Incorporações de codificadores para Object2Vec

Otimização de GPU: incorporações de codificador

Uma incorporação é um mapeamento de objetos discretos, como palavras, para vetores de números reais.

Devido à falta de memória de GPU, a variável de ambiente INFERENCE_PREFERRED_MODE pode ser especificada para otimização se Formatos de dados para inferência em Object2Vec ou a rede de inferência de incorporação de codificador for carregada na GPU. Se a maior parte da inferência for para incorporações de codificador, especifique INFERENCE_PREFERRED_MODE=embedding. Veja a seguir um exemplo de transformação em lotes usando 4 instâncias de p3.2xlarge que otimiza para inferência de incorporação de codificador:

transformer = o2v.transformer(instance_count=4,
                              instance_type="ml.p2.xlarge",
                              max_concurrent_transforms=2,
                              max_payload=1,  # 1MB
                              strategy='MultiRecord',
                              env={'INFERENCE_PREFERRED_MODE': 'embedding'},  # only useful with GPU
                              output_path=output_s3_path)

Entrada: incorporações de codificador

Content-type: application/json; infer_max_seqlens=<FWD-LENGTH>,<BCK-LENGTH>

Em que <FWD-LENGTH> e <BCK-LENGTH> são inteiros no intervalo [1,5000] e definem os comprimentos máximos de sequência para o codificador para a frente e para trás.

{
  "instances" : [
    {"in0": [6, 17, 606, 19, 53, 67, 52, 12, 5, 10, 15, 10178, 7, 33, 652, 80, 15, 69, 821, 4]},
    {"in0": [22, 1016, 32, 13, 25, 11, 5, 64, 573, 45, 5, 80, 15, 67, 21, 7, 9, 107, 4]},
    {"in0": [774, 14, 21, 206]}
  ]
}

Content-type: application/jsonlines; infer_max_seqlens=<FWD-LENGTH>,<BCK-LENGTH>

Em que <FWD-LENGTH> e <BCK-LENGTH> são inteiros no intervalo [1,5000] e definem os comprimentos máximos de sequência para o codificador para a frente e para trás.

{"in0": [6, 17, 606, 19, 53, 67, 52, 12, 5, 10, 15, 10178, 7, 33, 652, 80, 15, 69, 821, 4]}
{"in0": [22, 1016, 32, 13, 25, 11, 5, 64, 573, 45, 5, 80, 15, 67, 21, 7, 9, 107, 4]}
{"in0": [774, 14, 21, 206]}

Em ambos os formatos, você especifica apenas um tipo de entrada, ou “in0” ou “in1.”. O serviço de inferência chama o codificador correspondente e gera as incorporações para cada uma das instâncias.

Saída: incorporações de codificador

Content-type: application/json

{
  "predictions": [
    {"embeddings":[0.057368703186511,0.030703511089086,0.099890425801277,0.063688032329082,0.026327300816774,0.003637571120634,0.021305780857801,0.004316598642617,0.0,0.003397724591195,0.0,0.000378780066967,0.0,0.0,0.0,0.007419463712722]},
    {"embeddings":[0.150190666317939,0.05145975202322,0.098204270005226,0.064249359071254,0.056249320507049,0.01513972133398,0.047553978860378,0.0,0.0,0.011533712036907,0.011472506448626,0.010696629062294,0.0,0.0,0.0,0.008508535102009]}
  ]
}

Content-type: application/jsonlines

{"embeddings":[0.057368703186511,0.030703511089086,0.099890425801277,0.063688032329082,0.026327300816774,0.003637571120634,0.021305780857801,0.004316598642617,0.0,0.003397724591195,0.0,0.000378780066967,0.0,0.0,0.0,0.007419463712722]}
{"embeddings":[0.150190666317939,0.05145975202322,0.098204270005226,0.064249359071254,0.056249320507049,0.01513972133398,0.047553978860378,0.0,0.0,0.011533712036907,0.011472506448626,0.010696629062294,0.0,0.0,0.0,0.008508535102009]}

O comprimento de vetor das incorporações geradas pelo serviço de inferência é igual ao valor de um dos hiperparâmetros a seguir, que você especifica na ocasião do treinamento: enc0_token_embedding_dim, enc1_token_embedding_dim ou enc_dim.