publish trace-repro-deep-ar-s42-xlc

trace-repro-deep-ar-s42-xlc/best/config.json

@@ -0,0 +1,19 @@
+{
+  "input_dim": 768,
+  "hidden_dim": 768,
+  "output_dim": 768,
+  "num_layers": 12,
+  "num_heads": 12,
+  "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
+}

trace-repro-deep-ar-s42-xlc/best/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51ad86aa4b8eed5d9e9975e8b58a7d2aa3e4c7c113400d44175e85f00e72b523
+size 349928328

trace-repro-deep-ar-s42-xlc/config.json

@@ -0,0 +1,19 @@
+{
+  "input_dim": 768,
+  "hidden_dim": 768,
+  "output_dim": 768,
+  "num_layers": 12,
+  "num_heads": 12,
+  "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
+}

trace-repro-deep-ar-s42-xlc/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1866401845b1e7978f26268abd7e1b0337b16044c7aa8b4c841a9b982dc6aafb
+size 349928328

trace-repro-deep-ar-s42-xlc/trainlog.jsonl

@@ -0,0 +1,6 @@
+{"model": "deep", "mode": "ar", "seed": 42, "lr": 0.0001, "canary": false, "init": "/home/jfernandez/trace/models/contextual-turn-encoder-base-deep-ar-xlc-tm-repro/best", "epoch": 1, "train_loss": 3.77569, "val_loss": 4.86123, "epoch_sec": 221.9}
+{"model": "deep", "mode": "ar", "seed": 42, "lr": 0.0001, "canary": false, "init": "/home/jfernandez/trace/models/contextual-turn-encoder-base-deep-ar-xlc-tm-repro/best", "epoch": 2, "train_loss": 3.1655, "val_loss": 4.6401, "epoch_sec": 206.6}
+{"model": "deep", "mode": "ar", "seed": 42, "lr": 0.0001, "canary": false, "init": "/home/jfernandez/trace/models/contextual-turn-encoder-base-deep-ar-xlc-tm-repro/best", "epoch": 3, "train_loss": 2.72945, "val_loss": 4.54262, "epoch_sec": 205.9}
+{"model": "deep", "mode": "ar", "seed": 42, "lr": 0.0001, "canary": false, "init": "/home/jfernandez/trace/models/contextual-turn-encoder-base-deep-ar-xlc-tm-repro/best", "epoch": 4, "train_loss": 2.33792, "val_loss": 4.53047, "epoch_sec": 206.9}
+{"model": "deep", "mode": "ar", "seed": 42, "lr": 0.0001, "canary": false, "init": "/home/jfernandez/trace/models/contextual-turn-encoder-base-deep-ar-xlc-tm-repro/best", "epoch": 5, "train_loss": 2.01279, "val_loss": 4.55166, "epoch_sec": 206.0}
+{"model": "deep", "mode": "ar", "seed": 42, "lr": 0.0001, "canary": false, "init": "/home/jfernandez/trace/models/contextual-turn-encoder-base-deep-ar-xlc-tm-repro/best", "epoch": 6, "train_loss": 1.75724, "val_loss": 4.58946, "epoch_sec": 206.5}