Skip to content

change default cache_interal_miltiplier [test_doc_files_windows=demos… #4314

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

Open
dtrawins wants to merge 3 commits into
base: main
Choose a base branch
from
Open

change default cache_interal_miltiplier [test_doc_files_windows=demos… #4314

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
96fb9e4
change default cache_interal_miltiplier [test_doc_files_windows=demos…
dtrawins Jun 22, 2026
91ad990
fix tests
dtrawins Jun 22, 2026
e894177
Merge branch 'main' into CVS-188061-64
dtrawins Jun 25, 2026
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
1 change: 1 addition & 0 deletions docs/parameters.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -137,6 +137,7 @@ Task specific parameters for different tasks (text generation/image generation/e
| `--reasoning_parser` | `string` | Type of parser to use for reasoning content extraction from model output. Currently supported: [qwen3, gptoss, gemma4] |
| `--tool_parser` | `string` | Type of parser to use for tool calls extraction from model output. Currently supported: [llama3, phi4, hermes3, mistral, qwen3coder, gptoss, devstral, lfm2, gemma4] |
| `--enable_tool_guided_generation` | `bool` | Enables enforcing tool schema during generation. Requires setting response parser. Default: false. |
| `--cache_interval_multiplier` | `integer` | Multiplier for the KV cache block interval. Controls the granularity of cache allocation. Applicable only for models with linear attention. Default: 64. |

### Image generation
| option | Value format | Description |
Expand Down
4 changes: 2 additions & 2 deletions src/graph_export/graph_cli_parser.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,8 +82,8 @@ void GraphCLIParser::createOptions() {
cxxopts::value<std::string>()->default_value("false"),
"ENABLE_TOOL_GUIDED_GENERATION")
("cache_interval_multiplier",
"Multiplier for the KV cache block interval. Controls the granularity of cache allocation. Default: unset.",
cxxopts::value<uint64_t>(),
"Multiplier for the KV cache block interval. Controls the granularity of cache allocation. Applicable only for models with linear attention. Default: 64.",
cxxopts::value<uint64_t>()->default_value("64"),
"CACHE_INTERVAL_MULTIPLIER");
Comment on lines 84 to 87

options->add_options("plugin config")
Expand Down
4 changes: 1 addition & 3 deletions src/llm/language_model/continuous_batching/servable_initializer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,9 +171,7 @@ Status ContinuousBatchingServableInitializer::initialize(std::shared_ptr<GenAiSe
properties->schedulerConfig.dynamic_split_fuse = nodeOptions.dynamic_split_fuse();
properties->schedulerConfig.max_num_seqs = nodeOptions.max_num_seqs();
properties->schedulerConfig.enable_prefix_caching = nodeOptions.enable_prefix_caching();
if (nodeOptions.has_cache_interval_multiplier()) {
properties->schedulerConfig.cache_interval_multiplier = nodeOptions.cache_interval_multiplier();
}
properties->schedulerConfig.cache_interval_multiplier = nodeOptions.cache_interval_multiplier();

if (nodeOptions.has_cache_eviction_config()) {
properties->schedulerConfig.cache_eviction_config = prepareCacheEvictionConfig(nodeOptions);
Expand Down
3 changes: 2 additions & 1 deletion src/llm/llm_calculator.proto
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,7 +136,8 @@ message LLMCalculatorOptions {

optional SparseAttentionConfig sparse_attention_config = 24;

optional uint64 cache_interval_multiplier = 25;
// Applicable only for models with linear attention.
optional uint64 cache_interval_multiplier = 25 [default = 64];
Comment on lines +139 to +140

@dtrawins dtrawins Jun 29, 2026

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

confirmed that empty value results with correct default 64. has_cache_interval_multiplier is not used anymore


enum ChatTemplateMode {
// Use GenAI's apply_chat_template (minja-based).
Expand Down
2 changes: 2 additions & 0 deletions src/test/llm/llmnode_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4249,6 +4249,8 @@ void TestLLMNodeOptionsCheckDefault(std::string& modelsPath) {
ASSERT_EQ(properties->schedulerConfig.dynamic_split_fuse, true);
ASSERT_EQ(properties->schedulerConfig.max_num_seqs, 256);
ASSERT_EQ(properties->schedulerConfig.enable_prefix_caching, false);
ASSERT_TRUE(properties->schedulerConfig.cache_interval_multiplier.has_value());

@atobiszei atobiszei Jul 8, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If I recall even if we have default value set in Proto then has_value returns false (way of checking if option was manually set). Or is it set here since we export the graph, and we set default value in CLI and thus the value in pbtxt is populated anyway?

ASSERT_EQ(properties->schedulerConfig.cache_interval_multiplier.value(), 64);
Comment on lines +4252 to +4253
ASSERT_EQ(properties->device, "CPU");
// CPU default properties (inference_num_threads, enable_cpu_pinning) are automatically
// added to pluginConfig for CPU device; verify no user-specified entries are present.
Expand Down