65 lines
1.7 KiB
Python
65 lines
1.7 KiB
Python
from __future__ import annotations
|
|
|
|
import json
|
|
|
|
|
|
|
|
def test_simple_defaults_accept_tensor_split(load_nodes_module, monkeypatch, tmp_path):
|
|
module = load_nodes_module()
|
|
config_path = tmp_path / "simple_defaults.json"
|
|
config_path.write_text(
|
|
json.dumps(
|
|
{
|
|
"schema_version": 1,
|
|
"n_gpu_layers": -1,
|
|
"tensor_split": [1.0, 0.0],
|
|
}
|
|
),
|
|
encoding="utf-8",
|
|
)
|
|
|
|
defaults = module._load_simple_defaults(str(config_path))
|
|
|
|
assert defaults["tensor_split"] == [1.0, 0.0]
|
|
|
|
|
|
def test_simple_defaults_ignore_invalid_tensor_split(load_nodes_module, monkeypatch, tmp_path):
|
|
module = load_nodes_module()
|
|
config_path = tmp_path / "simple_defaults.json"
|
|
config_path.write_text(
|
|
json.dumps(
|
|
{
|
|
"schema_version": 1,
|
|
"tensor_split": [0.0, 0.0],
|
|
}
|
|
),
|
|
encoding="utf-8",
|
|
)
|
|
|
|
defaults = module._load_simple_defaults(str(config_path))
|
|
|
|
assert defaults["tensor_split"] is None
|
|
|
|
|
|
def test_model_manager_passes_tensor_split_to_llama(load_nodes_module, monkeypatch):
|
|
module = load_nodes_module()
|
|
calls = []
|
|
|
|
class DummyLlama:
|
|
def __init__(self, **kwargs):
|
|
calls.append(kwargs)
|
|
|
|
monkeypatch.setattr(module, "LLAMA_CPP_AVAILABLE", True)
|
|
monkeypatch.setattr(module, "Llama", DummyLlama)
|
|
|
|
manager = module.GGUFModelManager()
|
|
manager.load_model(
|
|
model_path="C:/models/test.gguf",
|
|
mmproj_path=module._MMPROJ_NOT_REQUIRED,
|
|
n_ctx=1024,
|
|
n_gpu_layers=-1,
|
|
tensor_split=[1.0, 0.0],
|
|
)
|
|
|
|
assert calls[0]["tensor_split"] == [1.0, 0.0]
|