add sweep_g1_w256_lr0.001 (completed, step 3000)

sweep_g1_w256_lr0.001/config.json

@@ -0,0 +1,19 @@
+{
+ "arch": "aspect-d bidirectional masked-diffusion transformer over Mimi tokens",
+ "config_id": "proxy_d12_w256",
+ "seed": 0,
+ "width": 256,
+ "depth": 12,
+ "heads": 4,
+ "nonembed_params": 9443584,
+ "total_params": 17876992,
+ "mup_base_width": 256,
+ "lr_base": 0.001,
+ "steps": 3000,
+ "status": "completed",
+ "codec": "kyutai/mimi (12.5 Hz, 8 codebooks, 2048+MASK+PAD)",
+ "phoneme_vocab_file": "phone_vocab.json",
+ "final_val_loss": 5.611729965079002,
+ "gpu_hours": 0.22675419390201568,
+ "dtype": "bfloat16"
+}

sweep_g1_w256_lr0.001/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d3af716d0d75b4253db4c0d1a9348622510d648e936c0ffb4a2732688c6d729
+size 35762192

sweep_g1_w256_lr0.001/run.json

@@ -0,0 +1,41 @@
+{
+ "config": "proxy_d12_w256",
+ "seed": 0,
+ "lr": 0.001,
+ "steps": 3000,
+ "width": 256,
+ "depth": 12,
+ "heads": 4,
+ "nonembed_params": 9443584,
+ "total_params": 17876992,
+ "micro_batch": 256,
+ "accum": 1,
+ "device": "NVIDIA B200",
+ "started": 1785971287.77931,
+ "status": "completed",
+ "wall_seconds": 816.3150980472565,
+ "gpu_hours": 0.22675419390201568,
+ "wall_seconds_this_segment": 816.3151023387909,
+ "final_step": 3000,
+ "val_hist": [
+  {
+   "step": 1000,
+   "val_loss": 6.158987213275466,
+   "wall": 277.7207260131836
+  },
+  {
+   "step": 2000,
+   "val_loss": 5.768409265687088,
+   "wall": 544.9150099754333
+  },
+  {
+   "step": 3000,
+   "val_loss": 5.611729965079002,
+   "wall": 812.6518595218658
+  }
+ ],
+ "tokens_seen": 695101232,
+ "final_train_loss": 5.651921023516699,
+ "final_val_loss": 5.611729965079002,
+ "finished": 1785972104.2436078
+}