stale
concordance-v5
34.16M parameter concordance model — bpe-8k tokenizer, 8L/512D/8H
Overview
34.16M
Parameters
1.1244
Final Loss
1.1277
Best Val Loss
3.1
Perplexity
61,849,600
Tokens Processed
1.8
Tokens/Param
7,168 tok/s
Avg Throughput
2h 23m
Training Time
Training Progress3,775 / 100,000 steps (3.8%)
Loss reduced by 87.6% from initial 9.0988
Dataset & Training
Domainconcordance
Tokenizerbpe-8k
Total Iterations100,000
Batch Size32
Context Length512 tokens
Tokens per Batch16,384
Dataset Passes~77
Effective Tokens61,849,600
Training Pipeline
Warmupsteps 1–1,000
Learning rate warmup — model weights adjusting to data distribution
Loss: 9.099 → 5.636Linear 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)
Parameters34.16M
Layers8
Embedding Dim512
Attention Heads8
Head Dim64
FFN Dim2048
FFN Activationswiglu
Vocab Size8,000
Context Length512 tokens
Dropout0
Training Configuration
Optimizeradamw
Learning Rate0.0006
LR Min0.00006
LR ScheduleCosine decay
Warmup Steps200
Batch Size32
Grad Accum Steps1
Effective Batch32
Grad Clip1
Weight Decay0.1
Backendhelios
Tokenizerbpe-8k
Seed42
Layer Structure
Token Embed
8,000×512
Pos Embed
512×512
Block 0
Attn+FFN
Block 1
Attn+FFN
Block 2
Attn+FFN
Block 3
Attn+FFN
Block 4
Attn+FFN
Block 5
Attn+FFN
LayerNorm
512
LM Head
512×8,000
Generated Samples
Step 0 — Mar 9, 2026 5:00 AM
Prompt: The
The Raut make um ; I Spanish 23sis, henS 04risUA6esbody dsnSool has AGeno Ho sudSzestós aros si 817 Loperse, seriknowledor the EquZon Tjo ltarpicomanddu
Step 1 — Mar 9, 2026 5:00 AM
Prompt: When in the Course of human events
When in the Course of human eventse smilusu. It was n t 't they mako Henganhae ą opkariil, Hel. Hotquay, Los Sa?, beavy Sedset sermemation of er ( Strwe3 back DUe forf
i all aggo ina icenderihim
Step 2 — Mar 9, 2026 5:00 AM
Prompt: It was a dark and stormy
It was a dark and stormydintkswashondo green oktrits (safasand ahaman, erwe lasople of are killemica biamitęcmandentesēot ent en Blfriown ongimfrdoaus S. L pbifiall Udback T
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: