diff --git a/exp12/000-module.3.input_layernorm/events.out.tfevents.1638219661.r6i6n8.58256.4 b/exp12-1/000-module.3.input_layernorm/events.out.tfevents.1638219661.r6i6n8.58256.4 similarity index 100% rename from exp12/000-module.3.input_layernorm/events.out.tfevents.1638219661.r6i6n8.58256.4 rename to exp12-1/000-module.3.input_layernorm/events.out.tfevents.1638219661.r6i6n8.58256.4 diff --git a/exp12/000-module.3.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.12 b/exp12-1/000-module.3.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.12 similarity index 100% rename from exp12/000-module.3.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.12 rename to exp12-1/000-module.3.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.12 diff --git a/exp12/000-module.3.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.11 b/exp12-1/000-module.3.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.11 similarity index 100% rename from exp12/000-module.3.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.11 rename to exp12-1/000-module.3.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.11 diff --git a/exp12/000-module.3.mlp/events.out.tfevents.1638219662.r6i6n8.58256.13 b/exp12-1/000-module.3.mlp/events.out.tfevents.1638219662.r6i6n8.58256.13 similarity index 100% rename from exp12/000-module.3.mlp/events.out.tfevents.1638219662.r6i6n8.58256.13 rename to exp12-1/000-module.3.mlp/events.out.tfevents.1638219662.r6i6n8.58256.13 diff --git a/exp12/000-module.3.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.10 b/exp12-1/000-module.3.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.10 similarity index 100% rename from exp12/000-module.3.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.10 rename to exp12-1/000-module.3.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.10 diff --git a/exp12/000-module.3.self_attention.attention_dropout/events.out.tfevents.1638219661.r6i6n8.58256.7 b/exp12-1/000-module.3.self_attention.attention_dropout/events.out.tfevents.1638219661.r6i6n8.58256.7 similarity index 100% rename from exp12/000-module.3.self_attention.attention_dropout/events.out.tfevents.1638219661.r6i6n8.58256.7 rename to exp12-1/000-module.3.self_attention.attention_dropout/events.out.tfevents.1638219661.r6i6n8.58256.7 diff --git a/exp12/000-module.3.self_attention.dense/events.out.tfevents.1638219661.r6i6n8.58256.8 b/exp12-1/000-module.3.self_attention.dense/events.out.tfevents.1638219661.r6i6n8.58256.8 similarity index 100% rename from exp12/000-module.3.self_attention.dense/events.out.tfevents.1638219661.r6i6n8.58256.8 rename to exp12-1/000-module.3.self_attention.dense/events.out.tfevents.1638219661.r6i6n8.58256.8 diff --git a/exp12/000-module.3.self_attention.query_key_value/events.out.tfevents.1638219661.r6i6n8.58256.5 b/exp12-1/000-module.3.self_attention.query_key_value/events.out.tfevents.1638219661.r6i6n8.58256.5 similarity index 100% rename from exp12/000-module.3.self_attention.query_key_value/events.out.tfevents.1638219661.r6i6n8.58256.5 rename to exp12-1/000-module.3.self_attention.query_key_value/events.out.tfevents.1638219661.r6i6n8.58256.5 diff --git a/exp12/000-module.3.self_attention.scale_mask_softmax/events.out.tfevents.1638219661.r6i6n8.58256.6 b/exp12-1/000-module.3.self_attention.scale_mask_softmax/events.out.tfevents.1638219661.r6i6n8.58256.6 similarity index 100% rename from exp12/000-module.3.self_attention.scale_mask_softmax/events.out.tfevents.1638219661.r6i6n8.58256.6 rename to exp12-1/000-module.3.self_attention.scale_mask_softmax/events.out.tfevents.1638219661.r6i6n8.58256.6 diff --git a/exp12/000-module.3.self_attention/events.out.tfevents.1638219661.r6i6n8.58256.9 b/exp12-1/000-module.3.self_attention/events.out.tfevents.1638219661.r6i6n8.58256.9 similarity index 100% rename from exp12/000-module.3.self_attention/events.out.tfevents.1638219661.r6i6n8.58256.9 rename to exp12-1/000-module.3.self_attention/events.out.tfevents.1638219661.r6i6n8.58256.9 diff --git a/exp12/000-module.3/events.out.tfevents.1638219662.r6i6n8.58256.14 b/exp12-1/000-module.3/events.out.tfevents.1638219662.r6i6n8.58256.14 similarity index 100% rename from exp12/000-module.3/events.out.tfevents.1638219662.r6i6n8.58256.14 rename to exp12-1/000-module.3/events.out.tfevents.1638219662.r6i6n8.58256.14 diff --git a/exp12/000-module.4.input_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.15 b/exp12-1/000-module.4.input_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.15 similarity index 100% rename from exp12/000-module.4.input_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.15 rename to exp12-1/000-module.4.input_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.15 diff --git a/exp12/000-module.4.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.23 b/exp12-1/000-module.4.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.23 similarity index 100% rename from exp12/000-module.4.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.23 rename to exp12-1/000-module.4.mlp.dense_4h_to_h/events.out.tfevents.1638219662.r6i6n8.58256.23 diff --git a/exp12/000-module.4.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.22 b/exp12-1/000-module.4.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.22 similarity index 100% rename from exp12/000-module.4.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.22 rename to exp12-1/000-module.4.mlp.dense_h_to_4h/events.out.tfevents.1638219662.r6i6n8.58256.22 diff --git a/exp12/000-module.4.mlp/events.out.tfevents.1638219662.r6i6n8.58256.24 b/exp12-1/000-module.4.mlp/events.out.tfevents.1638219662.r6i6n8.58256.24 similarity index 100% rename from exp12/000-module.4.mlp/events.out.tfevents.1638219662.r6i6n8.58256.24 rename to exp12-1/000-module.4.mlp/events.out.tfevents.1638219662.r6i6n8.58256.24 diff --git a/exp12/000-module.4.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.21 b/exp12-1/000-module.4.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.21 similarity index 100% rename from exp12/000-module.4.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.21 rename to exp12-1/000-module.4.post_attention_layernorm/events.out.tfevents.1638219662.r6i6n8.58256.21 diff --git a/exp12/000-module.4.self_attention.attention_dropout/events.out.tfevents.1638219662.r6i6n8.58256.18 b/exp12-1/000-module.4.self_attention.attention_dropout/events.out.tfevents.1638219662.r6i6n8.58256.18 similarity index 100% rename from exp12/000-module.4.self_attention.attention_dropout/events.out.tfevents.1638219662.r6i6n8.58256.18 rename to exp12-1/000-module.4.self_attention.attention_dropout/events.out.tfevents.1638219662.r6i6n8.58256.18 diff --git a/exp12/000-module.4.self_attention.dense/events.out.tfevents.1638219662.r6i6n8.58256.19 b/exp12-1/000-module.4.self_attention.dense/events.out.tfevents.1638219662.r6i6n8.58256.19 similarity index 100% rename from exp12/000-module.4.self_attention.dense/events.out.tfevents.1638219662.r6i6n8.58256.19 rename to exp12-1/000-module.4.self_attention.dense/events.out.tfevents.1638219662.r6i6n8.58256.19 diff --git a/exp12/000-module.4.self_attention.query_key_value/events.out.tfevents.1638219662.r6i6n8.58256.16 b/exp12-1/000-module.4.self_attention.query_key_value/events.out.tfevents.1638219662.r6i6n8.58256.16 similarity index 100% rename from exp12/000-module.4.self_attention.query_key_value/events.out.tfevents.1638219662.r6i6n8.58256.16 rename to exp12-1/000-module.4.self_attention.query_key_value/events.out.tfevents.1638219662.r6i6n8.58256.16 diff --git a/exp12/000-module.4.self_attention.scale_mask_softmax/events.out.tfevents.1638219662.r6i6n8.58256.17 b/exp12-1/000-module.4.self_attention.scale_mask_softmax/events.out.tfevents.1638219662.r6i6n8.58256.17 similarity index 100% rename from exp12/000-module.4.self_attention.scale_mask_softmax/events.out.tfevents.1638219662.r6i6n8.58256.17 rename to exp12-1/000-module.4.self_attention.scale_mask_softmax/events.out.tfevents.1638219662.r6i6n8.58256.17 diff --git a/exp12/000-module.4.self_attention/events.out.tfevents.1638219662.r6i6n8.58256.20 b/exp12-1/000-module.4.self_attention/events.out.tfevents.1638219662.r6i6n8.58256.20 similarity index 100% rename from exp12/000-module.4.self_attention/events.out.tfevents.1638219662.r6i6n8.58256.20 rename to exp12-1/000-module.4.self_attention/events.out.tfevents.1638219662.r6i6n8.58256.20 diff --git a/exp12/000-module.4/events.out.tfevents.1638219662.r6i6n8.58256.25 b/exp12-1/000-module.4/events.out.tfevents.1638219662.r6i6n8.58256.25 similarity index 100% rename from exp12/000-module.4/events.out.tfevents.1638219662.r6i6n8.58256.25 rename to exp12-1/000-module.4/events.out.tfevents.1638219662.r6i6n8.58256.25 diff --git a/exp12/000-module.tied_modules.embed.embedding_dropout/events.out.tfevents.1638219661.r6i6n8.58256.2 b/exp12-1/000-module.tied_modules.embed.embedding_dropout/events.out.tfevents.1638219661.r6i6n8.58256.2 similarity index 100% rename from exp12/000-module.tied_modules.embed.embedding_dropout/events.out.tfevents.1638219661.r6i6n8.58256.2 rename to exp12-1/000-module.tied_modules.embed.embedding_dropout/events.out.tfevents.1638219661.r6i6n8.58256.2 diff --git a/exp12/000-module.tied_modules.embed.position_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.1 b/exp12-1/000-module.tied_modules.embed.position_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.1 similarity index 100% rename from exp12/000-module.tied_modules.embed.position_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.1 rename to exp12-1/000-module.tied_modules.embed.position_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.1 diff --git a/exp12/000-module.tied_modules.embed.word_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.0 b/exp12-1/000-module.tied_modules.embed.word_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.0 similarity index 100% rename from exp12/000-module.tied_modules.embed.word_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.0 rename to exp12-1/000-module.tied_modules.embed.word_embeddings/events.out.tfevents.1638219661.r6i6n8.58256.0 diff --git a/exp12/000-module.tied_modules.embed/events.out.tfevents.1638219661.r6i6n8.58256.3 b/exp12-1/000-module.tied_modules.embed/events.out.tfevents.1638219661.r6i6n8.58256.3 similarity index 100% rename from exp12/000-module.tied_modules.embed/events.out.tfevents.1638219661.r6i6n8.58256.3 rename to exp12-1/000-module.tied_modules.embed/events.out.tfevents.1638219661.r6i6n8.58256.3 diff --git a/exp12/000-module/events.out.tfevents.1638219662.r6i6n8.58256.26 b/exp12-1/000-module/events.out.tfevents.1638219662.r6i6n8.58256.26 similarity index 100% rename from exp12/000-module/events.out.tfevents.1638219662.r6i6n8.58256.26 rename to exp12-1/000-module/events.out.tfevents.1638219662.r6i6n8.58256.26 diff --git a/exp12/028-module.17.input_layernorm/events.out.tfevents.1638219673.r6i7n6.58182.0 b/exp12-1/028-module.17.input_layernorm/events.out.tfevents.1638219673.r6i7n6.58182.0 similarity index 100% rename from exp12/028-module.17.input_layernorm/events.out.tfevents.1638219673.r6i7n6.58182.0 rename to exp12-1/028-module.17.input_layernorm/events.out.tfevents.1638219673.r6i7n6.58182.0 diff --git a/exp12/028-module.17.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.8 b/exp12-1/028-module.17.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.8 similarity index 100% rename from exp12/028-module.17.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.8 rename to exp12-1/028-module.17.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.8 diff --git a/exp12/028-module.17.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.7 b/exp12-1/028-module.17.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.7 similarity index 100% rename from exp12/028-module.17.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.7 rename to exp12-1/028-module.17.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.7 diff --git a/exp12/028-module.17.mlp/events.out.tfevents.1638219674.r6i7n6.58182.9 b/exp12-1/028-module.17.mlp/events.out.tfevents.1638219674.r6i7n6.58182.9 similarity index 100% rename from exp12/028-module.17.mlp/events.out.tfevents.1638219674.r6i7n6.58182.9 rename to exp12-1/028-module.17.mlp/events.out.tfevents.1638219674.r6i7n6.58182.9 diff --git a/exp12/028-module.17.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.6 b/exp12-1/028-module.17.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.6 similarity index 100% rename from exp12/028-module.17.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.6 rename to exp12-1/028-module.17.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.6 diff --git a/exp12/028-module.17.self_attention.attention_dropout/events.out.tfevents.1638219673.r6i7n6.58182.3 b/exp12-1/028-module.17.self_attention.attention_dropout/events.out.tfevents.1638219673.r6i7n6.58182.3 similarity index 100% rename from exp12/028-module.17.self_attention.attention_dropout/events.out.tfevents.1638219673.r6i7n6.58182.3 rename to exp12-1/028-module.17.self_attention.attention_dropout/events.out.tfevents.1638219673.r6i7n6.58182.3 diff --git a/exp12/028-module.17.self_attention.dense/events.out.tfevents.1638219673.r6i7n6.58182.4 b/exp12-1/028-module.17.self_attention.dense/events.out.tfevents.1638219673.r6i7n6.58182.4 similarity index 100% rename from exp12/028-module.17.self_attention.dense/events.out.tfevents.1638219673.r6i7n6.58182.4 rename to exp12-1/028-module.17.self_attention.dense/events.out.tfevents.1638219673.r6i7n6.58182.4 diff --git a/exp12/028-module.17.self_attention.query_key_value/events.out.tfevents.1638219673.r6i7n6.58182.1 b/exp12-1/028-module.17.self_attention.query_key_value/events.out.tfevents.1638219673.r6i7n6.58182.1 similarity index 100% rename from exp12/028-module.17.self_attention.query_key_value/events.out.tfevents.1638219673.r6i7n6.58182.1 rename to exp12-1/028-module.17.self_attention.query_key_value/events.out.tfevents.1638219673.r6i7n6.58182.1 diff --git a/exp12/028-module.17.self_attention.scale_mask_softmax/events.out.tfevents.1638219673.r6i7n6.58182.2 b/exp12-1/028-module.17.self_attention.scale_mask_softmax/events.out.tfevents.1638219673.r6i7n6.58182.2 similarity index 100% rename from exp12/028-module.17.self_attention.scale_mask_softmax/events.out.tfevents.1638219673.r6i7n6.58182.2 rename to exp12-1/028-module.17.self_attention.scale_mask_softmax/events.out.tfevents.1638219673.r6i7n6.58182.2 diff --git a/exp12/028-module.17.self_attention/events.out.tfevents.1638219673.r6i7n6.58182.5 b/exp12-1/028-module.17.self_attention/events.out.tfevents.1638219673.r6i7n6.58182.5 similarity index 100% rename from exp12/028-module.17.self_attention/events.out.tfevents.1638219673.r6i7n6.58182.5 rename to exp12-1/028-module.17.self_attention/events.out.tfevents.1638219673.r6i7n6.58182.5 diff --git a/exp12/028-module.17/events.out.tfevents.1638219674.r6i7n6.58182.10 b/exp12-1/028-module.17/events.out.tfevents.1638219674.r6i7n6.58182.10 similarity index 100% rename from exp12/028-module.17/events.out.tfevents.1638219674.r6i7n6.58182.10 rename to exp12-1/028-module.17/events.out.tfevents.1638219674.r6i7n6.58182.10 diff --git a/exp12/028-module.18.input_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.11 b/exp12-1/028-module.18.input_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.11 similarity index 100% rename from exp12/028-module.18.input_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.11 rename to exp12-1/028-module.18.input_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.11 diff --git a/exp12/028-module.18.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.19 b/exp12-1/028-module.18.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.19 similarity index 100% rename from exp12/028-module.18.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.19 rename to exp12-1/028-module.18.mlp.dense_4h_to_h/events.out.tfevents.1638219674.r6i7n6.58182.19 diff --git a/exp12/028-module.18.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.18 b/exp12-1/028-module.18.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.18 similarity index 100% rename from exp12/028-module.18.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.18 rename to exp12-1/028-module.18.mlp.dense_h_to_4h/events.out.tfevents.1638219674.r6i7n6.58182.18 diff --git a/exp12/028-module.18.mlp/events.out.tfevents.1638219674.r6i7n6.58182.20 b/exp12-1/028-module.18.mlp/events.out.tfevents.1638219674.r6i7n6.58182.20 similarity index 100% rename from exp12/028-module.18.mlp/events.out.tfevents.1638219674.r6i7n6.58182.20 rename to exp12-1/028-module.18.mlp/events.out.tfevents.1638219674.r6i7n6.58182.20 diff --git a/exp12/028-module.18.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.17 b/exp12-1/028-module.18.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.17 similarity index 100% rename from exp12/028-module.18.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.17 rename to exp12-1/028-module.18.post_attention_layernorm/events.out.tfevents.1638219674.r6i7n6.58182.17 diff --git a/exp12/028-module.18.self_attention.attention_dropout/events.out.tfevents.1638219674.r6i7n6.58182.14 b/exp12-1/028-module.18.self_attention.attention_dropout/events.out.tfevents.1638219674.r6i7n6.58182.14 similarity index 100% rename from exp12/028-module.18.self_attention.attention_dropout/events.out.tfevents.1638219674.r6i7n6.58182.14 rename to exp12-1/028-module.18.self_attention.attention_dropout/events.out.tfevents.1638219674.r6i7n6.58182.14 diff --git a/exp12/028-module.18.self_attention.dense/events.out.tfevents.1638219674.r6i7n6.58182.15 b/exp12-1/028-module.18.self_attention.dense/events.out.tfevents.1638219674.r6i7n6.58182.15 similarity index 100% rename from exp12/028-module.18.self_attention.dense/events.out.tfevents.1638219674.r6i7n6.58182.15 rename to exp12-1/028-module.18.self_attention.dense/events.out.tfevents.1638219674.r6i7n6.58182.15 diff --git a/exp12/028-module.18.self_attention.query_key_value/events.out.tfevents.1638219674.r6i7n6.58182.12 b/exp12-1/028-module.18.self_attention.query_key_value/events.out.tfevents.1638219674.r6i7n6.58182.12 similarity index 100% rename from exp12/028-module.18.self_attention.query_key_value/events.out.tfevents.1638219674.r6i7n6.58182.12 rename to exp12-1/028-module.18.self_attention.query_key_value/events.out.tfevents.1638219674.r6i7n6.58182.12 diff --git a/exp12/028-module.18.self_attention.scale_mask_softmax/events.out.tfevents.1638219674.r6i7n6.58182.13 b/exp12-1/028-module.18.self_attention.scale_mask_softmax/events.out.tfevents.1638219674.r6i7n6.58182.13 similarity index 100% rename from exp12/028-module.18.self_attention.scale_mask_softmax/events.out.tfevents.1638219674.r6i7n6.58182.13 rename to exp12-1/028-module.18.self_attention.scale_mask_softmax/events.out.tfevents.1638219674.r6i7n6.58182.13 diff --git a/exp12/028-module.18.self_attention/events.out.tfevents.1638219674.r6i7n6.58182.16 b/exp12-1/028-module.18.self_attention/events.out.tfevents.1638219674.r6i7n6.58182.16 similarity index 100% rename from exp12/028-module.18.self_attention/events.out.tfevents.1638219674.r6i7n6.58182.16 rename to exp12-1/028-module.18.self_attention/events.out.tfevents.1638219674.r6i7n6.58182.16 diff --git a/exp12/028-module.18/events.out.tfevents.1638219674.r6i7n6.58182.21 b/exp12-1/028-module.18/events.out.tfevents.1638219674.r6i7n6.58182.21 similarity index 100% rename from exp12/028-module.18/events.out.tfevents.1638219674.r6i7n6.58182.21 rename to exp12-1/028-module.18/events.out.tfevents.1638219674.r6i7n6.58182.21 diff --git a/exp12/028-module/events.out.tfevents.1638219674.r6i7n6.58182.22 b/exp12-1/028-module/events.out.tfevents.1638219674.r6i7n6.58182.22 similarity index 100% rename from exp12/028-module/events.out.tfevents.1638219674.r6i7n6.58182.22 rename to exp12-1/028-module/events.out.tfevents.1638219674.r6i7n6.58182.22 diff --git a/exp12/060-module.33.input_layernorm/events.out.tfevents.1638219686.r8i4n2.3275712.0 b/exp12-1/060-module.33.input_layernorm/events.out.tfevents.1638219686.r8i4n2.3275712.0 similarity index 100% rename from exp12/060-module.33.input_layernorm/events.out.tfevents.1638219686.r8i4n2.3275712.0 rename to exp12-1/060-module.33.input_layernorm/events.out.tfevents.1638219686.r8i4n2.3275712.0 diff --git a/exp12/060-module.33.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.8 b/exp12-1/060-module.33.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.8 similarity index 100% rename from exp12/060-module.33.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.8 rename to exp12-1/060-module.33.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.8 diff --git a/exp12/060-module.33.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.7 b/exp12-1/060-module.33.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.7 similarity index 100% rename from exp12/060-module.33.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.7 rename to exp12-1/060-module.33.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.7 diff --git a/exp12/060-module.33.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.9 b/exp12-1/060-module.33.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.9 similarity index 100% rename from exp12/060-module.33.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.9 rename to exp12-1/060-module.33.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.9 diff --git a/exp12/060-module.33.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.6 b/exp12-1/060-module.33.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.6 similarity index 100% rename from exp12/060-module.33.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.6 rename to exp12-1/060-module.33.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.6 diff --git a/exp12/060-module.33.self_attention.attention_dropout/events.out.tfevents.1638219686.r8i4n2.3275712.3 b/exp12-1/060-module.33.self_attention.attention_dropout/events.out.tfevents.1638219686.r8i4n2.3275712.3 similarity index 100% rename from exp12/060-module.33.self_attention.attention_dropout/events.out.tfevents.1638219686.r8i4n2.3275712.3 rename to exp12-1/060-module.33.self_attention.attention_dropout/events.out.tfevents.1638219686.r8i4n2.3275712.3 diff --git a/exp12/060-module.33.self_attention.dense/events.out.tfevents.1638219686.r8i4n2.3275712.4 b/exp12-1/060-module.33.self_attention.dense/events.out.tfevents.1638219686.r8i4n2.3275712.4 similarity index 100% rename from exp12/060-module.33.self_attention.dense/events.out.tfevents.1638219686.r8i4n2.3275712.4 rename to exp12-1/060-module.33.self_attention.dense/events.out.tfevents.1638219686.r8i4n2.3275712.4 diff --git a/exp12/060-module.33.self_attention.query_key_value/events.out.tfevents.1638219686.r8i4n2.3275712.1 b/exp12-1/060-module.33.self_attention.query_key_value/events.out.tfevents.1638219686.r8i4n2.3275712.1 similarity index 100% rename from exp12/060-module.33.self_attention.query_key_value/events.out.tfevents.1638219686.r8i4n2.3275712.1 rename to exp12-1/060-module.33.self_attention.query_key_value/events.out.tfevents.1638219686.r8i4n2.3275712.1 diff --git a/exp12/060-module.33.self_attention.scale_mask_softmax/events.out.tfevents.1638219686.r8i4n2.3275712.2 b/exp12-1/060-module.33.self_attention.scale_mask_softmax/events.out.tfevents.1638219686.r8i4n2.3275712.2 similarity index 100% rename from exp12/060-module.33.self_attention.scale_mask_softmax/events.out.tfevents.1638219686.r8i4n2.3275712.2 rename to exp12-1/060-module.33.self_attention.scale_mask_softmax/events.out.tfevents.1638219686.r8i4n2.3275712.2 diff --git a/exp12/060-module.33.self_attention/events.out.tfevents.1638219686.r8i4n2.3275712.5 b/exp12-1/060-module.33.self_attention/events.out.tfevents.1638219686.r8i4n2.3275712.5 similarity index 100% rename from exp12/060-module.33.self_attention/events.out.tfevents.1638219686.r8i4n2.3275712.5 rename to exp12-1/060-module.33.self_attention/events.out.tfevents.1638219686.r8i4n2.3275712.5 diff --git a/exp12/060-module.33/events.out.tfevents.1638219687.r8i4n2.3275712.10 b/exp12-1/060-module.33/events.out.tfevents.1638219687.r8i4n2.3275712.10 similarity index 100% rename from exp12/060-module.33/events.out.tfevents.1638219687.r8i4n2.3275712.10 rename to exp12-1/060-module.33/events.out.tfevents.1638219687.r8i4n2.3275712.10 diff --git a/exp12/060-module.34.input_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.11 b/exp12-1/060-module.34.input_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.11 similarity index 100% rename from exp12/060-module.34.input_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.11 rename to exp12-1/060-module.34.input_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.11 diff --git a/exp12/060-module.34.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.19 b/exp12-1/060-module.34.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.19 similarity index 100% rename from exp12/060-module.34.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.19 rename to exp12-1/060-module.34.mlp.dense_4h_to_h/events.out.tfevents.1638219687.r8i4n2.3275712.19 diff --git a/exp12/060-module.34.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.18 b/exp12-1/060-module.34.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.18 similarity index 100% rename from exp12/060-module.34.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.18 rename to exp12-1/060-module.34.mlp.dense_h_to_4h/events.out.tfevents.1638219687.r8i4n2.3275712.18 diff --git a/exp12/060-module.34.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.20 b/exp12-1/060-module.34.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.20 similarity index 100% rename from exp12/060-module.34.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.20 rename to exp12-1/060-module.34.mlp/events.out.tfevents.1638219687.r8i4n2.3275712.20 diff --git a/exp12/060-module.34.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.17 b/exp12-1/060-module.34.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.17 similarity index 100% rename from exp12/060-module.34.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.17 rename to exp12-1/060-module.34.post_attention_layernorm/events.out.tfevents.1638219687.r8i4n2.3275712.17 diff --git a/exp12/060-module.34.self_attention.attention_dropout/events.out.tfevents.1638219687.r8i4n2.3275712.14 b/exp12-1/060-module.34.self_attention.attention_dropout/events.out.tfevents.1638219687.r8i4n2.3275712.14 similarity index 100% rename from exp12/060-module.34.self_attention.attention_dropout/events.out.tfevents.1638219687.r8i4n2.3275712.14 rename to exp12-1/060-module.34.self_attention.attention_dropout/events.out.tfevents.1638219687.r8i4n2.3275712.14 diff --git a/exp12/060-module.34.self_attention.dense/events.out.tfevents.1638219687.r8i4n2.3275712.15 b/exp12-1/060-module.34.self_attention.dense/events.out.tfevents.1638219687.r8i4n2.3275712.15 similarity index 100% rename from exp12/060-module.34.self_attention.dense/events.out.tfevents.1638219687.r8i4n2.3275712.15 rename to exp12-1/060-module.34.self_attention.dense/events.out.tfevents.1638219687.r8i4n2.3275712.15 diff --git a/exp12/060-module.34.self_attention.query_key_value/events.out.tfevents.1638219687.r8i4n2.3275712.12 b/exp12-1/060-module.34.self_attention.query_key_value/events.out.tfevents.1638219687.r8i4n2.3275712.12 similarity index 100% rename from exp12/060-module.34.self_attention.query_key_value/events.out.tfevents.1638219687.r8i4n2.3275712.12 rename to exp12-1/060-module.34.self_attention.query_key_value/events.out.tfevents.1638219687.r8i4n2.3275712.12 diff --git a/exp12/060-module.34.self_attention.scale_mask_softmax/events.out.tfevents.1638219687.r8i4n2.3275712.13 b/exp12-1/060-module.34.self_attention.scale_mask_softmax/events.out.tfevents.1638219687.r8i4n2.3275712.13 similarity index 100% rename from exp12/060-module.34.self_attention.scale_mask_softmax/events.out.tfevents.1638219687.r8i4n2.3275712.13 rename to exp12-1/060-module.34.self_attention.scale_mask_softmax/events.out.tfevents.1638219687.r8i4n2.3275712.13 diff --git a/exp12/060-module.34.self_attention/events.out.tfevents.1638219687.r8i4n2.3275712.16 b/exp12-1/060-module.34.self_attention/events.out.tfevents.1638219687.r8i4n2.3275712.16 similarity index 100% rename from exp12/060-module.34.self_attention/events.out.tfevents.1638219687.r8i4n2.3275712.16 rename to exp12-1/060-module.34.self_attention/events.out.tfevents.1638219687.r8i4n2.3275712.16 diff --git a/exp12/060-module.34/events.out.tfevents.1638219687.r8i4n2.3275712.21 b/exp12-1/060-module.34/events.out.tfevents.1638219687.r8i4n2.3275712.21 similarity index 100% rename from exp12/060-module.34/events.out.tfevents.1638219687.r8i4n2.3275712.21 rename to exp12-1/060-module.34/events.out.tfevents.1638219687.r8i4n2.3275712.21 diff --git a/exp12/060-module/events.out.tfevents.1638219687.r8i4n2.3275712.22 b/exp12-1/060-module/events.out.tfevents.1638219687.r8i4n2.3275712.22 similarity index 100% rename from exp12/060-module/events.out.tfevents.1638219687.r8i4n2.3275712.22 rename to exp12-1/060-module/events.out.tfevents.1638219687.r8i4n2.3275712.22 diff --git a/exp12/092-module.49.input_layernorm/events.out.tfevents.1638219698.r8i5n7.607539.0 b/exp12-1/092-module.49.input_layernorm/events.out.tfevents.1638219698.r8i5n7.607539.0 similarity index 100% rename from exp12/092-module.49.input_layernorm/events.out.tfevents.1638219698.r8i5n7.607539.0 rename to exp12-1/092-module.49.input_layernorm/events.out.tfevents.1638219698.r8i5n7.607539.0 diff --git a/exp12/092-module.49.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.8 b/exp12-1/092-module.49.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.8 similarity index 100% rename from exp12/092-module.49.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.8 rename to exp12-1/092-module.49.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.8 diff --git a/exp12/092-module.49.mlp.dense_h_to_4h/events.out.tfevents.1638219699.r8i5n7.607539.7 b/exp12-1/092-module.49.mlp.dense_h_to_4h/events.out.tfevents.1638219699.r8i5n7.607539.7 similarity index 100% rename from exp12/092-module.49.mlp.dense_h_to_4h/events.out.tfevents.1638219699.r8i5n7.607539.7 rename to exp12-1/092-module.49.mlp.dense_h_to_4h/events.out.tfevents.1638219699.r8i5n7.607539.7 diff --git a/exp12/092-module.49.mlp/events.out.tfevents.1638219700.r8i5n7.607539.9 b/exp12-1/092-module.49.mlp/events.out.tfevents.1638219700.r8i5n7.607539.9 similarity index 100% rename from exp12/092-module.49.mlp/events.out.tfevents.1638219700.r8i5n7.607539.9 rename to exp12-1/092-module.49.mlp/events.out.tfevents.1638219700.r8i5n7.607539.9 diff --git a/exp12/092-module.49.post_attention_layernorm/events.out.tfevents.1638219699.r8i5n7.607539.6 b/exp12-1/092-module.49.post_attention_layernorm/events.out.tfevents.1638219699.r8i5n7.607539.6 similarity index 100% rename from exp12/092-module.49.post_attention_layernorm/events.out.tfevents.1638219699.r8i5n7.607539.6 rename to exp12-1/092-module.49.post_attention_layernorm/events.out.tfevents.1638219699.r8i5n7.607539.6 diff --git a/exp12/092-module.49.self_attention.attention_dropout/events.out.tfevents.1638219699.r8i5n7.607539.3 b/exp12-1/092-module.49.self_attention.attention_dropout/events.out.tfevents.1638219699.r8i5n7.607539.3 similarity index 100% rename from exp12/092-module.49.self_attention.attention_dropout/events.out.tfevents.1638219699.r8i5n7.607539.3 rename to exp12-1/092-module.49.self_attention.attention_dropout/events.out.tfevents.1638219699.r8i5n7.607539.3 diff --git a/exp12/092-module.49.self_attention.dense/events.out.tfevents.1638219699.r8i5n7.607539.4 b/exp12-1/092-module.49.self_attention.dense/events.out.tfevents.1638219699.r8i5n7.607539.4 similarity index 100% rename from exp12/092-module.49.self_attention.dense/events.out.tfevents.1638219699.r8i5n7.607539.4 rename to exp12-1/092-module.49.self_attention.dense/events.out.tfevents.1638219699.r8i5n7.607539.4 diff --git a/exp12/092-module.49.self_attention.query_key_value/events.out.tfevents.1638219699.r8i5n7.607539.1 b/exp12-1/092-module.49.self_attention.query_key_value/events.out.tfevents.1638219699.r8i5n7.607539.1 similarity index 100% rename from exp12/092-module.49.self_attention.query_key_value/events.out.tfevents.1638219699.r8i5n7.607539.1 rename to exp12-1/092-module.49.self_attention.query_key_value/events.out.tfevents.1638219699.r8i5n7.607539.1 diff --git a/exp12/092-module.49.self_attention.scale_mask_softmax/events.out.tfevents.1638219699.r8i5n7.607539.2 b/exp12-1/092-module.49.self_attention.scale_mask_softmax/events.out.tfevents.1638219699.r8i5n7.607539.2 similarity index 100% rename from exp12/092-module.49.self_attention.scale_mask_softmax/events.out.tfevents.1638219699.r8i5n7.607539.2 rename to exp12-1/092-module.49.self_attention.scale_mask_softmax/events.out.tfevents.1638219699.r8i5n7.607539.2 diff --git a/exp12/092-module.49.self_attention/events.out.tfevents.1638219699.r8i5n7.607539.5 b/exp12-1/092-module.49.self_attention/events.out.tfevents.1638219699.r8i5n7.607539.5 similarity index 100% rename from exp12/092-module.49.self_attention/events.out.tfevents.1638219699.r8i5n7.607539.5 rename to exp12-1/092-module.49.self_attention/events.out.tfevents.1638219699.r8i5n7.607539.5 diff --git a/exp12/092-module.49/events.out.tfevents.1638219700.r8i5n7.607539.10 b/exp12-1/092-module.49/events.out.tfevents.1638219700.r8i5n7.607539.10 similarity index 100% rename from exp12/092-module.49/events.out.tfevents.1638219700.r8i5n7.607539.10 rename to exp12-1/092-module.49/events.out.tfevents.1638219700.r8i5n7.607539.10 diff --git a/exp12/092-module.50.input_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.11 b/exp12-1/092-module.50.input_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.11 similarity index 100% rename from exp12/092-module.50.input_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.11 rename to exp12-1/092-module.50.input_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.11 diff --git a/exp12/092-module.50.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.19 b/exp12-1/092-module.50.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.19 similarity index 100% rename from exp12/092-module.50.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.19 rename to exp12-1/092-module.50.mlp.dense_4h_to_h/events.out.tfevents.1638219700.r8i5n7.607539.19 diff --git a/exp12/092-module.50.mlp.dense_h_to_4h/events.out.tfevents.1638219700.r8i5n7.607539.18 b/exp12-1/092-module.50.mlp.dense_h_to_4h/events.out.tfevents.1638219700.r8i5n7.607539.18 similarity index 100% rename from exp12/092-module.50.mlp.dense_h_to_4h/events.out.tfevents.1638219700.r8i5n7.607539.18 rename to exp12-1/092-module.50.mlp.dense_h_to_4h/events.out.tfevents.1638219700.r8i5n7.607539.18 diff --git a/exp12/092-module.50.mlp/events.out.tfevents.1638219700.r8i5n7.607539.20 b/exp12-1/092-module.50.mlp/events.out.tfevents.1638219700.r8i5n7.607539.20 similarity index 100% rename from exp12/092-module.50.mlp/events.out.tfevents.1638219700.r8i5n7.607539.20 rename to exp12-1/092-module.50.mlp/events.out.tfevents.1638219700.r8i5n7.607539.20 diff --git a/exp12/092-module.50.post_attention_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.17 b/exp12-1/092-module.50.post_attention_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.17 similarity index 100% rename from exp12/092-module.50.post_attention_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.17 rename to exp12-1/092-module.50.post_attention_layernorm/events.out.tfevents.1638219700.r8i5n7.607539.17 diff --git a/exp12/092-module.50.self_attention.attention_dropout/events.out.tfevents.1638219700.r8i5n7.607539.14 b/exp12-1/092-module.50.self_attention.attention_dropout/events.out.tfevents.1638219700.r8i5n7.607539.14 similarity index 100% rename from exp12/092-module.50.self_attention.attention_dropout/events.out.tfevents.1638219700.r8i5n7.607539.14 rename to exp12-1/092-module.50.self_attention.attention_dropout/events.out.tfevents.1638219700.r8i5n7.607539.14 diff --git a/exp12/092-module.50.self_attention.dense/events.out.tfevents.1638219700.r8i5n7.607539.15 b/exp12-1/092-module.50.self_attention.dense/events.out.tfevents.1638219700.r8i5n7.607539.15 similarity index 100% rename from exp12/092-module.50.self_attention.dense/events.out.tfevents.1638219700.r8i5n7.607539.15 rename to exp12-1/092-module.50.self_attention.dense/events.out.tfevents.1638219700.r8i5n7.607539.15 diff --git a/exp12/092-module.50.self_attention.query_key_value/events.out.tfevents.1638219700.r8i5n7.607539.12 b/exp12-1/092-module.50.self_attention.query_key_value/events.out.tfevents.1638219700.r8i5n7.607539.12 similarity index 100% rename from exp12/092-module.50.self_attention.query_key_value/events.out.tfevents.1638219700.r8i5n7.607539.12 rename to exp12-1/092-module.50.self_attention.query_key_value/events.out.tfevents.1638219700.r8i5n7.607539.12 diff --git a/exp12/092-module.50.self_attention.scale_mask_softmax/events.out.tfevents.1638219700.r8i5n7.607539.13 b/exp12-1/092-module.50.self_attention.scale_mask_softmax/events.out.tfevents.1638219700.r8i5n7.607539.13 similarity index 100% rename from exp12/092-module.50.self_attention.scale_mask_softmax/events.out.tfevents.1638219700.r8i5n7.607539.13 rename to exp12-1/092-module.50.self_attention.scale_mask_softmax/events.out.tfevents.1638219700.r8i5n7.607539.13 diff --git a/exp12/092-module.50.self_attention/events.out.tfevents.1638219700.r8i5n7.607539.16 b/exp12-1/092-module.50.self_attention/events.out.tfevents.1638219700.r8i5n7.607539.16 similarity index 100% rename from exp12/092-module.50.self_attention/events.out.tfevents.1638219700.r8i5n7.607539.16 rename to exp12-1/092-module.50.self_attention/events.out.tfevents.1638219700.r8i5n7.607539.16 diff --git a/exp12/092-module.50/events.out.tfevents.1638219700.r8i5n7.607539.21 b/exp12-1/092-module.50/events.out.tfevents.1638219700.r8i5n7.607539.21 similarity index 100% rename from exp12/092-module.50/events.out.tfevents.1638219700.r8i5n7.607539.21 rename to exp12-1/092-module.50/events.out.tfevents.1638219700.r8i5n7.607539.21 diff --git a/exp12/092-module/events.out.tfevents.1638219700.r8i5n7.607539.22 b/exp12-1/092-module/events.out.tfevents.1638219700.r8i5n7.607539.22 similarity index 100% rename from exp12/092-module/events.out.tfevents.1638219700.r8i5n7.607539.22 rename to exp12-1/092-module/events.out.tfevents.1638219700.r8i5n7.607539.22 diff --git a/exp12/124-module.65.input_layernorm/events.out.tfevents.1638219712.r8i6n6.751585.0 b/exp12-1/124-module.65.input_layernorm/events.out.tfevents.1638219712.r8i6n6.751585.0 similarity index 100% rename from exp12/124-module.65.input_layernorm/events.out.tfevents.1638219712.r8i6n6.751585.0 rename to exp12-1/124-module.65.input_layernorm/events.out.tfevents.1638219712.r8i6n6.751585.0 diff --git a/exp12/124-module.65.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.8 b/exp12-1/124-module.65.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.8 similarity index 100% rename from exp12/124-module.65.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.8 rename to exp12-1/124-module.65.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.8 diff --git a/exp12/124-module.65.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.7 b/exp12-1/124-module.65.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.7 similarity index 100% rename from exp12/124-module.65.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.7 rename to exp12-1/124-module.65.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.7 diff --git a/exp12/124-module.65.mlp/events.out.tfevents.1638219713.r8i6n6.751585.9 b/exp12-1/124-module.65.mlp/events.out.tfevents.1638219713.r8i6n6.751585.9 similarity index 100% rename from exp12/124-module.65.mlp/events.out.tfevents.1638219713.r8i6n6.751585.9 rename to exp12-1/124-module.65.mlp/events.out.tfevents.1638219713.r8i6n6.751585.9 diff --git a/exp12/124-module.65.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.6 b/exp12-1/124-module.65.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.6 similarity index 100% rename from exp12/124-module.65.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.6 rename to exp12-1/124-module.65.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.6 diff --git a/exp12/124-module.65.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.3 b/exp12-1/124-module.65.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.3 similarity index 100% rename from exp12/124-module.65.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.3 rename to exp12-1/124-module.65.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.3 diff --git a/exp12/124-module.65.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.4 b/exp12-1/124-module.65.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.4 similarity index 100% rename from exp12/124-module.65.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.4 rename to exp12-1/124-module.65.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.4 diff --git a/exp12/124-module.65.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.1 b/exp12-1/124-module.65.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.1 similarity index 100% rename from exp12/124-module.65.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.1 rename to exp12-1/124-module.65.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.1 diff --git a/exp12/124-module.65.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.2 b/exp12-1/124-module.65.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.2 similarity index 100% rename from exp12/124-module.65.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.2 rename to exp12-1/124-module.65.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.2 diff --git a/exp12/124-module.65.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.5 b/exp12-1/124-module.65.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.5 similarity index 100% rename from exp12/124-module.65.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.5 rename to exp12-1/124-module.65.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.5 diff --git a/exp12/124-module.65/events.out.tfevents.1638219713.r8i6n6.751585.10 b/exp12-1/124-module.65/events.out.tfevents.1638219713.r8i6n6.751585.10 similarity index 100% rename from exp12/124-module.65/events.out.tfevents.1638219713.r8i6n6.751585.10 rename to exp12-1/124-module.65/events.out.tfevents.1638219713.r8i6n6.751585.10 diff --git a/exp12/124-module.66.input_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.11 b/exp12-1/124-module.66.input_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.11 similarity index 100% rename from exp12/124-module.66.input_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.11 rename to exp12-1/124-module.66.input_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.11 diff --git a/exp12/124-module.66.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.19 b/exp12-1/124-module.66.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.19 similarity index 100% rename from exp12/124-module.66.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.19 rename to exp12-1/124-module.66.mlp.dense_4h_to_h/events.out.tfevents.1638219713.r8i6n6.751585.19 diff --git a/exp12/124-module.66.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.18 b/exp12-1/124-module.66.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.18 similarity index 100% rename from exp12/124-module.66.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.18 rename to exp12-1/124-module.66.mlp.dense_h_to_4h/events.out.tfevents.1638219713.r8i6n6.751585.18 diff --git a/exp12/124-module.66.mlp/events.out.tfevents.1638219713.r8i6n6.751585.20 b/exp12-1/124-module.66.mlp/events.out.tfevents.1638219713.r8i6n6.751585.20 similarity index 100% rename from exp12/124-module.66.mlp/events.out.tfevents.1638219713.r8i6n6.751585.20 rename to exp12-1/124-module.66.mlp/events.out.tfevents.1638219713.r8i6n6.751585.20 diff --git a/exp12/124-module.66.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.17 b/exp12-1/124-module.66.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.17 similarity index 100% rename from exp12/124-module.66.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.17 rename to exp12-1/124-module.66.post_attention_layernorm/events.out.tfevents.1638219713.r8i6n6.751585.17 diff --git a/exp12/124-module.66.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.14 b/exp12-1/124-module.66.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.14 similarity index 100% rename from exp12/124-module.66.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.14 rename to exp12-1/124-module.66.self_attention.attention_dropout/events.out.tfevents.1638219713.r8i6n6.751585.14 diff --git a/exp12/124-module.66.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.15 b/exp12-1/124-module.66.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.15 similarity index 100% rename from exp12/124-module.66.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.15 rename to exp12-1/124-module.66.self_attention.dense/events.out.tfevents.1638219713.r8i6n6.751585.15 diff --git a/exp12/124-module.66.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.12 b/exp12-1/124-module.66.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.12 similarity index 100% rename from exp12/124-module.66.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.12 rename to exp12-1/124-module.66.self_attention.query_key_value/events.out.tfevents.1638219713.r8i6n6.751585.12 diff --git a/exp12/124-module.66.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.13 b/exp12-1/124-module.66.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.13 similarity index 100% rename from exp12/124-module.66.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.13 rename to exp12-1/124-module.66.self_attention.scale_mask_softmax/events.out.tfevents.1638219713.r8i6n6.751585.13 diff --git a/exp12/124-module.66.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.16 b/exp12-1/124-module.66.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.16 similarity index 100% rename from exp12/124-module.66.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.16 rename to exp12-1/124-module.66.self_attention/events.out.tfevents.1638219713.r8i6n6.751585.16 diff --git a/exp12/124-module.66/events.out.tfevents.1638219713.r8i6n6.751585.21 b/exp12-1/124-module.66/events.out.tfevents.1638219713.r8i6n6.751585.21 similarity index 100% rename from exp12/124-module.66/events.out.tfevents.1638219713.r8i6n6.751585.21 rename to exp12-1/124-module.66/events.out.tfevents.1638219713.r8i6n6.751585.21 diff --git a/exp12/124-module.68/events.out.tfevents.1638219713.r8i6n6.751585.22 b/exp12-1/124-module.68/events.out.tfevents.1638219713.r8i6n6.751585.22 similarity index 100% rename from exp12/124-module.68/events.out.tfevents.1638219713.r8i6n6.751585.22 rename to exp12-1/124-module.68/events.out.tfevents.1638219713.r8i6n6.751585.22 diff --git a/exp12/124-module/events.out.tfevents.1638219713.r8i6n6.751585.23 b/exp12-1/124-module/events.out.tfevents.1638219713.r8i6n6.751585.23 similarity index 100% rename from exp12/124-module/events.out.tfevents.1638219713.r8i6n6.751585.23 rename to exp12-1/124-module/events.out.tfevents.1638219713.r8i6n6.751585.23