diff --git "a/logs/training.log" "b/logs/training.log" new file mode 100644--- /dev/null +++ "b/logs/training.log" @@ -0,0 +1,2382 @@ +W0824 23:13:21.569000 5193 torch/distributed/run.py:774] +W0824 23:13:21.569000 5193 torch/distributed/run.py:774] ***************************************** +W0824 23:13:21.569000 5193 torch/distributed/run.py:774] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0824 23:13:21.569000 5193 torch/distributed/run.py:774] ***************************************** +Overriding config with configs/training/minimal-en-125m-4ep.py: +# Approximately 125M-parameter pretraining run on Minimal EN Corpus 2.5B. + +# I/O +out_dir = "../runs/minimal-en-125m-4ep" +eval_interval = 1000 +eval_iters = 100 +log_interval = 10 +always_save_checkpoint = True +init_from = "scratch" + +# Logging +wandb_log = True +wandb_project = "minimal-en" +wandb_run_name = "125m-12k-4ep" + +# Data +dataset = "minimal-en-corpus-2.5b" +gradient_accumulation_steps = 40 +batch_size = 12 +block_size = 1024 + +# Model (127.6M parameters with the 12,288-token vocabulary) +n_layer = 12 +n_head = 14 +n_embd = 896 +dropout = 0.0 +bias = False + +# AdamW +learning_rate = 3e-4 +max_iters = 22003 +weight_decay = 0.1 +beta1 = 0.9 +beta2 = 0.95 +grad_clip = 1.0 + +# Cosine learning-rate schedule +decay_lr = True +warmup_iters = 440 +lr_decay_iters = 22003 +min_lr = 3e-5 + +# Runtime +backend = "nccl" +device = "cuda" +dtype = "bfloat16" +compile = True + +Overriding config with configs/training/minimal-en-125m-4ep.py: +# Approximately 125M-parameter pretraining run on Minimal EN Corpus 2.5B. + +# I/O +out_dir = "../runs/minimal-en-125m-4ep" +eval_interval = 1000 +eval_iters = 100 +log_interval = 10 +always_save_checkpoint = True +init_from = "scratch" + +# Logging +wandb_log = True +wandb_project = "minimal-en" +wandb_run_name = "125m-12k-4ep" + +# Data +dataset = "minimal-en-corpus-2.5b" +gradient_accumulation_steps = 40 +batch_size = 12 +block_size = 1024 + +# Model (127.6M parameters with the 12,288-token vocabulary) +n_layer = 12 +n_head = 14 +n_embd = 896 +dropout = 0.0 +bias = False + +# AdamW +learning_rate = 3e-4 +max_iters = 22003 +weight_decay = 0.1 +beta1 = 0.9 +beta2 = 0.95 +grad_clip = 1.0 + +# Cosine learning-rate schedule +decay_lr = True +warmup_iters = 440 +lr_decay_iters = 22003 +min_lr = 3e-5 + +# Runtime +backend = "nccl" +device = "cuda" +dtype = "bfloat16" +compile = True + +tokens per iteration will be: 491,520 +found vocab_size = 12288 (inside data/minimal-en-corpus-2.5b/meta.pkl) +Initializing a new model from scratch +tokens per iteration will be: 491,520 +found vocab_size = 12288 (inside data/minimal-en-corpus-2.5b/meta.pkl) +Initializing a new model from scratch +number of parameters: 126.64M +number of parameters: 126.64M +/workspace/nanogpt/train.py:196: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = torch.cuda.amp.GradScaler(enabled=(dtype == 'float16')) +num decayed parameter tensors: 50, with 127,533,056 parameters +num non-decayed parameter tensors: 25, with 22,400 parameters +/workspace/nanogpt/train.py:196: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = torch.cuda.amp.GradScaler(enabled=(dtype == 'float16')) +num decayed parameter tensors: 50, with 127,533,056 parameters +num non-decayed parameter tensors: 25, with 22,400 parameters +using fused AdamW: True +compiling the model... (takes a ~minute) +using fused AdamW: True +compiling the model... (takes a ~minute) +NCCL version 2.27.3+cuda12.9 +wandb: [wandb.login()] Loaded credentials for https://api.wandb.ai from /root/.netrc. +wandb: Currently logged in as: dawidmajewski (dawidmajewski-ezotic) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin +wandb: setting up run hlbpqpoe +wandb: Tracking run with wandb version 0.28.2 +wandb: Run data is saved locally in /workspace/logs/wandb/wandb/run-20260824_231331-hlbpqpoe +wandb: Run `wandb offline` to turn off syncing. +wandb: Syncing run 125m-12k-4ep +wandb: ⭐️ View project at https://wandb.ai/dawidmajewski-ezotic/minimal-en +wandb: 🚀 View run at https://wandb.ai/dawidmajewski-ezotic/minimal-en/runs/hlbpqpoe +step 0: train loss 9.5797, val loss 9.5732 +iter 0: loss 9.5701, time 36799.30ms, mfu -100.00% +iter 10: loss 8.6515, time 1993.32ms, mfu 35.25% +iter 20: loss 8.4448, time 2017.84ms, mfu 35.20% +iter 30: loss 8.2446, time 2039.18ms, mfu 35.13% +iter 40: loss 7.9333, time 2044.29ms, mfu 35.05% +iter 50: loss 7.8167, time 2061.04ms, mfu 34.96% +iter 60: loss 7.4723, time 2066.40ms, mfu 34.86% +iter 70: loss 7.1812, time 2067.45ms, mfu 34.77% +iter 80: loss 7.0585, time 2085.67ms, mfu 34.66% +iter 90: loss 6.8916, time 2071.91ms, mfu 34.59% +iter 100: loss 6.9671, time 2082.79ms, mfu 34.50% +iter 110: loss 6.6913, time 2083.09ms, mfu 34.43% +iter 120: loss 6.4896, time 2070.43ms, mfu 34.38% +iter 130: loss 6.4626, time 2069.20ms, mfu 34.33% +iter 140: loss 6.3726, time 2095.37ms, mfu 34.25% +iter 150: loss 6.2077, time 2093.91ms, mfu 34.18% +iter 160: loss 6.1422, time 2093.72ms, mfu 34.12% +iter 170: loss 5.9801, time 2101.12ms, mfu 34.05% +iter 180: loss 6.2115, time 2087.54ms, mfu 34.01% +iter 190: loss 5.8301, time 2103.76ms, mfu 33.95% +iter 200: loss 5.9559, time 2084.75ms, mfu 33.93% +iter 210: loss 5.8527, time 2090.91ms, mfu 33.89% +iter 220: loss 5.8517, time 2100.90ms, mfu 33.85% +iter 230: loss 5.6927, time 2080.43ms, mfu 33.84% +iter 240: loss 5.5679, time 2077.56ms, mfu 33.84% +iter 250: loss 5.6053, time 2106.90ms, mfu 33.79% +iter 260: loss 5.4759, time 2077.33ms, mfu 33.79% +iter 270: loss 5.3678, time 2076.06ms, mfu 33.80% +iter 280: loss 5.4428, time 2097.76ms, mfu 33.77% +iter 290: loss 5.2989, time 2086.06ms, mfu 33.76% +iter 300: loss 5.0415, time 2098.62ms, mfu 33.73% +iter 310: loss 4.9178, time 2078.58ms, mfu 33.74% +iter 320: loss 5.1031, time 2080.53ms, mfu 33.74% +iter 330: loss 5.3207, time 2104.51ms, mfu 33.70% +iter 340: loss 5.2468, time 2104.70ms, mfu 33.67% +iter 350: loss 5.2147, time 2086.30ms, mfu 33.67% +iter 360: loss 5.1747, time 2085.95ms, mfu 33.67% +iter 370: loss 5.0355, time 2079.08ms, mfu 33.69% +iter 380: loss 5.0886, time 2089.18ms, mfu 33.68% +iter 390: loss 5.0313, time 2095.34ms, mfu 33.67% +iter 400: loss 4.9797, time 2078.22ms, mfu 33.68% +iter 410: loss 5.0387, time 2069.48ms, mfu 33.71% +iter 420: loss 4.8122, time 2077.42ms, mfu 33.72% +iter 430: loss 5.1012, time 2091.23ms, mfu 33.71% +iter 440: loss 5.0285, time 2090.85ms, mfu 33.70% +iter 450: loss 4.7803, time 2085.61ms, mfu 33.69% +iter 460: loss 4.9936, time 2100.78ms, mfu 33.67% +iter 470: loss 4.2048, time 2130.72ms, mfu 33.60% +iter 480: loss 4.6369, time 2100.40ms, mfu 33.58% +iter 490: loss 4.6676, time 2104.42ms, mfu 33.56% +iter 500: loss 4.8083, time 2109.12ms, mfu 33.54% +iter 510: loss 4.4331, time 2074.72ms, mfu 33.57% +iter 520: loss 4.6347, time 2106.84ms, mfu 33.55% +iter 530: loss 4.7749, time 2081.79ms, mfu 33.57% +iter 540: loss 4.6217, time 2099.14ms, mfu 33.56% +iter 550: loss 4.4999, time 2101.37ms, mfu 33.55% +iter 560: loss 4.3056, time 2070.50ms, mfu 33.59% +iter 570: loss 4.4181, time 2104.49ms, mfu 33.57% +iter 580: loss 4.3402, time 2083.84ms, mfu 33.58% +iter 590: loss 4.5584, time 2103.17ms, mfu 33.56% +iter 600: loss 4.4462, time 2078.00ms, mfu 33.59% +iter 610: loss 4.2681, time 2104.35ms, mfu 33.57% +iter 620: loss 3.8997, time 2119.20ms, mfu 33.53% +iter 630: loss 4.1709, time 2084.69ms, mfu 33.54% +iter 640: loss 4.0962, time 2104.52ms, mfu 33.53% +iter 650: loss 3.9719, time 2088.32ms, mfu 33.54% +iter 660: loss 4.1625, time 2102.31ms, mfu 33.53% +iter 670: loss 4.0754, time 2086.50ms, mfu 33.54% +iter 680: loss 3.7406, time 2094.57ms, mfu 33.54% +iter 690: loss 3.9339, time 2085.20ms, mfu 33.56% +iter 700: loss 3.8437, time 2074.83ms, mfu 33.59% +iter 710: loss 3.9143, time 2098.15ms, mfu 33.58% +iter 720: loss 3.9684, time 2100.45ms, mfu 33.56% +iter 730: loss 3.5712, time 2098.23ms, mfu 33.56% +iter 740: loss 4.0678, time 2093.66ms, mfu 33.56% +iter 750: loss 4.0130, time 2084.70ms, mfu 33.57% +iter 760: loss 3.8476, time 2105.70ms, mfu 33.55% +iter 770: loss 3.9298, time 2086.56ms, mfu 33.56% +iter 780: loss 3.6444, time 2093.14ms, mfu 33.56% +iter 790: loss 3.5690, time 2090.08ms, mfu 33.57% +iter 800: loss 3.4821, time 2103.30ms, mfu 33.55% +iter 810: loss 3.7268, time 2082.37ms, mfu 33.57% +iter 820: loss 3.5891, time 2085.35ms, mfu 33.58% +iter 830: loss 3.7140, time 2083.72ms, mfu 33.60% +iter 840: loss 3.2428, time 2111.91ms, mfu 33.56% +iter 850: loss 3.3661, time 2104.27ms, mfu 33.55% +iter 860: loss 3.4586, time 2098.45ms, mfu 33.54% +iter 870: loss 3.6945, time 2090.80ms, mfu 33.55% +iter 880: loss 3.4731, time 2101.89ms, mfu 33.53% +iter 890: loss 3.5809, time 2074.95ms, mfu 33.57% +iter 900: loss 3.6004, time 2099.76ms, mfu 33.56% +iter 910: loss 3.5649, time 2095.65ms, mfu 33.55% +iter 920: loss 3.7382, time 2105.51ms, mfu 33.53% +iter 930: loss 3.5281, time 2083.98ms, mfu 33.55% +iter 940: loss 3.8531, time 2115.73ms, mfu 33.52% +iter 950: loss 3.6362, time 2108.05ms, mfu 33.50% +iter 960: loss 3.4495, time 2112.52ms, mfu 33.47% +iter 970: loss 3.5198, time 2104.74ms, mfu 33.47% +iter 980: loss 3.3078, time 2089.55ms, mfu 33.48% +iter 990: loss 3.4061, time 2092.22ms, mfu 33.49% +step 1000: train loss 3.4769, val loss 3.5219 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 1000: loss 3.3850, time 9006.11ms, mfu 30.92% +iter 1010: loss 3.2370, time 2078.63ms, mfu 31.21% +iter 1020: loss 3.3580, time 2100.29ms, mfu 31.43% +iter 1030: loss 3.6355, time 2092.06ms, mfu 31.65% +iter 1040: loss 3.2284, time 2087.74ms, mfu 31.85% +iter 1050: loss 3.3297, time 2097.62ms, mfu 32.01% +iter 1060: loss 3.2152, time 2086.47ms, mfu 32.18% +iter 1070: loss 3.5770, time 2076.36ms, mfu 32.35% +iter 1080: loss 3.4146, time 2106.38ms, mfu 32.45% +iter 1090: loss 2.9954, time 2085.45ms, mfu 32.57% +iter 1100: loss 3.6749, time 2109.49ms, mfu 32.64% +iter 1110: loss 3.2327, time 2106.02ms, mfu 32.72% +iter 1120: loss 3.2230, time 2088.08ms, mfu 32.81% +iter 1130: loss 3.4330, time 2106.66ms, mfu 32.86% +iter 1140: loss 3.2531, time 2090.60ms, mfu 32.94% +iter 1150: loss 3.3821, time 2092.74ms, mfu 33.00% +iter 1160: loss 3.4423, time 2105.26ms, mfu 33.04% +iter 1170: loss 3.3262, time 2087.35ms, mfu 33.10% +iter 1180: loss 3.5381, time 2091.57ms, mfu 33.15% +iter 1190: loss 3.2797, time 2110.25ms, mfu 33.16% +iter 1200: loss 3.3246, time 2075.26ms, mfu 33.23% +iter 1210: loss 3.4333, time 2094.02ms, mfu 33.26% +iter 1220: loss 3.4259, time 2088.27ms, mfu 33.30% +iter 1230: loss 3.2118, time 2091.79ms, mfu 33.33% +iter 1240: loss 2.9138, time 2097.08ms, mfu 33.35% +iter 1250: loss 3.2667, time 2088.42ms, mfu 33.38% +iter 1260: loss 2.9204, time 2099.43ms, mfu 33.39% +iter 1270: loss 3.0572, time 2086.39ms, mfu 33.42% +iter 1280: loss 3.1568, time 2117.32ms, mfu 33.39% +iter 1290: loss 3.4388, time 2086.31ms, mfu 33.42% +iter 1300: loss 3.1613, time 2086.44ms, mfu 33.45% +iter 1310: loss 3.4910, time 2088.56ms, mfu 33.47% +iter 1320: loss 3.1674, time 2080.59ms, mfu 33.50% +iter 1330: loss 3.0806, time 2086.68ms, mfu 33.51% +iter 1340: loss 3.3800, time 2108.42ms, mfu 33.49% +iter 1350: loss 3.3195, time 2089.45ms, mfu 33.51% +iter 1360: loss 3.5254, time 2096.98ms, mfu 33.51% +iter 1370: loss 3.3237, time 2108.79ms, mfu 33.49% +iter 1380: loss 3.2951, time 2104.57ms, mfu 33.48% +iter 1390: loss 3.2490, time 2087.63ms, mfu 33.50% +iter 1400: loss 3.2035, time 2099.25ms, mfu 33.49% +iter 1410: loss 3.2028, time 2096.14ms, mfu 33.49% +iter 1420: loss 2.9834, time 2091.12ms, mfu 33.51% +iter 1430: loss 3.1010, time 2106.93ms, mfu 33.49% +iter 1440: loss 3.0424, time 2083.61ms, mfu 33.51% +iter 1450: loss 2.9155, time 2098.67ms, mfu 33.51% +iter 1460: loss 3.1797, time 2084.02ms, mfu 33.53% +iter 1470: loss 3.1364, time 2088.68ms, mfu 33.54% +iter 1480: loss 3.5031, time 2089.95ms, mfu 33.55% +iter 1490: loss 3.3129, time 2098.83ms, mfu 33.54% +iter 1500: loss 3.3140, time 2078.42ms, mfu 33.57% +iter 1510: loss 3.3269, time 2096.52ms, mfu 33.56% +iter 1520: loss 3.1147, time 2094.67ms, mfu 33.56% +iter 1530: loss 3.5681, time 2081.67ms, mfu 33.58% +iter 1540: loss 3.3100, time 2091.29ms, mfu 33.58% +iter 1550: loss 3.1636, time 2082.59ms, mfu 33.60% +iter 1560: loss 3.2076, time 2081.72ms, mfu 33.61% +iter 1570: loss 3.2339, time 2082.92ms, mfu 33.62% +iter 1580: loss 2.9632, time 2081.89ms, mfu 33.64% +iter 1590: loss 3.2501, time 2090.25ms, mfu 33.63% +iter 1600: loss 3.0978, time 2086.36ms, mfu 33.64% +iter 1610: loss 3.0183, time 2095.82ms, mfu 33.63% +iter 1620: loss 3.2033, time 2103.42ms, mfu 33.60% +iter 1630: loss 3.3264, time 2082.80ms, mfu 33.62% +iter 1640: loss 2.8962, time 2080.24ms, mfu 33.63% +iter 1650: loss 3.2974, time 2082.40ms, mfu 33.64% +iter 1660: loss 3.1878, time 2077.11ms, mfu 33.66% +iter 1670: loss 2.9475, time 2088.90ms, mfu 33.66% +iter 1680: loss 3.1013, time 2079.50ms, mfu 33.67% +iter 1690: loss 3.0734, time 2073.39ms, mfu 33.69% +iter 1700: loss 3.5066, time 2079.11ms, mfu 33.70% +iter 1710: loss 3.1908, time 2088.10ms, mfu 33.70% +iter 1720: loss 3.2536, time 2108.03ms, mfu 33.66% +iter 1730: loss 3.2540, time 2107.71ms, mfu 33.63% +iter 1740: loss 2.9262, time 2088.94ms, mfu 33.63% +iter 1750: loss 3.2834, time 2091.66ms, mfu 33.62% +iter 1760: loss 3.4747, time 2087.66ms, mfu 33.63% +iter 1770: loss 3.2402, time 2081.42ms, mfu 33.64% +iter 1780: loss 2.9697, time 2075.73ms, mfu 33.66% +iter 1790: loss 2.9834, time 2087.18ms, mfu 33.66% +iter 1800: loss 3.1758, time 2095.63ms, mfu 33.65% +iter 1810: loss 3.0526, time 2105.01ms, mfu 33.62% +iter 1820: loss 2.9602, time 2089.51ms, mfu 33.62% +iter 1830: loss 2.8927, time 2103.48ms, mfu 33.60% +iter 1840: loss 3.2547, time 2103.57ms, mfu 33.58% +iter 1850: loss 3.1535, time 2078.31ms, mfu 33.60% +iter 1860: loss 3.1056, time 2091.17ms, mfu 33.60% +iter 1870: loss 3.2162, time 2073.97ms, mfu 33.63% +iter 1880: loss 3.0082, time 2083.10ms, mfu 33.64% +iter 1890: loss 3.0484, time 2075.17ms, mfu 33.66% +iter 1900: loss 3.1529, time 2081.50ms, mfu 33.67% +iter 1910: loss 3.0617, time 2108.79ms, mfu 33.63% +iter 1920: loss 2.8841, time 2093.34ms, mfu 33.63% +iter 1930: loss 2.9809, time 2109.44ms, mfu 33.60% +iter 1940: loss 3.0346, time 2087.30ms, mfu 33.60% +iter 1950: loss 3.2623, time 2075.74ms, mfu 33.63% +iter 1960: loss 3.0727, time 2084.41ms, mfu 33.63% +iter 1970: loss 3.2350, time 2091.24ms, mfu 33.63% +iter 1980: loss 2.7521, time 2094.86ms, mfu 33.62% +iter 1990: loss 3.0140, time 2087.20ms, mfu 33.63% +step 2000: train loss 3.0943, val loss 3.1266 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 2000: loss 3.0097, time 9108.18ms, mfu 31.03% +iter 2010: loss 2.9544, time 2086.93ms, mfu 31.30% +iter 2020: loss 3.1589, time 2078.56ms, mfu 31.55% +iter 2030: loss 3.1263, time 2097.99ms, mfu 31.74% +iter 2040: loss 2.9587, time 2099.60ms, mfu 31.91% +iter 2050: loss 3.0223, time 2097.16ms, mfu 32.07% +iter 2060: loss 2.9487, time 2088.10ms, mfu 32.23% +iter 2070: loss 2.9620, time 2082.66ms, mfu 32.38% +iter 2080: loss 3.0641, time 2087.66ms, mfu 32.51% +iter 2090: loss 2.7195, time 2074.76ms, mfu 32.64% +iter 2100: loss 3.2061, time 2072.21ms, mfu 32.77% +iter 2110: loss 2.8790, time 2098.70ms, mfu 32.84% +iter 2120: loss 3.2879, time 2075.23ms, mfu 32.94% +iter 2130: loss 3.1826, time 2090.97ms, mfu 33.01% +iter 2140: loss 3.4425, time 2082.97ms, mfu 33.08% +iter 2150: loss 2.9741, time 2072.72ms, mfu 33.16% +iter 2160: loss 3.2976, time 2078.42ms, mfu 33.23% +iter 2170: loss 2.9592, time 2085.85ms, mfu 33.27% +iter 2180: loss 3.4325, time 2083.51ms, mfu 33.32% +iter 2190: loss 3.1819, time 2098.03ms, mfu 33.33% +iter 2200: loss 3.1337, time 2071.30ms, mfu 33.39% +iter 2210: loss 3.0284, time 2103.11ms, mfu 33.39% +iter 2220: loss 2.9793, time 2081.02ms, mfu 33.43% +iter 2230: loss 3.0734, time 2094.47ms, mfu 33.44% +iter 2240: loss 3.1064, time 2074.89ms, mfu 33.48% +iter 2250: loss 3.2566, time 2083.70ms, mfu 33.51% +iter 2260: loss 2.8806, time 2086.37ms, mfu 33.52% +iter 2270: loss 3.0763, time 2091.94ms, mfu 33.53% +iter 2280: loss 3.0993, time 2082.17ms, mfu 33.55% +iter 2290: loss 2.7687, time 2103.77ms, mfu 33.54% +iter 2300: loss 3.1544, time 2086.01ms, mfu 33.55% +iter 2310: loss 3.0680, time 2081.71ms, mfu 33.57% +iter 2320: loss 2.9684, time 2097.08ms, mfu 33.56% +iter 2330: loss 2.7440, time 2096.52ms, mfu 33.56% +iter 2340: loss 3.0904, time 2076.54ms, mfu 33.59% +iter 2350: loss 3.0910, time 2079.44ms, mfu 33.61% +iter 2360: loss 3.1517, time 2098.28ms, mfu 33.59% +iter 2370: loss 3.2139, time 2067.59ms, mfu 33.63% +iter 2380: loss 3.2552, time 2086.13ms, mfu 33.64% +iter 2390: loss 2.9476, time 2096.71ms, mfu 33.62% +iter 2400: loss 3.0786, time 2093.13ms, mfu 33.62% +iter 2410: loss 2.8561, time 2068.74ms, mfu 33.65% +iter 2420: loss 2.9681, time 2079.62ms, mfu 33.67% +iter 2430: loss 3.0639, time 2077.07ms, mfu 33.68% +iter 2440: loss 2.9306, time 2074.17ms, mfu 33.70% +iter 2450: loss 2.7985, time 2096.54ms, mfu 33.68% +iter 2460: loss 2.9853, time 2068.47ms, mfu 33.71% +iter 2470: loss 2.9712, time 2067.52ms, mfu 33.74% +iter 2480: loss 2.7578, time 2075.79ms, mfu 33.75% +iter 2490: loss 2.9264, time 2094.64ms, mfu 33.73% +iter 2500: loss 2.6382, time 2072.98ms, mfu 33.74% +iter 2510: loss 2.9940, time 2086.30ms, mfu 33.74% +iter 2520: loss 2.7323, time 2069.35ms, mfu 33.76% +iter 2530: loss 2.6686, time 2079.04ms, mfu 33.76% +iter 2540: loss 2.7341, time 2090.39ms, mfu 33.75% +iter 2550: loss 2.7594, time 2081.52ms, mfu 33.75% +iter 2560: loss 2.8333, time 2086.59ms, mfu 33.74% +iter 2570: loss 2.9103, time 2095.76ms, mfu 33.72% +iter 2580: loss 2.8612, time 2070.09ms, mfu 33.74% +iter 2590: loss 2.8191, time 2077.10ms, mfu 33.75% +iter 2600: loss 2.7326, time 2089.63ms, mfu 33.74% +iter 2610: loss 2.7020, time 2068.17ms, mfu 33.76% +iter 2620: loss 2.7047, time 2095.00ms, mfu 33.74% +iter 2630: loss 3.0022, time 2082.40ms, mfu 33.74% +iter 2640: loss 3.3161, time 2085.93ms, mfu 33.73% +iter 2650: loss 3.1273, time 2084.28ms, mfu 33.73% +iter 2660: loss 2.9416, time 2096.91ms, mfu 33.71% +iter 2670: loss 2.7499, time 2079.56ms, mfu 33.72% +iter 2680: loss 3.1701, time 2088.46ms, mfu 33.71% +iter 2690: loss 2.8368, time 2081.15ms, mfu 33.71% +iter 2700: loss 2.7531, time 2099.55ms, mfu 33.69% +iter 2710: loss 3.0352, time 2079.42ms, mfu 33.70% +iter 2720: loss 2.8548, time 2082.25ms, mfu 33.70% +iter 2730: loss 2.9761, time 2073.71ms, mfu 33.72% +iter 2740: loss 3.1069, time 2095.52ms, mfu 33.70% +iter 2750: loss 2.7885, time 2105.93ms, mfu 33.67% +iter 2760: loss 2.9944, time 2073.42ms, mfu 33.69% +iter 2770: loss 3.1238, time 2092.19ms, mfu 33.68% +iter 2780: loss 2.9192, time 2088.67ms, mfu 33.67% +iter 2790: loss 3.1387, time 2093.42ms, mfu 33.66% +iter 2800: loss 3.0954, time 2092.68ms, mfu 33.65% +iter 2810: loss 3.2072, time 2093.31ms, mfu 33.64% +iter 2820: loss 2.9308, time 2083.50ms, mfu 33.65% +iter 2830: loss 2.9661, time 2090.68ms, mfu 33.65% +iter 2840: loss 2.6713, time 2076.45ms, mfu 33.67% +iter 2850: loss 2.7229, time 2075.23ms, mfu 33.69% +iter 2860: loss 3.2801, time 2084.72ms, mfu 33.69% +iter 2870: loss 2.9257, time 2102.12ms, mfu 33.66% +iter 2880: loss 2.8821, time 2082.41ms, mfu 33.67% +iter 2890: loss 2.9680, time 2079.90ms, mfu 33.68% +iter 2900: loss 3.0326, time 2074.65ms, mfu 33.70% +iter 2910: loss 2.5116, time 2080.17ms, mfu 33.71% +iter 2920: loss 3.1245, time 2065.37ms, mfu 33.74% +iter 2930: loss 3.0239, time 2092.04ms, mfu 33.72% +iter 2940: loss 3.0774, time 2081.26ms, mfu 33.73% +iter 2950: loss 2.8134, time 2075.49ms, mfu 33.74% +iter 2960: loss 2.9250, time 2097.11ms, mfu 33.71% +iter 2970: loss 3.1224, time 2079.51ms, mfu 33.72% +iter 2980: loss 2.8348, time 2084.81ms, mfu 33.72% +iter 2990: loss 3.0060, time 2109.60ms, mfu 33.68% +step 3000: train loss 2.9281, val loss 2.9812 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 3000: loss 3.0601, time 9151.48ms, mfu 31.08% +iter 3010: loss 2.7150, time 2084.09ms, mfu 31.34% +iter 3020: loss 2.7070, time 2070.79ms, mfu 31.60% +iter 3030: loss 2.7966, time 2095.53ms, mfu 31.79% +iter 3040: loss 2.9238, time 2077.16ms, mfu 32.00% +iter 3050: loss 2.6938, time 2068.39ms, mfu 32.19% +iter 3060: loss 2.8782, time 2080.72ms, mfu 32.35% +iter 3070: loss 2.9914, time 2080.87ms, mfu 32.49% +iter 3080: loss 2.7994, time 2076.33ms, mfu 32.63% +iter 3090: loss 2.9092, time 2083.13ms, mfu 32.74% +iter 3100: loss 3.0294, time 2086.56ms, mfu 32.83% +iter 3110: loss 3.1543, time 2089.83ms, mfu 32.91% +iter 3120: loss 2.8346, time 2076.61ms, mfu 33.00% +iter 3130: loss 2.6862, time 2070.32ms, mfu 33.09% +iter 3140: loss 3.1379, time 2081.03ms, mfu 33.16% +iter 3150: loss 2.5061, time 2074.49ms, mfu 33.23% +iter 3160: loss 2.6616, time 2080.76ms, mfu 33.29% +iter 3170: loss 2.9740, time 2069.50ms, mfu 33.35% +iter 3180: loss 2.6214, time 2063.92ms, mfu 33.42% +iter 3190: loss 3.0916, time 2091.78ms, mfu 33.44% +iter 3200: loss 2.9823, time 2078.97ms, mfu 33.47% +iter 3210: loss 2.7964, time 2079.92ms, mfu 33.50% +iter 3220: loss 3.0369, time 2099.36ms, mfu 33.50% +iter 3230: loss 2.4957, time 2081.40ms, mfu 33.53% +iter 3240: loss 2.8929, time 2063.21ms, mfu 33.58% +iter 3250: loss 2.8439, time 2064.79ms, mfu 33.62% +iter 3260: loss 3.0494, time 2096.57ms, mfu 33.61% +iter 3270: loss 2.9273, time 2080.16ms, mfu 33.63% +iter 3280: loss 2.9881, time 2088.97ms, mfu 33.63% +iter 3290: loss 3.3021, time 2068.27ms, mfu 33.66% +iter 3300: loss 2.7670, time 2067.16ms, mfu 33.70% +iter 3310: loss 2.9800, time 2097.20ms, mfu 33.68% +iter 3320: loss 3.2683, time 2071.10ms, mfu 33.70% +iter 3330: loss 3.1901, time 2068.84ms, mfu 33.73% +iter 3340: loss 2.9526, time 2092.34ms, mfu 33.71% +iter 3350: loss 2.8863, time 2079.09ms, mfu 33.72% +iter 3360: loss 2.8820, time 2072.72ms, mfu 33.74% +iter 3370: loss 3.0160, time 2082.08ms, mfu 33.74% +iter 3380: loss 3.1090, time 2088.08ms, mfu 33.73% +iter 3390: loss 2.9832, time 2076.39ms, mfu 33.74% +iter 3400: loss 3.0788, time 2095.70ms, mfu 33.72% +iter 3410: loss 3.0431, time 2084.59ms, mfu 33.72% +iter 3420: loss 3.0189, time 2100.32ms, mfu 33.69% +iter 3430: loss 2.8774, time 2076.34ms, mfu 33.71% +iter 3440: loss 2.7842, time 2084.61ms, mfu 33.70% +iter 3450: loss 3.1838, time 2082.34ms, mfu 33.71% +iter 3460: loss 2.8264, time 2072.92ms, mfu 33.73% +iter 3470: loss 2.9276, time 2100.59ms, mfu 33.70% +iter 3480: loss 3.0675, time 2094.04ms, mfu 33.68% +iter 3490: loss 3.2108, time 2086.26ms, mfu 33.68% +iter 3500: loss 2.5512, time 2073.21ms, mfu 33.70% +iter 3510: loss 3.1138, time 2076.55ms, mfu 33.72% +iter 3520: loss 2.9521, time 2058.69ms, mfu 33.76% +iter 3530: loss 2.6554, time 2067.60ms, mfu 33.78% +iter 3540: loss 2.8287, time 2069.90ms, mfu 33.80% +iter 3550: loss 2.9961, time 2085.06ms, mfu 33.79% +iter 3560: loss 2.7529, time 2085.13ms, mfu 33.78% +iter 3570: loss 2.8463, time 2071.72ms, mfu 33.79% +iter 3580: loss 2.8391, time 2074.18ms, mfu 33.80% +iter 3590: loss 2.6227, time 2063.66ms, mfu 33.82% +iter 3600: loss 3.0554, time 2079.61ms, mfu 33.82% +iter 3610: loss 2.6545, time 2068.14ms, mfu 33.83% +iter 3620: loss 3.1233, time 2077.02ms, mfu 33.83% +iter 3630: loss 2.8129, time 2068.63ms, mfu 33.85% +iter 3640: loss 2.6925, time 2079.76ms, mfu 33.84% +iter 3650: loss 2.7285, time 2072.39ms, mfu 33.85% +iter 3660: loss 2.9807, time 2066.40ms, mfu 33.86% +iter 3670: loss 3.0230, time 2080.69ms, mfu 33.85% +iter 3680: loss 2.6756, time 2064.03ms, mfu 33.87% +iter 3690: loss 2.8071, time 2081.33ms, mfu 33.86% +iter 3700: loss 2.6854, time 2070.27ms, mfu 33.87% +iter 3710: loss 2.9006, time 2075.38ms, mfu 33.87% +iter 3720: loss 2.9650, time 2086.78ms, mfu 33.85% +iter 3730: loss 2.8955, time 2081.14ms, mfu 33.84% +iter 3740: loss 3.1014, time 2082.24ms, mfu 33.83% +iter 3750: loss 3.1095, time 2070.20ms, mfu 33.84% +iter 3760: loss 2.9557, time 2071.86ms, mfu 33.85% +iter 3770: loss 2.8755, time 2075.61ms, mfu 33.85% +iter 3780: loss 2.7802, time 2073.01ms, mfu 33.85% +iter 3790: loss 3.0165, time 2086.68ms, mfu 33.83% +iter 3800: loss 2.9937, time 2086.96ms, mfu 33.82% +iter 3810: loss 2.9377, time 2077.09ms, mfu 33.82% +iter 3820: loss 2.9562, time 2097.84ms, mfu 33.78% +iter 3830: loss 2.8494, time 2062.39ms, mfu 33.81% +iter 3840: loss 2.8790, time 2076.07ms, mfu 33.82% +iter 3850: loss 2.9325, time 2067.72ms, mfu 33.83% +iter 3860: loss 3.0470, time 2082.84ms, mfu 33.82% +iter 3870: loss 2.6130, time 2075.21ms, mfu 33.83% +iter 3880: loss 2.3148, time 2070.67ms, mfu 33.84% +iter 3890: loss 2.8490, time 2068.97ms, mfu 33.85% +iter 3900: loss 2.5053, time 2081.71ms, mfu 33.84% +iter 3910: loss 2.6167, time 2075.10ms, mfu 33.84% +iter 3920: loss 2.8977, time 2068.89ms, mfu 33.85% +iter 3930: loss 2.9963, time 2055.53ms, mfu 33.88% +iter 3940: loss 2.7006, time 2073.11ms, mfu 33.89% +iter 3950: loss 3.0808, time 2071.63ms, mfu 33.89% +iter 3960: loss 2.7748, time 2074.35ms, mfu 33.89% +iter 3970: loss 2.9384, time 2067.47ms, mfu 33.90% +iter 3980: loss 3.1498, time 2081.32ms, mfu 33.88% +iter 3990: loss 3.2189, time 2082.19ms, mfu 33.87% +step 4000: train loss 2.8648, val loss 2.9117 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 4000: loss 2.8254, time 9065.40ms, mfu 31.26% +iter 4010: loss 2.9850, time 2068.52ms, mfu 31.53% +iter 4020: loss 3.1507, time 2066.73ms, mfu 31.77% +iter 4030: loss 3.0200, time 2070.10ms, mfu 31.99% +iter 4040: loss 2.9688, time 2061.25ms, mfu 32.20% +iter 4050: loss 2.8385, time 2062.97ms, mfu 32.39% +iter 4060: loss 3.1681, time 2061.49ms, mfu 32.56% +iter 4070: loss 2.8123, time 2065.62ms, mfu 32.70% +iter 4080: loss 2.6837, time 2082.52ms, mfu 32.80% +iter 4090: loss 2.9108, time 2066.67ms, mfu 32.92% +iter 4100: loss 2.9533, time 2074.16ms, mfu 33.02% +iter 4110: loss 2.4704, time 2074.39ms, mfu 33.10% +iter 4120: loss 3.0496, time 2075.79ms, mfu 33.18% +iter 4130: loss 2.8591, time 2069.94ms, mfu 33.25% +iter 4140: loss 2.9128, time 2077.01ms, mfu 33.31% +iter 4150: loss 3.1376, time 2076.29ms, mfu 33.36% +iter 4160: loss 3.0003, time 2061.50ms, mfu 33.44% +iter 4170: loss 2.7843, time 2064.36ms, mfu 33.50% +iter 4180: loss 2.5587, time 2066.67ms, mfu 33.55% +iter 4190: loss 2.6196, time 2064.97ms, mfu 33.59% +iter 4200: loss 2.8213, time 2064.89ms, mfu 33.64% +iter 4210: loss 3.1516, time 2062.18ms, mfu 33.68% +iter 4220: loss 2.5794, time 2079.40ms, mfu 33.69% +iter 4230: loss 3.1111, time 2068.03ms, mfu 33.72% +iter 4240: loss 2.7717, time 2067.29ms, mfu 33.75% +iter 4250: loss 2.6470, time 2063.30ms, mfu 33.78% +iter 4260: loss 2.6162, time 2074.66ms, mfu 33.79% +iter 4270: loss 2.8615, time 2071.04ms, mfu 33.80% +iter 4280: loss 2.7625, time 2061.29ms, mfu 33.83% +iter 4290: loss 3.0987, time 2059.62ms, mfu 33.86% +iter 4300: loss 2.8871, time 2067.75ms, mfu 33.87% +iter 4310: loss 2.7844, time 2055.51ms, mfu 33.90% +iter 4320: loss 2.8825, time 2054.72ms, mfu 33.93% +iter 4330: loss 2.8165, time 2076.42ms, mfu 33.92% +iter 4340: loss 3.0207, time 2062.89ms, mfu 33.93% +iter 4350: loss 2.8004, time 2073.50ms, mfu 33.93% +iter 4360: loss 2.9181, time 2074.57ms, mfu 33.92% +iter 4370: loss 2.8804, time 2069.34ms, mfu 33.93% +iter 4380: loss 2.9741, time 2074.38ms, mfu 33.92% +iter 4390: loss 2.8343, time 2063.98ms, mfu 33.93% +iter 4400: loss 2.7576, time 2063.77ms, mfu 33.94% +iter 4410: loss 2.7447, time 2075.10ms, mfu 33.93% +iter 4420: loss 2.8071, time 2080.23ms, mfu 33.92% +iter 4430: loss 2.6349, time 2066.90ms, mfu 33.93% +iter 4440: loss 2.7790, time 2079.09ms, mfu 33.91% +iter 4450: loss 2.4973, time 2074.52ms, mfu 33.91% +iter 4460: loss 2.7604, time 2054.86ms, mfu 33.94% +iter 4470: loss 2.9552, time 2061.55ms, mfu 33.95% +iter 4480: loss 2.6731, time 2060.41ms, mfu 33.97% +iter 4490: loss 2.7743, time 2072.27ms, mfu 33.96% +iter 4500: loss 2.8437, time 2064.80ms, mfu 33.97% +iter 4510: loss 2.5555, time 2064.05ms, mfu 33.97% +iter 4520: loss 2.7099, time 2058.34ms, mfu 33.99% +iter 4530: loss 2.9083, time 2062.84ms, mfu 34.00% +iter 4540: loss 2.6443, time 2063.40ms, mfu 34.00% +iter 4550: loss 2.7532, time 2059.47ms, mfu 34.01% +iter 4560: loss 2.7980, time 2064.34ms, mfu 34.02% +iter 4570: loss 2.6229, time 2062.99ms, mfu 34.02% +iter 4580: loss 2.8752, time 2055.44ms, mfu 34.04% +iter 4590: loss 2.9128, time 2066.59ms, mfu 34.03% +iter 4600: loss 2.8879, time 2053.29ms, mfu 34.05% +iter 4610: loss 2.8667, time 2075.48ms, mfu 34.03% +iter 4620: loss 2.9390, time 2066.45ms, mfu 34.03% +iter 4630: loss 2.6457, time 2065.42ms, mfu 34.03% +iter 4640: loss 3.0657, time 2078.80ms, mfu 34.00% +iter 4650: loss 2.9372, time 2067.81ms, mfu 34.00% +iter 4660: loss 2.6715, time 2074.73ms, mfu 33.99% +iter 4670: loss 2.9651, time 2063.95ms, mfu 33.99% +iter 4680: loss 2.8179, time 2063.05ms, mfu 34.00% +iter 4690: loss 2.6349, time 2066.46ms, mfu 34.00% +iter 4700: loss 2.8116, time 2061.80ms, mfu 34.01% +iter 4710: loss 2.9767, time 2075.57ms, mfu 33.99% +iter 4720: loss 2.9309, time 2063.79ms, mfu 34.00% +iter 4730: loss 2.8098, time 2067.92ms, mfu 33.99% +iter 4740: loss 2.5890, time 2057.94ms, mfu 34.01% +iter 4750: loss 2.7399, time 2067.88ms, mfu 34.01% +iter 4760: loss 2.5801, time 2071.37ms, mfu 34.00% +iter 4770: loss 2.9965, time 2062.90ms, mfu 34.00% +iter 4780: loss 2.3116, time 2065.53ms, mfu 34.00% +iter 4790: loss 2.4977, time 2070.60ms, mfu 34.00% +iter 4800: loss 3.0623, time 2053.57ms, mfu 34.02% +iter 4810: loss 2.9972, time 2078.39ms, mfu 34.00% +iter 4820: loss 2.8860, time 2062.83ms, mfu 34.00% +iter 4830: loss 2.7598, time 2070.66ms, mfu 34.00% +iter 4840: loss 2.6594, time 2077.25ms, mfu 33.98% +iter 4850: loss 2.8354, time 2073.79ms, mfu 33.97% +iter 4860: loss 2.6458, time 2059.65ms, mfu 33.98% +iter 4870: loss 2.5955, time 2055.48ms, mfu 34.00% +iter 4880: loss 2.5666, time 2069.63ms, mfu 34.00% +iter 4890: loss 2.2673, time 2078.16ms, mfu 33.98% +iter 4900: loss 2.9341, time 2063.25ms, mfu 33.99% +iter 4910: loss 2.7086, time 2076.25ms, mfu 33.97% +iter 4920: loss 2.8378, time 2059.09ms, mfu 33.99% +iter 4930: loss 2.6511, time 2059.68ms, mfu 34.00% +iter 4940: loss 2.8945, time 2070.19ms, mfu 33.99% +iter 4950: loss 2.9020, time 2056.77ms, mfu 34.01% +iter 4960: loss 2.6619, time 2065.04ms, mfu 34.01% +iter 4970: loss 3.1589, time 2063.18ms, mfu 34.01% +iter 4980: loss 2.7789, time 2055.75ms, mfu 34.03% +iter 4990: loss 3.0120, time 2066.03ms, mfu 34.03% +step 5000: train loss 2.8030, val loss 2.8569 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 5000: loss 2.8127, time 9037.83ms, mfu 31.40% +iter 5010: loss 2.7815, time 2060.24ms, mfu 31.67% +iter 5020: loss 2.5290, time 2063.72ms, mfu 31.91% +iter 5030: loss 2.4893, time 2060.45ms, mfu 32.13% +iter 5040: loss 2.9599, time 2052.24ms, mfu 32.34% +iter 5050: loss 2.9941, time 2061.26ms, mfu 32.51% +iter 5060: loss 2.4154, time 2072.39ms, mfu 32.65% +iter 5070: loss 2.5486, time 2064.66ms, mfu 32.79% +iter 5080: loss 2.3005, time 2068.79ms, mfu 32.91% +iter 5090: loss 2.7054, time 2076.96ms, mfu 33.00% +iter 5100: loss 2.6714, time 2059.32ms, mfu 33.11% +iter 5110: loss 3.0298, time 2061.06ms, mfu 33.21% +iter 5120: loss 2.6713, time 2057.03ms, mfu 33.30% +iter 5130: loss 3.1154, time 2051.57ms, mfu 33.40% +iter 5140: loss 3.0023, time 2052.74ms, mfu 33.48% +iter 5150: loss 2.9078, time 2068.32ms, mfu 33.53% +iter 5160: loss 2.8363, time 2063.19ms, mfu 33.58% +iter 5170: loss 2.8335, time 2059.94ms, mfu 33.63% +iter 5180: loss 2.4896, time 2078.49ms, mfu 33.65% +iter 5190: loss 3.0122, time 2061.53ms, mfu 33.69% +iter 5200: loss 2.6836, time 2068.42ms, mfu 33.72% +iter 5210: loss 2.1724, time 2069.70ms, mfu 33.74% +iter 5220: loss 2.6539, time 2058.02ms, mfu 33.78% +iter 5230: loss 2.6254, time 2059.58ms, mfu 33.82% +iter 5240: loss 2.7223, time 2072.75ms, mfu 33.82% +iter 5250: loss 2.6877, time 2071.79ms, mfu 33.83% +iter 5260: loss 2.5440, time 2066.79ms, mfu 33.85% +iter 5270: loss 2.8340, time 2062.85ms, mfu 33.87% +iter 5280: loss 2.6948, time 2075.01ms, mfu 33.87% +iter 5290: loss 2.7700, time 2063.62ms, mfu 33.89% +iter 5300: loss 2.7649, time 2061.02ms, mfu 33.91% +iter 5310: loss 2.4702, time 2062.59ms, mfu 33.92% +iter 5320: loss 2.7136, time 2061.96ms, mfu 33.94% +iter 5330: loss 2.7587, time 2068.73ms, mfu 33.94% +iter 5340: loss 2.8885, time 2061.52ms, mfu 33.95% +iter 5350: loss 2.9620, time 2058.17ms, mfu 33.97% +iter 5360: loss 2.5015, time 2059.90ms, mfu 33.99% +iter 5370: loss 2.6350, time 2052.63ms, mfu 34.01% +iter 5380: loss 2.8041, time 2065.43ms, mfu 34.01% +iter 5390: loss 2.6625, time 2053.61ms, mfu 34.03% +iter 5400: loss 2.9459, time 2060.95ms, mfu 34.04% +iter 5410: loss 2.8226, time 2071.94ms, mfu 34.02% +iter 5420: loss 2.5892, time 2061.57ms, mfu 34.03% +iter 5430: loss 2.8050, time 2059.07ms, mfu 34.04% +iter 5440: loss 2.5097, time 2059.87ms, mfu 34.05% +iter 5450: loss 2.6736, time 2062.97ms, mfu 34.05% +iter 5460: loss 2.8148, time 2062.16ms, mfu 34.05% +iter 5470: loss 2.7791, time 2059.89ms, mfu 34.05% +iter 5480: loss 2.7999, time 2059.54ms, mfu 34.06% +iter 5490: loss 3.2725, time 2068.44ms, mfu 34.05% +iter 5500: loss 2.8687, time 2063.56ms, mfu 34.05% +iter 5510: loss 2.7539, time 2060.16ms, mfu 34.06% +iter 5520: loss 3.0550, time 2068.15ms, mfu 34.05% +iter 5530: loss 2.7933, time 2066.53ms, mfu 34.04% +iter 5540: loss 2.6721, time 2061.05ms, mfu 34.05% +iter 5550: loss 2.9439, time 2073.22ms, mfu 34.03% +iter 5560: loss 2.6863, time 2062.04ms, mfu 34.04% +iter 5570: loss 2.6809, time 2059.43ms, mfu 34.04% +iter 5580: loss 2.5534, time 2058.25ms, mfu 34.05% +iter 5590: loss 2.7018, time 2064.93ms, mfu 34.05% +iter 5600: loss 2.8506, time 2062.74ms, mfu 34.05% +iter 5610: loss 3.0196, time 2082.08ms, mfu 34.02% +iter 5620: loss 2.6995, time 2072.20ms, mfu 34.01% +iter 5630: loss 3.0107, time 2059.27ms, mfu 34.02% +iter 5640: loss 2.4634, time 2059.52ms, mfu 34.03% +iter 5650: loss 2.7511, time 2055.32ms, mfu 34.04% +iter 5660: loss 2.6317, time 2060.02ms, mfu 34.05% +iter 5670: loss 2.8164, time 2064.27ms, mfu 34.05% +iter 5680: loss 2.7845, time 2066.83ms, mfu 34.04% +iter 5690: loss 2.7309, time 2053.46ms, mfu 34.06% +iter 5700: loss 2.6763, time 2064.71ms, mfu 34.06% +iter 5710: loss 2.4668, time 2059.13ms, mfu 34.06% +iter 5720: loss 2.8844, time 2064.02ms, mfu 34.06% +iter 5730: loss 2.5106, time 2058.15ms, mfu 34.07% +iter 5740: loss 2.7284, time 2070.48ms, mfu 34.06% +iter 5750: loss 2.8526, time 2071.04ms, mfu 34.04% +iter 5760: loss 2.8057, time 2062.77ms, mfu 34.04% +iter 5770: loss 2.7825, time 2072.32ms, mfu 34.03% +iter 5780: loss 2.7251, time 2055.06ms, mfu 34.05% +iter 5790: loss 2.7657, time 2066.28ms, mfu 34.04% +iter 5800: loss 2.8040, time 2060.52ms, mfu 34.05% +iter 5810: loss 2.8236, time 2060.50ms, mfu 34.05% +iter 5820: loss 2.5736, time 2074.36ms, mfu 34.03% +iter 5830: loss 2.6721, time 2061.64ms, mfu 34.04% +iter 5840: loss 2.7983, time 2063.89ms, mfu 34.04% +iter 5850: loss 2.8966, time 2061.17ms, mfu 34.04% +iter 5860: loss 2.8243, time 2059.42ms, mfu 34.05% +iter 5870: loss 2.7769, time 2071.73ms, mfu 34.04% +iter 5880: loss 2.8519, time 2064.61ms, mfu 34.04% +iter 5890: loss 2.9137, time 2059.39ms, mfu 34.04% +iter 5900: loss 2.6517, time 2070.95ms, mfu 34.03% +iter 5910: loss 2.9979, time 2076.52ms, mfu 34.01% +iter 5920: loss 2.5328, time 2070.57ms, mfu 34.00% +iter 5930: loss 2.4874, time 2055.01ms, mfu 34.02% +iter 5940: loss 2.6012, time 2069.76ms, mfu 34.02% +iter 5950: loss 2.7926, time 2058.31ms, mfu 34.03% +iter 5960: loss 2.8399, time 2062.83ms, mfu 34.03% +iter 5970: loss 2.4334, time 2063.89ms, mfu 34.03% +iter 5980: loss 2.4871, time 2061.36ms, mfu 34.04% +iter 5990: loss 2.7603, time 2075.80ms, mfu 34.02% +step 6000: train loss 2.7348, val loss 2.7800 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 6000: loss 2.7937, time 9040.42ms, mfu 31.39% +iter 6010: loss 2.8561, time 2051.49ms, mfu 31.68% +iter 6020: loss 2.9252, time 2078.77ms, mfu 31.89% +iter 6030: loss 2.7826, time 2057.91ms, mfu 32.12% +iter 6040: loss 2.6314, time 2064.55ms, mfu 32.31% +iter 6050: loss 2.7288, time 2066.30ms, mfu 32.48% +iter 6060: loss 2.8380, time 2071.90ms, mfu 32.62% +iter 6070: loss 2.7656, time 2062.80ms, mfu 32.76% +iter 6080: loss 2.8201, time 2063.82ms, mfu 32.89% +iter 6090: loss 2.9309, time 2065.50ms, mfu 33.00% +iter 6100: loss 2.7605, time 2068.20ms, mfu 33.10% +iter 6110: loss 2.8659, time 2054.06ms, mfu 33.21% +iter 6120: loss 2.8095, time 2064.98ms, mfu 33.29% +iter 6130: loss 2.6604, time 2065.05ms, mfu 33.37% +iter 6140: loss 2.5700, time 2061.70ms, mfu 33.44% +iter 6150: loss 2.5395, time 2055.06ms, mfu 33.51% +iter 6160: loss 2.7895, time 2062.48ms, mfu 33.57% +iter 6170: loss 2.7476, time 2054.03ms, mfu 33.63% +iter 6180: loss 2.3769, time 2056.50ms, mfu 33.68% +iter 6190: loss 2.7728, time 2062.22ms, mfu 33.72% +iter 6200: loss 2.8410, time 2056.67ms, mfu 33.77% +iter 6210: loss 2.4412, time 2059.55ms, mfu 33.80% +iter 6220: loss 2.8561, time 2063.53ms, mfu 33.83% +iter 6230: loss 2.7499, time 2064.29ms, mfu 33.85% +iter 6240: loss 3.0280, time 2059.13ms, mfu 33.87% +iter 6250: loss 2.8489, time 2060.44ms, mfu 33.90% +iter 6260: loss 2.5052, time 2057.64ms, mfu 33.92% +iter 6270: loss 2.4484, time 2068.48ms, mfu 33.93% +iter 6280: loss 2.8751, time 2059.17ms, mfu 33.95% +iter 6290: loss 2.8386, time 2072.31ms, mfu 33.94% +iter 6300: loss 2.8937, time 2059.09ms, mfu 33.96% +iter 6310: loss 2.9598, time 2074.39ms, mfu 33.95% +iter 6320: loss 2.9603, time 2070.23ms, mfu 33.95% +iter 6330: loss 2.5865, time 2066.53ms, mfu 33.95% +iter 6340: loss 2.8333, time 2059.97ms, mfu 33.97% +iter 6350: loss 3.0807, time 2073.29ms, mfu 33.96% +iter 6360: loss 2.6060, time 2060.75ms, mfu 33.97% +iter 6370: loss 2.7680, time 2062.30ms, mfu 33.98% +iter 6380: loss 2.7461, time 2067.87ms, mfu 33.98% +iter 6390: loss 2.5055, time 2047.10ms, mfu 34.02% +iter 6400: loss 2.8759, time 2060.80ms, mfu 34.02% +iter 6410: loss 2.7861, time 2071.88ms, mfu 34.01% +iter 6420: loss 2.8208, time 2073.05ms, mfu 34.00% +iter 6430: loss 2.4515, time 2065.19ms, mfu 34.00% +iter 6440: loss 2.9032, time 2067.03ms, mfu 34.00% +iter 6450: loss 2.7120, time 2070.47ms, mfu 33.99% +iter 6460: loss 2.6434, time 2064.10ms, mfu 34.00% +iter 6470: loss 2.7141, time 2059.00ms, mfu 34.01% +iter 6480: loss 2.3601, time 2056.58ms, mfu 34.03% +iter 6490: loss 2.7366, time 2065.14ms, mfu 34.03% +iter 6500: loss 2.7211, time 2078.13ms, mfu 34.00% +iter 6510: loss 2.8686, time 2072.11ms, mfu 33.99% +iter 6520: loss 2.8704, time 2073.99ms, mfu 33.98% +iter 6530: loss 2.7231, time 2069.10ms, mfu 33.98% +iter 6540: loss 2.5677, time 2061.74ms, mfu 33.99% +iter 6550: loss 2.9674, time 2061.93ms, mfu 34.00% +iter 6560: loss 2.6548, time 2062.25ms, mfu 34.01% +iter 6570: loss 2.8212, time 2065.32ms, mfu 34.01% +iter 6580: loss 2.8856, time 2069.61ms, mfu 34.00% +iter 6590: loss 2.5839, time 2078.52ms, mfu 33.98% +iter 6600: loss 2.7059, time 2084.07ms, mfu 33.95% +iter 6610: loss 2.8822, time 2072.34ms, mfu 33.95% +iter 6620: loss 2.6400, time 2052.48ms, mfu 33.98% +iter 6630: loss 3.0263, time 2064.04ms, mfu 33.98% +iter 6640: loss 2.8466, time 2076.14ms, mfu 33.97% +iter 6650: loss 2.6849, time 2068.52ms, mfu 33.97% +iter 6660: loss 3.0016, time 2059.73ms, mfu 33.98% +iter 6670: loss 2.7598, time 2058.34ms, mfu 34.00% +iter 6680: loss 2.6723, time 2069.23ms, mfu 33.99% +iter 6690: loss 2.6609, time 2067.45ms, mfu 33.99% +iter 6700: loss 2.7810, time 2072.90ms, mfu 33.98% +iter 6710: loss 2.5192, time 2074.93ms, mfu 33.97% +iter 6720: loss 2.3313, time 2066.86ms, mfu 33.97% +iter 6730: loss 2.8297, time 2079.93ms, mfu 33.95% +iter 6740: loss 2.3553, time 2071.79ms, mfu 33.95% +iter 6750: loss 2.7652, time 2066.70ms, mfu 33.95% +iter 6760: loss 2.7459, time 2057.07ms, mfu 33.97% +iter 6770: loss 2.5465, time 2065.00ms, mfu 33.98% +iter 6780: loss 2.2827, time 2064.57ms, mfu 33.98% +iter 6790: loss 2.7563, time 2060.09ms, mfu 34.00% +iter 6800: loss 2.5942, time 2067.08ms, mfu 34.00% +iter 6810: loss 2.4001, time 2075.86ms, mfu 33.98% +iter 6820: loss 2.7329, time 2087.28ms, mfu 33.95% +iter 6830: loss 2.6670, time 2080.82ms, mfu 33.93% +iter 6840: loss 2.7223, time 2073.69ms, mfu 33.92% +iter 6850: loss 2.4119, time 2067.73ms, mfu 33.93% +iter 6860: loss 2.4829, time 2066.01ms, mfu 33.94% +iter 6870: loss 2.7631, time 2066.09ms, mfu 33.94% +iter 6880: loss 2.6818, time 2068.91ms, mfu 33.95% +iter 6890: loss 2.8367, time 2078.08ms, mfu 33.93% +iter 6900: loss 2.9548, time 2065.18ms, mfu 33.94% +iter 6910: loss 2.7495, time 2058.83ms, mfu 33.96% +iter 6920: loss 2.6917, time 2065.74ms, mfu 33.96% +iter 6930: loss 3.0313, time 2065.94ms, mfu 33.97% +iter 6940: loss 2.9945, time 2071.37ms, mfu 33.96% +iter 6950: loss 2.9768, time 2068.13ms, mfu 33.96% +iter 6960: loss 2.2144, time 2062.39ms, mfu 33.97% +iter 6970: loss 2.5947, time 2068.89ms, mfu 33.97% +iter 6980: loss 2.1534, time 2074.16ms, mfu 33.96% +iter 6990: loss 2.8031, time 2066.51ms, mfu 33.97% +step 7000: train loss 2.7199, val loss 2.7451 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 7000: loss 2.6378, time 8990.41ms, mfu 31.35% +iter 7010: loss 2.3748, time 2058.15ms, mfu 31.63% +iter 7020: loss 2.9583, time 2065.35ms, mfu 31.87% +iter 7030: loss 2.8537, time 2073.01ms, mfu 32.07% +iter 7040: loss 2.6749, time 2068.63ms, mfu 32.26% +iter 7050: loss 2.8980, time 2066.23ms, mfu 32.43% +iter 7060: loss 2.6195, time 2070.61ms, mfu 32.58% +iter 7070: loss 2.6558, time 2060.36ms, mfu 32.74% +iter 7080: loss 2.7641, time 2064.97ms, mfu 32.86% +iter 7090: loss 2.4308, time 2063.20ms, mfu 32.98% +iter 7100: loss 2.4519, time 2074.19ms, mfu 33.07% +iter 7110: loss 2.7545, time 2067.47ms, mfu 33.16% +iter 7120: loss 2.8048, time 2077.71ms, mfu 33.23% +iter 7130: loss 2.9350, time 2064.97ms, mfu 33.31% +iter 7140: loss 2.6718, time 2060.87ms, mfu 33.39% +iter 7150: loss 2.8571, time 2054.19ms, mfu 33.47% +iter 7160: loss 2.7824, time 2063.02ms, mfu 33.53% +iter 7170: loss 2.7879, time 2062.80ms, mfu 33.58% +iter 7180: loss 2.7715, time 2056.40ms, mfu 33.64% +iter 7190: loss 2.6512, time 2047.85ms, mfu 33.71% +iter 7200: loss 2.6282, time 2068.08ms, mfu 33.73% +iter 7210: loss 2.6781, time 2080.54ms, mfu 33.74% +iter 7220: loss 2.3595, time 2061.42ms, mfu 33.77% +iter 7230: loss 2.8317, time 2056.37ms, mfu 33.81% +iter 7240: loss 2.8356, time 2069.20ms, mfu 33.82% +iter 7250: loss 2.6854, time 2068.84ms, mfu 33.84% +iter 7260: loss 3.0189, time 2066.52ms, mfu 33.85% +iter 7270: loss 2.3311, time 2062.97ms, mfu 33.87% +iter 7280: loss 2.6729, time 2073.55ms, mfu 33.88% +iter 7290: loss 2.4065, time 2062.07ms, mfu 33.89% +iter 7300: loss 2.7326, time 2066.48ms, mfu 33.91% +iter 7310: loss 2.8379, time 2071.99ms, mfu 33.91% +iter 7320: loss 2.9087, time 2072.50ms, mfu 33.91% +iter 7330: loss 2.8323, time 2061.21ms, mfu 33.92% +iter 7340: loss 2.5362, time 2066.00ms, mfu 33.93% +iter 7350: loss 2.6603, time 2055.68ms, mfu 33.96% +iter 7360: loss 2.4863, time 2072.29ms, mfu 33.95% +iter 7370: loss 2.6952, time 2059.11ms, mfu 33.97% +iter 7380: loss 2.9313, time 2063.75ms, mfu 33.98% +iter 7390: loss 2.8076, time 2080.66ms, mfu 33.95% +iter 7400: loss 2.7884, time 2062.11ms, mfu 33.97% +iter 7410: loss 2.2030, time 2060.66ms, mfu 33.98% +iter 7420: loss 2.8226, time 2051.75ms, mfu 34.01% +iter 7430: loss 2.7441, time 2056.97ms, mfu 34.02% +iter 7440: loss 2.6438, time 2065.42ms, mfu 34.02% +iter 7450: loss 2.8727, time 2064.33ms, mfu 34.02% +iter 7460: loss 2.3760, time 2063.12ms, mfu 34.02% +iter 7470: loss 2.6915, time 2072.23ms, mfu 34.01% +iter 7480: loss 2.3677, time 2071.93ms, mfu 34.00% +iter 7490: loss 2.8116, time 2063.97ms, mfu 34.01% +iter 7500: loss 2.8886, time 2063.60ms, mfu 34.01% +iter 7510: loss 2.5551, time 2062.11ms, mfu 34.02% +iter 7520: loss 2.5777, time 2056.48ms, mfu 34.03% +iter 7530: loss 2.7493, time 2078.79ms, mfu 34.01% +iter 7540: loss 2.6058, time 2052.31ms, mfu 34.03% +iter 7550: loss 2.7032, time 2065.04ms, mfu 34.03% +iter 7560: loss 2.5870, time 2056.16ms, mfu 34.04% +iter 7570: loss 2.7618, time 2056.58ms, mfu 34.06% +iter 7580: loss 2.6164, time 2076.06ms, mfu 34.03% +iter 7590: loss 2.7801, time 2064.83ms, mfu 34.03% +iter 7600: loss 2.7909, time 2062.50ms, mfu 34.04% +iter 7610: loss 2.8960, time 2058.75ms, mfu 34.05% +iter 7620: loss 2.8551, time 2055.24ms, mfu 34.06% +iter 7630: loss 2.7798, time 2069.51ms, mfu 34.05% +iter 7640: loss 2.6871, time 2072.71ms, mfu 34.03% +iter 7650: loss 3.0158, time 2062.60ms, mfu 34.04% +iter 7660: loss 2.8931, time 2065.47ms, mfu 34.03% +iter 7670: loss 2.3321, time 2062.72ms, mfu 34.04% +iter 7680: loss 2.2162, time 2054.44ms, mfu 34.05% +iter 7690: loss 2.7187, time 2073.70ms, mfu 34.04% +iter 7700: loss 3.0990, time 2075.42ms, mfu 34.02% +iter 7710: loss 3.0998, time 2082.62ms, mfu 33.99% +iter 7720: loss 2.8787, time 2061.16ms, mfu 34.00% +iter 7730: loss 2.7363, time 2062.32ms, mfu 34.01% +iter 7740: loss 2.7613, time 2058.37ms, mfu 34.02% +iter 7750: loss 2.4897, time 2068.29ms, mfu 34.01% +iter 7760: loss 2.8700, time 2069.80ms, mfu 34.01% +iter 7770: loss 2.6892, time 2063.09ms, mfu 34.01% +iter 7780: loss 2.5989, time 2064.17ms, mfu 34.01% +iter 7790: loss 2.5908, time 2050.02ms, mfu 34.04% +iter 7800: loss 2.5284, time 2078.61ms, mfu 34.02% +iter 7810: loss 2.4190, time 2067.64ms, mfu 34.01% +iter 7820: loss 2.7623, time 2060.49ms, mfu 34.02% +iter 7830: loss 2.6298, time 2072.14ms, mfu 34.01% +iter 7840: loss 2.5627, time 2054.42ms, mfu 34.03% +iter 7850: loss 2.6072, time 2069.88ms, mfu 34.02% +iter 7860: loss 2.7139, time 2059.51ms, mfu 34.03% +iter 7870: loss 2.5188, time 2068.34ms, mfu 34.02% +iter 7880: loss 2.2874, time 2070.74ms, mfu 34.01% +iter 7890: loss 2.3243, time 2073.84ms, mfu 34.00% +iter 7900: loss 2.6967, time 2062.62ms, mfu 34.01% +iter 7910: loss 2.6026, time 2062.92ms, mfu 34.01% +iter 7920: loss 2.7556, time 2077.83ms, mfu 33.99% +iter 7930: loss 2.8487, time 2064.57ms, mfu 34.00% +iter 7940: loss 2.5658, time 2071.63ms, mfu 33.99% +iter 7950: loss 2.5790, time 2065.73ms, mfu 33.99% +iter 7960: loss 2.8130, time 2065.42ms, mfu 33.99% +iter 7970: loss 2.5048, time 2064.73ms, mfu 34.00% +iter 7980: loss 2.6489, time 2077.72ms, mfu 33.98% +iter 7990: loss 2.6368, time 2064.54ms, mfu 33.98% +step 8000: train loss 2.6810, val loss 2.7095 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 8000: loss 2.4861, time 12696.51ms, mfu 31.14% +iter 8010: loss 2.7566, time 2065.55ms, mfu 31.43% +iter 8020: loss 2.7151, time 2069.67ms, mfu 31.68% +iter 8030: loss 2.8548, time 2070.45ms, mfu 31.90% +iter 8040: loss 2.7300, time 2069.78ms, mfu 32.11% +iter 8050: loss 2.1881, time 2060.55ms, mfu 32.31% +iter 8060: loss 2.5641, time 2072.43ms, mfu 32.47% +iter 8070: loss 2.7639, time 2056.83ms, mfu 32.64% +iter 8080: loss 2.7717, time 2062.66ms, mfu 32.78% +iter 8090: loss 2.2702, time 2061.05ms, mfu 32.91% +iter 8100: loss 2.4686, time 2060.00ms, mfu 33.03% +iter 8110: loss 2.6002, time 2058.27ms, mfu 33.14% +iter 8120: loss 2.7896, time 2064.73ms, mfu 33.23% +iter 8130: loss 2.5712, time 2077.51ms, mfu 33.29% +iter 8140: loss 2.6600, time 2061.58ms, mfu 33.37% +iter 8150: loss 2.5645, time 2064.34ms, mfu 33.43% +iter 8160: loss 2.7058, time 2070.11ms, mfu 33.48% +iter 8170: loss 2.6112, time 2065.09ms, mfu 33.54% +iter 8180: loss 2.8279, time 2056.39ms, mfu 33.60% +iter 8190: loss 2.7758, time 2061.77ms, mfu 33.65% +iter 8200: loss 2.8485, time 2073.25ms, mfu 33.67% +iter 8210: loss 2.7109, time 2070.40ms, mfu 33.70% +iter 8220: loss 2.7037, time 2065.21ms, mfu 33.73% +iter 8230: loss 2.6159, time 2066.61ms, mfu 33.76% +iter 8240: loss 2.4814, time 2059.59ms, mfu 33.79% +iter 8250: loss 2.5850, time 2065.16ms, mfu 33.82% +iter 8260: loss 2.7522, time 2065.04ms, mfu 33.84% +iter 8270: loss 2.8145, time 2071.00ms, mfu 33.84% +iter 8280: loss 2.7633, time 2061.92ms, mfu 33.87% +iter 8290: loss 2.8925, time 2072.70ms, mfu 33.87% +iter 8300: loss 2.4420, time 2072.57ms, mfu 33.87% +iter 8310: loss 2.5096, time 2062.87ms, mfu 33.89% +iter 8320: loss 2.3306, time 2058.83ms, mfu 33.92% +iter 8330: loss 2.7891, time 2063.07ms, mfu 33.93% +iter 8340: loss 2.4590, time 2068.68ms, mfu 33.93% +iter 8350: loss 2.6073, time 2061.93ms, mfu 33.95% +iter 8360: loss 2.6111, time 2064.37ms, mfu 33.96% +iter 8370: loss 2.9273, time 2060.10ms, mfu 33.97% +iter 8380: loss 2.6358, time 2068.16ms, mfu 33.97% +iter 8390: loss 2.4288, time 2073.60ms, mfu 33.96% +iter 8400: loss 2.5959, time 2066.53ms, mfu 33.97% +iter 8410: loss 2.5588, time 2062.63ms, mfu 33.98% +iter 8420: loss 2.7146, time 2063.75ms, mfu 33.98% +iter 8430: loss 2.7121, time 2070.44ms, mfu 33.98% +iter 8440: loss 2.1057, time 2065.77ms, mfu 33.98% +iter 8450: loss 2.8603, time 2077.68ms, mfu 33.96% +iter 8460: loss 2.6322, time 2067.53ms, mfu 33.97% +iter 8470: loss 2.8530, time 2073.25ms, mfu 33.96% +iter 8480: loss 2.8399, time 2055.41ms, mfu 33.98% +iter 8490: loss 2.8382, time 2066.71ms, mfu 33.98% +iter 8500: loss 2.4481, time 2067.90ms, mfu 33.98% +iter 8510: loss 2.3293, time 2068.23ms, mfu 33.98% +iter 8520: loss 2.5659, time 2056.64ms, mfu 34.00% +iter 8530: loss 2.8179, time 2066.04ms, mfu 34.00% +iter 8540: loss 2.6658, time 2058.01ms, mfu 34.01% +iter 8550: loss 2.7417, time 2056.76ms, mfu 34.03% +iter 8560: loss 2.5201, time 2068.36ms, mfu 34.02% +iter 8570: loss 2.4454, time 2058.80ms, mfu 34.03% +iter 8580: loss 2.7683, time 2066.59ms, mfu 34.03% +iter 8590: loss 2.5503, time 2057.12ms, mfu 34.04% +iter 8600: loss 2.5039, time 2073.37ms, mfu 34.03% +iter 8610: loss 2.6278, time 2069.51ms, mfu 34.02% +iter 8620: loss 2.5305, time 2064.83ms, mfu 34.02% +iter 8630: loss 2.8835, time 2074.82ms, mfu 34.00% +iter 8640: loss 2.6614, time 2068.06ms, mfu 34.00% +iter 8650: loss 2.7330, time 2073.04ms, mfu 33.99% +iter 8660: loss 2.0691, time 2072.65ms, mfu 33.98% +iter 8670: loss 2.8066, time 2068.17ms, mfu 33.98% +iter 8680: loss 2.2184, time 2057.42ms, mfu 34.00% +iter 8690: loss 2.6948, time 2067.88ms, mfu 33.99% +iter 8700: loss 2.6029, time 2067.33ms, mfu 33.99% +iter 8710: loss 2.4013, time 2057.17ms, mfu 34.01% +iter 8720: loss 2.6765, time 2067.40ms, mfu 34.01% +iter 8730: loss 2.5796, time 2074.69ms, mfu 33.99% +iter 8740: loss 2.4355, time 2060.88ms, mfu 34.00% +iter 8750: loss 2.4995, time 2064.73ms, mfu 34.00% +iter 8760: loss 2.7665, time 2073.26ms, mfu 33.99% +iter 8770: loss 3.0022, time 2061.41ms, mfu 34.00% +iter 8780: loss 2.8570, time 2062.83ms, mfu 34.01% +iter 8790: loss 2.4886, time 2067.07ms, mfu 34.01% +iter 8800: loss 2.5721, time 2062.35ms, mfu 34.01% +iter 8810: loss 2.7232, time 2060.39ms, mfu 34.02% +iter 8820: loss 2.6539, time 2063.56ms, mfu 34.02% +iter 8830: loss 2.6905, time 2064.49ms, mfu 34.02% +iter 8840: loss 2.1432, time 2064.24ms, mfu 34.03% +iter 8850: loss 2.6969, time 2067.57ms, mfu 34.02% +iter 8860: loss 2.6095, time 2069.66ms, mfu 34.01% +iter 8870: loss 2.8346, time 2058.43ms, mfu 34.03% +iter 8880: loss 2.2024, time 2060.10ms, mfu 34.03% +iter 8890: loss 2.5171, time 2077.34ms, mfu 34.01% +iter 8900: loss 2.3834, time 2048.69ms, mfu 34.04% +iter 8910: loss 2.4685, time 2065.86ms, mfu 34.04% +iter 8920: loss 2.7455, time 2063.03ms, mfu 34.04% +iter 8930: loss 2.0023, time 2065.21ms, mfu 34.04% +iter 8940: loss 2.5953, time 2061.96ms, mfu 34.04% +iter 8950: loss 2.4106, time 2062.46ms, mfu 34.04% +iter 8960: loss 2.4227, time 2065.19ms, mfu 34.04% +iter 8970: loss 2.8596, time 2071.85ms, mfu 34.03% +iter 8980: loss 2.3502, time 2061.79ms, mfu 34.03% +iter 8990: loss 2.9678, time 2058.52ms, mfu 34.04% +step 9000: train loss 2.6725, val loss 2.7046 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 9000: loss 2.6708, time 9067.21ms, mfu 31.41% +iter 9010: loss 2.3761, time 2059.76ms, mfu 31.68% +iter 9020: loss 2.9535, time 2065.20ms, mfu 31.92% +iter 9030: loss 2.8212, time 2062.28ms, mfu 32.13% +iter 9040: loss 2.7287, time 2067.86ms, mfu 32.32% +iter 9050: loss 2.7450, time 2075.15ms, mfu 32.47% +iter 9060: loss 2.8350, time 2059.45ms, mfu 32.63% +iter 9070: loss 2.8669, time 2062.13ms, mfu 32.78% +iter 9080: loss 2.4298, time 2059.04ms, mfu 32.91% +iter 9090: loss 2.7568, time 2075.61ms, mfu 33.01% +iter 9100: loss 2.7852, time 2067.10ms, mfu 33.10% +iter 9110: loss 2.3842, time 2064.67ms, mfu 33.20% +iter 9120: loss 2.4844, time 2069.41ms, mfu 33.27% +iter 9130: loss 2.7053, time 2053.77ms, mfu 33.37% +iter 9140: loss 2.5114, time 2071.96ms, mfu 33.42% +iter 9150: loss 2.7582, time 2069.12ms, mfu 33.47% +iter 9160: loss 2.8697, time 2052.09ms, mfu 33.55% +iter 9170: loss 2.7488, time 2057.87ms, mfu 33.61% +iter 9180: loss 2.5352, time 2058.42ms, mfu 33.66% +iter 9190: loss 2.7650, time 2071.84ms, mfu 33.69% +iter 9200: loss 2.5589, time 2060.48ms, mfu 33.73% +iter 9210: loss 2.8538, time 2055.66ms, mfu 33.77% +iter 9220: loss 2.6922, time 2073.17ms, mfu 33.78% +iter 9230: loss 2.7160, time 2058.79ms, mfu 33.82% +iter 9240: loss 2.7395, time 2063.07ms, mfu 33.84% +iter 9250: loss 2.5759, time 2068.71ms, mfu 33.85% +iter 9260: loss 2.8481, time 2058.15ms, mfu 33.88% +iter 9270: loss 2.5700, time 2063.53ms, mfu 33.90% +iter 9280: loss 2.7680, time 2060.54ms, mfu 33.92% +iter 9290: loss 2.5290, time 2069.19ms, mfu 33.92% +iter 9300: loss 2.3819, time 2054.04ms, mfu 33.95% +iter 9310: loss 2.4058, time 2082.99ms, mfu 33.93% +iter 9320: loss 2.8427, time 2060.19ms, mfu 33.95% +iter 9330: loss 2.5775, time 2072.85ms, mfu 33.94% +iter 9340: loss 2.5913, time 2067.09ms, mfu 33.95% +iter 9350: loss 3.0575, time 2059.60ms, mfu 33.96% +iter 9360: loss 2.5959, time 2063.73ms, mfu 33.97% +iter 9370: loss 2.9189, time 2061.32ms, mfu 33.98% +iter 9380: loss 3.0476, time 2058.14ms, mfu 34.00% +iter 9390: loss 2.8034, time 2072.76ms, mfu 33.99% +iter 9400: loss 2.6706, time 2068.00ms, mfu 33.99% +iter 9410: loss 2.6080, time 2057.80ms, mfu 34.00% +iter 9420: loss 2.6711, time 2069.01ms, mfu 34.00% +iter 9430: loss 2.5529, time 2058.13ms, mfu 34.01% +iter 9440: loss 2.5636, time 2069.97ms, mfu 34.00% +iter 9450: loss 2.4920, time 2059.26ms, mfu 34.02% +iter 9460: loss 2.5836, time 2079.10ms, mfu 33.99% +iter 9470: loss 2.6364, time 2054.41ms, mfu 34.01% +iter 9480: loss 2.6912, time 2062.91ms, mfu 34.02% +iter 9490: loss 2.6827, time 2054.98ms, mfu 34.04% +iter 9500: loss 2.7192, time 2077.00ms, mfu 34.01% +iter 9510: loss 2.8513, time 2062.92ms, mfu 34.02% +iter 9520: loss 2.7043, time 2068.20ms, mfu 34.01% +iter 9530: loss 2.5075, time 2064.79ms, mfu 34.02% +iter 9540: loss 2.8808, time 2063.65ms, mfu 34.02% +iter 9550: loss 2.6072, time 2062.87ms, mfu 34.02% +iter 9560: loss 2.6233, time 2068.67ms, mfu 34.02% +iter 9570: loss 2.6678, time 2059.40ms, mfu 34.03% +iter 9580: loss 2.3900, time 2059.01ms, mfu 34.04% +iter 9590: loss 2.6057, time 2069.21ms, mfu 34.03% +iter 9600: loss 2.7250, time 2059.72ms, mfu 34.04% +iter 9610: loss 2.7833, time 2067.90ms, mfu 34.03% +iter 9620: loss 2.4287, time 2063.40ms, mfu 34.03% +iter 9630: loss 2.6121, time 2071.65ms, mfu 34.02% +iter 9640: loss 2.6881, time 2063.89ms, mfu 34.02% +iter 9650: loss 2.5513, time 2058.03ms, mfu 34.03% +iter 9660: loss 2.5789, time 2063.16ms, mfu 34.04% +iter 9670: loss 2.5896, time 2058.13ms, mfu 34.05% +iter 9680: loss 2.9033, time 2077.73ms, mfu 34.02% +iter 9690: loss 2.5954, time 2054.36ms, mfu 34.04% +iter 9700: loss 2.7855, time 2065.53ms, mfu 34.04% +iter 9710: loss 2.7659, time 2060.47ms, mfu 34.04% +iter 9720: loss 2.9010, time 2077.33ms, mfu 34.02% +iter 9730: loss 2.5085, time 2060.36ms, mfu 34.03% +iter 9740: loss 2.6956, time 2059.17ms, mfu 34.04% +iter 9750: loss 2.5603, time 2066.01ms, mfu 34.04% +iter 9760: loss 2.2655, time 2061.20ms, mfu 34.04% +iter 9770: loss 2.3644, time 2061.35ms, mfu 34.04% +iter 9780: loss 2.6883, time 2074.92ms, mfu 34.03% +iter 9790: loss 2.6941, time 2060.25ms, mfu 34.03% +iter 9800: loss 2.6390, time 2076.50ms, mfu 34.01% +iter 9810: loss 2.5607, time 2055.54ms, mfu 34.03% +iter 9820: loss 2.5934, time 2065.19ms, mfu 34.03% +iter 9830: loss 2.4772, time 2068.38ms, mfu 34.02% +iter 9840: loss 2.4179, time 2063.82ms, mfu 34.03% +iter 9850: loss 2.5912, time 2060.43ms, mfu 34.03% +iter 9860: loss 2.4204, time 2061.86ms, mfu 34.04% +iter 9870: loss 2.7484, time 2053.80ms, mfu 34.05% +iter 9880: loss 3.0405, time 2063.22ms, mfu 34.05% +iter 9890: loss 2.5587, time 2059.87ms, mfu 34.06% +iter 9900: loss 2.5990, time 2065.75ms, mfu 34.05% +iter 9910: loss 2.6287, time 2058.45ms, mfu 34.06% +iter 9920: loss 2.4566, time 2066.16ms, mfu 34.06% +iter 9930: loss 2.9590, time 2060.24ms, mfu 34.06% +iter 9940: loss 2.5088, time 2070.80ms, mfu 34.05% +iter 9950: loss 2.6071, time 2054.06ms, mfu 34.06% +iter 9960: loss 2.6405, time 2060.80ms, mfu 34.07% +iter 9970: loss 2.6527, time 2069.78ms, mfu 34.05% +iter 9980: loss 2.2369, time 2057.63ms, mfu 34.06% +iter 9990: loss 2.9309, time 2059.60ms, mfu 34.07% +step 10000: train loss 2.6098, val loss 2.6928 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 10000: loss 2.6818, time 9066.47ms, mfu 31.44% +iter 10010: loss 2.4153, time 2060.50ms, mfu 31.70% +iter 10020: loss 2.8921, time 2064.69ms, mfu 31.93% +iter 10030: loss 2.4925, time 2049.52ms, mfu 32.17% +iter 10040: loss 2.7314, time 2056.18ms, mfu 32.37% +iter 10050: loss 2.6600, time 2066.31ms, mfu 32.53% +iter 10060: loss 2.4391, time 2057.98ms, mfu 32.69% +iter 10070: loss 2.8679, time 2079.04ms, mfu 32.80% +iter 10080: loss 2.8293, time 2060.80ms, mfu 32.93% +iter 10090: loss 2.7463, time 2074.05ms, mfu 33.03% +iter 10100: loss 2.4889, time 2059.03ms, mfu 33.14% +iter 10110: loss 2.7040, time 2070.97ms, mfu 33.22% +iter 10120: loss 2.6523, time 2073.87ms, mfu 33.28% +iter 10130: loss 2.8288, time 2058.71ms, mfu 33.37% +iter 10140: loss 2.3883, time 2069.49ms, mfu 33.42% +iter 10150: loss 2.8258, time 2062.30ms, mfu 33.49% +iter 10160: loss 2.4561, time 2067.90ms, mfu 33.54% +iter 10170: loss 2.4972, time 2054.33ms, mfu 33.60% +iter 10180: loss 2.6207, time 2065.53ms, mfu 33.64% +iter 10190: loss 2.7256, time 2056.09ms, mfu 33.70% +iter 10200: loss 2.4992, time 2065.36ms, mfu 33.73% +iter 10210: loss 2.3864, time 2062.48ms, mfu 33.76% +iter 10220: loss 2.3153, time 2059.83ms, mfu 33.80% +iter 10230: loss 2.8782, time 2057.51ms, mfu 33.83% +iter 10240: loss 2.7194, time 2053.70ms, mfu 33.87% +iter 10250: loss 2.5983, time 2052.90ms, mfu 33.91% +iter 10260: loss 2.2427, time 2064.02ms, mfu 33.92% +iter 10270: loss 2.4142, time 2060.09ms, mfu 33.94% +iter 10280: loss 2.7083, time 2062.23ms, mfu 33.95% +iter 10290: loss 2.7991, time 2064.34ms, mfu 33.96% +iter 10300: loss 2.2950, time 2074.24ms, mfu 33.95% +iter 10310: loss 2.7618, time 2064.57ms, mfu 33.96% +iter 10320: loss 2.6323, time 2057.95ms, mfu 33.98% +iter 10330: loss 2.7193, time 2060.09ms, mfu 33.99% +iter 10340: loss 2.3563, time 2061.61ms, mfu 34.00% +iter 10350: loss 2.2564, time 2064.29ms, mfu 34.00% +iter 10360: loss 2.6965, time 2063.59ms, mfu 34.01% +iter 10370: loss 2.4466, time 2058.57ms, mfu 34.02% +iter 10380: loss 2.6256, time 2071.88ms, mfu 34.01% +iter 10390: loss 2.6320, time 2070.80ms, mfu 34.00% +iter 10400: loss 2.7519, time 2053.04ms, mfu 34.02% +iter 10410: loss 2.4551, time 2060.26ms, mfu 34.03% +iter 10420: loss 2.2128, time 2059.74ms, mfu 34.04% +iter 10430: loss 2.6083, time 2055.56ms, mfu 34.05% +iter 10440: loss 2.9739, time 2067.86ms, mfu 34.04% +iter 10450: loss 2.4810, time 2059.90ms, mfu 34.05% +iter 10460: loss 2.6777, time 2047.67ms, mfu 34.08% +iter 10470: loss 2.3237, time 2056.70ms, mfu 34.09% +iter 10480: loss 2.2072, time 2061.85ms, mfu 34.08% +iter 10490: loss 2.5535, time 2055.57ms, mfu 34.09% +iter 10500: loss 2.7508, time 2071.99ms, mfu 34.08% +iter 10510: loss 2.4628, time 2060.19ms, mfu 34.08% +iter 10520: loss 2.7934, time 2058.71ms, mfu 34.08% +iter 10530: loss 2.2086, time 2059.99ms, mfu 34.09% +iter 10540: loss 2.4975, time 2057.92ms, mfu 34.09% +iter 10550: loss 2.6391, time 2068.16ms, mfu 34.08% +iter 10560: loss 2.4180, time 2068.61ms, mfu 34.07% +iter 10570: loss 2.4721, time 2058.13ms, mfu 34.07% +iter 10580: loss 2.5195, time 2072.43ms, mfu 34.06% +iter 10590: loss 2.5897, time 2063.70ms, mfu 34.06% +iter 10600: loss 2.6969, time 2057.46ms, mfu 34.06% +iter 10610: loss 2.7318, time 2062.76ms, mfu 34.06% +iter 10620: loss 2.5522, time 2062.53ms, mfu 34.06% +iter 10630: loss 2.4850, time 2068.16ms, mfu 34.06% +iter 10640: loss 2.5061, time 2059.72ms, mfu 34.06% +iter 10650: loss 2.8947, time 2062.64ms, mfu 34.06% +iter 10660: loss 2.6385, time 2064.40ms, mfu 34.06% +iter 10670: loss 2.3477, time 2070.80ms, mfu 34.05% +iter 10680: loss 2.7086, time 2069.82ms, mfu 34.03% +iter 10690: loss 2.9617, time 2059.70ms, mfu 34.04% +iter 10700: loss 2.8884, time 2053.54ms, mfu 34.06% +iter 10710: loss 2.4545, time 2063.79ms, mfu 34.06% +iter 10720: loss 2.3016, time 2071.36ms, mfu 34.04% +iter 10730: loss 2.4445, time 2069.64ms, mfu 34.03% +iter 10740: loss 2.4914, time 2061.34ms, mfu 34.04% +iter 10750: loss 2.2444, time 2061.84ms, mfu 34.04% +iter 10760: loss 2.8815, time 2059.65ms, mfu 34.05% +iter 10770: loss 2.5330, time 2055.11ms, mfu 34.06% +iter 10780: loss 2.7568, time 2051.29ms, mfu 34.08% +iter 10790: loss 2.4291, time 2055.94ms, mfu 34.09% +iter 10800: loss 2.5251, time 2071.08ms, mfu 34.07% +iter 10810: loss 2.4847, time 2059.14ms, mfu 34.08% +iter 10820: loss 2.7133, time 2073.96ms, mfu 34.06% +iter 10830: loss 2.6447, time 2073.83ms, mfu 34.04% +iter 10840: loss 2.5646, time 2064.62ms, mfu 34.04% +iter 10850: loss 2.5846, time 2056.03ms, mfu 34.05% +iter 10860: loss 2.3809, time 2058.69ms, mfu 34.06% +iter 10870: loss 2.6342, time 2066.15ms, mfu 34.05% +iter 10880: loss 2.5886, time 2062.46ms, mfu 34.06% +iter 10890: loss 2.5468, time 2058.63ms, mfu 34.06% +iter 10900: loss 2.6735, time 2076.87ms, mfu 34.04% +iter 10910: loss 2.8591, time 2071.32ms, mfu 34.03% +iter 10920: loss 2.4669, time 2058.01ms, mfu 34.04% +iter 10930: loss 2.4665, time 2061.06ms, mfu 34.04% +iter 10940: loss 2.2181, time 2079.22ms, mfu 34.02% +iter 10950: loss 2.5276, time 2060.87ms, mfu 34.03% +iter 10960: loss 2.6678, time 2067.74ms, mfu 34.02% +iter 10970: loss 2.5568, time 2060.92ms, mfu 34.03% +iter 10980: loss 2.6901, time 2072.64ms, mfu 34.01% +iter 10990: loss 2.6687, time 2057.86ms, mfu 34.03% +step 11000: train loss 2.6080, val loss 2.6397 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 11000: loss 2.4670, time 9053.47ms, mfu 31.40% +iter 11010: loss 2.6667, time 2060.61ms, mfu 31.67% +iter 11020: loss 2.3796, time 2061.45ms, mfu 31.91% +iter 11030: loss 2.8439, time 2069.49ms, mfu 32.12% +iter 11040: loss 2.5498, time 2060.99ms, mfu 32.31% +iter 11050: loss 2.9064, time 2063.02ms, mfu 32.49% +iter 11060: loss 2.5818, time 2061.54ms, mfu 32.65% +iter 11070: loss 2.8234, time 2069.07ms, mfu 32.78% +iter 11080: loss 2.7696, time 2063.83ms, mfu 32.90% +iter 11090: loss 2.9317, time 2063.52ms, mfu 33.02% +iter 11100: loss 3.0274, time 2064.01ms, mfu 33.12% +iter 11110: loss 2.8279, time 2056.49ms, mfu 33.22% +iter 11120: loss 2.7370, time 2077.21ms, mfu 33.28% +iter 11130: loss 2.5241, time 2052.89ms, mfu 33.38% +iter 11140: loss 2.5037, time 2058.84ms, mfu 33.45% +iter 11150: loss 2.5966, time 2065.71ms, mfu 33.51% +iter 11160: loss 2.6222, time 2052.02ms, mfu 33.58% +iter 11170: loss 2.6213, time 2059.76ms, mfu 33.63% +iter 11180: loss 2.5192, time 2062.17ms, mfu 33.68% +iter 11190: loss 2.2646, time 2062.04ms, mfu 33.72% +iter 11200: loss 2.5076, time 2061.29ms, mfu 33.75% +iter 11210: loss 2.4897, time 2059.35ms, mfu 33.79% +iter 11220: loss 2.4618, time 2052.26ms, mfu 33.83% +iter 11230: loss 2.4812, time 2055.71ms, mfu 33.87% +iter 11240: loss 2.6200, time 2058.84ms, mfu 33.89% +iter 11250: loss 2.6613, time 2049.23ms, mfu 33.93% +iter 11260: loss 2.4928, time 2061.67ms, mfu 33.95% +iter 11270: loss 3.0143, time 2060.88ms, mfu 33.96% +iter 11280: loss 2.6406, time 2065.26ms, mfu 33.97% +iter 11290: loss 2.6204, time 2060.38ms, mfu 33.98% +iter 11300: loss 2.7274, time 2069.37ms, mfu 33.98% +iter 11310: loss 2.8454, time 2054.26ms, mfu 34.00% +iter 11320: loss 2.2316, time 2065.30ms, mfu 34.00% +iter 11330: loss 2.5909, time 2057.67ms, mfu 34.02% +iter 11340: loss 2.7343, time 2062.24ms, mfu 34.02% +iter 11350: loss 2.6025, time 2058.97ms, mfu 34.03% +iter 11360: loss 2.7927, time 2061.03ms, mfu 34.04% +iter 11370: loss 2.5694, time 2060.10ms, mfu 34.04% +iter 11380: loss 2.9913, time 2065.86ms, mfu 34.04% +iter 11390: loss 2.7212, time 2075.13ms, mfu 34.02% +iter 11400: loss 2.3514, time 2065.52ms, mfu 34.02% +iter 11410: loss 2.3657, time 2054.95ms, mfu 34.04% +iter 11420: loss 2.5710, time 2068.00ms, mfu 34.03% +iter 11430: loss 2.6336, time 2065.74ms, mfu 34.03% +iter 11440: loss 2.9689, time 2058.63ms, mfu 34.04% +iter 11450: loss 2.8207, time 2068.54ms, mfu 34.03% +iter 11460: loss 2.8212, time 2060.36ms, mfu 34.04% +iter 11470: loss 2.6242, time 2064.60ms, mfu 34.04% +iter 11480: loss 2.4707, time 2069.76ms, mfu 34.03% +iter 11490: loss 2.5390, time 2058.07ms, mfu 34.04% +iter 11500: loss 2.6867, time 2057.60ms, mfu 34.05% +iter 11510: loss 2.8560, time 2059.79ms, mfu 34.06% +iter 11520: loss 2.4891, time 2054.33ms, mfu 34.07% +iter 11530: loss 2.7639, time 2050.21ms, mfu 34.09% +iter 11540: loss 2.6526, time 2064.52ms, mfu 34.08% +iter 11550: loss 2.7234, time 2058.79ms, mfu 34.09% +iter 11560: loss 2.6222, time 2048.64ms, mfu 34.11% +iter 11570: loss 2.5293, time 2059.88ms, mfu 34.11% +iter 11580: loss 2.5126, time 2062.10ms, mfu 34.11% +iter 11590: loss 2.6322, time 2063.48ms, mfu 34.10% +iter 11600: loss 2.5095, time 2049.66ms, mfu 34.12% +iter 11610: loss 2.9998, time 2071.40ms, mfu 34.10% +iter 11620: loss 2.4878, time 2054.63ms, mfu 34.11% +iter 11630: loss 2.3961, time 2063.01ms, mfu 34.10% +iter 11640: loss 2.5189, time 2065.25ms, mfu 34.09% +iter 11650: loss 2.7439, time 2058.58ms, mfu 34.10% +iter 11660: loss 2.2963, time 2064.10ms, mfu 34.09% +iter 11670: loss 2.2873, time 2066.71ms, mfu 34.08% +iter 11680: loss 2.5790, time 2053.27ms, mfu 34.10% +iter 11690: loss 2.5777, time 2058.24ms, mfu 34.10% +iter 11700: loss 2.5350, time 2064.83ms, mfu 34.09% +iter 11710: loss 2.7743, time 2052.43ms, mfu 34.11% +iter 11720: loss 2.2890, time 2063.76ms, mfu 34.10% +iter 11730: loss 2.7664, time 2073.01ms, mfu 34.08% +iter 11740: loss 2.7230, time 2056.49ms, mfu 34.09% +iter 11750: loss 2.5505, time 2060.73ms, mfu 34.09% +iter 11760: loss 2.4322, time 2067.34ms, mfu 34.08% +iter 11770: loss 2.4931, time 2076.52ms, mfu 34.05% +iter 11780: loss 2.5660, time 2066.01ms, mfu 34.05% +iter 11790: loss 2.6750, time 2060.29ms, mfu 34.05% +iter 11800: loss 3.0233, time 2055.14ms, mfu 34.07% +iter 11810: loss 2.3940, time 2061.12ms, mfu 34.07% +iter 11820: loss 2.5639, time 2065.28ms, mfu 34.06% +iter 11830: loss 2.4781, time 2068.13ms, mfu 34.05% +iter 11840: loss 2.2906, time 2058.33ms, mfu 34.06% +iter 11850: loss 2.6401, time 2051.42ms, mfu 34.08% +iter 11860: loss 2.6692, time 2055.97ms, mfu 34.09% +iter 11870: loss 2.5525, time 2059.25ms, mfu 34.09% +iter 11880: loss 2.6553, time 2065.60ms, mfu 34.09% +iter 11890: loss 2.2409, time 2056.33ms, mfu 34.09% +iter 11900: loss 2.4609, time 2055.12ms, mfu 34.10% +iter 11910: loss 2.6352, time 2058.76ms, mfu 34.11% +iter 11920: loss 2.6475, time 2056.36ms, mfu 34.11% +iter 11930: loss 2.4512, time 2072.67ms, mfu 34.09% +iter 11940: loss 2.5475, time 2062.44ms, mfu 34.09% +iter 11950: loss 2.4088, time 2065.88ms, mfu 34.08% +iter 11960: loss 2.7781, time 2060.43ms, mfu 34.08% +iter 11970: loss 2.7867, time 2079.31ms, mfu 34.05% +iter 11980: loss 2.5822, time 2056.45ms, mfu 34.06% +iter 11990: loss 2.4477, time 2059.48ms, mfu 34.07% +step 12000: train loss 2.5605, val loss 2.6301 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 12000: loss 2.0776, time 9091.03ms, mfu 31.43% +iter 12010: loss 2.3315, time 2054.17ms, mfu 31.71% +iter 12020: loss 2.6457, time 2059.31ms, mfu 31.95% +iter 12030: loss 2.4206, time 2055.05ms, mfu 32.18% +iter 12040: loss 2.6454, time 2064.45ms, mfu 32.36% +iter 12050: loss 2.7072, time 2071.03ms, mfu 32.52% +iter 12060: loss 2.5767, time 2051.99ms, mfu 32.69% +iter 12070: loss 2.7401, time 2055.95ms, mfu 32.84% +iter 12080: loss 2.4669, time 2062.49ms, mfu 32.96% +iter 12090: loss 2.3963, time 2064.36ms, mfu 33.07% +iter 12100: loss 2.7748, time 2060.38ms, mfu 33.17% +iter 12110: loss 2.6146, time 2055.91ms, mfu 33.27% +iter 12120: loss 2.4939, time 2050.86ms, mfu 33.37% +iter 12130: loss 2.6030, time 2050.41ms, mfu 33.46% +iter 12140: loss 2.5353, time 2060.08ms, mfu 33.52% +iter 12150: loss 2.5083, time 2071.09ms, mfu 33.56% +iter 12160: loss 2.5298, time 2060.78ms, mfu 33.62% +iter 12170: loss 2.6107, time 2056.81ms, mfu 33.67% +iter 12180: loss 2.5145, time 2064.40ms, mfu 33.71% +iter 12190: loss 2.8468, time 2063.68ms, mfu 33.74% +iter 12200: loss 2.6968, time 2062.90ms, mfu 33.77% +iter 12210: loss 2.6488, time 2061.54ms, mfu 33.80% +iter 12220: loss 2.3838, time 2065.22ms, mfu 33.83% +iter 12230: loss 2.3998, time 2054.90ms, mfu 33.86% +iter 12240: loss 2.7740, time 2050.77ms, mfu 33.90% +iter 12250: loss 2.6679, time 2063.03ms, mfu 33.92% +iter 12260: loss 2.3619, time 2060.71ms, mfu 33.93% +iter 12270: loss 2.9184, time 2054.22ms, mfu 33.96% +iter 12280: loss 2.2216, time 2066.02ms, mfu 33.97% +iter 12290: loss 2.7737, time 2062.23ms, mfu 33.98% +iter 12300: loss 2.2813, time 2066.06ms, mfu 33.98% +iter 12310: loss 2.7578, time 2054.62ms, mfu 34.00% +iter 12320: loss 2.4254, time 2062.75ms, mfu 34.01% +iter 12330: loss 2.4140, time 2051.90ms, mfu 34.03% +iter 12340: loss 2.3865, time 2059.76ms, mfu 34.04% +iter 12350: loss 2.8641, time 2070.56ms, mfu 34.03% +iter 12360: loss 2.6863, time 2055.26ms, mfu 34.04% +iter 12370: loss 2.3716, time 2052.03ms, mfu 34.06% +iter 12380: loss 2.5410, time 2074.35ms, mfu 34.04% +iter 12390: loss 2.6395, time 2059.74ms, mfu 34.05% +iter 12400: loss 2.3718, time 2059.30ms, mfu 34.06% +iter 12410: loss 2.3765, time 2054.23ms, mfu 34.07% +iter 12420: loss 2.3483, time 2066.84ms, mfu 34.06% +iter 12430: loss 2.5009, time 2058.70ms, mfu 34.07% +iter 12440: loss 2.7279, time 2057.16ms, mfu 34.08% +iter 12450: loss 2.7081, time 2064.47ms, mfu 34.07% +iter 12460: loss 3.0000, time 2056.11ms, mfu 34.08% +iter 12470: loss 2.6775, time 2054.56ms, mfu 34.09% +iter 12480: loss 2.7992, time 2049.64ms, mfu 34.11% +iter 12490: loss 2.7590, time 2065.50ms, mfu 34.10% +iter 12500: loss 2.4465, time 2033.22ms, mfu 34.15% +iter 12510: loss 2.5414, time 2051.11ms, mfu 34.16% +iter 12520: loss 2.7769, time 2053.03ms, mfu 34.17% +iter 12530: loss 2.5233, time 2067.70ms, mfu 34.15% +iter 12540: loss 2.8246, time 2049.87ms, mfu 34.16% +iter 12550: loss 2.5609, time 2060.79ms, mfu 34.15% +iter 12560: loss 2.6976, time 2072.47ms, mfu 34.13% +iter 12570: loss 2.4856, time 2065.39ms, mfu 34.12% +iter 12580: loss 2.8740, time 2073.80ms, mfu 34.09% +iter 12590: loss 2.6110, time 2063.77ms, mfu 34.09% +iter 12600: loss 2.6411, time 2056.00ms, mfu 34.10% +iter 12610: loss 2.5130, time 2063.49ms, mfu 34.09% +iter 12620: loss 2.7267, time 2057.86ms, mfu 34.10% +iter 12630: loss 2.1115, time 2063.43ms, mfu 34.09% +iter 12640: loss 2.5355, time 2063.19ms, mfu 34.09% +iter 12650: loss 2.3345, time 2068.63ms, mfu 34.08% +iter 12660: loss 2.6651, time 2062.40ms, mfu 34.07% +iter 12670: loss 2.2534, time 2048.00ms, mfu 34.10% +iter 12680: loss 2.3468, time 2060.68ms, mfu 34.10% +iter 12690: loss 2.4591, time 2074.64ms, mfu 34.07% +iter 12700: loss 2.5695, time 2053.39ms, mfu 34.09% +iter 12710: loss 2.2450, time 2048.97ms, mfu 34.11% +iter 12720: loss 2.5286, time 2058.98ms, mfu 34.11% +iter 12730: loss 2.6772, time 2075.76ms, mfu 34.08% +iter 12740: loss 2.7052, time 2057.47ms, mfu 34.09% +iter 12750: loss 2.6637, time 2058.86ms, mfu 34.09% +iter 12760: loss 2.5787, time 2070.42ms, mfu 34.08% +iter 12770: loss 2.8703, time 2069.21ms, mfu 34.07% +iter 12780: loss 2.7567, time 2072.70ms, mfu 34.05% +iter 12790: loss 2.6592, time 2053.78ms, mfu 34.06% +iter 12800: loss 2.4989, time 2054.01ms, mfu 34.08% +iter 12810: loss 2.4378, time 2074.32ms, mfu 34.06% +iter 12820: loss 2.7438, time 2057.89ms, mfu 34.07% +iter 12830: loss 2.7508, time 2055.97ms, mfu 34.08% +iter 12840: loss 2.6588, time 2058.76ms, mfu 34.08% +iter 12850: loss 2.4466, time 2074.52ms, mfu 34.06% +iter 12860: loss 2.4130, time 2061.19ms, mfu 34.06% +iter 12870: loss 2.4161, time 2066.53ms, mfu 34.06% +iter 12880: loss 2.6381, time 2053.29ms, mfu 34.07% +iter 12890: loss 2.4429, time 2064.49ms, mfu 34.07% +iter 12900: loss 2.5636, time 2057.16ms, mfu 34.08% +iter 12910: loss 2.5343, time 2071.58ms, mfu 34.06% +iter 12920: loss 2.5449, time 2059.03ms, mfu 34.07% +iter 12930: loss 2.8290, time 2053.00ms, mfu 34.08% +iter 12940: loss 2.3646, time 2067.89ms, mfu 34.07% +iter 12950: loss 2.6786, time 2061.23ms, mfu 34.07% +iter 12960: loss 2.7049, time 2062.13ms, mfu 34.07% +iter 12970: loss 2.4932, time 2068.48ms, mfu 34.06% +iter 12980: loss 2.3409, time 2053.80ms, mfu 34.08% +iter 12990: loss 2.4000, time 2066.20ms, mfu 34.07% +step 13000: train loss 2.5864, val loss 2.6138 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 13000: loss 2.7380, time 9066.45ms, mfu 31.44% +iter 13010: loss 2.3439, time 2049.57ms, mfu 31.72% +iter 13020: loss 2.6481, time 2077.55ms, mfu 31.93% +iter 13030: loss 2.5613, time 2069.23ms, mfu 32.13% +iter 13040: loss 2.3326, time 2054.73ms, mfu 32.34% +iter 13050: loss 2.6436, time 2071.79ms, mfu 32.50% +iter 13060: loss 2.7501, time 2052.48ms, mfu 32.67% +iter 13070: loss 2.6416, time 2046.99ms, mfu 32.84% +iter 13080: loss 2.7759, time 2070.79ms, mfu 32.94% +iter 13090: loss 2.7115, time 2055.40ms, mfu 33.07% +iter 13100: loss 2.6570, time 2058.22ms, mfu 33.18% +iter 13110: loss 2.6462, time 2075.60ms, mfu 33.24% +iter 13120: loss 2.4610, time 2052.49ms, mfu 33.34% +iter 13130: loss 2.4683, time 2061.66ms, mfu 33.42% +iter 13140: loss 2.6229, time 2051.07ms, mfu 33.50% +iter 13150: loss 2.5957, time 2078.38ms, mfu 33.53% +iter 13160: loss 2.6516, time 2057.44ms, mfu 33.59% +iter 13170: loss 2.3856, time 2058.19ms, mfu 33.65% +iter 13180: loss 2.4373, time 2058.86ms, mfu 33.69% +iter 13190: loss 2.5548, time 2065.81ms, mfu 33.73% +iter 13200: loss 2.3040, time 2062.69ms, mfu 33.76% +iter 13210: loss 2.7417, time 2062.70ms, mfu 33.79% +iter 13220: loss 2.4150, time 2053.57ms, mfu 33.83% +iter 13230: loss 2.5343, time 2057.92ms, mfu 33.86% +iter 13240: loss 2.3495, time 2072.88ms, mfu 33.87% +iter 13250: loss 2.4724, time 2058.65ms, mfu 33.89% +iter 13260: loss 2.7842, time 2057.68ms, mfu 33.92% +iter 13270: loss 2.3401, time 2058.67ms, mfu 33.94% +iter 13280: loss 2.6965, time 2073.04ms, mfu 33.93% +iter 13290: loss 2.2799, time 2062.26ms, mfu 33.95% +iter 13300: loss 2.7876, time 2060.75ms, mfu 33.96% +iter 13310: loss 2.4272, time 2067.31ms, mfu 33.96% +iter 13320: loss 2.5021, time 2046.23ms, mfu 34.00% +iter 13330: loss 2.8174, time 2054.58ms, mfu 34.02% +iter 13340: loss 2.4047, time 2057.83ms, mfu 34.03% +iter 13350: loss 2.5204, time 2075.18ms, mfu 34.02% +iter 13360: loss 2.8097, time 2055.31ms, mfu 34.03% +iter 13370: loss 2.6610, time 2050.86ms, mfu 34.05% +iter 13380: loss 2.4978, time 2066.50ms, mfu 34.05% +iter 13390: loss 2.4170, time 2063.99ms, mfu 34.05% +iter 13400: loss 2.8377, time 2066.82ms, mfu 34.04% +iter 13410: loss 2.5655, time 2051.32ms, mfu 34.06% +iter 13420: loss 2.4013, time 2055.35ms, mfu 34.08% +iter 13430: loss 2.7592, time 2065.76ms, mfu 34.07% +iter 13440: loss 2.5101, time 2064.50ms, mfu 34.07% +iter 13450: loss 2.8504, time 2055.05ms, mfu 34.08% +iter 13460: loss 2.6912, time 2080.14ms, mfu 34.05% +iter 13470: loss 2.4065, time 2060.37ms, mfu 34.05% +iter 13480: loss 2.5997, time 2045.47ms, mfu 34.08% +iter 13490: loss 2.8609, time 2071.01ms, mfu 34.07% +iter 13500: loss 2.4951, time 2067.84ms, mfu 34.06% +iter 13510: loss 2.4402, time 2062.77ms, mfu 34.06% +iter 13520: loss 2.6831, time 2064.39ms, mfu 34.06% +iter 13530: loss 2.8145, time 2060.25ms, mfu 34.06% +iter 13540: loss 2.5684, time 2066.46ms, mfu 34.05% +iter 13550: loss 2.6011, time 2057.54ms, mfu 34.06% +iter 13560: loss 2.3702, time 2061.98ms, mfu 34.06% +iter 13570: loss 2.7817, time 2061.32ms, mfu 34.07% +iter 13580: loss 2.6540, time 2052.62ms, mfu 34.08% +iter 13590: loss 2.4682, time 2073.13ms, mfu 34.06% +iter 13600: loss 2.4975, time 2061.92ms, mfu 34.06% +iter 13610: loss 2.6230, time 2063.33ms, mfu 34.06% +iter 13620: loss 2.3349, time 2052.95ms, mfu 34.08% +iter 13630: loss 2.2930, time 2087.13ms, mfu 34.04% +iter 13640: loss 2.5781, time 2061.86ms, mfu 34.04% +iter 13650: loss 2.7901, time 2063.06ms, mfu 34.04% +iter 13660: loss 2.5238, time 2057.47ms, mfu 34.05% +iter 13670: loss 2.2911, time 2063.01ms, mfu 34.05% +iter 13680: loss 2.6762, time 2061.15ms, mfu 34.06% +iter 13690: loss 2.4528, time 2056.75ms, mfu 34.07% +iter 13700: loss 2.9699, time 2067.16ms, mfu 34.06% +iter 13710: loss 2.4235, time 2062.99ms, mfu 34.06% +iter 13720: loss 2.5405, time 2064.34ms, mfu 34.06% +iter 13730: loss 2.8400, time 2064.87ms, mfu 34.05% +iter 13740: loss 2.7712, time 2060.92ms, mfu 34.06% +iter 13750: loss 2.8090, time 2056.13ms, mfu 34.07% +iter 13760: loss 2.4001, time 2056.18ms, mfu 34.08% +iter 13770: loss 2.5907, time 2054.68ms, mfu 34.09% +iter 13780: loss 2.7545, time 2053.94ms, mfu 34.10% +iter 13790: loss 2.4430, time 2053.55ms, mfu 34.11% +iter 13800: loss 2.8071, time 2049.99ms, mfu 34.13% +iter 13810: loss 2.5336, time 2064.05ms, mfu 34.12% +iter 13820: loss 2.3503, time 2048.92ms, mfu 34.14% +iter 13830: loss 2.5504, time 2049.13ms, mfu 34.15% +iter 13840: loss 2.6415, time 2053.36ms, mfu 34.16% +iter 13850: loss 2.8027, time 2055.39ms, mfu 34.16% +iter 13860: loss 2.5493, time 2062.19ms, mfu 34.15% +iter 13870: loss 2.5127, time 2048.00ms, mfu 34.17% +iter 13880: loss 2.4871, time 2059.22ms, mfu 34.16% +iter 13890: loss 2.5747, time 2057.28ms, mfu 34.16% +iter 13900: loss 2.8729, time 2056.06ms, mfu 34.16% +iter 13910: loss 2.7369, time 2049.10ms, mfu 34.17% +iter 13920: loss 2.4248, time 2067.21ms, mfu 34.16% +iter 13930: loss 2.7975, time 2063.91ms, mfu 34.14% +iter 13940: loss 2.3047, time 2061.30ms, mfu 34.14% +iter 13950: loss 2.6183, time 2060.41ms, mfu 34.13% +iter 13960: loss 2.5239, time 2062.41ms, mfu 34.13% +iter 13970: loss 2.3967, time 2057.85ms, mfu 34.13% +iter 13980: loss 2.3231, time 2054.23ms, mfu 34.14% +iter 13990: loss 2.6802, time 2054.23ms, mfu 34.14% +step 14000: train loss 2.5575, val loss 2.6164 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 14000: loss 2.8029, time 9061.69ms, mfu 31.50% +iter 14010: loss 2.2803, time 2058.57ms, mfu 31.77% +iter 14020: loss 2.4520, time 2053.19ms, mfu 32.01% +iter 14030: loss 2.6593, time 2058.83ms, mfu 32.22% +iter 14040: loss 2.6270, time 2062.11ms, mfu 32.41% +iter 14050: loss 2.6535, time 2063.15ms, mfu 32.57% +iter 14060: loss 2.4947, time 2051.61ms, mfu 32.74% +iter 14070: loss 2.5535, time 2059.72ms, mfu 32.88% +iter 14080: loss 2.7918, time 2057.74ms, mfu 33.00% +iter 14090: loss 2.6818, time 2059.67ms, mfu 33.11% +iter 14100: loss 2.4939, time 2055.59ms, mfu 33.22% +iter 14110: loss 2.3787, time 2061.76ms, mfu 33.31% +iter 14120: loss 2.7219, time 2068.15ms, mfu 33.37% +iter 14130: loss 2.7953, time 2050.43ms, mfu 33.46% +iter 14140: loss 2.5326, time 2051.66ms, mfu 33.54% +iter 14150: loss 2.6087, time 2057.54ms, mfu 33.60% +iter 14160: loss 2.0256, time 2054.80ms, mfu 33.66% +iter 14170: loss 2.6912, time 2055.35ms, mfu 33.71% +iter 14180: loss 2.6032, time 2053.06ms, mfu 33.76% +iter 14190: loss 2.3300, time 2058.39ms, mfu 33.80% +iter 14200: loss 2.6319, time 2057.47ms, mfu 33.83% +iter 14210: loss 2.5641, time 2054.45ms, mfu 33.87% +iter 14220: loss 2.2662, time 2059.74ms, mfu 33.90% +iter 14230: loss 2.5128, time 2064.64ms, mfu 33.91% +iter 14240: loss 2.3486, time 2056.29ms, mfu 33.93% +iter 14250: loss 1.9805, time 2056.54ms, mfu 33.96% +iter 14260: loss 2.2195, time 2047.35ms, mfu 33.99% +iter 14270: loss 2.3904, time 2054.91ms, mfu 34.01% +iter 14280: loss 2.6537, time 2052.87ms, mfu 34.03% +iter 14290: loss 2.6837, time 2053.81ms, mfu 34.05% +iter 14300: loss 2.4102, time 2043.07ms, mfu 34.09% +iter 14310: loss 2.7811, time 2054.35ms, mfu 34.10% +iter 14320: loss 2.1844, time 2051.38ms, mfu 34.11% +iter 14330: loss 2.8172, time 2067.50ms, mfu 34.10% +iter 14340: loss 2.6297, time 2051.32ms, mfu 34.11% +iter 14350: loss 2.4724, time 2061.10ms, mfu 34.11% +iter 14360: loss 2.2456, time 2057.96ms, mfu 34.11% +iter 14370: loss 2.7333, time 2061.22ms, mfu 34.11% +iter 14380: loss 2.3020, time 2052.53ms, mfu 34.12% +iter 14390: loss 2.8970, time 2056.63ms, mfu 34.13% +iter 14400: loss 2.5623, time 2059.27ms, mfu 34.13% +iter 14410: loss 2.5717, time 2065.48ms, mfu 34.12% +iter 14420: loss 2.2594, time 2054.26ms, mfu 34.12% +iter 14430: loss 2.7837, time 2051.33ms, mfu 34.14% +iter 14440: loss 2.3503, time 2061.86ms, mfu 34.13% +iter 14450: loss 2.5512, time 2068.87ms, mfu 34.11% +iter 14460: loss 2.4546, time 2067.24ms, mfu 34.10% +iter 14470: loss 2.5551, time 2057.82ms, mfu 34.10% +iter 14480: loss 2.8855, time 2052.69ms, mfu 34.12% +iter 14490: loss 2.6255, time 2053.91ms, mfu 34.13% +iter 14500: loss 2.3376, time 2055.56ms, mfu 34.13% +iter 14510: loss 2.4140, time 2044.11ms, mfu 34.16% +iter 14520: loss 2.5269, time 2054.41ms, mfu 34.16% +iter 14530: loss 2.7492, time 2052.74ms, mfu 34.17% +iter 14540: loss 2.7673, time 2060.31ms, mfu 34.16% +iter 14550: loss 2.4833, time 2051.53ms, mfu 34.17% +iter 14560: loss 2.6350, time 2068.51ms, mfu 34.15% +iter 14570: loss 2.6695, time 2058.86ms, mfu 34.15% +iter 14580: loss 2.3509, time 2054.67ms, mfu 34.15% +iter 14590: loss 2.4199, time 2051.26ms, mfu 34.16% +iter 14600: loss 2.7392, time 2049.65ms, mfu 34.17% +iter 14610: loss 2.4420, time 2059.19ms, mfu 34.17% +iter 14620: loss 2.6778, time 2046.55ms, mfu 34.18% +iter 14630: loss 2.4819, time 2057.19ms, mfu 34.18% +iter 14640: loss 2.3190, time 2052.14ms, mfu 34.19% +iter 14650: loss 2.6133, time 2058.33ms, mfu 34.18% +iter 14660: loss 2.4233, time 2050.75ms, mfu 34.19% +iter 14670: loss 2.3562, time 2057.31ms, mfu 34.18% +iter 14680: loss 2.5500, time 2054.87ms, mfu 34.19% +iter 14690: loss 2.5178, time 2063.88ms, mfu 34.17% +iter 14700: loss 2.3900, time 2055.78ms, mfu 34.17% +iter 14710: loss 2.4925, time 2046.91ms, mfu 34.19% +iter 14720: loss 2.5132, time 2048.60ms, mfu 34.20% +iter 14730: loss 2.5137, time 2055.82ms, mfu 34.20% +iter 14740: loss 2.7357, time 2054.62ms, mfu 34.20% +iter 14750: loss 2.8404, time 2052.81ms, mfu 34.20% +iter 14760: loss 2.3811, time 2050.69ms, mfu 34.20% +iter 14770: loss 2.6199, time 2050.52ms, mfu 34.21% +iter 14780: loss 2.3716, time 2062.08ms, mfu 34.20% +iter 14790: loss 2.2393, time 2047.75ms, mfu 34.21% +iter 14800: loss 2.9103, time 2065.51ms, mfu 34.19% +iter 14810: loss 2.7354, time 2047.65ms, mfu 34.20% +iter 14820: loss 2.2770, time 2050.80ms, mfu 34.21% +iter 14830: loss 2.4562, time 2053.93ms, mfu 34.21% +iter 14840: loss 2.4282, time 2055.70ms, mfu 34.20% +iter 14850: loss 2.7176, time 2055.31ms, mfu 34.20% +iter 14860: loss 2.6067, time 2059.74ms, mfu 34.19% +iter 14870: loss 2.7508, time 2050.48ms, mfu 34.20% +iter 14880: loss 2.7051, time 2052.87ms, mfu 34.20% +iter 14890: loss 2.5682, time 2051.48ms, mfu 34.21% +iter 14900: loss 2.6220, time 2053.99ms, mfu 34.21% +iter 14910: loss 2.6087, time 2052.58ms, mfu 34.21% +iter 14920: loss 2.5139, time 2044.17ms, mfu 34.22% +iter 14930: loss 2.6384, time 2048.19ms, mfu 34.23% +iter 14940: loss 2.3015, time 2049.53ms, mfu 34.24% +iter 14950: loss 2.7948, time 2049.86ms, mfu 34.24% +iter 14960: loss 2.6191, time 2052.49ms, mfu 34.24% +iter 14970: loss 2.6581, time 2053.43ms, mfu 34.24% +iter 14980: loss 2.7812, time 2054.09ms, mfu 34.23% +iter 14990: loss 2.5137, time 2062.48ms, mfu 34.22% +step 15000: train loss 2.5469, val loss 2.6056 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 15000: loss 2.5324, time 9050.78ms, mfu 31.57% +iter 15010: loss 2.3895, time 2047.69ms, mfu 31.85% +iter 15020: loss 2.8188, time 2045.50ms, mfu 32.10% +iter 15030: loss 2.7421, time 2061.15ms, mfu 32.30% +iter 15040: loss 2.6073, time 2057.51ms, mfu 32.48% +iter 15050: loss 2.5282, time 2044.93ms, mfu 32.67% +iter 15060: loss 2.6307, time 2050.35ms, mfu 32.83% +iter 15070: loss 2.3521, time 2041.18ms, mfu 32.99% +iter 15080: loss 2.6511, time 2055.11ms, mfu 33.11% +iter 15090: loss 2.5863, time 2047.23ms, mfu 33.23% +iter 15100: loss 2.3589, time 2053.05ms, mfu 33.33% +iter 15110: loss 2.1197, time 2048.22ms, mfu 33.42% +iter 15120: loss 2.3586, time 2050.31ms, mfu 33.51% +iter 15130: loss 2.1581, time 2057.07ms, mfu 33.57% +iter 15140: loss 2.3600, time 2052.52ms, mfu 33.64% +iter 15150: loss 2.4316, time 2050.59ms, mfu 33.70% +iter 15160: loss 2.3814, time 2052.94ms, mfu 33.75% +iter 15170: loss 2.5108, time 2047.81ms, mfu 33.81% +iter 15180: loss 2.6854, time 2056.53ms, mfu 33.84% +iter 15190: loss 2.4573, time 2052.01ms, mfu 33.88% +iter 15200: loss 2.5847, time 2045.68ms, mfu 33.93% +iter 15210: loss 2.5517, time 2033.83ms, mfu 33.99% +iter 15220: loss 2.6235, time 2062.86ms, mfu 34.00% +iter 15230: loss 2.6293, time 2050.30ms, mfu 34.03% +iter 15240: loss 2.5032, time 2045.70ms, mfu 34.06% +iter 15250: loss 2.5022, time 2047.63ms, mfu 34.08% +iter 15260: loss 2.4612, time 2048.19ms, mfu 34.10% +iter 15270: loss 2.5695, time 2051.40ms, mfu 34.12% +iter 15280: loss 2.5566, time 2058.64ms, mfu 34.12% +iter 15290: loss 2.5804, time 2060.08ms, mfu 34.12% +iter 15300: loss 2.1813, time 2061.19ms, mfu 34.12% +iter 15310: loss 2.6232, time 2057.42ms, mfu 34.12% +iter 15320: loss 2.6546, time 2063.45ms, mfu 34.11% +iter 15330: loss 2.4853, time 2045.05ms, mfu 34.14% +iter 15340: loss 2.3950, time 2054.40ms, mfu 34.14% +iter 15350: loss 2.5932, time 2054.46ms, mfu 34.15% +iter 15360: loss 2.6560, time 2047.94ms, mfu 34.16% +iter 15370: loss 2.6596, time 2048.36ms, mfu 34.18% +iter 15380: loss 2.3597, time 2063.25ms, mfu 34.16% +iter 15390: loss 2.5653, time 2053.59ms, mfu 34.17% +iter 15400: loss 2.5883, time 2052.95ms, mfu 34.17% +iter 15410: loss 2.8086, time 2060.44ms, mfu 34.17% +iter 15420: loss 2.3315, time 2048.61ms, mfu 34.18% +iter 15430: loss 2.3299, time 2055.70ms, mfu 34.18% +iter 15440: loss 2.6439, time 2048.96ms, mfu 34.19% +iter 15450: loss 2.3613, time 2046.56ms, mfu 34.20% +iter 15460: loss 2.4900, time 2049.88ms, mfu 34.21% +iter 15470: loss 2.3410, time 2045.58ms, mfu 34.23% +iter 15480: loss 3.0208, time 2052.42ms, mfu 34.23% +iter 15490: loss 2.1029, time 2049.41ms, mfu 34.23% +iter 15500: loss 2.5235, time 2037.47ms, mfu 34.26% +iter 15510: loss 2.4992, time 2050.61ms, mfu 34.26% +iter 15520: loss 2.7876, time 2052.40ms, mfu 34.25% +iter 15530: loss 2.4952, time 2044.66ms, mfu 34.27% +iter 15540: loss 2.6472, time 2052.98ms, mfu 34.26% +iter 15550: loss 2.7036, time 2049.75ms, mfu 34.26% +iter 15560: loss 2.2222, time 2049.50ms, mfu 34.26% +iter 15570: loss 2.8318, time 2051.03ms, mfu 34.26% +iter 15580: loss 2.5763, time 2065.18ms, mfu 34.24% +iter 15590: loss 2.6690, time 2049.73ms, mfu 34.24% +iter 15600: loss 2.5214, time 2041.67ms, mfu 34.26% +iter 15610: loss 2.3846, time 2057.31ms, mfu 34.25% +iter 15620: loss 2.4082, time 2056.73ms, mfu 34.24% +iter 15630: loss 2.7194, time 2054.98ms, mfu 34.23% +iter 15640: loss 2.3063, time 2056.83ms, mfu 34.23% +iter 15650: loss 2.3059, time 2052.67ms, mfu 34.23% +iter 15660: loss 2.5217, time 2052.38ms, mfu 34.23% +iter 15670: loss 2.3697, time 2057.60ms, mfu 34.22% +iter 15680: loss 2.7433, time 2060.98ms, mfu 34.21% +iter 15690: loss 2.3187, time 2048.51ms, mfu 34.22% +iter 15700: loss 2.6030, time 2053.68ms, mfu 34.22% +iter 15710: loss 2.6533, time 2043.51ms, mfu 34.23% +iter 15720: loss 2.8647, time 2057.78ms, mfu 34.22% +iter 15730: loss 2.2071, time 2062.18ms, mfu 34.21% +iter 15740: loss 2.1813, time 2049.82ms, mfu 34.21% +iter 15750: loss 2.6227, time 2048.93ms, mfu 34.22% +iter 15760: loss 2.7295, time 2058.90ms, mfu 34.21% +iter 15770: loss 2.2505, time 2065.11ms, mfu 34.19% +iter 15780: loss 2.6953, time 2058.72ms, mfu 34.19% +iter 15790: loss 2.5725, time 2048.19ms, mfu 34.20% +iter 15800: loss 2.4904, time 2047.66ms, mfu 34.21% +iter 15810: loss 2.3056, time 2051.80ms, mfu 34.21% +iter 15820: loss 2.2874, time 2060.33ms, mfu 34.20% +iter 15830: loss 2.4724, time 2054.06ms, mfu 34.20% +iter 15840: loss 2.3941, time 2058.38ms, mfu 34.19% +iter 15850: loss 2.5230, time 2053.16ms, mfu 34.20% +iter 15860: loss 2.5530, time 2050.85ms, mfu 34.20% +iter 15870: loss 2.2016, time 2056.78ms, mfu 34.20% +iter 15880: loss 2.4216, time 2060.16ms, mfu 34.19% +iter 15890: loss 2.6163, time 2054.45ms, mfu 34.19% +iter 15900: loss 2.2982, time 2054.87ms, mfu 34.19% +iter 15910: loss 2.2620, time 2049.58ms, mfu 34.20% +iter 15920: loss 2.6602, time 2048.36ms, mfu 34.21% +iter 15930: loss 2.4288, time 2050.69ms, mfu 34.21% +iter 15940: loss 2.6208, time 2045.64ms, mfu 34.23% +iter 15950: loss 2.7025, time 2050.76ms, mfu 34.23% +iter 15960: loss 2.3091, time 2049.07ms, mfu 34.24% +iter 15970: loss 2.6048, time 2051.71ms, mfu 34.24% +iter 15980: loss 2.5497, time 2060.04ms, mfu 34.22% +iter 15990: loss 2.6321, time 2044.96ms, mfu 34.24% +step 16000: train loss 2.5108, val loss 2.5968 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 16000: loss 2.4993, time 9021.45ms, mfu 31.59% +iter 16010: loss 2.8461, time 2053.27ms, mfu 31.85% +iter 16020: loss 2.2652, time 2054.36ms, mfu 32.09% +iter 16030: loss 2.6752, time 2051.28ms, mfu 32.31% +iter 16040: loss 2.6377, time 2046.22ms, mfu 32.51% +iter 16050: loss 2.7656, time 2060.81ms, mfu 32.67% +iter 16060: loss 2.5904, time 2050.13ms, mfu 32.83% +iter 16070: loss 2.5344, time 2050.30ms, mfu 32.97% +iter 16080: loss 2.4649, time 2054.32ms, mfu 33.09% +iter 16090: loss 2.3970, time 2050.45ms, mfu 33.21% +iter 16100: loss 2.7187, time 2045.18ms, mfu 33.33% +iter 16110: loss 2.7439, time 2057.61ms, mfu 33.41% +iter 16120: loss 2.4232, time 2058.49ms, mfu 33.48% +iter 16130: loss 2.7324, time 2056.35ms, mfu 33.55% +iter 16140: loss 2.8502, time 2048.49ms, mfu 33.62% +iter 16150: loss 2.5180, time 2044.42ms, mfu 33.70% +iter 16160: loss 2.5804, time 2053.10ms, mfu 33.75% +iter 16170: loss 2.5761, time 2046.38ms, mfu 33.81% +iter 16180: loss 2.4343, time 2043.24ms, mfu 33.87% +iter 16190: loss 2.3515, time 2047.65ms, mfu 33.91% +iter 16200: loss 2.2920, time 2047.42ms, mfu 33.95% +iter 16210: loss 2.9033, time 2058.19ms, mfu 33.97% +iter 16220: loss 2.3696, time 2049.86ms, mfu 34.00% +iter 16230: loss 2.4057, time 2045.33ms, mfu 34.03% +iter 16240: loss 2.6864, time 2058.83ms, mfu 34.04% +iter 16250: loss 2.5651, time 2052.33ms, mfu 34.06% +iter 16260: loss 2.6559, time 2043.90ms, mfu 34.09% +iter 16270: loss 2.3451, time 2051.73ms, mfu 34.11% +iter 16280: loss 2.6381, time 2053.28ms, mfu 34.12% +iter 16290: loss 2.7059, time 2050.64ms, mfu 34.13% +iter 16300: loss 2.7849, time 2042.69ms, mfu 34.16% +iter 16310: loss 2.5320, time 2043.28ms, mfu 34.18% +iter 16320: loss 2.3190, time 2055.18ms, mfu 34.18% +iter 16330: loss 2.5818, time 2044.84ms, mfu 34.20% +iter 16340: loss 2.7277, time 2055.69ms, mfu 34.20% +iter 16350: loss 2.7512, time 2055.99ms, mfu 34.20% +iter 16360: loss 2.4944, time 2055.38ms, mfu 34.19% +iter 16370: loss 2.7119, time 2056.66ms, mfu 34.19% +iter 16380: loss 2.7095, time 2050.27ms, mfu 34.20% +iter 16390: loss 2.6410, time 2061.16ms, mfu 34.19% +iter 16400: loss 2.5953, time 2053.21ms, mfu 34.19% +iter 16410: loss 2.4840, time 2053.07ms, mfu 34.19% +iter 16420: loss 2.5949, time 2048.99ms, mfu 34.20% +iter 16430: loss 2.7740, time 2047.46ms, mfu 34.21% +iter 16440: loss 2.8673, time 2054.82ms, mfu 34.21% +iter 16450: loss 2.6198, time 2045.64ms, mfu 34.23% +iter 16460: loss 2.8139, time 2047.91ms, mfu 34.23% +iter 16470: loss 2.4340, time 2045.09ms, mfu 34.25% +iter 16480: loss 2.3495, time 2047.67ms, mfu 34.25% +iter 16490: loss 2.4162, time 2051.58ms, mfu 34.25% +iter 16500: loss 2.6461, time 2043.42ms, mfu 34.26% +iter 16510: loss 2.5307, time 2048.68ms, mfu 34.27% +iter 16520: loss 2.4841, time 2042.90ms, mfu 34.28% +iter 16530: loss 2.5116, time 2041.76ms, mfu 34.29% +iter 16540: loss 2.3272, time 2042.52ms, mfu 34.30% +iter 16550: loss 2.2739, time 2053.35ms, mfu 34.29% +iter 16560: loss 2.6956, time 2052.60ms, mfu 34.29% +iter 16570: loss 2.8273, time 2034.91ms, mfu 34.31% +iter 16580: loss 2.5294, time 2044.89ms, mfu 34.32% +iter 16590: loss 2.7706, time 2048.31ms, mfu 34.32% +iter 16600: loss 2.5464, time 2048.27ms, mfu 34.31% +iter 16610: loss 2.6033, time 2044.75ms, mfu 34.32% +iter 16620: loss 2.6606, time 2047.78ms, mfu 34.32% +iter 16630: loss 2.7447, time 2044.96ms, mfu 34.32% +iter 16640: loss 2.6759, time 2058.74ms, mfu 34.30% +iter 16650: loss 2.6810, time 2051.58ms, mfu 34.30% +iter 16660: loss 2.3872, time 2056.85ms, mfu 34.28% +iter 16670: loss 2.2233, time 2044.02ms, mfu 34.29% +iter 16680: loss 2.3473, time 2061.95ms, mfu 34.27% +iter 16690: loss 2.6947, time 2045.75ms, mfu 34.28% +iter 16700: loss 2.4462, time 2053.62ms, mfu 34.27% +iter 16710: loss 2.4761, time 2063.45ms, mfu 34.25% +iter 16720: loss 2.6136, time 2033.21ms, mfu 34.28% +iter 16730: loss 2.4234, time 2043.31ms, mfu 34.29% +iter 16740: loss 2.2819, time 2041.39ms, mfu 34.30% +iter 16750: loss 2.4658, time 2057.88ms, mfu 34.29% +iter 16760: loss 2.5526, time 2048.49ms, mfu 34.29% +iter 16770: loss 2.4635, time 2056.37ms, mfu 34.28% +iter 16780: loss 2.6369, time 2055.27ms, mfu 34.27% +iter 16790: loss 2.4946, time 2046.38ms, mfu 34.27% +iter 16800: loss 2.4449, time 2051.69ms, mfu 34.27% +iter 16810: loss 2.8056, time 2050.97ms, mfu 34.27% +iter 16820: loss 2.6816, time 2050.53ms, mfu 34.27% +iter 16830: loss 2.6181, time 2045.86ms, mfu 34.28% +iter 16840: loss 2.6604, time 2051.87ms, mfu 34.27% +iter 16850: loss 2.6324, time 2044.74ms, mfu 34.28% +iter 16860: loss 2.4470, time 2043.88ms, mfu 34.29% +iter 16870: loss 2.4891, time 2046.77ms, mfu 34.29% +iter 16880: loss 2.2405, time 2050.77ms, mfu 34.29% +iter 16890: loss 2.3522, time 2057.29ms, mfu 34.28% +iter 16900: loss 2.7712, time 2048.22ms, mfu 34.28% +iter 16910: loss 2.5849, time 2062.88ms, mfu 34.26% +iter 16920: loss 2.2291, time 2047.28ms, mfu 34.26% +iter 16930: loss 2.3130, time 2047.19ms, mfu 34.27% +iter 16940: loss 2.6223, time 2047.89ms, mfu 34.27% +iter 16950: loss 2.3072, time 2040.53ms, mfu 34.29% +iter 16960: loss 2.6404, time 2048.33ms, mfu 34.29% +iter 16970: loss 2.5679, time 2051.82ms, mfu 34.28% +iter 16980: loss 2.6659, time 2045.36ms, mfu 34.29% +iter 16990: loss 2.4610, time 2046.12ms, mfu 34.30% +step 17000: train loss 2.5031, val loss 2.6176 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 17000: loss 2.5025, time 9049.82ms, mfu 31.64% +iter 17010: loss 2.4492, time 2046.85ms, mfu 31.91% +iter 17020: loss 2.7681, time 2048.29ms, mfu 32.15% +iter 17030: loss 2.6648, time 2058.91ms, mfu 32.35% +iter 17040: loss 2.6501, time 2043.28ms, mfu 32.55% +iter 17050: loss 2.6715, time 2043.62ms, mfu 32.73% +iter 17060: loss 2.8489, time 2053.03ms, mfu 32.88% +iter 17070: loss 2.3205, time 2046.36ms, mfu 33.03% +iter 17080: loss 2.1749, time 2049.32ms, mfu 33.15% +iter 17090: loss 2.8133, time 2052.82ms, mfu 33.26% +iter 17100: loss 2.6915, time 2047.16ms, mfu 33.37% +iter 17110: loss 2.7509, time 2049.01ms, mfu 33.46% +iter 17120: loss 2.8725, time 2061.15ms, mfu 33.52% +iter 17130: loss 2.7440, time 2050.09ms, mfu 33.60% +iter 17140: loss 2.5501, time 2052.00ms, mfu 33.66% +iter 17150: loss 2.5734, time 2055.80ms, mfu 33.71% +iter 17160: loss 2.3107, time 2051.93ms, mfu 33.76% +iter 17170: loss 2.6111, time 2065.28ms, mfu 33.79% +iter 17180: loss 2.5572, time 2047.61ms, mfu 33.84% +iter 17190: loss 2.2599, time 2059.02ms, mfu 33.87% +iter 17200: loss 2.5295, time 2059.74ms, mfu 33.89% +iter 17210: loss 2.6658, time 2049.26ms, mfu 33.93% +iter 17220: loss 2.3482, time 2055.35ms, mfu 33.96% +iter 17230: loss 2.3992, time 2046.48ms, mfu 34.00% +iter 17240: loss 2.7894, time 2057.81ms, mfu 34.01% +iter 17250: loss 2.6257, time 2055.51ms, mfu 34.03% +iter 17260: loss 2.3506, time 2067.69ms, mfu 34.02% +iter 17270: loss 2.4402, time 2069.62ms, mfu 34.01% +iter 17280: loss 2.5134, time 2048.38ms, mfu 34.04% +iter 17290: loss 2.5276, time 2041.84ms, mfu 34.08% +iter 17300: loss 2.2971, time 2048.46ms, mfu 34.10% +iter 17310: loss 2.6004, time 2064.06ms, mfu 34.10% +iter 17320: loss 2.5300, time 2038.85ms, mfu 34.13% +iter 17330: loss 2.4263, time 2051.50ms, mfu 34.14% +iter 17340: loss 2.3857, time 2048.43ms, mfu 34.16% +iter 17350: loss 2.0726, time 2052.51ms, mfu 34.17% +iter 17360: loss 2.4436, time 2057.21ms, mfu 34.16% +iter 17370: loss 2.5092, time 2059.29ms, mfu 34.16% +iter 17380: loss 2.7875, time 2053.31ms, mfu 34.17% +iter 17390: loss 2.6140, time 2044.99ms, mfu 34.18% +iter 17400: loss 2.6837, time 2034.19ms, mfu 34.22% +iter 17410: loss 2.4962, time 2046.45ms, mfu 34.23% +iter 17420: loss 2.4321, time 2043.09ms, mfu 34.25% +iter 17430: loss 2.7111, time 2051.59ms, mfu 34.25% +iter 17440: loss 2.5206, time 2040.64ms, mfu 34.27% +iter 17450: loss 2.5912, time 2037.67ms, mfu 34.29% +iter 17460: loss 2.7268, time 2053.88ms, mfu 34.28% +iter 17470: loss 2.3265, time 2035.62ms, mfu 34.30% +iter 17480: loss 2.6156, time 2048.18ms, mfu 34.30% +iter 17490: loss 2.4059, time 2049.73ms, mfu 34.30% +iter 17500: loss 2.2469, time 2057.49ms, mfu 34.28% +iter 17510: loss 2.5060, time 2052.37ms, mfu 34.28% +iter 17520: loss 2.7370, time 2059.90ms, mfu 34.26% +iter 17530: loss 2.3551, time 2055.96ms, mfu 34.25% +iter 17540: loss 2.4487, time 2036.45ms, mfu 34.28% +iter 17550: loss 2.8264, time 2050.90ms, mfu 34.28% +iter 17560: loss 2.5364, time 2040.23ms, mfu 34.29% +iter 17570: loss 2.5814, time 2036.22ms, mfu 34.31% +iter 17580: loss 2.5693, time 2041.48ms, mfu 34.32% +iter 17590: loss 2.6920, time 2047.07ms, mfu 34.32% +iter 17600: loss 2.6617, time 2039.36ms, mfu 34.34% +iter 17610: loss 2.3513, time 2029.56ms, mfu 34.36% +iter 17620: loss 2.4541, time 2029.78ms, mfu 34.39% +iter 17630: loss 2.5524, time 2027.45ms, mfu 34.42% +iter 17640: loss 2.5926, time 2043.88ms, mfu 34.41% +iter 17650: loss 2.5012, time 2049.50ms, mfu 34.40% +iter 17660: loss 2.7000, time 2044.81ms, mfu 34.39% +iter 17670: loss 2.4874, time 2045.13ms, mfu 34.39% +iter 17680: loss 2.3603, time 2053.44ms, mfu 34.37% +iter 17690: loss 2.6509, time 2047.91ms, mfu 34.37% +iter 17700: loss 2.2015, time 2057.21ms, mfu 34.34% +iter 17710: loss 2.4685, time 2046.04ms, mfu 34.34% +iter 17720: loss 2.4874, time 2048.66ms, mfu 34.34% +iter 17730: loss 2.7454, time 2047.60ms, mfu 34.34% +iter 17740: loss 2.6389, time 2048.97ms, mfu 34.33% +iter 17750: loss 2.6719, time 2032.72ms, mfu 34.36% +iter 17760: loss 2.3490, time 2048.04ms, mfu 34.35% +iter 17770: loss 2.5283, time 2044.39ms, mfu 34.35% +iter 17780: loss 2.6234, time 2049.51ms, mfu 34.34% +iter 17790: loss 2.9732, time 2043.98ms, mfu 34.35% +iter 17800: loss 2.2693, time 2041.98ms, mfu 34.35% +iter 17810: loss 2.4541, time 2032.48ms, mfu 34.37% +iter 17820: loss 2.5366, time 2041.11ms, mfu 34.38% +iter 17830: loss 2.4183, time 2050.35ms, mfu 34.37% +iter 17840: loss 2.5554, time 2039.38ms, mfu 34.38% +iter 17850: loss 2.6485, time 2060.44ms, mfu 34.35% +iter 17860: loss 2.4161, time 2033.29ms, mfu 34.37% +iter 17870: loss 2.5290, time 2053.53ms, mfu 34.35% +iter 17880: loss 2.5156, time 2052.25ms, mfu 34.34% +iter 17890: loss 2.4510, time 2046.29ms, mfu 34.34% +iter 17900: loss 2.6007, time 2051.44ms, mfu 34.33% +iter 17910: loss 2.4240, time 2049.71ms, mfu 34.33% +iter 17920: loss 2.7542, time 2053.78ms, mfu 34.31% +iter 17930: loss 2.0915, time 2054.13ms, mfu 34.30% +iter 17940: loss 2.5501, time 2037.62ms, mfu 34.32% +iter 17950: loss 2.6879, time 2049.09ms, mfu 34.32% +iter 17960: loss 2.5390, time 2050.53ms, mfu 34.31% +iter 17970: loss 2.1091, time 2037.76ms, mfu 34.33% +iter 17980: loss 2.6417, time 2054.78ms, mfu 34.32% +iter 17990: loss 2.4377, time 2049.14ms, mfu 34.31% +step 18000: train loss 2.5104, val loss 2.5790 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 18000: loss 2.6239, time 9213.51ms, mfu 31.64% +iter 18010: loss 2.5675, time 2042.40ms, mfu 31.92% +iter 18020: loss 2.2252, time 2040.96ms, mfu 32.17% +iter 18030: loss 2.2279, time 2045.03ms, mfu 32.39% +iter 18040: loss 2.5594, time 2035.03ms, mfu 32.60% +iter 18050: loss 2.1532, time 2031.29ms, mfu 32.80% +iter 18060: loss 2.7848, time 2048.00ms, mfu 32.95% +iter 18070: loss 2.4558, time 2055.00ms, mfu 33.07% +iter 18080: loss 2.5108, time 2041.13ms, mfu 33.21% +iter 18090: loss 2.5548, time 2050.17ms, mfu 33.32% +iter 18100: loss 2.6184, time 2035.04ms, mfu 33.44% +iter 18110: loss 2.6283, time 2048.13ms, mfu 33.52% +iter 18120: loss 2.3918, time 2058.51ms, mfu 33.58% +iter 18130: loss 2.7367, time 2038.65ms, mfu 33.67% +iter 18140: loss 2.4815, time 2032.35ms, mfu 33.76% +iter 18150: loss 2.4789, time 2038.03ms, mfu 33.83% +iter 18160: loss 2.3502, time 2043.69ms, mfu 33.89% +iter 18170: loss 2.5408, time 2061.57ms, mfu 33.91% +iter 18180: loss 2.6044, time 2051.94ms, mfu 33.94% +iter 18190: loss 2.4899, time 2030.59ms, mfu 34.01% +iter 18200: loss 2.4462, time 2037.56ms, mfu 34.05% +iter 18210: loss 2.3983, time 2049.02ms, mfu 34.08% +iter 18220: loss 2.4561, time 2051.31ms, mfu 34.09% +iter 18230: loss 2.5420, time 2044.02ms, mfu 34.12% +iter 18240: loss 2.6179, time 2023.95ms, mfu 34.18% +iter 18250: loss 2.6346, time 2048.23ms, mfu 34.19% +iter 18260: loss 2.3404, time 2047.15ms, mfu 34.21% +iter 18270: loss 2.4838, time 2041.89ms, mfu 34.23% +iter 18280: loss 2.4960, time 2044.54ms, mfu 34.24% +iter 18290: loss 2.4660, time 2045.21ms, mfu 34.25% +iter 18300: loss 2.6292, time 2043.23ms, mfu 34.26% +iter 18310: loss 2.3600, time 2042.28ms, mfu 34.28% +iter 18320: loss 2.6103, time 2048.39ms, mfu 34.28% +iter 18330: loss 2.9168, time 2040.72ms, mfu 34.30% +iter 18340: loss 2.3151, time 2048.41ms, mfu 34.30% +iter 18350: loss 2.5969, time 2049.46ms, mfu 34.29% +iter 18360: loss 2.4688, time 2045.08ms, mfu 34.30% +iter 18370: loss 2.3587, time 2025.80ms, mfu 34.34% +iter 18380: loss 2.5908, time 2035.95ms, mfu 34.36% +iter 18390: loss 2.6701, time 2038.69ms, mfu 34.37% +iter 18400: loss 2.3106, time 2020.63ms, mfu 34.41% +iter 18410: loss 2.6747, time 2047.00ms, mfu 34.40% +iter 18420: loss 2.3913, time 2030.54ms, mfu 34.42% +iter 18430: loss 2.6270, time 2035.82ms, mfu 34.43% +iter 18440: loss 1.9692, time 2030.70ms, mfu 34.44% +iter 18450: loss 2.6884, time 2034.75ms, mfu 34.45% +iter 18460: loss 2.3653, time 2022.74ms, mfu 34.48% +iter 18470: loss 2.6160, time 2046.20ms, mfu 34.47% +iter 18480: loss 2.4401, time 2029.32ms, mfu 34.48% +iter 18490: loss 2.4347, time 2050.40ms, mfu 34.46% +iter 18500: loss 2.3109, time 2027.22ms, mfu 34.48% +iter 18510: loss 2.6085, time 2029.05ms, mfu 34.49% +iter 18520: loss 2.5804, time 2045.09ms, mfu 34.48% +iter 18530: loss 2.5533, time 2041.40ms, mfu 34.47% +iter 18540: loss 2.6045, time 2033.03ms, mfu 34.48% +iter 18550: loss 2.5358, time 2041.20ms, mfu 34.48% +iter 18560: loss 2.5098, time 2028.49ms, mfu 34.49% +iter 18570: loss 2.4624, time 2039.25ms, mfu 34.49% +iter 18580: loss 2.5906, time 2034.30ms, mfu 34.49% +iter 18590: loss 2.4067, time 2042.36ms, mfu 34.48% +iter 18600: loss 2.5380, time 2042.66ms, mfu 34.48% +iter 18610: loss 2.4452, time 2032.47ms, mfu 34.48% +iter 18620: loss 2.4393, time 2056.12ms, mfu 34.45% +iter 18630: loss 2.2584, time 2045.95ms, mfu 34.44% +iter 18640: loss 2.6107, time 2045.91ms, mfu 34.43% +iter 18650: loss 2.7307, time 2042.87ms, mfu 34.43% +iter 18660: loss 2.5931, time 2036.80ms, mfu 34.43% +iter 18670: loss 2.3195, time 2046.09ms, mfu 34.42% +iter 18680: loss 2.4437, time 2052.15ms, mfu 34.41% +iter 18690: loss 2.4574, time 2049.43ms, mfu 34.39% +iter 18700: loss 2.4084, time 2036.04ms, mfu 34.40% +iter 18710: loss 2.8515, time 2032.45ms, mfu 34.42% +iter 18720: loss 2.5933, time 2023.79ms, mfu 34.45% +iter 18730: loss 2.4207, time 2053.02ms, mfu 34.43% +iter 18740: loss 2.4620, time 2046.36ms, mfu 34.42% +iter 18750: loss 2.1568, time 2046.31ms, mfu 34.41% +iter 18760: loss 2.5149, time 2039.52ms, mfu 34.41% +iter 18770: loss 2.4324, time 2050.11ms, mfu 34.40% +iter 18780: loss 2.4186, time 2045.70ms, mfu 34.39% +iter 18790: loss 2.5736, time 2055.88ms, mfu 34.37% +iter 18800: loss 2.6184, time 2041.83ms, mfu 34.38% +iter 18810: loss 2.2102, time 2046.84ms, mfu 34.37% +iter 18820: loss 2.4944, time 2051.02ms, mfu 34.36% +iter 18830: loss 2.3915, time 2050.44ms, mfu 34.35% +iter 18840: loss 2.5534, time 2048.20ms, mfu 34.34% +iter 18850: loss 2.8659, time 2051.50ms, mfu 34.34% +iter 18860: loss 2.3966, time 2045.36ms, mfu 34.34% +iter 18870: loss 2.6404, time 2042.98ms, mfu 34.34% +iter 18880: loss 2.7463, time 2040.94ms, mfu 34.35% +iter 18890: loss 2.7846, time 2046.15ms, mfu 34.35% +iter 18900: loss 2.3818, time 2048.30ms, mfu 34.34% +iter 18910: loss 2.5840, time 2053.99ms, mfu 34.33% +iter 18920: loss 2.5176, time 2048.05ms, mfu 34.33% +iter 18930: loss 2.7594, time 2043.07ms, mfu 34.33% +iter 18940: loss 2.4353, time 2034.33ms, mfu 34.35% +iter 18950: loss 2.5070, time 2048.95ms, mfu 34.35% +iter 18960: loss 2.3331, time 2042.46ms, mfu 34.35% +iter 18970: loss 2.2222, time 2039.34ms, mfu 34.36% +iter 18980: loss 2.4629, time 2059.51ms, mfu 34.34% +iter 18990: loss 2.8133, time 2073.84ms, mfu 34.29% +step 19000: train loss 2.4960, val loss 2.5795 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 19000: loss 2.6946, time 9157.05ms, mfu 31.63% +iter 19010: loss 2.6494, time 2067.11ms, mfu 31.87% +iter 19020: loss 2.2805, time 2074.73ms, mfu 32.07% +iter 19030: loss 2.6235, time 2094.20ms, mfu 32.21% +iter 19040: loss 2.5598, time 2070.01ms, mfu 32.39% +iter 19050: loss 2.2728, time 2073.01ms, mfu 32.54% +iter 19060: loss 2.3224, time 2066.86ms, mfu 32.68% +iter 19070: loss 2.2549, time 2048.33ms, mfu 32.84% +iter 19080: loss 2.5840, time 2075.74ms, mfu 32.94% +iter 19090: loss 2.3142, time 2059.30ms, mfu 33.06% +iter 19100: loss 2.4708, time 2078.70ms, mfu 33.14% +iter 19110: loss 2.5081, time 2069.21ms, mfu 33.22% +iter 19120: loss 2.6420, time 2067.26ms, mfu 33.29% +iter 19130: loss 2.4997, time 2087.81ms, mfu 33.33% +iter 19140: loss 2.5713, time 2106.13ms, mfu 33.33% +iter 19150: loss 2.3247, time 2104.61ms, mfu 33.34% +iter 19160: loss 2.5173, time 2090.53ms, mfu 33.36% +iter 19170: loss 2.1094, time 2096.41ms, mfu 33.38% +iter 19180: loss 2.5944, time 2114.51ms, mfu 33.36% +iter 19190: loss 2.4135, time 2113.27ms, mfu 33.35% +iter 19200: loss 2.2291, time 2077.70ms, mfu 33.40% +iter 19210: loss 2.3812, time 2088.56ms, mfu 33.42% +iter 19220: loss 2.3152, time 2115.48ms, mfu 33.40% +iter 19230: loss 2.6011, time 2098.44ms, mfu 33.41% +iter 19240: loss 2.1657, time 2077.72ms, mfu 33.45% +iter 19250: loss 2.3054, time 2076.12ms, mfu 33.49% +iter 19260: loss 2.6090, time 2105.13ms, mfu 33.48% +iter 19270: loss 2.2419, time 2105.72ms, mfu 33.47% +iter 19280: loss 2.2283, time 2125.57ms, mfu 33.43% +iter 19290: loss 2.3726, time 2105.62ms, mfu 33.42% +iter 19300: loss 2.4235, time 2105.86ms, mfu 33.41% +iter 19310: loss 2.7795, time 2076.56ms, mfu 33.46% +iter 19320: loss 2.7724, time 2073.76ms, mfu 33.50% +iter 19330: loss 2.2650, time 2079.17ms, mfu 33.53% +iter 19340: loss 2.6037, time 2106.95ms, mfu 33.51% +iter 19350: loss 2.4660, time 2120.26ms, mfu 33.47% +iter 19360: loss 2.5757, time 2101.19ms, mfu 33.47% +iter 19370: loss 2.2094, time 2082.51ms, mfu 33.50% +iter 19380: loss 2.6828, time 2124.18ms, mfu 33.45% +iter 19390: loss 2.7281, time 2109.57ms, mfu 33.44% +iter 19400: loss 2.4323, time 2078.23ms, mfu 33.48% +iter 19410: loss 2.2160, time 2111.44ms, mfu 33.46% +iter 19420: loss 2.3757, time 2097.52ms, mfu 33.46% +iter 19430: loss 2.6591, time 2083.16ms, mfu 33.49% +iter 19440: loss 2.5653, time 2119.07ms, mfu 33.45% +iter 19450: loss 2.3441, time 2072.10ms, mfu 33.50% +iter 19460: loss 2.6150, time 2081.46ms, mfu 33.52% +iter 19470: loss 2.4495, time 2105.86ms, mfu 33.51% +iter 19480: loss 2.1716, time 2121.41ms, mfu 33.47% +iter 19490: loss 2.7254, time 2068.88ms, mfu 33.52% +iter 19500: loss 2.5382, time 2086.21ms, mfu 33.53% +iter 19510: loss 2.3775, time 2092.54ms, mfu 33.54% +iter 19520: loss 2.4431, time 2124.95ms, mfu 33.49% +iter 19530: loss 2.3789, time 2102.55ms, mfu 33.48% +iter 19540: loss 2.5554, time 2091.72ms, mfu 33.49% +iter 19550: loss 2.2595, time 2123.87ms, mfu 33.45% +iter 19560: loss 2.3262, time 2086.51ms, mfu 33.47% +iter 19570: loss 2.2374, time 2075.02ms, mfu 33.51% +iter 19580: loss 2.4286, time 2079.22ms, mfu 33.54% +iter 19590: loss 2.4578, time 2078.14ms, mfu 33.57% +iter 19600: loss 2.4941, time 2082.04ms, mfu 33.59% +iter 19610: loss 2.2544, time 2109.73ms, mfu 33.56% +iter 19620: loss 2.9105, time 2125.76ms, mfu 33.51% +iter 19630: loss 2.7088, time 2124.45ms, mfu 33.46% +iter 19640: loss 2.4173, time 2110.34ms, mfu 33.45% +iter 19650: loss 2.5965, time 2122.06ms, mfu 33.41% +iter 19660: loss 2.7698, time 2095.87ms, mfu 33.42% +iter 19670: loss 2.6804, time 2106.50ms, mfu 33.42% +iter 19680: loss 2.5901, time 2107.80ms, mfu 33.41% +iter 19690: loss 2.2906, time 2097.26ms, mfu 33.42% +iter 19700: loss 2.5577, time 2069.51ms, mfu 33.47% +iter 19710: loss 2.8713, time 2073.22ms, mfu 33.51% +iter 19720: loss 2.6003, time 2070.65ms, mfu 33.55% +iter 19730: loss 2.4759, time 2066.23ms, mfu 33.60% +iter 19740: loss 2.7348, time 2056.79ms, mfu 33.65% +iter 19750: loss 2.3621, time 2056.41ms, mfu 33.71% +iter 19760: loss 2.6156, time 2045.84ms, mfu 33.77% +iter 19770: loss 2.3359, time 2040.40ms, mfu 33.84% +iter 19780: loss 2.5365, time 2083.39ms, mfu 33.82% +iter 19790: loss 2.7236, time 2097.98ms, mfu 33.79% +iter 19800: loss 2.3659, time 2081.01ms, mfu 33.79% +iter 19810: loss 2.0193, time 2075.69ms, mfu 33.79% +iter 19820: loss 2.4093, time 2061.28ms, mfu 33.82% +iter 19830: loss 2.5405, time 2071.63ms, mfu 33.83% +iter 19840: loss 2.2666, time 2059.57ms, mfu 33.86% +iter 19850: loss 2.1745, time 2078.81ms, mfu 33.85% +iter 19860: loss 2.7285, time 2082.36ms, mfu 33.84% +iter 19870: loss 2.5300, time 2077.24ms, mfu 33.84% +iter 19880: loss 2.5143, time 2110.61ms, mfu 33.79% +iter 19890: loss 2.4780, time 2128.99ms, mfu 33.71% +iter 19900: loss 2.7815, time 2108.06ms, mfu 33.67% +iter 19910: loss 2.9184, time 2086.88ms, mfu 33.67% +iter 19920: loss 2.3199, time 2097.57ms, mfu 33.65% +iter 19930: loss 2.3362, time 2079.79ms, mfu 33.66% +iter 19940: loss 2.4887, time 2091.47ms, mfu 33.66% +iter 19950: loss 2.4897, time 2113.30ms, mfu 33.62% +iter 19960: loss 2.7871, time 2098.47ms, mfu 33.60% +iter 19970: loss 2.4831, time 2113.39ms, mfu 33.57% +iter 19980: loss 2.3700, time 2090.87ms, mfu 33.57% +iter 19990: loss 2.5672, time 2111.66ms, mfu 33.54% +step 20000: train loss 2.4705, val loss 2.5600 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 20000: loss 2.3049, time 9127.17ms, mfu 30.96% +iter 20010: loss 2.5311, time 2082.86ms, mfu 31.23% +iter 20020: loss 2.7243, time 2070.26ms, mfu 31.50% +iter 20030: loss 2.7753, time 2078.03ms, mfu 31.73% +iter 20040: loss 2.1705, time 2072.72ms, mfu 31.95% +iter 20050: loss 2.4184, time 2083.52ms, mfu 32.13% +iter 20060: loss 2.4823, time 2109.43ms, mfu 32.25% +iter 20070: loss 2.3417, time 2107.97ms, mfu 32.35% +iter 20080: loss 2.4189, time 2099.63ms, mfu 32.46% +iter 20090: loss 2.5853, time 2092.77ms, mfu 32.58% +iter 20100: loss 2.5596, time 2096.81ms, mfu 32.67% +iter 20110: loss 2.6586, time 2126.40ms, mfu 32.71% +iter 20120: loss 2.6058, time 2117.85ms, mfu 32.75% +iter 20130: loss 2.2927, time 2080.97ms, mfu 32.85% +iter 20140: loss 2.8337, time 2070.70ms, mfu 32.96% +iter 20150: loss 2.4707, time 2109.90ms, mfu 33.00% +iter 20160: loss 2.6414, time 2122.88ms, mfu 33.01% +iter 20170: loss 2.3587, time 2113.06ms, mfu 33.03% +iter 20180: loss 2.5295, time 2085.31ms, mfu 33.10% +iter 20190: loss 2.5012, time 2062.77ms, mfu 33.19% +iter 20200: loss 2.4359, time 2088.61ms, mfu 33.24% +iter 20210: loss 2.5483, time 2093.39ms, mfu 33.27% +iter 20220: loss 2.6374, time 2075.75ms, mfu 33.33% +iter 20230: loss 2.0785, time 2072.26ms, mfu 33.38% +iter 20240: loss 1.9792, time 2073.14ms, mfu 33.44% +iter 20250: loss 2.4951, time 2077.57ms, mfu 33.47% +iter 20260: loss 2.3644, time 2073.81ms, mfu 33.51% +iter 20270: loss 2.5195, time 2048.74ms, mfu 33.59% +iter 20280: loss 2.7400, time 2078.72ms, mfu 33.61% +iter 20290: loss 2.7078, time 2056.61ms, mfu 33.67% +iter 20300: loss 2.3227, time 2051.42ms, mfu 33.73% +iter 20310: loss 2.5757, time 2076.06ms, mfu 33.74% +iter 20320: loss 2.6827, time 2079.06ms, mfu 33.74% +iter 20330: loss 2.5830, time 2090.59ms, mfu 33.73% +iter 20340: loss 2.6910, time 2079.09ms, mfu 33.74% +iter 20350: loss 2.4816, time 2069.08ms, mfu 33.76% +iter 20360: loss 2.4796, time 2085.73ms, mfu 33.75% +iter 20370: loss 2.5446, time 2064.10ms, mfu 33.78% +iter 20380: loss 2.3412, time 2065.52ms, mfu 33.80% +iter 20390: loss 2.5178, time 2047.76ms, mfu 33.85% +iter 20400: loss 2.7551, time 2061.34ms, mfu 33.88% +iter 20410: loss 2.5763, time 2073.31ms, mfu 33.88% +iter 20420: loss 2.7298, time 2076.27ms, mfu 33.87% +iter 20430: loss 2.3009, time 2065.80ms, mfu 33.89% +iter 20440: loss 2.3277, time 2098.70ms, mfu 33.85% +iter 20450: loss 2.1856, time 2098.19ms, mfu 33.81% +iter 20460: loss 2.4148, time 2092.47ms, mfu 33.79% +iter 20470: loss 2.8183, time 2112.99ms, mfu 33.73% +iter 20480: loss 2.6513, time 2095.29ms, mfu 33.71% +iter 20490: loss 2.5901, time 2095.69ms, mfu 33.69% +iter 20500: loss 2.3461, time 2101.42ms, mfu 33.67% +iter 20510: loss 2.0833, time 2112.99ms, mfu 33.63% +iter 20520: loss 2.2939, time 2115.17ms, mfu 33.59% +iter 20530: loss 2.5580, time 2096.05ms, mfu 33.58% +iter 20540: loss 2.7201, time 2099.78ms, mfu 33.57% +iter 20550: loss 2.4388, time 2110.35ms, mfu 33.54% +iter 20560: loss 2.5888, time 2111.72ms, mfu 33.51% +iter 20570: loss 2.0955, time 2076.86ms, mfu 33.54% +iter 20580: loss 2.6397, time 2109.87ms, mfu 33.52% +iter 20590: loss 2.6569, time 2084.79ms, mfu 33.54% +iter 20600: loss 2.1208, time 2110.78ms, mfu 33.51% +iter 20610: loss 2.3751, time 2116.68ms, mfu 33.48% +iter 20620: loss 2.4391, time 2078.04ms, mfu 33.51% +iter 20630: loss 2.3905, time 2083.16ms, mfu 33.53% +iter 20640: loss 2.6307, time 2086.27ms, mfu 33.55% +iter 20650: loss 2.5851, time 2084.46ms, mfu 33.56% +iter 20660: loss 2.6328, time 2118.20ms, mfu 33.53% +iter 20670: loss 2.7644, time 2079.58ms, mfu 33.55% +iter 20680: loss 2.4270, time 2067.97ms, mfu 33.59% +iter 20690: loss 2.5857, time 2076.95ms, mfu 33.62% +iter 20700: loss 2.5030, time 2092.26ms, mfu 33.61% +iter 20710: loss 2.2983, time 2079.15ms, mfu 33.63% +iter 20720: loss 2.7840, time 2073.03ms, mfu 33.66% +iter 20730: loss 2.7703, time 2072.05ms, mfu 33.68% +iter 20740: loss 2.3082, time 2115.14ms, mfu 33.64% +iter 20750: loss 2.5007, time 2090.94ms, mfu 33.63% +iter 20760: loss 2.3180, time 2090.88ms, mfu 33.63% +iter 20770: loss 2.3490, time 2074.07ms, mfu 33.65% +iter 20780: loss 2.7984, time 2130.72ms, mfu 33.59% +iter 20790: loss 2.5980, time 2119.87ms, mfu 33.54% +iter 20800: loss 2.3243, time 2106.92ms, mfu 33.52% +iter 20810: loss 2.6890, time 2102.07ms, mfu 33.51% +iter 20820: loss 2.3951, time 2086.62ms, mfu 33.53% +iter 20830: loss 2.3490, time 2079.88ms, mfu 33.55% +iter 20840: loss 2.1059, time 2085.54ms, mfu 33.57% +iter 20850: loss 2.5789, time 2090.74ms, mfu 33.57% +iter 20860: loss 2.5478, time 2110.53ms, mfu 33.54% +iter 20870: loss 2.7613, time 2082.73ms, mfu 33.56% +iter 20880: loss 2.5769, time 2107.56ms, mfu 33.54% +iter 20890: loss 2.2841, time 2092.34ms, mfu 33.54% +iter 20900: loss 2.6098, time 2091.61ms, mfu 33.55% +iter 20910: loss 2.9469, time 2090.03ms, mfu 33.55% +iter 20920: loss 2.4519, time 2090.17ms, mfu 33.56% +iter 20930: loss 2.1891, time 2088.59ms, mfu 33.57% +iter 20940: loss 2.4173, time 2101.59ms, mfu 33.55% +iter 20950: loss 2.5453, time 2101.07ms, mfu 33.54% +iter 20960: loss 2.3601, time 2097.66ms, mfu 33.54% +iter 20970: loss 2.4965, time 2076.41ms, mfu 33.57% +iter 20980: loss 2.3732, time 2082.07ms, mfu 33.59% +iter 20990: loss 2.5255, time 2081.84ms, mfu 33.60% +step 21000: train loss 2.4877, val loss 2.5653 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 21000: loss 2.3033, time 9065.20ms, mfu 31.02% +iter 21010: loss 2.7702, time 2081.10ms, mfu 31.29% +iter 21020: loss 2.2506, time 2122.56ms, mfu 31.47% +iter 21030: loss 2.4540, time 2108.63ms, mfu 31.66% +iter 21040: loss 2.5078, time 2084.83ms, mfu 31.86% +iter 21050: loss 2.4593, time 2083.02ms, mfu 32.05% +iter 21060: loss 2.2905, time 2111.12ms, mfu 32.17% +iter 21070: loss 2.6290, time 2090.70ms, mfu 32.31% +iter 21080: loss 2.4797, time 2077.13ms, mfu 32.47% +iter 21090: loss 2.3769, time 2074.53ms, mfu 32.61% +iter 21100: loss 2.4844, time 2087.15ms, mfu 32.71% +iter 21110: loss 2.1747, time 2074.18ms, mfu 32.83% +iter 21120: loss 2.5464, time 2059.90ms, mfu 32.96% +iter 21130: loss 2.6820, time 2064.25ms, mfu 33.06% +iter 21140: loss 2.6739, time 2054.59ms, mfu 33.18% +iter 21150: loss 2.6393, time 2069.52ms, mfu 33.25% +iter 21160: loss 2.7563, time 2065.32ms, mfu 33.33% +iter 21170: loss 2.2612, time 2048.20ms, mfu 33.43% +iter 21180: loss 2.5129, time 2059.80ms, mfu 33.50% +iter 21190: loss 2.6525, time 2048.66ms, mfu 33.58% +iter 21200: loss 2.4726, time 2044.12ms, mfu 33.66% +iter 21210: loss 2.5289, time 2064.42ms, mfu 33.69% +iter 21220: loss 2.7360, time 2047.37ms, mfu 33.76% +iter 21230: loss 2.4160, time 2047.87ms, mfu 33.81% +iter 21240: loss 2.3951, time 2065.11ms, mfu 33.83% +iter 21250: loss 2.5662, time 2046.70ms, mfu 33.88% +iter 21260: loss 2.3221, time 2054.57ms, mfu 33.91% +iter 21270: loss 2.5679, time 2052.95ms, mfu 33.94% +iter 21280: loss 2.6328, time 2045.00ms, mfu 33.98% +iter 21290: loss 2.6171, time 2042.36ms, mfu 34.03% +iter 21300: loss 1.9673, time 2047.65ms, mfu 34.05% +iter 21310: loss 2.4585, time 2055.41ms, mfu 34.07% +iter 21320: loss 2.2799, time 2057.06ms, mfu 34.08% +iter 21330: loss 2.3813, time 2045.73ms, mfu 34.10% +iter 21340: loss 2.5388, time 2043.79ms, mfu 34.13% +iter 21350: loss 2.4416, time 2049.61ms, mfu 34.15% +iter 21360: loss 2.2244, time 2053.65ms, mfu 34.15% +iter 21370: loss 2.4727, time 2044.09ms, mfu 34.17% +iter 21380: loss 2.4452, time 2045.23ms, mfu 34.19% +iter 21390: loss 2.3373, time 2046.00ms, mfu 34.21% +iter 21400: loss 2.5141, time 2048.81ms, mfu 34.21% +iter 21410: loss 2.5042, time 2054.62ms, mfu 34.21% +iter 21420: loss 2.3967, time 2048.00ms, mfu 34.22% +iter 21430: loss 2.3168, time 2046.91ms, mfu 34.23% +iter 21440: loss 2.5505, time 2044.47ms, mfu 34.25% +iter 21450: loss 2.6172, time 2045.61ms, mfu 34.26% +iter 21460: loss 2.1205, time 2042.59ms, mfu 34.27% +iter 21470: loss 2.3121, time 2047.03ms, mfu 34.28% +iter 21480: loss 2.6566, time 2039.52ms, mfu 34.29% +iter 21490: loss 2.3667, time 2051.38ms, mfu 34.29% +iter 21500: loss 2.5575, time 2044.32ms, mfu 34.30% +iter 21510: loss 2.4720, time 2046.98ms, mfu 34.30% +iter 21520: loss 2.5616, time 2030.81ms, mfu 34.33% +iter 21530: loss 2.3708, time 2025.87ms, mfu 34.36% +iter 21540: loss 2.6299, time 2053.08ms, mfu 34.35% +iter 21550: loss 2.3262, time 2058.04ms, mfu 34.33% +iter 21560: loss 2.3968, time 2025.53ms, mfu 34.36% +iter 21570: loss 2.5163, time 2037.70ms, mfu 34.38% +iter 21580: loss 2.3235, time 2044.13ms, mfu 34.38% +iter 21590: loss 2.5240, time 2062.71ms, mfu 34.34% +iter 21600: loss 2.6779, time 2048.24ms, mfu 34.34% +iter 21610: loss 2.4958, time 2041.75ms, mfu 34.35% +iter 21620: loss 2.1344, time 2048.44ms, mfu 34.34% +iter 21630: loss 2.7206, time 2035.28ms, mfu 34.36% +iter 21640: loss 2.7461, time 2049.99ms, mfu 34.35% +iter 21650: loss 2.6700, time 2051.29ms, mfu 34.34% +iter 21660: loss 2.4399, time 2033.07ms, mfu 34.36% +iter 21670: loss 2.6212, time 2052.43ms, mfu 34.35% +iter 21680: loss 2.7213, time 2046.05ms, mfu 34.35% +iter 21690: loss 2.5390, time 2053.74ms, mfu 34.33% +iter 21700: loss 2.6351, time 2035.97ms, mfu 34.35% +iter 21710: loss 2.4779, time 2042.10ms, mfu 34.36% +iter 21720: loss 2.3776, time 2054.87ms, mfu 34.34% +iter 21730: loss 2.3918, time 2048.39ms, mfu 34.34% +iter 21740: loss 2.4109, time 2034.45ms, mfu 34.36% +iter 21750: loss 2.3127, time 2041.94ms, mfu 34.36% +iter 21760: loss 2.4471, time 2051.70ms, mfu 34.35% +iter 21770: loss 2.3663, time 2039.56ms, mfu 34.36% +iter 21780: loss 2.7221, time 2055.04ms, mfu 34.34% +iter 21790: loss 2.6982, time 2039.90ms, mfu 34.35% +iter 21800: loss 2.4300, time 2056.98ms, mfu 34.33% +iter 21810: loss 2.6521, time 2058.72ms, mfu 34.31% +iter 21820: loss 2.5549, time 2062.07ms, mfu 34.29% +iter 21830: loss 2.4077, time 2044.30ms, mfu 34.30% +iter 21840: loss 2.4835, time 2047.11ms, mfu 34.30% +iter 21850: loss 2.5915, time 2043.43ms, mfu 34.31% +iter 21860: loss 2.4817, time 2046.34ms, mfu 34.31% +iter 21870: loss 2.5733, time 2049.69ms, mfu 34.31% +iter 21880: loss 2.4169, time 2035.20ms, mfu 34.33% +iter 21890: loss 2.7048, time 2037.58ms, mfu 34.34% +iter 21900: loss 2.1925, time 2051.87ms, mfu 34.33% +iter 21910: loss 2.6805, time 2058.14ms, mfu 34.31% +iter 21920: loss 2.5104, time 2054.06ms, mfu 34.30% +iter 21930: loss 2.3913, time 2040.68ms, mfu 34.32% +iter 21940: loss 2.4296, time 2046.36ms, mfu 34.32% +iter 21950: loss 2.3566, time 2027.26ms, mfu 34.35% +iter 21960: loss 2.5261, time 2042.07ms, mfu 34.36% +iter 21970: loss 2.7742, time 2060.12ms, mfu 34.33% +iter 21980: loss 2.9902, time 2060.84ms, mfu 34.31% +iter 21990: loss 2.8476, time 2034.09ms, mfu 34.33% +step 22000: train loss 2.4545, val loss 2.5775 +saving checkpoint to ../runs/minimal-en-125m-4ep +iter 22000: loss 2.6384, time 9025.38ms, mfu 31.68% +saving final checkpoint to ../runs/minimal-en-125m-4ep +wandb: +wandb: 🚀 View run 125m-12k-4ep at: https://wandb.ai/dawidmajewski-ezotic/minimal-en/runs/hlbpqpoe +wandb: Find logs at: ../logs/wandb/wandb/run-20260824_231331-hlbpqpoe/logs