Machine Learning Calculators

How do I calculate llm training cost?

LLM Training Cost

Estimate total cost using C=6PD formula, GPU type, cloud pricing. Based on Chinchilla scaling laws.

How do I calculate gpu vram?

GPU VRAM

Calculate memory requirements for training & inference across FP32/FP16/INT8/INT4 precisions.

How do I calculate chinchilla scaling?

Chinchilla Scaling

Find compute-optimal model size and training tokens given a compute budget. Chinchilla & beyond.

How do I calculate lora / qlora fine-tuning?

LoRA / QLoRA Fine-Tuning

Calculate trainable parameters, memory savings, and adapter sizes for LoRA/QLoRA fine-tuning.

How do I calculate token count & api cost?

Token Count & API Cost

Compare token costs across OpenAI, Anthropic, Google, and Mistral. Input vs output pricing.

How do I calculate neural network flops?

Neural Network FLOPs

Calculate floating-point operations per layer type: Linear, Conv2D, Attention, LayerNorm.

How do I calculate model quantization?

Model Quantization

Compare GPTQ/AWQ/GGUF quantization: memory savings, speed gains, accuracy tradeoffs.

How do I calculate kv cache size?

KV Cache Size

Calculate KV cache memory for MHA/MQA/GQA attention types. Based on PagedAttention research.

How do I calculate context window cost?

Context Window Cost

Analyze quadratic attention scaling costs. Compare standard vs Flash Attention memory.

How do I calculate inference throughput?

Inference Throughput

Estimate tokens/sec, TTFT, and inter-token latency for LLM serving on various GPUs.

How do I calculate model distillation?

Model Distillation

Plan teacher-to-student compression: size ratios, accuracy retention, training tokens needed.

How do I calculate gradient accumulation?

Gradient Accumulation

Calculate accumulation steps to achieve target effective batch size on limited GPU memory.

How do I calculate activation memory?

Activation Memory

Estimate activation memory with/without gradient checkpointing. Based on NVIDIA research.

How do I calculate confusion matrix?

Confusion Matrix

Compute Accuracy, Precision, Recall, F1, MCC, ROC-AUC from TP/FP/TN/FN values.

How do I calculate batch size & learning rate?

Batch Size & Learning Rate

Linear & sqrt LR scaling rules, warmup estimation, cosine/linear schedule visualization.

How do I calculate neural network parameter counter?

Neural Network Parameter Counter

Count parameters for Linear, Conv2D, Embedding, LayerNorm, and Attention layers.

How do I calculate rag?

RAG

Calculate chunk sizes, vector store memory, token budgets for retrieval-augmented generation.

How do I calculate training data size?

Training Data Size

How much data do you need? Chinchilla ratios, fine-tuning guidelines, LIMA insights.

How do I calculate embedding dimension?

Embedding Dimension

Optimal embedding dimensions for LLMs, RAG, classification. Memory vs expressiveness.

How do I calculate cross-validation sample?

Cross-Validation Sample

Minimum samples for reliable k-fold CV with stratification and class imbalance.

How do I calculate attention head?

Attention Head

Configure MHA/MQA/GQA attention: head counts, dimensions, memory per attention type.

How do I calculate moe efficiency?

MoE Efficiency

Total vs active parameters for Mixture of Experts. Mixtral, DeepSeek, Switch Transformer.

How do I calculate ml carbon footprint?

ML Carbon Footprint

Estimate CO2 emissions from GPU training. Energy consumption, carbon equivalents.

How do I calculate differential privacy?

Differential Privacy

Calculate epsilon via RDP accounting. Apple, Google, and Meta privacy budget standards.

How do I calculate ai fairness & bias?

AI Fairness & Bias

Demographic parity, equalized odds, disparate impact. IBM AIF360 & Microsoft Fairlearn metrics.