fine_corpus_clean_v10stalefine_corpus20.51M params41m 52s elapsed · Updated 36d ago
8L / 384D / 6H · helios · bpe-8k · adamw· Created Mar 8, 2026 1:03 PM
Step 17,941 / 150,00012.0%
6.9952
Loss?
6.7717
Best Loss?
-22.7% from start
7.9269
Val Loss?
best: 7.2329
2.99e-5
Learning Rate?
8,082
Throughput?
tok/s (avg)
254
Speed?
ms/iter (avg)
1.487
Grad Norm?
avg: 0.074
20.48M
Tokens
processed
Loss Curve ? click any chart to add markers
?
?
?
?
Architecture
Layers?8
Embedding?384
Heads?6
Vocab?8,000
Context?512
Dropout?0
Parameters?20.51M
Training Config
Total iters?150,000
Batch size?4
Max LR?0.00003
Optimizer?adamw
Backend?helios
Tokenizer?bpe-8k
Seed?42
Weight decay?0.1
Grad clip?0
Eval interval?2000
Throughput (tok/s)
Step Time (ms/iter)
GPU & VRAM
Perplexity
Train/Val Gap
Learning Rate
Grad Norm
Smoothed Loss (EMA)
Loss Velocity
Gradient Clipping
GPU Operations
Step Time Breakdown
No timing data
Timing Phase Lines
No timing data
Backward / Forward Ratio
No timing data
Transformer Layer Analysis
Gradient Norm Heatmap
Per-Layer Gradient Evolution
Checkpoints (0) ?
No checkpoints saved
Sample Generations (1)
#CheckpointPrompt (preview)Generated
1-The 36d ago
Prompt
The
Output
The C
s. He would which was Nowed a M
of
Model Config (JSON)
{
"vocabSize": 8000,
"blockSize": 512,
"nLayer": 8,
"nEmbd": 384,
"nHead": 6,
"dropout": 0,
"ffnActivation": "gelu"
}Training Config (JSON)
{
"iters": 150000,
"batchSize": 4,
"lr": 0.00003,
"lrMin": 0.000003,
"warmupIters": 5000,
"beta1": 0.9,
"beta2": 0.999,
"eps": 1e-8,
"weightDecay": 0.1,
"gradClip": 0,
"evalInterval": 2000,
"evalIters": 10,
"seed": 42,
"backend": "helios",
"tokenizer": "bpe-8k",
"optimizer": "adamw",
"logLevel": "info",
"logEvery": 25,
"trace": false,
"gradAccumSteps": 1,
"sampleInterval": 10000,
"spikeThreshold": 0,
"embGradScale": 0.1,
"syncEvery": 1,
"gcEvery": 1,
"packed": true,
"symbio": false,
"symbioConfig": null
}