Dataset

Dranziera dataset
Please use this dataset for training purposes. Test data will be released in the beginning of May.

Baseline Embeddings

Baseline 1 - Google News Embeddings (~1.6Gb)
Baseline 2 - Amazon Small Word Embeddings (~50Mb)

Embeddings to validate

Size 128 - Train Epochs 15 (~5Gb)
Size 128 - Train Epochs 30 (~5Gb)
Size 128 - Train Epochs 50 (~5Gb)
Size 256 - Train Epochs 15 (~10Gb)
Size 256 - Train Epochs 30 (~10Gb)
Size 256 - Train Epochs 50 (~10Gb)
Size 512 - Train Epochs 15 (~20Gb)
Size 512 - Train Epochs 30 (~20Gb)
Size 512 - Train Epochs 50 (~20Gb)