Skip to content

Commit e3b1bd6

Browse files
ggerganovggerganov
committed
minor : code style
1 parent 91a2a56 commit e3b1bd6

File tree

4 files changed

+42
-35
lines changed

4 files changed

+42
-35
lines changed

‎common/chat.h‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -33,8 +33,8 @@ struct common_chat_msg_content_part {
3333
struct common_chat_msg {
3434
std::string role;
3535
std::string content;
36-
std::vector<common_chat_msg_content_part> content_parts = {};
37-
std::vector<common_chat_tool_call> tool_calls = {};
36+
std::vector<common_chat_msg_content_part> content_parts;
37+
std::vector<common_chat_tool_call> tool_calls;
3838
std::string reasoning_content;
3939
std::string tool_name;
4040
std::string tool_call_id;
@@ -44,7 +44,7 @@ struct common_chat_msg {
4444
bool empty() const {
4545
return content.empty() && content_parts.empty() && tool_calls.empty() && reasoning_content.empty() && tool_name.empty() && tool_call_id.empty();
4646
}
47-
void ensure_tool_call_ids_set(std::vector<std::string> & ids_cache, const std::function<std::string()> & gen_tool_call_id) {
47+
void set_tool_call_ids(std::vector<std::string> & ids_cache, const std::function<std::string()> & gen_tool_call_id) {
4848
for (auto i = 0u; i < tool_calls.size(); i++) {
4949
if (ids_cache.size() <= i) {
5050
auto id = tool_calls[i].id;

‎src/llama-kv-cache.cpp‎

Lines changed: 2 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -123,11 +123,8 @@ llama_kv_cache::llama_kv_cache(
123123
throw std::runtime_error("failed to create ggml context for kv cache");
124124
}
125125

126-
ggml_tensor * k;
127-
ggml_tensor * v;
128-
129-
k = ggml_new_tensor_3d(ctx, type_k, n_embd_k_gqa, kv_size, n_stream);
130-
v = ggml_new_tensor_3d(ctx, type_v, n_embd_v_gqa, kv_size, n_stream);
126+
ggml_tensor * k = ggml_new_tensor_3d(ctx, type_k, n_embd_k_gqa, kv_size, n_stream);
127+
ggml_tensor * v = ggml_new_tensor_3d(ctx, type_v, n_embd_v_gqa, kv_size, n_stream);
131128

132129
ggml_format_name(k, "cache_k_l%d", il);
133130
ggml_format_name(v, "cache_v_l%d", il);

‎tools/server/server.cpp‎

Lines changed: 7 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -1619,7 +1619,7 @@ struct server_slot {
16191619
/* is_partial= */ stop != STOP_TYPE_EOS,
16201620
params.oaicompat_chat_syntax);
16211621
if (!new_msg.empty()) {
1622-
new_msg.ensure_tool_call_ids_set(generated_tool_call_ids, gen_tool_call_id);
1622+
new_msg.set_tool_call_ids(generated_tool_call_ids, gen_tool_call_id);
16231623
chat_msg = new_msg;
16241624
diffs = common_chat_msg_diff::compute_diffs(previous_msg, new_msg.empty() ? previous_msg : new_msg);
16251625
}
@@ -2749,7 +2749,7 @@ struct server_context {
27492749
}
27502750

27512751
// if multimodal is enabled, send an error and return false
2752-
bool ensure_no_mtmd(const int id_task) {
2752+
bool check_no_mtmd(const int id_task) {
27532753
if (mctx) {
27542754
send_error(id_task, "This feature is not supported by multimodal", ERROR_TYPE_NOT_SUPPORTED);
27552755
return false;
@@ -3121,7 +3121,7 @@ struct server_context {
31213121
} break;
31223122
case SERVER_TASK_TYPE_SLOT_SAVE:
31233123
{
3124-
if (!ensure_no_mtmd(task.id)) {
3124+
if (!check_no_mtmd(task.id)) {
31253125
break;
31263126
}
31273127

@@ -3162,7 +3162,7 @@ struct server_context {
31623162
} break;
31633163
case SERVER_TASK_TYPE_SLOT_RESTORE:
31643164
{
3165-
if (!ensure_no_mtmd(task.id)) break;
3165+
if (!check_no_mtmd(task.id)) break;
31663166
int id_slot = task.slot_action.slot_id;
31673167
server_slot * slot = get_slot_by_id(id_slot);
31683168
if (slot == nullptr) {
@@ -3209,7 +3209,9 @@ struct server_context {
32093209
} break;
32103210
case SERVER_TASK_TYPE_SLOT_ERASE:
32113211
{
3212-
if (!ensure_no_mtmd(task.id)) break;
3212+
if (!check_no_mtmd(task.id)) {
3213+
break;
3214+
}
32133215
int id_slot = task.slot_action.slot_id;
32143216
server_slot * slot = get_slot_by_id(id_slot);
32153217
if (slot == nullptr) {

‎tools/server/utils.hpp‎

Lines changed: 30 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -1144,9 +1144,8 @@ struct server_tokens {
11441144
auto it = map_pos_to_media.find(pos);
11451145
if (it != map_pos_to_media.end()) {
11461146
return it->second;
1147-
} else {
1148-
throw std::runtime_error("Chunk not found");
11491147
}
1148+
throw std::runtime_error("Chunk not found");
11501149
}
11511150

11521151
void push_back(llama_token tok) {
@@ -1170,7 +1169,7 @@ struct server_tokens {
11701169
map_pos_to_media[start_pos] = std::move(new_chunk);
11711170
} else if (type == MTMD_INPUT_CHUNK_TYPE_TEXT) {
11721171
size_t n_tokens;
1173-
auto text_tokens = mtmd_input_chunk_get_tokens_text(chunk, &n_tokens);
1172+
const auto * text_tokens = mtmd_input_chunk_get_tokens_text(chunk, &n_tokens);
11741173
for (size_t i = 0; i < n_tokens; ++i) {
11751174
push_back(text_tokens[i]);
11761175
}
@@ -1190,7 +1189,7 @@ struct server_tokens {
11901189
// We could also just check, but this will prevent silently dropping MTMD data.
11911190
GGML_ASSERT(has_mtmd);
11921191
for (auto it = tokens.map_pos_to_media.begin(); it != tokens.map_pos_to_media.end(); ) {
1193-
auto chunk = tokens.map_pos_to_media[it->first].get();
1192+
auto * chunk = tokens.map_pos_to_media[it->first].get();
11941193
mtmd::input_chunk_ptr new_chunk(mtmd_input_chunk_copy(chunk));
11951194
map_pos_to_media[start_pos+it->first] = std::move(new_chunk);
11961195
}
@@ -1271,33 +1270,42 @@ struct server_tokens {
12711270
}
12721271

12731272
size_t get_common_prefix(const server_tokens & b) const {
1274-
size_t max_idx = std::min(tokens.size(), b.tokens.size());
1273+
const size_t max_idx = std::min(tokens.size(), b.tokens.size());
1274+
12751275
for (size_t i = 0; i < max_idx; ++i) {
1276-
auto & ai = tokens[i];
1277-
auto & bi = b.tokens[i];
1276+
const llama_token ai = tokens[i];
1277+
const llama_token bi = b.tokens[i];
12781278

12791279
if (ai == LLAMA_TOKEN_NULL && bi == LLAMA_TOKEN_NULL) {
12801280
GGML_ASSERT(has_mtmd);
1281+
12811282
const auto & a_chunk = find_chunk(i);
12821283
const auto & b_chunk = b.find_chunk(i);
1284+
12831285
GGML_ASSERT(a_chunk && b_chunk);
1284-
std::string ai_id = mtmd_input_chunk_get_id(a_chunk.get());
1285-
std::string bi_id = mtmd_input_chunk_get_id(b_chunk.get());
1286-
size_t a_pos = mtmd_input_chunk_get_n_pos(a_chunk.get());
1287-
size_t b_pos = mtmd_input_chunk_get_n_pos(b_chunk.get());
1288-
if (ai_id == bi_id && a_pos == b_pos) {
1289-
GGML_ASSERT(a_pos > 0 && "Invalid media chunk"); // should never happen
1290-
i += a_pos - 1; // will be +1 by the for loop
1286+
1287+
const std::string id_ai = mtmd_input_chunk_get_id(a_chunk.get());
1288+
const std::string id_bi = mtmd_input_chunk_get_id(b_chunk.get());
1289+
1290+
const size_t pos_a = mtmd_input_chunk_get_n_pos(a_chunk.get());
1291+
const size_t pos_b = mtmd_input_chunk_get_n_pos(b_chunk.get());
1292+
1293+
if (id_ai == id_bi && pos_a == pos_b) {
1294+
GGML_ASSERT(pos_a > 0 && "Invalid media chunk"); // should never happen
1295+
i += pos_a - 1; // will be +1 by the for loop
12911296
continue;
1292-
} else {
1293-
return i;
12941297
}
1295-
} else if (ai == bi) {
1296-
continue;
1297-
} else {
1298+
12981299
return i;
12991300
}
1301+
1302+
if (ai == bi) {
1303+
continue;
1304+
}
1305+
1306+
return i;
13001307
}
1308+
13011309
return max_idx; // all tokens are equal
13021310
}
13031311

@@ -1308,7 +1316,7 @@ struct server_tokens {
13081316
const int32_t n_vocab = llama_vocab_n_tokens(vocab);
13091317

13101318
for (size_t i = 0; i < tokens.size(); ++i) {
1311-
auto & t = tokens[i];
1319+
const auto & t = tokens[i];
13121320
if (t == LLAMA_TOKEN_NULL) {
13131321
try {
13141322
const auto & chunk = find_chunk(i);
@@ -1330,8 +1338,8 @@ struct server_tokens {
13301338
mtmd_context * mctx,
13311339
llama_pos n_past,
13321340
int32_t seq_id,
1333-
llama_pos & n_pos_out) {
1334-
auto & chunk = find_chunk(n_past);
1341+
llama_pos & n_pos_out) const {
1342+
const auto & chunk = find_chunk(n_past);
13351343
const char * name = mtmd_input_chunk_get_type(chunk.get()) == MTMD_INPUT_CHUNK_TYPE_IMAGE
13361344
? "image" : "audio";
13371345
SRV_INF("processing %s...\n", name);

0 commit comments

Comments
 (0)