From 7d6ff64c133353d69f0d1a9c8e8eba2d5d72d97a Mon Sep 17 00:00:00 2001 From: David Corvoysier Date: Wed, 19 Feb 2025 14:51:43 +0000 Subject: [PATCH] test(neuron): use smaller llama model --- integration-tests/fixtures/neuron/model.py | 2 +- integration-tests/neuron/integration/test_generate.py | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/integration-tests/fixtures/neuron/model.py b/integration-tests/fixtures/neuron/model.py index 676cb093..eb109121 100644 --- a/integration-tests/fixtures/neuron/model.py +++ b/integration-tests/fixtures/neuron/model.py @@ -32,7 +32,7 @@ MODEL_CONFIGURATIONS = { "export_kwargs": {"batch_size": 4, "sequence_length": 1024, "num_cores": 2, "auto_cast_type": "fp16"}, }, "llama": { - "model_id": "NousResearch/Hermes-2-Theta-Llama-3-8B", + "model_id": "unsloth/Llama-3.2-1B-Instruct", "export_kwargs": {"batch_size": 4, "sequence_length": 2048, "num_cores": 2, "auto_cast_type": "fp16"}, }, "mistral": { diff --git a/integration-tests/neuron/integration/test_generate.py b/integration-tests/neuron/integration/test_generate.py index 80c83623..96f09838 100644 --- a/integration-tests/neuron/integration/test_generate.py +++ b/integration-tests/neuron/integration/test_generate.py @@ -21,7 +21,7 @@ async def test_model_single_request(tgi_service): assert response.details.generated_tokens == 17 greedy_expectations = { "gpt2": "\n\nDeep learning is a new field of research that has been around for a while", - "llama": " A Beginner’s Guide\nDeep learning is a subset of machine learning that involves the use", + "llama": " and How Does it Work?\nDeep learning is a subset of machine learning that uses artificial", "mistral": "\nWhat is Deep Learning?\nDeep Learning is a type of machine learning that", "qwen2": " - Part 1\n\nDeep Learning is a subset of Machine Learning that is based on", "granite": "\n\nDeep Learning is a subset of Machine Learning, which is a branch of Art", @@ -82,7 +82,7 @@ async def test_model_multiple_requests(tgi_service, neuron_generate_load): assert len(responses) == 4 expectations = { "gpt2": "Deep learning is a new field of research that has been around for a while", - "llama": "Deep learning is a subset of machine learning that involves the use", + "llama": "Deep learning is a subset of machine learning that uses artificial", "mistral": "Deep Learning is a type of machine learning that", "qwen2": "Deep Learning is a subset of Machine Learning that is based on", "granite": "Deep Learning is a subset of Machine Learning, which is a branch of Art",