A
Alpha
stale

fine_corpus_clean_v10

20.51M parameter fine_corpus model — bpe-8k tokenizer, 8L/384D/6H

Overview

20.51M
Parameters
6.9952
Final Loss
7.2329
Best Val Loss
1091.4
Perplexity
36,743,168
Tokens Processed
1.8
Tokens/Param
8,172 tok/s
Avg Throughput
41m 52s
Training Time
Training Progress17,941 / 150,000 steps (12.0%)
Loss reduced by 22.7% from initial 9.0450

Dataset & Training

Domainfine_corpus
Tokenizerbpe-8k
Total Iterations150,000
Batch Size4
Context Length512 tokens
Tokens per Batch2,048
Dataset Passes~46
Effective Tokens36,743,168

Training Pipeline

Warmupsteps 11,500

Learning rate warmup — model weights adjusting to data distribution

Loss: 9.0457.666Linear LR warmup, gradient clipping

Training Metrics

Loss Curve
?
?
?
?
Smoothed Loss
Perplexity
Learning Rate
Gradient Norm
Throughput (tok/s)
Timing Breakdown
No Telemetry

Model Architecture

Model Configuration

ArchitectureGPT (decoder-only transformer)
Parameters20.51M
Layers8
Embedding Dim384
Attention Heads6
Head Dim64
FFN Dim1536
FFN Activationgelu
Vocab Size8,000
Context Length512 tokens
Dropout0

Training Configuration

Optimizeradamw
Learning Rate0.00003
LR Min0.000003
LR ScheduleCosine decay
Warmup Steps5,000
Batch Size4
Grad Accum Steps1
Effective Batch4
Grad Clip0
Weight Decay0.1
Backendhelios
Tokenizerbpe-8k
Seed42

Layer Structure

Token Embed
8,000×384
Pos Embed
512×384
Block 0
Attn+FFN
Block 1
Attn+FFN
Block 2
Attn+FFN
Block 3
Attn+FFN
Block 4
Attn+FFN
Block 5
Attn+FFN
...2 more
LayerNorm
384
LM Head
384×8,000

Generated Samples

Step 0Mar 8, 2026 1:46 PM
Prompt: The
The C s. He would which was Nowed a M of

Checkpoints

No checkpoints saved yet.

Chat with Model

Send a message to chat with this model
Generated Invalid Date Invalid Date — Alpha Training SystemConfig hash: