Raj Lab Piscis models
Collection
1 item • Updated
Piscis ps_20240419_112256
PyTorch conversion of the legacy Piscis JAX checkpoint
wniu/ps_20240419_112256,
for Flax-free inference in NimbusImage's Piscis Predict and Train workers.
Conversion validation
The Piscis 1.0.0 JAX-to-PyTorch converter was first validated against the
published 20230905 JAX and PyTorch checkpoint pair. All 583 tensors matched
exactly in key, shape, dtype, and value; the maximum absolute difference was
0.0. The same validated converter produced this checkpoint.
- Source checkpoint SHA-256:
28fea47454c108dac5e2899f5c4f84fd489792f28fbd5d6b3e002236516952e2 - Converted checkpoint SHA-256:
f281a239d1d5075ee2342c4772ba7167c441f878e5345c632436caf3b43d320a - Converted checkpoint size:
30,083,712bytes - State dictionary: 583 tensors
- Metadata:
adjustment=standardize,input_size=(256, 256),dilation_iterations=1,channels=1
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support