A
Alpha
20260225011732_xk5ostalechat17.44M params2h 52m elapsed · Updated 48d ago
8L / 384D / 8H · helios · bpe-4k · adamw· Created Feb 25, 2026 1:18 AM
Step 4,935 / 50,0009.9%
4.4456
Loss?
4.1969
Best Loss?
-46.9% from start
5.2024
Val Loss?
best: 5.2024
4.93e-5
Learning Rate?
4,775
Throughput?
tok/s (avg)
2,145
Speed?
ms/iter (avg)
2.547
Grad Norm?
avg: 1.551
50.43M
Tokens
processed
1034ms
Forward
48% of step
975ms
Backward
45% of step
45ms
GPU Sync
2% of step
935
GPU Ops
per step
1.6%
MFU
model FLOPS util
0.9x
Bwd/Fwd
ratio
Loss Curve ? click any chart to add markers
?
?
?
?
Architecture
Layers?8
Embedding?384
Heads?8
Vocab?4,000
Context?512
Dropout?0.1
Parameters?17.44M
Training Config
Total iters?50,000
Batch size?20
Max LR?0.00005
Optimizer?adamw
Backend?helios
Tokenizer?bpe-4k
Seed?42
Weight decay?0.1
Grad clip?1
Eval interval?1000
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
Step Time Breakdown
Forward
Backward
Grad Norm
Optimizer
GPU Sync
Data
Timing Phase Lines
Backward / Forward Ratio
Checkpoints (0) ?
No checkpoints saved
Sample Generations (3)
#CheckpointPrompt (preview)Generated
1-<|user|> Hello, how are you? <|assistant|>48d ago
Prompt
<|user|> Hello, how are you? <|assistant|>
Output
<|user|> Hello, how are you? <|assistant|>—in with and inct to a centas not to a with that sto unns, or rather than a I t and and which an that st and unct to its is ines unm, yet to I maintain that the s
2-<|user|> What do you like to do for fun? <|assistant|>48d ago
Prompt
<|user|> What do you like to do for fun? <|assistant|>
Output
<|user|> What do you like to do for fun? <|assistant|>of their reds to dto un, in in k, we es our by des, in unduys, not all must and dilies it; that the sch; their what of un—and ps, an to st
3-<|user|> Tell me about yourself. <|assistant|>48d ago
Prompt
<|user|> Tell me about yourself. <|assistant|>
Output
<|user|> Tell me about yourself. <|assistant|>yaas in our to ; the selgs, as may inal gs with to fals, the nature of but are where p ors, of the a the s: it that ins of the sacli—that patt ace
Model Config (JSON)
{
  "vocabSize": 4000,
  "blockSize": 512,
  "nLayer": 8,
  "nEmbd": 384,
  "nHead": 8,
  "dropout": 0.1
}
Training Config (JSON)
{
  "iters": 50000,
  "batchSize": 20,
  "lr": 0.00005,
  "lrMin": 0.000005,
  "warmupIters": 1000,
  "beta1": 0.9,
  "beta2": 0.95,
  "eps": 0.000001,
  "weightDecay": 0.1,
  "gradClip": 1,
  "evalInterval": 1000,
  "evalIters": 10,
  "seed": 42,
  "backend": "helios",
  "tokenizer": "bpe-4k",
  "optimizer": "adamw",
  "logLevel": "info",
  "trace": false,
  "gradAccumSteps": 1,
  "sampleInterval": 500,
  "spikeThreshold": 0
}