poprox_recommender.model

Functions

load_checkpoint([device])
load_model(checkpoint, device)

Classes

ModelConfig([num_epochs, ...])
RecommenderComponents(tokenizer, model, device)

class poprox_recommender.model.ModelConfig(num_epochs: 'float' = 10, num_clicked_news_a_user: 'float' = 50, word_freq_threshold: 'float' = 1, dropout_probability: 'float' = 0.2, word_embedding_dim: 'float' = 300, category_embedding_dim: 'float' = 100, query_vector_dim: 'float' = 200, additive_attn_hidden_dim: 'float' = 200, num_attention_heads: 'float' = 16, hidden_size: 'int' = 768)

Bases: object

Parameters:

• num_epochs (float)

• num_clicked_news_a_user (float)

• word_freq_threshold (float)

• dropout_probability (float)

• word_embedding_dim (float)

• category_embedding_dim (float)

• query_vector_dim (float)

• additive_attn_hidden_dim (float)

• num_attention_heads (float)

• hidden_size (int)

class poprox_recommender.model.RecommenderComponents(tokenizer: 'AutoTokenizer', model: 'NRMS', device: 'str | None')

Bases: object

Parameters:

• tokenizer (AutoTokenizer)

• model (NRMS)

• device (str | None)

Modules

general
nrms