Archive MaleCNS research sources, provenance, bibliography and immutable data inventory
This commit is contained in:
@@ -0,0 +1,20 @@
|
||||
# MaleCNS LIF Phase 4A
|
||||
|
||||
Auditable, single-RTX-4090 whole-connectome LIF baseline. It consumes only the
|
||||
validated Phase 3B neuron-only products and never writes to the dataset tree.
|
||||
|
||||
Run with the exact GPU UUID to prevent accidental V100 or mixed-GPU use:
|
||||
|
||||
```bash
|
||||
CUDA_VISIBLE_DEVICES=GPU-79f3a85d-adce-7833-7003-6488140ca596 \
|
||||
.venv/bin/python phase4a.py \
|
||||
--config /data/digital-drosophila/experiments/phase4a-baseline/config/baseline.yaml
|
||||
```
|
||||
|
||||
The raw `int64` synapse-count array remains read-only. The configured transform
|
||||
creates a separate signed `float32` effective-weight array. Spike output is a
|
||||
sparse Parquet event table, never a dense time-by-neuron matrix.
|
||||
|
||||
The baseline parameters are FlyWire-inspired modeling defaults, not MaleCNS
|
||||
single-cell measurements. Synthetic drive amplitudes are deterministic smoke-test
|
||||
inputs and are not a biological calibration.
|
||||
Reference in New Issue
Block a user