[V1] LoRA Support (#10957)

Signed-off-by: Varun Sundar Rabindranath <varun@neuralmagic.com>
Co-authored-by: Varun Sundar Rabindranath <varun@neuralmagic.com>
This commit is contained in:
Varun Sundar Rabindranath
2025-02-06 09:32:51 -08:00
committed by GitHub
co-authored by Varun Sundar Rabindranath
parent 8108ac841d
commit 467a96a541
16 changed files with 453 additions and 56 deletions
+17
View File
@@ -306,3 +306,20 @@ def llama_2_7b_engine_extra_embeddings():
def llama_2_7b_model_extra_embeddings(llama_2_7b_engine_extra_embeddings):
yield (llama_2_7b_engine_extra_embeddings.model_executor.driver_worker.
model_runner.model)
@pytest.fixture(params=[True, False])
def run_with_both_engines_lora(request, monkeypatch):
# Automatically runs tests twice, once with V1 and once without
use_v1 = request.param
# Tests decorated with `@skip_v1` are only run without v1
skip_v1 = request.node.get_closest_marker("skip_v1")
if use_v1:
if skip_v1:
pytest.skip("Skipping test on vllm V1")
monkeypatch.setenv('VLLM_USE_V1', '1')
else:
monkeypatch.setenv('VLLM_USE_V1', '0')
yield