backup contextual-turn-encoder-base-lite-ar-1m

contextual-turn-encoder-base-lite-ar-1m/best/config.json

@@ -0,0 +1,19 @@
+{
+  "input_dim": 768,
+  "hidden_dim": 768,
+  "output_dim": 768,
+  "num_layers": 4,
+  "num_heads": 8,
+  "dropout": 0.1,
+  "max_turns": 64,
+  "attention_mode": "autoregressive",
+  "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-ar-1m/best/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1826f8f0dec4636db46f8bd511fec770fd4f557c53bc7e09fe1d4605cb03397
+size 123101328

contextual-turn-encoder-base-lite-ar-1m/best/training_args.json

@@ -0,0 +1,8 @@
+{
+  "arch": "v2",
+  "mode": "autoregressive",
+  "epoch": 10,
+  "train_loss": 1.89073,
+  "val_loss": 4.4073,
+  "epoch_sec": 263.8
+}

contextual-turn-encoder-base-lite-ar-1m/config.json

@@ -0,0 +1,19 @@
+{
+  "input_dim": 768,
+  "hidden_dim": 768,
+  "output_dim": 768,
+  "num_layers": 4,
+  "num_heads": 8,
+  "dropout": 0.1,
+  "max_turns": 64,
+  "attention_mode": "autoregressive",
+  "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-ar-1m/config.yaml

@@ -0,0 +1,66 @@
+model:
+  input_dim: 768
+  hidden_dim: 768
+  output_dim: 768
+  num_layers: 4
+  num_heads: 8
+  dropout: 0.1
+  max_turns: 64
+  attention_mode: autoregressive
+  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: 10
+  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-ar-1m/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1826f8f0dec4636db46f8bd511fec770fd4f557c53bc7e09fe1d4605cb03397
+size 123101328

contextual-turn-encoder-base-lite-ar-1m/training_args.json

@@ -0,0 +1,8 @@
+{
+  "arch": "v2",
+  "mode": "autoregressive",
+  "epoch": 10,
+  "train_loss": 1.89073,
+  "val_loss": 4.4073,
+  "epoch_sec": 263.8
+}

contextual-turn-encoder-base-lite-ar-1m/trainlog.jsonl

@@ -0,0 +1,10 @@
+{"arch": "v2", "mode": "autoregressive", "epoch": 1, "train_loss": 4.53824, "val_loss": 4.79324, "epoch_sec": 285.4}
+{"arch": "v2", "mode": "autoregressive", "epoch": 2, "train_loss": 3.5848, "val_loss": 4.57696, "epoch_sec": 273.6}
+{"arch": "v2", "mode": "autoregressive", "epoch": 3, "train_loss": 3.31053, "val_loss": 4.47773, "epoch_sec": 268.0}
+{"arch": "v2", "mode": "autoregressive", "epoch": 4, "train_loss": 3.06479, "val_loss": 4.43022, "epoch_sec": 264.7}
+{"arch": "v2", "mode": "autoregressive", "epoch": 5, "train_loss": 2.78972, "val_loss": 4.42681, "epoch_sec": 282.1}
+{"arch": "v2", "mode": "autoregressive", "epoch": 6, "train_loss": 2.53113, "val_loss": 4.42761, "epoch_sec": 266.2}
+{"arch": "v2", "mode": "autoregressive", "epoch": 7, "train_loss": 2.30905, "val_loss": 4.43798, "epoch_sec": 269.8}
+{"arch": "v2", "mode": "autoregressive", "epoch": 8, "train_loss": 2.1279, "val_loss": 4.45396, "epoch_sec": 265.3}
+{"arch": "v2", "mode": "autoregressive", "epoch": 9, "train_loss": 1.9868, "val_loss": 4.43622, "epoch_sec": 263.1}
+{"arch": "v2", "mode": "autoregressive", "epoch": 10, "train_loss": 1.89073, "val_loss": 4.4073, "epoch_sec": 263.8}