20260222104034_9q15staleconcordance172.23M params24m 3s elapsed · Updated 36d ago
12L / 768D / 12H · helios · bpe-64k · adamw· Created Feb 22, 2026 10:40 AM
Step 435 / 2,00021.8%
7.5514
Loss?
6.5595
Best Loss?
-23.7% from start
7.6278
Val Loss?
best: 7.6278
3.58e-5
Learning Rate?
1,385
Throughput?
tok/s (avg)
2,957
Speed?
ms/iter (avg)
6.632
Grad Norm?
avg: 68.782
1.99M
Tokens
processed
Loss Curve ? click any chart to add markers
?
?
?
?
Architecture
Layers?12
Embedding?768
Heads?12
Vocab?19,777
Context?256
Dropout?0
Parameters?172.23M
Training Config
Total iters?2,000
Batch size?16
Max LR?0.00004
Optimizer?adamw
Backend?helios
Tokenizer?bpe-64k
Seed?42
Weight decay?0.01
Grad clip?1
Eval interval?100
Throughput (tok/s)
Step Time (ms/iter)
GPU & VRAM
Perplexity
Train/Val Gap
Learning Rate
Grad Norm
Smoothed Loss (EMA)
Loss Velocity
Gradient Clipping
No clipping data
GPU Operations
No GPU ops data
Step Time Breakdown
No timing data
Timing Phase Lines
No timing data
Backward / Forward Ratio
No timing data
Model Config (JSON)
{
"vocabSize": 19777,
"blockSize": 256,
"nLayer": 12,
"nEmbd": 768,
"nHead": 12,
"dropout": 0
}Training Config (JSON)
{
"iters": 2000,
"batchSize": 16,
"lr": 0.00004,
"beta1": 0.9,
"beta2": 0.999,
"eps": 1e-8,
"weightDecay": 0.01,
"gradClip": 1,
"evalInterval": 100,
"evalIters": 10,
"seed": 42,
"backend": "helios",
"tokenizer": "bpe-64k",
"optimizer": "adamw",
"logLevel": "info",
"trace": false
}