completed
super_chat_20260307010642_qe06
9.99M parameter super_chat model — bpe-chat-4k tokenizer, 6L/256D/8H
Overview
9.99M
Parameters
5.2090
Final Loss
6.5487
Best Val Loss
182.9
Perplexity
15,360,000
Tokens Processed
1.5
Tokens/Param
27,161 tok/s
Avg Throughput
13m 45s
Training Time
Training Progress15,000 / 15,000 steps (100.0%)
Loss reduced by 37.6% from initial 8.3457
Dataset & Training
Domainsuper_chat
Tokenizerbpe-chat-4k
Total Iterations15,000
Batch Size4
Context Length256 tokens
Tokens per Batch1,024
Dataset Passes~38
Effective Tokens15,360,000
Training Pipeline
Warmupsteps 1–150
Learning rate warmup — model weights adjusting to data distribution
Loss: 8.346 → 7.377Linear LR warmup, gradient clipping
Rapid Descentsteps 150–4,501
Steepest loss reduction — model learning primary patterns
Loss: 7.377 → 6.370Cosine 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)
Parameters9.99M
Layers6
Embedding Dim256
Attention Heads8
Head Dim32
FFN Dim1024
FFN Activationgelu
Vocab Size4,000
Context Length256 tokens
Dropout0.1
Training Configuration
Optimizeradamw
Learning Rate0.0001
LR Min0.00001
LR ScheduleCosine decay
Warmup Steps500
Batch Size4
Grad Accum Steps2
Effective Batch8
Grad Clip1
Weight Decay0.1
Backendhelios
Tokenizerbpe-chat-4k
Seed42
Layer Structure
Token Embed
4,000×256
Pos Embed
256×256
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
256
LM Head
256×4,000
Generated Samples
Step 0 — Mar 7, 2026 1:35 AM
Prompt: <|user|> Hello, how are you? <|assistant|>
<|user|> Hello, how are you? <|assistant|>? ? it<|end_of_text|>? s <|end_of_text|>s ? it? <|end_of_text|><|end_of_text|>
Checkpoints
| Step | File | Size | Date |
|---|---|---|---|
| 15,000 | checkpoint-15000.json | 27.6 MB | Mar 8, 2026 4:44 AM |
Chat with Model
Send a message to chat with this model
Generated Invalid Date Invalid Date — Alpha Training SystemConfig hash: