Bobazooba's workspace
Runs
65
Name
1 visualized
State
Notes
User
Tags
Created
Runtime
Sweep
aggregation/class_path
aggregation/parameters/dropout
aggregation/parameters/inner_dim
aggregation/parameters/length_scaling
aggregation/parameters/model_dim
aggregation/parameters/normalization_type
aggregation/parameters/num_heads
aggregation/parameters/pooling_type
aggregation/parameters/scaling_square_root
backbone/class_path
backbone/parameters/activation_type
backbone/parameters/dropout
backbone/parameters/feed_forward_dim
backbone/parameters/feed_forward_normalization_type
backbone/parameters/head_dim
backbone/parameters/max_relative_position
backbone/parameters/model_dim
backbone/parameters/normalization_type
backbone/parameters/num_heads
backbone/parameters/num_layers
backbone/parameters/shared_relative_positions
backbone/parameters/use_attention_bias
backbone/parameters/use_bias_positions
backbone/parameters/use_relative_positions
datamodule/batch_size
datamodule/class_path
datamodule/num_workers
datamodule/parameters/batch_size
datamodule/parameters/file_path
datamodule/parameters/max_lines
datamodule/parameters/valid_limit
datamodule/train_chunk_size
datamodule/train_dataset_module_path
datamodule/train_file_path
datamodule/valid_chunk_size
datamodule/valid_dataset_keys
datamodule/valid_dataset_module_paths
datamodule/valid_datasets_module_paths
datamodule/valid_file_paths
embedding/class_path
embedding/parameters/dropout
embedding/parameters/embedding_dim
embedding/parameters/model_dim
embedding/parameters/n_positions
Finished
-
bobazooba
31m 58s
-
conversation.modeling.nn.models.aggregation.ResidualAttentionAggregation
0.1
256
true
768
rms
12
mean
true
conversation.modeling.nn.models.backbone.Transformer
geglu
0.1
1536
rms
None
10
768
rms
12
8
true
false
true
true
192
-
-
-
-
-
-
1000000
conversation.modeling.data.dataset.PairsDataset
../../data/pairs/sample.jsonl
-1
-
conversation.modeling.data.dataset.RelevancePairsDataset
-
../../data/relevance/all_v1.jsonl
conversation.modeling.nn.models.embedding.TransformerEmbedding
0.1
384
768
24
1-1
of 1