Training Status · EchoNest-1
We're training our first model.
EchoNest-1 is Frontier's inaugural foundation model — a hybrid recurrent-attention architecture designed to run efficiently on CPU and edge hardware. Here's where we are.
Training Phases
01
Data pipeline & tokenizer
Built and validated the tokenizer corpus, preprocessing pipeline, and data loader. Training set: Alpaca (52k) + Python coding (18k) = ~70k instruction-response pairs.
02
Training
Waiting for first epoch to start…
03
Evaluation & release
Benchmark evaluation, safety review, and staged API rollout. Early access invites will go out before general availability.
Overall Progress
Fine-tuning run · 10 epochs
—
Current Metrics
Train Loss
—
current batch
Val Loss
—
last epoch end
Val Perplexity
—
lower is better
Grad Norm
—
gradient L2 norm
Learning Rate
—
current LR
Global Step
—
optimizer steps
Loss History
Batch training loss (last 200 batches)
no data yet
—