huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-04-21 14:52:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

test(neuron): use smaller llama model

Browse Source
This commit is contained in:
David Corvoysier 2025-02-19 14:51:43 +00:00
parent 6f92198eb9
commit b5e98a6d5a
2 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
integration-tests/fixtures/neuron/model.py
View File

@ -32,7 +32,7 @@ MODEL_CONFIGURATIONS = {
"export_kwargs": {"batch_size": 4, "sequence_length": 1024, "num_cores": 2, "auto_cast_type": "fp16"}, "export_kwargs": {"batch_size": 4, "sequence_length": 1024, "num_cores": 2, "auto_cast_type": "fp16"},
}, },
"llama": { "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"}, "export_kwargs": {"batch_size": 4, "sequence_length": 2048, "num_cores": 2, "auto_cast_type": "fp16"},
}, },
"mistral": { "mistral": {

4
integration-tests/neuron/integration/test_generate.py
View File

@ -21,7 +21,7 @@ async def test_model_single_request(tgi_service):
assert response.details.generated_tokens == 17 assert response.details.generated_tokens == 17
greedy_expectations = { greedy_expectations = {
"gpt2": "\n\nDeep learning is a new field of research that has been around for a while", "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", "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", "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", "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 assert len(responses) == 4
expectations = { expectations = {
"gpt2": "Deep learning is a new field of research that has been around for a while", "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", "mistral": "Deep Learning is a type of machine learning that",
"qwen2": "Deep Learning is a subset of Machine Learning that is based on", "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", "granite": "Deep Learning is a subset of Machine Learning, which is a branch of Art",