fixed tests

Signed-off-by: Onkar Chougule <ochougul@qti.qualcomm.com>

Author: ochougul

tests/peft/lora/test_lora_model.py

@@ -222,7 +222,7 @@ def test_auto_lora_model_for_causal_lm_noncb_export_compile_generate(
 
     # export
     start = perf_counter()
-    qeff_model.export(export_dir=tmp_path)
+    onnx_path = qeff_model.export(export_dir=tmp_path)
     end = perf_counter()
     export_time_0 = end - start
     model_path = tmp_path.with_name(tmp_path.name + "-" + qeff_model.export_hash)
@@ -237,7 +237,7 @@ def test_auto_lora_model_for_causal_lm_noncb_export_compile_generate(
     assert export_time_1 < export_time_0
 
     # test compile
-    qeff_model.compile(prefill_seq_len=32, ctx_len=64)
+    qeff_model.compile(onnx_path=onnx_path, prefill_seq_len=32, ctx_len=64)
     assert Path(qeff_model.qpc_path).is_dir()
     assert os.path.isfile(os.path.join(os.path.dirname(qeff_model.qpc_path), "qconfig.json"))
 

tests/transformers/models/test_disagg_mode.py

@@ -56,9 +56,7 @@ def test_disagg_mode_prefill(model_id, prompt):
 
     undo_transformers_quantizers()
 
-    qeff_model = QEFFAutoModelForCausalLM.from_pretrained(
-        model_id, num_hidden_layers=2, max_position_embeddings=64 * 1024
-    )
+    qeff_model = QEFFAutoModelForCausalLM.from_pretrained(model_id, num_hidden_layers=2)
     qeff_model.prefill(True)
     config = qeff_model.model.config
     inputs = tokenizer(prompt, return_tensors="np", padding="max_length", max_length=padded_len)