Fix max nodes (again) (#1306)
This commit is contained in:
parent
2bb40f8c35
commit
68bd30d99c
|
|
@ -430,8 +430,8 @@ struct llama_model {
|
|||
if (arch == LLM_ARCH_QWEN3NEXT || arch == LLM_ARCH_QWEN35MOE) {
|
||||
return std::max<size_t>(n_tokens * 40, 32u * n_tensors);
|
||||
}
|
||||
return std::max<size_t>(1024, 8*n_tensors);
|
||||
//return 65536 * 2;
|
||||
//return std::max<size_t>(1024, 8*n_tensors);
|
||||
return 65536;
|
||||
}
|
||||
|
||||
bool has_tensor_overrides() const {
|
||||
|
|
|
|||
Loading…
Reference in New Issue