stale
minipile_english_20260311001814_bvwz
97.65M parameter unknown model — bpe-8k-chat tokenizer, 12L/768D/12H
Overview
97.65M
Parameters
4.8662
Final Loss
4.9547
Best Val Loss
129.8
Perplexity
13,312,000
Tokens Processed
0.1
Tokens/Param
2,043 tok/s
Avg Throughput
7h 14m
Training Time
Training Progress3,250 / 10,000 steps (32.5%)
Loss reduced by 46.7% from initial 9.1365
Dataset & Training
Domainunknown
Tokenizerbpe-8k-chat
Total Iterations10,000
Batch Size8
Context Length512 tokens
Tokens per Batch4,096
Dataset Passes~17
Effective Tokens13,312,000
Training Pipeline
Warmupsteps 1–100
Learning rate warmup — model weights adjusting to data distribution
Loss: 9.136 → 7.461Linear LR warmup, gradient clipping
Rapid Descentsteps 100–3,001
Steepest loss reduction — model learning primary patterns
Loss: 7.461 → 5.010Cosine LR schedule, AdamW optimization
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)
Parameters97.65M
Layers12
Embedding Dim768
Attention Heads12
Head Dim64
FFN Dim3072
FFN Activationgelu
Vocab Size8,000
Context Length512 tokens
Dropout0
Training Configuration
Optimizeradamw
Learning Rate0.0003
LR Min0.00003
LR ScheduleCosine decay
Warmup Steps500
Batch Size8
Grad Accum Steps4
Effective Batch32
Grad Clip1
Weight Decay0.1
Backendhelios
Tokenizerbpe-8k-chat
Seed42
Layer Structure
Token Embed
8,000×768
Pos Embed
512×768
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
768
LM Head
768×8,000
Generated Samples
No samples generated yet. Samples appear at configured intervals during training.
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: