Skip to content

Commit d8651c8

Browse files
committed
Merge remote-tracking branch 'origin/master' into gabe-l-hart/alora-support
* origin/master: (72 commits) metal : Add template specialization for mul_mm_id w/ ne20 == 10 (#15799) llama : set n_outputs to 1 to avoid 0 outputs mean-pooling (#15791) CANN: Refactor ND to NZ workspace to be per-device (#15763) server: add exceed_context_size_error type (#15780) Document the new max GPU layers default in help (#15771) ggml: add ops for WAN video model (cuda && cpu) (#15669) CANN: Fix precision issue on 310I DUO multi-devices (#15784) opencl: add hs=40 to FA (#15758) CANN: fix acl_rstd allocation size in ggml_cann_rms_norm (#15760) vulkan: fix mmv subgroup16 selection (#15775) vulkan: don't use std::string in load_shaders, to improve compile time (#15724) vulkan : update ggml_vk_instance_validation_ext_available (#15666) ggml vulkan: add hardsigmoid and hardswish operations (#15762) CUDA: Optimize `rms_norm_f32` kernel and its fused variants, giving 1-6% perf E2E (#15715) model-conversion : fix pyright errors (#15770) sampling : optimize dist sampler (#15704) llama : fix incorrect model type for Gemma 270M (#15764) model-conversion : remove hardcoded /bin/bash shebangs [no ci] (#15765) CANN: Add RoPE contiguous check for 310I DUP device (#15735) ggml-cpu : optimize RVV kernels (#15720) ...
2 parents f863c9f + 856ed09 commit d8651c8

File tree

127 files changed

+5472
-1580
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

127 files changed

+5472
-1580
lines changed

.clang-format

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -22,7 +22,7 @@ AllowShortIfStatementsOnASingleLine: Never
2222
AllowShortLambdasOnASingleLine: Inline
2323
AllowShortLoopsOnASingleLine: false
2424
AlwaysBreakBeforeMultilineStrings: true
25-
BinPackArguments: false
25+
BinPackArguments: true
2626
BinPackParameters: false # OnePerLine
2727
BitFieldColonSpacing: Both
2828
BreakBeforeBraces: Custom # Attach

README.md

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -137,6 +137,7 @@ Instructions for adding support for new models: [HOWTO-add-model.md](docs/develo
137137
- [X] [Trillion-7B-preview](https://huggingface.co/trillionlabs/Trillion-7B-preview)
138138
- [x] [Ling models](https://huggingface.co/collections/inclusionAI/ling-67c51c85b34a7ea0aba94c32)
139139
- [x] [LFM2 models](https://huggingface.co/collections/LiquidAI/lfm2-686d721927015b2ad73eaa38)
140+
- [x] [Hunyuan models](https://huggingface.co/collections/tencent/hunyuan-dense-model-6890632cda26b19119c9c5e7)
140141

141142
#### Multimodal
142143

ci/run.sh

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -386,10 +386,10 @@ function gg_run_open_llama_7b_v2 {
386386

387387
(time ./bin/llama-imatrix --model ${model_f16} -f ${wiki_test} -t 1 -ngl 99 -c 2048 -b 512 --chunks 4 ) 2>&1 | tee -a $OUT/${ci}-imatrix.log
388388

389-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
390-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 -fa ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
391-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
392-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
389+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 -fa off ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
390+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 -fa on ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
391+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa off ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
392+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa on ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
393393

394394
function check_ppl {
395395
qnt="$1"
@@ -520,8 +520,8 @@ function gg_run_pythia_1_4b {
520520

521521
(time ./bin/llama-imatrix --model ${model_f16} -f ${wiki_test_60} -ngl 99 -c 128 -b 128 --chunks 1 ) 2>&1 | tee -a $OUT/${ci}-imatrix.log
522522

523-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
524-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
523+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa off ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
524+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa on ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
525525

526526
function check_ppl {
527527
qnt="$1"
@@ -651,10 +651,10 @@ function gg_run_pythia_2_8b {
651651

652652
(time ./bin/llama-imatrix --model ${model_f16} -f ${wiki_test} -t 1 -ngl 99 -c 2048 -b 512 --chunks 4 ) 2>&1 | tee -a $OUT/${ci}-imatrix.log
653653

654-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
655-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 -fa ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
656-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
657-
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
654+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 -fa off ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
655+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 10 -c 0 -fa on ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
656+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa off ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
657+
(time ./bin/llama-save-load-state --model ${model_q4_0} -ngl 99 -c 0 -fa on ) 2>&1 | tee -a $OUT/${ci}-save-load-state.log
658658

659659
function check_ppl {
660660
qnt="$1"

common/arg.cpp

Lines changed: 20 additions & 26 deletions
Original file line numberDiff line numberDiff line change
@@ -1545,10 +1545,18 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
15451545
}
15461546
).set_examples({LLAMA_EXAMPLE_IMATRIX, LLAMA_EXAMPLE_PERPLEXITY, LLAMA_EXAMPLE_RETRIEVAL}));
15471547
add_opt(common_arg(
1548-
{"-fa", "--flash-attn"},
1549-
string_format("enable Flash Attention (default: %s)", params.flash_attn ? "enabled" : "disabled"),
1550-
[](common_params & params) {
1551-
params.flash_attn = true;
1548+
{"-fa", "--flash-attn"}, "FA",
1549+
string_format("set Flash Attention use ('on', 'off', or 'auto', default: '%s')", llama_flash_attn_type_name(params.flash_attn_type)),
1550+
[](common_params & params, const std::string & value) {
1551+
if (value == "on" || value == "enabled" || value == "1") {
1552+
params.flash_attn_type = LLAMA_FLASH_ATTN_TYPE_ENABLED;
1553+
} else if (value == "off" || value == "disabled" || value == "0") {
1554+
params.flash_attn_type = LLAMA_FLASH_ATTN_TYPE_DISABLED;
1555+
} else if (value == "auto" || value == "-1") {
1556+
params.flash_attn_type = LLAMA_FLASH_ATTN_TYPE_AUTO;
1557+
} else {
1558+
throw std::runtime_error(string_format("error: unkown value for --flash-attn: '%s'\n", value.c_str()));
1559+
}
15521560
}
15531561
).set_env("LLAMA_ARG_FLASH_ATTN"));
15541562
add_opt(common_arg(
@@ -2458,7 +2466,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
24582466
).set_examples({LLAMA_EXAMPLE_SPECULATIVE, LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_N_CPU_MOE_DRAFT"));
24592467
add_opt(common_arg(
24602468
{"-ngl", "--gpu-layers", "--n-gpu-layers"}, "N",
2461-
"number of layers to store in VRAM",
2469+
string_format("max. number of layers to store in VRAM (default: %d)", params.n_gpu_layers),
24622470
[](common_params & params, int value) {
24632471
params.n_gpu_layers = value;
24642472
if (!llama_supports_gpu_offload()) {
@@ -2954,20 +2962,20 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
29542962
params.endpoint_metrics = true;
29552963
}
29562964
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_ENDPOINT_METRICS"));
2957-
add_opt(common_arg(
2958-
{"--slots"},
2959-
string_format("enable slots monitoring endpoint (default: %s)", params.endpoint_slots ? "enabled" : "disabled"),
2960-
[](common_params & params) {
2961-
params.endpoint_slots = true;
2962-
}
2963-
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_ENDPOINT_SLOTS"));
29642965
add_opt(common_arg(
29652966
{"--props"},
29662967
string_format("enable changing global properties via POST /props (default: %s)", params.endpoint_props ? "enabled" : "disabled"),
29672968
[](common_params & params) {
29682969
params.endpoint_props = true;
29692970
}
29702971
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_ENDPOINT_PROPS"));
2972+
add_opt(common_arg(
2973+
{"--slots"},
2974+
string_format("enable slots monitoring endpoint (default: %s)", params.endpoint_slots ? "enabled" : "disabled"),
2975+
[](common_params & params) {
2976+
params.endpoint_slots = true;
2977+
}
2978+
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_ENDPOINT_SLOTS"));
29712979
add_opt(common_arg(
29722980
{"--no-slots"},
29732981
"disables slots monitoring endpoint",
@@ -3459,8 +3467,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
34593467
params.model.hf_repo = "ggml-org/Qwen2.5-Coder-1.5B-Q8_0-GGUF";
34603468
params.model.hf_file = "qwen2.5-coder-1.5b-q8_0.gguf";
34613469
params.port = 8012;
3462-
params.n_gpu_layers = 99;
3463-
params.flash_attn = true;
34643470
params.n_ubatch = 1024;
34653471
params.n_batch = 1024;
34663472
params.n_ctx = 0;
@@ -3475,8 +3481,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
34753481
params.model.hf_repo = "ggml-org/Qwen2.5-Coder-3B-Q8_0-GGUF";
34763482
params.model.hf_file = "qwen2.5-coder-3b-q8_0.gguf";
34773483
params.port = 8012;
3478-
params.n_gpu_layers = 99;
3479-
params.flash_attn = true;
34803484
params.n_ubatch = 1024;
34813485
params.n_batch = 1024;
34823486
params.n_ctx = 0;
@@ -3491,8 +3495,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
34913495
params.model.hf_repo = "ggml-org/Qwen2.5-Coder-7B-Q8_0-GGUF";
34923496
params.model.hf_file = "qwen2.5-coder-7b-q8_0.gguf";
34933497
params.port = 8012;
3494-
params.n_gpu_layers = 99;
3495-
params.flash_attn = true;
34963498
params.n_ubatch = 1024;
34973499
params.n_batch = 1024;
34983500
params.n_ctx = 0;
@@ -3508,10 +3510,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
35083510
params.model.hf_file = "qwen2.5-coder-7b-q8_0.gguf";
35093511
params.speculative.model.hf_repo = "ggml-org/Qwen2.5-Coder-0.5B-Q8_0-GGUF";
35103512
params.speculative.model.hf_file = "qwen2.5-coder-0.5b-q8_0.gguf";
3511-
params.speculative.n_gpu_layers = 99;
35123513
params.port = 8012;
3513-
params.n_gpu_layers = 99;
3514-
params.flash_attn = true;
35153514
params.n_ubatch = 1024;
35163515
params.n_batch = 1024;
35173516
params.n_ctx = 0;
@@ -3527,10 +3526,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
35273526
params.model.hf_file = "qwen2.5-coder-14b-q8_0.gguf";
35283527
params.speculative.model.hf_repo = "ggml-org/Qwen2.5-Coder-0.5B-Q8_0-GGUF";
35293528
params.speculative.model.hf_file = "qwen2.5-coder-0.5b-q8_0.gguf";
3530-
params.speculative.n_gpu_layers = 99;
35313529
params.port = 8012;
3532-
params.n_gpu_layers = 99;
3533-
params.flash_attn = true;
35343530
params.n_ubatch = 1024;
35353531
params.n_batch = 1024;
35363532
params.n_ctx = 0;
@@ -3545,8 +3541,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
35453541
params.model.hf_repo = "ggml-org/Qwen3-Coder-30B-A3B-Instruct-Q8_0-GGUF";
35463542
params.model.hf_file = "qwen3-coder-30b-a3b-instruct-q8_0.gguf";
35473543
params.port = 8012;
3548-
params.n_gpu_layers = 99;
3549-
params.flash_attn = true;
35503544
params.n_ubatch = 1024;
35513545
params.n_batch = 1024;
35523546
params.n_ctx = 0;

common/chat.cpp

Lines changed: 152 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -622,6 +622,7 @@ const char * common_chat_format_name(common_chat_format format) {
622622
case COMMON_CHAT_FORMAT_COMMAND_R7B: return "Command R7B";
623623
case COMMON_CHAT_FORMAT_GRANITE: return "Granite";
624624
case COMMON_CHAT_FORMAT_GPT_OSS: return "GPT-OSS";
625+
case COMMON_CHAT_FORMAT_SEED_OSS: return "Seed-OSS";
625626
default:
626627
throw std::runtime_error("Unknown chat format");
627628
}
@@ -2059,6 +2060,94 @@ static void common_chat_parse_granite(common_chat_msg_parser & builder) {
20592060
}
20602061
}
20612062

2063+
static void common_chat_parse_seed_oss(common_chat_msg_parser & builder) {
2064+
// Parse thinking tags first - this handles the main reasoning content
2065+
builder.try_parse_reasoning("<seed:think>", "</seed:think>");
2066+
2067+
if (!builder.syntax().parse_tool_calls) {
2068+
builder.add_content(builder.consume_rest());
2069+
return;
2070+
}
2071+
2072+
// Parse tool calls - Seed-OSS uses <seed:tool_call> format
2073+
static const common_regex tool_call_begin_regex("<seed:tool_call>");
2074+
static const common_regex tool_call_end_regex("</seed:tool_call>");
2075+
static const common_regex function_regex("<function=([^>]+)>");
2076+
static const common_regex param_regex("<parameter=([^>]+)>");
2077+
2078+
while (auto tool_res = builder.try_find_regex(tool_call_begin_regex)) {
2079+
builder.consume_spaces(); // Consume whitespace after <seed:tool_call>
2080+
2081+
// Look for function call inside tool call, ignore any content before it
2082+
if (auto func_res = builder.try_find_regex(function_regex, std::string::npos, false)) {
2083+
auto function_name = builder.str(func_res->groups[1]);
2084+
2085+
// Parse Seed-OSS parameters <parameter=name>value</parameter>
2086+
json args = json::object();
2087+
// Parse all parameters
2088+
while (auto param_res = builder.try_find_regex(param_regex, std::string::npos, false)) {
2089+
// again, ignore noise around parameters
2090+
auto param_name = builder.str(param_res->groups[1]);
2091+
builder.move_to(param_res->groups[0].end);
2092+
builder.consume_spaces(); // Consume whitespace after parameter
2093+
auto savedPos = builder.pos();
2094+
if (auto param_parse = builder.try_find_literal("</parameter>")) {
2095+
auto param = param_parse->prelude;
2096+
builder.move_to(savedPos);
2097+
try {
2098+
if (auto param_res = builder.try_consume_json()) {
2099+
args[param_name] = param_res->json;
2100+
} else {
2101+
args[param_name] = param;
2102+
}
2103+
} catch (json::exception &) {
2104+
args[param_name] = param;
2105+
}
2106+
} else {
2107+
throw common_chat_msg_partial_exception("Incomplete tool parameter");
2108+
}
2109+
}
2110+
// Look for closing function tag
2111+
auto end_func = builder.try_find_literal("</function>");
2112+
if (end_func) {
2113+
builder.move_to(end_func->groups[0].end);
2114+
builder.consume_spaces(); // Consume whitespace after </function>
2115+
2116+
// Add the tool call with parsed arguments, but only if we REALLY got the literal
2117+
auto eaten_fragment = builder.input().substr(end_func->groups[0].begin, end_func->groups[0].end);
2118+
auto funlen = std::string("</function>").length();
2119+
if (eaten_fragment.length() >= funlen && eaten_fragment.substr(0, funlen) == std::string("</function>")) {
2120+
if (!builder.add_tool_call(function_name, "", args.dump())) {
2121+
throw common_chat_msg_partial_exception("Incomplete tool call");
2122+
}
2123+
} else {
2124+
throw common_chat_msg_partial_exception("Incomplete tool call");
2125+
}
2126+
} else {
2127+
throw common_chat_msg_partial_exception("Incomplete tool call");
2128+
}
2129+
// Look for closing tool call tag
2130+
if (auto end_tool = builder.try_find_regex(tool_call_end_regex, std::string::npos, false)) {
2131+
builder.move_to(end_tool->groups[0].end);
2132+
builder.consume_spaces(); // Consume trailing whitespace after tool call
2133+
} else {
2134+
throw common_chat_msg_partial_exception("Incomplete tool call");
2135+
}
2136+
} else {
2137+
// No function found - don't consume content here, let it be handled at the end
2138+
break;
2139+
}
2140+
}
2141+
2142+
// Consume any remaining whitespace after all tool call processing
2143+
builder.consume_spaces();
2144+
auto remaining = builder.consume_rest();
2145+
// If there's any non-whitespace content remaining, add it as content
2146+
if (!string_strip(remaining).empty()) {
2147+
builder.add_content(remaining);
2148+
}
2149+
}
2150+
20622151
static common_chat_params common_chat_params_init_without_tools(const common_chat_template & tmpl, const struct templates_params & inputs) {
20632152
common_chat_params data;
20642153
data.prompt = apply(tmpl, inputs);
@@ -2075,8 +2164,62 @@ static common_chat_params common_chat_params_init_without_tools(const common_cha
20752164
return data;
20762165
}
20772166

2167+
static common_chat_params common_chat_params_init_seed_oss(
2168+
const common_chat_template & tmpl,
2169+
templates_params & params,
2170+
const common_chat_templates_inputs & inputs)
2171+
{
2172+
common_chat_params data;
2173+
data.prompt = apply(tmpl, params);
2174+
data.format = COMMON_CHAT_FORMAT_SEED_OSS;
2175+
if (string_ends_with(data.prompt, "<seed:think>")) {
2176+
if (!inputs.enable_thinking) {
2177+
data.prompt += "</seed:think>";
2178+
} else {
2179+
data.thinking_forced_open = true;
2180+
}
2181+
}
2182+
2183+
if (params.tools.is_array() && !params.tools.empty()) {
2184+
data.grammar_lazy = inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_REQUIRED;
2185+
data.grammar = build_grammar([&](const common_grammar_builder & builder) {
2186+
std::vector<std::string> tool_rules;
2187+
foreach_function(params.tools, [&](const json & tool) {
2188+
const auto & function = tool.at("function");
2189+
std::string name = function.at("name");
2190+
auto parameters = function.at("parameters");
2191+
builder.resolve_refs(parameters);
2192+
2193+
// Create rule for Seed-OSS function call format
2194+
std::string param_rules;
2195+
if (parameters.contains("properties")) {
2196+
for (const auto & [key, value] : parameters.at("properties").items()) {
2197+
param_rules += "\"<parameter=" + key + ">\"" + builder.add_schema(name + "-arg-" + key, value) +
2198+
"\"</parameter>\"";
2199+
}
2200+
}
2201+
2202+
tool_rules.push_back(builder.add_rule(name + "-call",
2203+
"\"<seed:tool_call>\" space \"<function=" + name + ">\" space " +
2204+
param_rules +
2205+
" \"</function>\" space \"</seed:tool_call>\""));
2206+
});
2207+
2208+
data.grammar_triggers.push_back({ COMMON_GRAMMAR_TRIGGER_TYPE_WORD, "<seed:tool_call>" });
2209+
2210+
data.preserved_tokens = {
2211+
"<seed:think>", "</seed:think>", "<seed:tool_call>", "</seed:tool_call>",
2212+
"<function=", "</function>", "<parameter=", "</parameter>",
2213+
};
2214+
2215+
builder.add_rule("root", string_join(tool_rules, " | "));
2216+
});
2217+
}
2218+
return data;
2219+
}
2220+
20782221
static common_chat_params common_chat_templates_apply_jinja(
2079-
const struct common_chat_templates * tmpls,
2222+
const struct common_chat_templates * tmpls,
20802223
const struct common_chat_templates_inputs & inputs)
20812224
{
20822225
templates_params params;
@@ -2145,6 +2288,11 @@ static common_chat_params common_chat_templates_apply_jinja(
21452288
return common_chat_params_init_gpt_oss(tmpl, params);
21462289
}
21472290

2291+
// Seed-OSS
2292+
if (src.find("<seed:think>") != std::string::npos) {
2293+
return common_chat_params_init_seed_oss(tmpl, params, inputs);
2294+
}
2295+
21482296
// Use generic handler when mixing tools + JSON schema.
21492297
// TODO: support that mix in handlers below.
21502298
if ((params.tools.is_array() && params.json_schema.is_object())) {
@@ -2303,6 +2451,9 @@ static void common_chat_parse(common_chat_msg_parser & builder) {
23032451
case COMMON_CHAT_FORMAT_GPT_OSS:
23042452
common_chat_parse_gpt_oss(builder);
23052453
break;
2454+
case COMMON_CHAT_FORMAT_SEED_OSS:
2455+
common_chat_parse_seed_oss(builder);
2456+
break;
23062457
default:
23072458
throw std::runtime_error(std::string("Unsupported format: ") + common_chat_format_name(builder.syntax().format));
23082459
}

common/chat.h

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -111,6 +111,7 @@ enum common_chat_format {
111111
COMMON_CHAT_FORMAT_COMMAND_R7B,
112112
COMMON_CHAT_FORMAT_GRANITE,
113113
COMMON_CHAT_FORMAT_GPT_OSS,
114+
COMMON_CHAT_FORMAT_SEED_OSS,
114115

115116
COMMON_CHAT_FORMAT_COUNT, // Not a format, just the # formats
116117
};

0 commit comments

Comments
 (0)