Mixed Precision Estimator

FP16 / BF16 / TF32, memory & speed vs FP32 baseline

e.g., 350M = 350
hidden dim estimated: 4096
Model memory,
Optimizer states,
Activations (est.),
Gradients,
Total VRAM,
Speedup vs FP32,
Memory savings,
ModeModelOptimizerActivationsTotal VRAMSpeedupSavings
# PyTorch AMP / autocast example (FP16 mixed)
import torch
from torch.cuda.amp import autocast, GradScaler

scaler = GradScaler()
for data, target in dataloader:
  with autocast(dtype=torch.float16):
    output = model(data)
    loss = loss_fn(output, target)
  scaler.scale(loss).backward()
  scaler.step(optimizer)
  scaler.update()
© 2025 heytensor.com

TENSOR PREFLIGHT · DOWNLOADABLE PYTORCH TOOLKIT

Take the next step in your own PyTorch model.

Trace a local forward pass, inspect tensor shapes in an HTML report, and turn a repair into a regression check. Includes eight broken-and-fixed workflows, source code and setup instructions.

Python 3.9+ · PyTorch 2.8+ · Runs locally

Get Tensor Preflight — $29 One-time payment · ZIP download See a sample report