);

const int kid = gguf_find_key(meta, key.c_str());

if (kid < 0) {
Expand All @@ -4041,6 +4041,10 @@ struct llama_model_loader {
return false;
}

if (n > N_MAX) {
throw std::runtime_error(format("n > N_MAX: %u > %u for key %s", (uint32_t) n, (uint32_t) N_MAX, key.c_str()));
}

if (gguf_get_kv_type(meta, kid) == GGUF_TYPE_ARRAY) {
struct GGUFMeta::ArrayInfo arr_info =
GGUFMeta::GKV<GGUFMeta::ArrayInfo>::get_kv(meta, kid);
Expand Down Expand Up @@ -19877,7 +19881,7 @@ size_t llama_state_get_size(const struct llama_context * ctx) {
);

// on session change it is very likely that the state size has changed - so we need to update this function
static_assert(LLAMA_SESSION_VERSION == 6, "So you just bumped the session version - good. But did you remember to update llama_state_get_size?");
static_assert(LLAMA_SESSION_VERSION == 7, "So you just bumped the session version - good. But did you remember to update llama_state_get_size?");

return s_total;
}
Expand Down
Loading