FORM ADV · TRAINING CONTROL
ELAPSED
--:--:--
ETA
--:--:--
RATE
,
step/s
LR
,
CONNECTING
--:--:--
Training Progress
0.0%
Step
,
/
,
Epoch
,
Loss
,
Best
,
LR
,
Loss Curve
System Status
GPU VRAM
,
GPU UTIL
,
GPU TEMP
,
SYS RAM
,
CPU
,
%
Training Dataset
Total Examples
,
Live Training Log
Connecting…
·
,
Waiting for dashboard…
Controls
▶ WARM RESUME
■ STOP
↺ RECONNECT
Saved Checkpoints
None yet
Model Config
Base Model
TinyLlama-1.1B
LoRA Rank / α
16 / 32
Trainable Params
12.6M (1.13%)
Max Seq Length
512 tokens
Quantization
4-bit NF4
Optimizer
AdamW 8-bit
Batch × Accum
1 × 8 = 8 eff.
Scheduler
Cosine + 100 warmup