result h75b1c5b6bff8_s1

sweeps/h75b1c5b6bff8_s1/metrics.json

@@ -0,0 +1,80 @@
+{
+  "label": "norm_typelayernorm",
+  "overrides": {
+    "norm_type": "layernorm"
+  },
+  "seed": 1,
+  "config_hash": "75b1c5b6bff8",
+  "config": {
+    "d_model": 384,
+    "n_layers": 6,
+    "n_heads": 6,
+    "n_kv_heads": 2,
+    "vocab_size": 8192,
+    "seq_len": 512,
+    "tie_embeddings": true,
+    "norm_type": "layernorm",
+    "pos_type": "rope",
+    "mlp_type": "swiglu",
+    "dropout": 0.0,
+    "lr": 0.0004,
+    "min_lr": 4e-05,
+    "warmup_steps": 120,
+    "max_steps": 2000,
+    "weight_decay": 0.1,
+    "betas": [
+      0.9,
+      0.95
+    ],
+    "grad_clip": 1.0,
+    "batch_size": 24,
+    "grad_accum_steps": 4,
+    "train_bin": "",
+    "val_bin": "",
+    "dtype": "float16",
+    "compile": false,
+    "seed": 1,
+    "ckpt_every": 500,
+    "eval_every": 500,
+    "eval_batches": 20,
+    "log_every": 50,
+    "wandb_project": "slm-lab",
+    "wandb_mode": "online"
+  },
+  "final_val_loss": 1.8144319081306457,
+  "perplexity": 6.137588472761583,
+  "bits_per_byte": 0.08368552061562498,
+  "val_tokens_scored": 614400,
+  "val_bytes": 19218350,
+  "wall_clock_s": 1196.1300036907196,
+  "device": {
+    "device": "Tesla T4",
+    "capability": [
+      7,
+      5
+    ],
+    "bf16": false,
+    "total_gb": 15.637086208
+  },
+  "performance": {
+    "device": "Tesla T4",
+    "params": 12592896,
+    "wall_clock_s": 1193.640834569931,
+    "tokens_seen": 98304000,
+    "session_tokens": 98304000,
+    "resumed_from_step": 0,
+    "mfu": 0.11366854181733042,
+    "tokens_per_sec": 82356.43329226106,
+    "flops_per_token": 89713152,
+    "peak_tflops": 65.0,
+    "peak_is_known": true,
+    "achieved_tflops": 7.388455218126477,
+    "peak_memory_mb": 3118.088192,
+    "estimated": {
+      "params": 12592896,
+      "optimizer_mb": 201.486336,
+      "activations_mb": 2661.285888,
+      "estimated_total_mb": 2862.772224
+    }
+  }
+}