Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 01:05:09 +02:00
Code Issues Packages Projects Releases Wiki Activity

does not work for e4b

Browse Source
This commit is contained in:
Concedo
2026-06-04 10:44:40 +08:00
parent 1f14c98d36 94a220cd67
commit 7fb55b1e32
106 changed files with 4684 additions and 1205 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/arg.cpp
+12 -5
View File
@@ -354,7 +354,6 @@ static handle_model_result common_params_handle_model(struct common_params_model
model.path = "";
}
common_download_opts hf_opts = opts;
hf_opts.download_mmproj = true; // also look for mmproj when downloading hf model
auto download_result = common_download_model(model, hf_opts);
if (download_result.model_path.empty()) {
@@ -442,10 +441,11 @@ bool common_params_handle_models(common_params & params, llama_example curr_ex)
COMMON_SPECULATIVE_TYPE_DRAFT_MTP) != params.speculative.types.end();
common_download_opts opts;
opts.bearer_token = params.hf_token;
opts.offline = params.offline;
opts.skip_download = params.skip_download;
opts.download_mtp = spec_type_draft_mtp;
opts.bearer_token = params.hf_token;
opts.offline = params.offline;
opts.skip_download = params.skip_download;
opts.download_mtp = spec_type_draft_mtp;
opts.download_mmproj = !params.no_mmproj;
try {
auto res = common_params_handle_model(params.model, opts);
@@ -3028,6 +3028,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
params.timeout_write = value;
}
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_TIMEOUT"));
add_opt(common_arg(
{"--sse-ping-interval"}, "N",
string_format("server SSE ping interval in seconds (-1 = disabled, default: %d)", params.sse_ping_interval),
[](common_params & params, int value) {
params.sse_ping_interval = value;
}
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_SSE_PING_INTERVAL"));
add_opt(common_arg(
{"--threads-http"}, "N",
string_format("number of threads used to process HTTP requests (default: %d)", params.n_threads_http),