Skip to content

Saved_models.md

Latest commit

 

History

History
117 lines (85 loc) · 3.06 KB

Saved_models.md

File metadata and controls

117 lines (85 loc) · 3.06 KB

Saved models

Transformer models

SloBERTa model (trainsmall)

The model is saved to the Wandb - to load it:

!pip install wandb
import wandb
wandb.login()

# Initialize Wandb
run = wandb.init(project="SLED-categorization", entity="tajak", name="testing-trained-model")

# Load the saved model
artifact = run.use_artifact('tajak/SLED-categorization/SLED-SloBERTa-trainsmall-classifier:v0', type='model')
artifact_dir = artifact.download()

# Loading a local save
model = ClassificationModel(
    "camembert", artifact_dir)

XLM-RoBERTa model (trainsmall)

The model is saved to the Wandb - to load it:

!pip install wandb
import wandb
wandb.login()

# Initialize Wandb
run = wandb.init(project="SLED-categorization", entity="tajak", name="testing-trained-model")

# Load the saved model
artifact = run.use_artifact('tajak/SLED-categorization/SLED-XLM-RoBERTa-trainsmall-classifier:v0', type='model')
artifact_dir = artifact.download()

# Loading a local save
model = ClassificationModel(
    "xlmroberta", artifact_dir)

FastText models

Trainsmall, no embeddings

The model is saved to the Wandb - to load it:

!pip install wandb
import wandb
wandb.login()

# Initialize Wandb
run = wandb.init(project="SLED-categorization", entity="tajak", name="testing-trained-model")

# Load the saved model
artifact = run.use_artifact('tajak/SLED-categorization/SLED-categorization-trainsmall-noembeddings-model:v0', type='model')
artifact_dir = artifact.download()

model = fasttext.load_model(f"{artifact_dir}/FastText-model-trainsmall-noembeddings.bin")

Trainsmall, embeddings

The model is saved to the Wandb - to load it:

!pip install wandb
import wandb
wandb.login()

# Initialize Wandb
run = wandb.init(project="SLED-categorization", entity="tajak", name="testing-trained-model")

# Load the saved model
artifact = run.use_artifact('tajak/SLED-categorization/SLED-categorization-trainsmall-embeddings-model:v0', type='model')
artifact_dir = artifact.download()

model = fasttext.load_model(f"{artifact_dir}/FastText-model-trainsmall-embeddings.bin")

Trainlarge, no embeddings

The model is saved to the Wandb - to load it:

!pip install wandb
import wandb
wandb.login()

# Initialize Wandb
run = wandb.init(project="SLED-categorization", entity="tajak", name="testing-trained-model")

# Load the saved model
artifact = run.use_artifact('tajak/SLED-categorization/SLED-categorization-trainlarge-noembeddings-model:v0', type='model')
artifact_dir = artifact.download()

model = fasttext.load_model(f"{artifact_dir}/FastText-model-trainlarge-noembeddings.bin")

Trainlarge, embeddings

The model is saved to the Wandb - to load it:

!pip install wandb
import wandb
wandb.login()

# Initialize Wandb
run = wandb.init(project="SLED-categorization", entity="tajak", name="testing-trained-model")

# Load the saved model
artifact = run.use_artifact('tajak/SLED-categorization/SLED-categorization-trainlarge-embeddings-model:v0', type='model')
artifact_dir = artifact.download()

model = fasttext.load_model(f"{artifact_dir}/FastText-model-trainlarge-embeddings.bin")