Last verified: 2026-07-13 — refresh when a new size-class anchor is validated or optimizer/dtype defaults change. Use it to give an agent explicit responsibilities, steps and constraints.
Memory Math
Snapshot 2026-08-04 16:17:00 UTC · version 1
Research document
Memory Math
Last verified: 2026-07-13 — refresh when a new size-class anchor is validated or optimizer/dtype defaults change. Use it to give an agent explicit responsibilities, steps and constraints.
Editorial note: curated source snapshot published by Collider.club under the MIT License. Source attribution is preserved in the front matter.
Source snapshot
Last verified: 2026-07-13 — refresh when a new size-class anchor is validated or optimizer/dtype defaults change.
Memory Math
A worksheet for estimating whether a model
size-class, method, and batch/pack combination
fits available memory before a run. This is
planning math, not a guarantee — leave headroom
rather than sizing to the byte. Base models are
never named here; every example is labeled by
size class only (for example, "8B-class LoRA
bf16"). See model-catalog.md for which actual
model to use at a given size class.
The Four Terms
Total footprint ≈ weights + optimizer states + gradients + activations, plus a near-zero term for LoRA/QLoRA adapters. Work each term from parameter count and dtype, then sum.
1. Weights
params × bytes/param, by dtype:
| dtype | bytes/param |
|---|---|
| fp32 | 4 |
| bf16 / fp16 | 2 |
| int8 | 1 |
| int4 (QLoRA NF4) | 0.5 |
This term dominates for full fine-tuning, and the
calculation (params × bytes/param) is the same
formula regardless of method — but the dtype, and
so the result, is not: bf16 LoRA loads weights at
2 bytes/param while int4 QLoRA loads the same
parameter count at 0.5 bytes/param, a 4x gap.
Reuse the formula across methods; never reuse the
resulting weight-memory number from one method's
dtype for another's.
2. Optimizer states
Full fine-tuning carries optimizer state for every trainable parameter; LoRA and QLoRA carry it only for the adapter parameters, which is why this term is negligible for them regardless of base model size.
| Optimizer | bytes/param (trainable only) |
|---|---|
| AdamW, fp32 states | 8 (4B momentum + 4B variance) |
| AdamW 8-bit | ≈2 (quantized momentum + variance) |
8-bit AdamW roughly quarters this term versus the fp32 variant for any run that isn't LoRA/QLoRA- adapter-only, where it's already negligible.
3. Gradients
Same dtype as compute precision — typically bf16, so 2 bytes/param — and, like optimizer state, only for trainable parameters. Full fine-tuning pays this for every weight; LoRA and QLoRA pay it only for the adapter, since frozen base weights never accumulate a gradient.
4. Activations
The hardest term to pin to a single number — it scales with batch size, sequence/packing length, and architecture, not just parameter count. Two levers matter more than exact estimation:
- Gradient checkpointing trades recompute for memory: expect roughly 30% savings on this term versus no checkpointing, at the cost of a recompute pass per checkpointed segment.
- Packing/sequence length is a more direct lever than batch size for this term.
5. LoRA/QLoRA adapter overhead
A rank-r adapter on a linear layer adds
r × (in + out) parameters — A is r×in and B
is out×r, so together they contribute
r·in + r·out. At normal rank sizes (1–32 for RL,
up to ~256 for SFT-at-scale),
this is a small fraction of a percent of base
model size — round it to zero in the worksheet
unless an unusually high rank is in play.
Worked Examples
8B-class LoRA, bf16
Weights dominate; optimizer state and gradients are adapter-only and small.
params = 8e9
weights_gb = params * 2 / 1e9 # bf16, step 1
adapter_gb = 0.2 # step 5, negligible
total_gb = weights_gb + adapter_gb # + activations
print(f"{total_gb:.0f}GB before activations")
Weights alone land around 16GB — the reference point for "an 8B-class model fits comfortably on a single high-memory GPU in bf16 LoRA."
8B-class QLoRA
Same parameter count, quantized weights:
params = 8e9
weights_gb = params * 0.5 / 1e9 # int4 NF4, step 1
adapter_gb = 0.2 # step 5, negligible
total_gb = weights_gb + adapter_gb # + activations
print(f"{total_gb:.0f}GB before activations")
Weights land around 4GB — roughly a 4x reduction versus bf16 LoRA, which is why QLoRA is the method that buys headroom for larger batch size or longer packing at the same size class, not just a way to fit bigger models.
70B-class QLoRA (≈40GB anchor)
params = 70e9
weights_gb = params * 0.5 / 1e9 # int4 NF4, step 1
adapter_gb = 0.5 # step 5, negligible
total_gb = weights_gb + adapter_gb # + activations
print(f"{total_gb:.0f}GB before activations")
The idealized formula lands weights at ≈35GB (decimal GB, weights only); treat ≈40GB as the real-world anchor once quantization metadata (NF4 double-quant constants) and runtime overhead are included — the reference point for "a 70B-class model is reachable via QLoRA, not bf16," where bf16 weights alone (≈140GB) would already exceed most single-device budgets before optimizer state, gradients, or activations are added. A plan estimating far above the ≈40GB anchor for the same size class is a signal to recheck dtype and method, not just add headroom.
Using These Numbers
- Pick the size class and method from
model-catalog.md. - Sum weights + optimizer + gradients from the tables above for that combination.
- Add activations, applying the ~30% gradient- checkpointing saving if it's enabled.
- Compare against the closest worked example or anchor above rather than trusting the estimate in isolation — a plan far off an anchor for the same size class and method is a signal to recheck inputs before assuming the hardware won't work.
About Collider.club
This card belongs to the curated knowledge base of Collider.club — a closed business club for entrepreneurs, engineers, investors and domain experts building projects for international markets. Members work across DeFi, AI/ML, FinTech, Web3, banking, hardware and venture capital, and the club runs closed sessions on high-margin niches with anonymous speakers.
- Club: https://collider.club
- Collection: Collider.club curated card library (
mdrss-card/v2) - Maintainer: Collider.club editorial team
License
MIT License — Copyright (c) 2026 Collider.club. Full text: LICENSE · https://opensource.org/licenses/MIT
Why MDRSS assigned this score
- evidence comes from multiple domains
- some evidence URLs look like primary-source hosts
Evidence (4)
concept:training-fine-tuningorg:collider-club Discussion 0
Sign in to join the discussion.