backup contextual-turn-encoder-base-lite-bidi-full

contextual-turn-encoder-base-lite-bidi-full/best/config.json

@@ -0,0 +1,19 @@
+{
+  "input_dim": 768,
+  "hidden_dim": 768,
+  "output_dim": 768,
+  "num_layers": 6,
+  "num_heads": 8,
+  "dropout": 0.1,
+  "max_turns": 64,
+  "attention_mode": "bidirectional",
+  "use_speaker_embeddings": true,
+  "num_speakers": 4,
+  "layer_norm": true,
+  "ff_dim": 3072,
+  "activation": "gelu",
+  "output_residual": true,
+  "arch": "v2",
+  "head_mode": "tied_continuous",
+  "head_transform": true
+}

contextual-turn-encoder-base-lite-bidi-full/best/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edeb7642dc377e93f33e4ce9c0f9e7b85ad8eb8f208fa3dfc1198ddeb6917c56
+size 179808064

contextual-turn-encoder-base-lite-bidi-full/best/training_args.json

@@ -0,0 +1,8 @@
+{
+  "arch": "v2",
+  "mode": "bidirectional",
+  "epoch": 11,
+  "train_loss": 0.41652,
+  "val_loss": 2.84498,
+  "epoch_sec": 918.6
+}

contextual-turn-encoder-base-lite-bidi-full/config.json

@@ -0,0 +1,19 @@
+{
+  "input_dim": 768,
+  "hidden_dim": 768,
+  "output_dim": 768,
+  "num_layers": 6,
+  "num_heads": 8,
+  "dropout": 0.1,
+  "max_turns": 64,
+  "attention_mode": "bidirectional",
+  "use_speaker_embeddings": true,
+  "num_speakers": 4,
+  "layer_norm": true,
+  "ff_dim": 3072,
+  "activation": "gelu",
+  "output_residual": true,
+  "arch": "v2",
+  "head_mode": "tied_continuous",
+  "head_transform": true
+}

contextual-turn-encoder-base-lite-bidi-full/config.yaml

@@ -0,0 +1,66 @@
+model:
+  input_dim: 768
+  hidden_dim: 768
+  output_dim: 768
+  num_layers: 6
+  num_heads: 8
+  dropout: 0.1
+  max_turns: 64
+  attention_mode: bidirectional
+  use_speaker_embeddings: true
+  num_speakers: 4
+  layer_norm: true
+  ff_dim: 3072
+  activation: gelu
+  output_residual: true
+  arch: v2
+  head_mode: tied_continuous
+  head_transform: true
+losses:
+  masked_reconstruction:
+    enabled: null
+    mask_prob: 0.15
+    weight: 1.0
+  next_turn_prediction:
+    enabled: null
+    weight: 1.0
+  embedding_retrieval:
+    enabled: true
+    weight: 1.0
+    temperature: 0.07
+    normalize: true
+    candidate_mode: in_batch
+    target: auto
+  lambda_cosine: 1.0
+training:
+  seed: 42
+  batch_size: 128
+  epochs: 15
+  learning_rate: 0.0002
+  weight_decay: 0.01
+  warmup_ratio: 0.05
+  gradient_clip_norm: 1.0
+  device: mps
+  mixed_precision: false
+  num_workers: 0
+  log_interval: 100
+  output_dir: models/contextual-turn-encoder-base
+data:
+  path: null
+  max_turns: 64
+  window: truncate
+  stride: 32
+  dialogue_id_col: dialogue_id
+  turn_id_col: turn_id
+  utterance_col: utterance
+  speaker_col: speaker
+  embedding_col: embedding
+  speaker_map: null
+base_encoder:
+  model_name: sergioburdisso/dialog2flow-joint-bert-base
+  backend: sentence_transformers
+  batch_size: 64
+  normalize: false
+  freeze: true
+  device: auto
+  cache_dir: null

contextual-turn-encoder-base-lite-bidi-full/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61e07b1f5f6445ffe0caa945b7ded92ef870691923b6eb0c1fad1270eecb988a
+size 179808064

contextual-turn-encoder-base-lite-bidi-full/training_args.json

@@ -0,0 +1,8 @@
+{
+  "arch": "v2",
+  "mode": "bidirectional",
+  "epoch": 15,
+  "train_loss": 0.29313,
+  "val_loss": 2.85316,
+  "epoch_sec": 899.5
+}

contextual-turn-encoder-base-lite-bidi-full/trainlog.jsonl

@@ -0,0 +1,15 @@
+{"arch": "v2", "mode": "bidirectional", "epoch": 1, "train_loss": 3.29382, "val_loss": 3.6867, "epoch_sec": 1050.6}
+{"arch": "v2", "mode": "bidirectional", "epoch": 2, "train_loss": 2.33961, "val_loss": 3.24572, "epoch_sec": 1155.6}
+{"arch": "v2", "mode": "bidirectional", "epoch": 3, "train_loss": 1.85397, "val_loss": 3.00495, "epoch_sec": 950.9}
+{"arch": "v2", "mode": "bidirectional", "epoch": 4, "train_loss": 1.50441, "val_loss": 2.92765, "epoch_sec": 905.6}
+{"arch": "v2", "mode": "bidirectional", "epoch": 5, "train_loss": 1.22799, "val_loss": 2.9019, "epoch_sec": 946.5}
+{"arch": "v2", "mode": "bidirectional", "epoch": 6, "train_loss": 1.00536, "val_loss": 2.90056, "epoch_sec": 913.7}
+{"arch": "v2", "mode": "bidirectional", "epoch": 7, "train_loss": 0.82941, "val_loss": 2.91215, "epoch_sec": 919.2}
+{"arch": "v2", "mode": "bidirectional", "epoch": 8, "train_loss": 0.68773, "val_loss": 2.8756, "epoch_sec": 905.9}
+{"arch": "v2", "mode": "bidirectional", "epoch": 9, "train_loss": 0.57565, "val_loss": 2.88066, "epoch_sec": 924.9}
+{"arch": "v2", "mode": "bidirectional", "epoch": 10, "train_loss": 0.4866, "val_loss": 2.87642, "epoch_sec": 901.8}
+{"arch": "v2", "mode": "bidirectional", "epoch": 11, "train_loss": 0.41652, "val_loss": 2.84498, "epoch_sec": 918.6}
+{"arch": "v2", "mode": "bidirectional", "epoch": 12, "train_loss": 0.36512, "val_loss": 2.86164, "epoch_sec": 899.1}
+{"arch": "v2", "mode": "bidirectional", "epoch": 13, "train_loss": 0.32902, "val_loss": 2.84941, "epoch_sec": 910.0}
+{"arch": "v2", "mode": "bidirectional", "epoch": 14, "train_loss": 0.30566, "val_loss": 2.85103, "epoch_sec": 903.2}
+{"arch": "v2", "mode": "bidirectional", "epoch": 15, "train_loss": 0.29313, "val_loss": 2.85316, "epoch_sec": 899.5}