PLAPT

Around 3.9 × 10²² FLOP, on NVIDIA GeForce RTX 4060 Ti. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.

None. PLAPT is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.

The licensing for PLAPT was never recorded in our source data. We treat unstated licensing as closed, because an unrecorded licence is not one to rely on.

PLAPT has 1.5M parameters. "Figure 3: The prediction module takes a 1792x1 feature vector as its input, which is then partitioned into two streams: The first stream processes the first 1024 indices of the feature vector through a 512-node protein-specific linear layer, followed by a ReLU activation function. Concurrently, the second stream processes the latter 768 feature indices through a similar 512-node moleculespecific linear layer, also followed by a ReLU activation. Outputs from both streams are concatenated into a single vector of 1024 elements. This combined vector is passed through a batch normalization layer with a momentum of 0.9 and epsilon of 0.001. The vector is then fed through the 512-node Linear Layer 1 and ReLU activation. A dropout layer with a probability of 20% is then applied to mitigate overfitting. Following the dropout layer, the prediction module continues to reduce the feature space with the 64-node Linear Layer 2 with ReLU activation, before reaching the single node Linear Layer 3. This layer outputs the predicted scalar value representing the normalized negative log10 affinity value, which is then un-normalized." (512 * 1024) + (512 * 768) + (512 * 1024) + (64 * 512) + 64 = 1474624. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.

PLAPT was published by Wolfram Research,ASC27,Newport High School,Sanskriti School, based in United States of America, categorised as industry,Industry,Academia.

PLAPT was published in February 2024. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

PLAPT works in Biology, and is recorded as handling protein-ligand binding affinity prediction. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.