ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

disable test_optimized_model.py temporarily due to out of memory on A730M(pr validation machine) (#9658)

Browse source 
* disable test_optimized_model.py

* disable seq2seq
This commit is contained in:
Xin Qiu 2023-12-12 17:13:52 +08:00 committed by GitHub
parent 59ce86d292
commit 0e639b920f
2 changed files with 19 additions and 19 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

36
python/llm/test/inference_gpu/test_transformers_api.py
View file

@ -50,24 +50,24 @@ def test_completion(Model, Tokenizer, model_path, prompt, answer):
assert answer in output_str assert answer in output_str
def test_transformers_auto_model_for_speech_seq2seq_int4(): #def test_transformers_auto_model_for_speech_seq2seq_int4():
from transformers import WhisperProcessor # from transformers import WhisperProcessor
from datasets import load_from_disk # from datasets import load_from_disk
model_path = os.environ.get('WHISPER_TINY_ORIGIN_PATH') # model_path = os.environ.get('WHISPER_TINY_ORIGIN_PATH')
dataset_path = os.environ.get('SPEECH_DATASET_PATH') # dataset_path = os.environ.get('SPEECH_DATASET_PATH')
processor = WhisperProcessor.from_pretrained(model_path) # processor = WhisperProcessor.from_pretrained(model_path)
ds = load_from_disk(dataset_path) # ds = load_from_disk(dataset_path)
sample = ds[0]["audio"] # sample = ds[0]["audio"]
input_features = processor(sample["array"], sampling_rate=sample["sampling_rate"], return_tensors="pt").input_features # input_features = processor(sample["array"], sampling_rate=sample["sampling_rate"], return_tensors="pt").input_features
input_features = input_features.to(device) # input_features = input_features.to(device)
model = AutoModelForSpeechSeq2Seq.from_pretrained(model_path, trust_remote_code=True, load_in_4bit=True, optimize_model=True) # model = AutoModelForSpeechSeq2Seq.from_pretrained(model_path, trust_remote_code=True, load_in_4bit=True, optimize_model=True)
model = model.to(device) # model = model.to(device)
predicted_ids = model.generate(input_features) # predicted_ids = model.generate(input_features)
# decode token ids to text # # decode token ids to text
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=False) # transcription = processor.batch_decode(predicted_ids, skip_special_tokens=False)
model.to('cpu') # model.to('cpu')
print('Output:', transcription) # print('Output:', transcription)
assert 'Mr. Quilter is the apostle of the middle classes and we are glad to welcome his gospel.' in transcription[0] # assert 'Mr. Quilter is the apostle of the middle classes and we are glad to welcome his gospel.' in transcription[0]
if __name__ == '__main__': if __name__ == '__main__':
pytest.main([__file__]) pytest.main([__file__])

2
python/llm/test/run-llm-inference-tests-gpu.sh
View file

@ -16,7 +16,7 @@ if [ -z "$THREAD_NUM" ]; then
THREAD_NUM=2 THREAD_NUM=2
fi fi
export OMP_NUM_THREADS=$THREAD_NUM export OMP_NUM_THREADS=$THREAD_NUM
pytest ${LLM_INFERENCE_TEST_DIR} -v -s pytest ${LLM_INFERENCE_TEST_DIR}/test_transformers_api.py -v -s
now=$(date "+%s") now=$(date "+%s")
time=$((now-start)) time=$((now-start))