oborchers

# Adapted
for i in xrange(len(sentences)):
  if len(sentences[i]) > 0:
    # Define a memory view of the sentence indices
    sentence_view = sentences[i]
    sentence_len = len(sentences[i])
    # Pass all arguments to the C-Loop
    sif_embeddings_cloop(size, sentence_view, sentence_len, i, vectors, sv)
return output

oborchers

def sif_embeddings(sentences, model):
    cdef int size = model.vector_size
    # Unsafe access via pointers
    cdef REAL_t *vectors = <REAL_t *>(np.PyArray_DATA(model.wv.sif_vectors))
    
    output = np.zeros((len(sentences), size), dtype=np.float32)
    cdef REAL_t *sv = <REAL_t *>(np.PyArray_DATA(output))
    
    cdef INT_t *sentence_view
    

oborchers

exclusion_matrix = random_sparse(
    m = replications,           # Number of subsampled datasets
    n = len(X),                 # Length of training data
    density = p_excluded/100.,  # Percentage of excluded observations
    format = "csr",
    dtype = bool,               # True = Excluded
    random_state=random_state,  
)

oborchers

exog_perf = [] # Exogenous performance

# Iterate rows of exclusion matrix and estimate model on reduced data
for row in trange(excl_mat.shape[0]):
    # Get exclusion mask and invert to inclusion mask (True = included)
    # Inversion is necessary for numpy indexing with boolean arrays
    mask = invert(squeeze(excl_mat[row].toarray()))

    # Fit arbtitrary estimator
    est = LinearRegression(n_jobs=-1)

oborchers

topn = 50
# Using -coef_ because we sort size
top_idx = argsort(-est.coef_)[-topn:]
print(
    len(
        set(top_idx).intersection(set(data["noise_index"]))
    )
)

oborchers

@misc{Borchers2020,
abstract = {This article develops the LASSO The Traitors (LTT) method. LTT filters out noisy observations from a dataset based on an exogenous performance metric. LTT significantly improves the performance of estimators based on the cleaned dataset. LTT is fast, easily applicable, and task agnostic.},
author = {Borchers, Oliver and Ringel, Daniel M.},
booktitle = {Towards Data Science},
title = {{Your Labels and Data are Noisy? LASSO The Traitors!}},
url = {https://medium.com/@oliverbor/lasso-the-traitors-dd33ea5942bc},
year = {2020}
}

oborchers

TY  - ICOMM
T1  - Your Labels and Data are Noisy? LASSO The Traitors!
A1  - Borchers, Oliver
A1  - Ringel, Daniel M.
Y1  - 2020///
JF  - Towards Data Science
UR  - https://medium.com/@oliverbor/lasso-the-traitors-dd33ea5942bc
N2  - This article develops the LASSO The Traitors (LTT) method. LTT filters out noisy observations from a dataset based on an exogenous performance metric. LTT significantly improves the performance of estimators based on the cleaned dataset. LTT is fast, easily applicable, and task agnostic.
ER  - 

oborchers

apt-get update

# Install ONNX ML
export ONNX_ML=1
pip install -U onnx

# Clone ONNX Runtime and build
git clone https://github.com/microsoft/onnxruntime.git --branch v1.7.1 --single-branch
/bin/sh onnxruntime/dockerfiles/scripts/install_common_deps.sh
cd onnxruntime/

oborchers

# Script to export a transformers model
model_name = "sentence-transformers/bert-base-nli-stsb-mean-tokens"
pipeline_name = "feature-extraction"
model_pth = Path(f"encoder/{model_name}.onnx")

nlp = transformers.pipeline(pipeline_name, model=model_name, tokenizer=model_name, device=0)
tokenizer = nlp.tokenizer

if model_pth.exists():
    model_pth.unlink()

oborchers

# We start by working with CUDA only
ONNX_PROVIDERS = ["CUDAExecutionProvider", "CPUExecutionProvider"]
opt = rt.SessionOptions()
sess = rt.InferenceSession(str(model_pth), opt, providers=ONNX_PROVIDERS)

model_input = tokenizer.encode_plus(span)
model_input = {name : np.atleast_2d(value) for name, value in model_input.items()}
onnx_result = sess.run(None, model_input)

print(onnx_result[0].shape)