Clarify latency comparability: base quantization must match, not just GPU

README.md

@@ -186,9 +186,12 @@ device** holding a complete model — the model is replicated, not sharded.
 |---|---|---|---|---|---|
 | this variant (7B) | NVIDIA H200 (gpuH200x8) | ~16 GB | 4-bit | **23.70 GB** | **1.63 s/image** |
 
-> **Different variants in this series were measured on different GPUs** (0.5B on
-> A100, 7B and 72B on H200), so do not compare latency across scales. Comparisons
-> *within* a scale are valid — those pairs ran on the same partition.
+> **Comparing these numbers across variants requires care.** Scales were measured
+> on different GPUs (0.5B on A100, 7B and 72B on H200), so **latency is not
+> comparable across scales**. Within a scale it is comparable only when the *base
+> loading* also matches: QLoRA vs QDoRA is a fair comparison (same GPU, both
+> 4-bit), but LoRA vs QLoRA is **not** — those differ in base quantization as well
+> as PEFT method, so the gap conflates the two.
 
 > **Why peak memory exceeds the weight size.** LLaVA-OneVision uses anyres tiling:
 > a large image expands into thousands of visual tokens (1024x768 -> ~5,100 tokens;