Skip Open Llama 3B from nightly due to DRAM limitations. (#1063)

Keeping Llama 3.2 1B as ref.

forge/test/mlir/llama/tests/test_llama_prefil.py

@@ -56,6 +56,9 @@ def test_llama_prefil_on_device_decode_on_cpu(model_path):
     - The first part is the prefilling of the model on the device.
     - The second part is the decoding of the model on the CPU without KV cache.
     """
+    if model_path == "openlm-research/open_llama_3b":
+        pytest.skip("Insufficient host DRAM to run this model (requires a bit more than 32 GB during compile time)")
+
     # Load Llama model and tokenizer
     model, tokenizer = load_model(model_path, return_dict=True)