llama: refer to segment documentation [no ci] (#29074)
This commit is contained in:
1 parent
f1cee9941b
commit
78e2964c23
2 files changed
+2
-1
No files matched your search
+1
-1
@@ -604,8 +604,8 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str
|
||||
return get_tensor_config_impl(GGML_BACKEND_SPLIT_AXIS_MIRRORED);
|
||||
};
|
||||
|
||||
// if a model has fused tensors they need to be separated into "segments", see the comments on ggml_backend_meta_split_state struct
|
||||
auto get_split_segments = [&](int axis, uint32_t il) -> std::vector<std::pair<int64_t, uint32_t>> {
|
||||
// TODO: clarify why this is necessary specifically for these models
|
||||
// TODO: deduplicate condition [TAG_SPLIT_QGATE_QWEN]
|
||||
if (ud->model->arch == LLM_ARCH_QWEN3NEXT || ud->model->arch == LLM_ARCH_QWEN35 || ud->model->arch == LLM_ARCH_QWEN35MOE ||
|
||||
ud->model->arch == LLM_ARCH_QWEN4EXP) {
|
||||
|
||||
Reference in new issue
Block a user