pimm export (config.json)

README.md

@@ -1,27 +1,16 @@
 ---
-license: mit
-datasets:
-- DeepLearnPhysics/PILArNet-M
 library_name: pimm
 tags:
-- particle
-- physics
-- 3D
-- lartpc
-- panda
-- point-transformer-v3
+  - particle-physics
+  - lartpc
+  - point-cloud
 ---
 
 # Panda — PointTransformerV3 encoder (pretrained)
 
-A PointTransformerV3 (`PT-v3m2`) encoder pretrained with masked
-point modeling on LArTPC (PILArNet) data. Intended as a backbone / warm-start for
-downstream segmentation and detection tasks (it produces multi-scale point
-features; it has no task head).
+PointTransformerV3 (`PT-v3m2`) encoder pretrained with masked point modeling on LArTPC (PILArNet) data. Backbone / warm-start for downstream tasks (no task head).
 
-- **pimm type:** `PT-v3m2`
-- **enc_depths:** (3, 3, 3, 9, 3) · **enc_channels:** (48, 96, 192, 384, 512)
-- **encoder-only** (`enc_mode=True`), masked-modeling token enabled.
+- **pimm type:** `PT-v3m2` · encoder-only · enc_channels (48,96,192,384,512)
 
 ## Loading
 
@@ -30,24 +19,9 @@ import pimm
 model = pimm.from_pretrained("hf://deeplearnphysics/panda-base")
 ```
 
-The architecture and all hyper-parameters are carried in `training_config.json`,
-so no config file is needed. Weights are bitwise-identical to the original
-checkpoint.
-
-## Input
-
-A pimm `Point` dict for LArTPC (PILArNet) point clouds: `coord` (xyz) + `energy`,
-collected as `feat_keys=("coord", "energy")` (4 input channels). Coordinates are
-normalized in the model forward.
-
-## Notes
-
-- Config sets `enable_flash=True`. On CPU or non-FlashAttention setups, override
-  it (e.g. pass `enable_flash=False` to `from_pretrained`, or for detectors set it
-  on the `backbone`).
+Architecture + hyper-parameters travel in `config.json`, so no config file is
+needed. Weights are bitwise-identical to the original checkpoint.
 
 ## Provenance
 
-Repackaged from the original [`DeepLearnPhysics/panda`](https://github.com/DeepLearnPhysics/panda) checkpoints into the
-[`pimm`](https://github.com/youngsm/particle-imaging-models) export format. Inherits
-the license of the source repository.
+Repackaged from the original [`panda`](https://github.com/DeepLearnPhysics/panda) checkpoints into the [`pimm`](https://github.com/youngsm/particle-imaging-models) export format. Inherits the source repo's license.

config.json

@@ -0,0 +1,64 @@
+{
+  "model": {
+    "type": "PT-v3m2",
+    "in_channels": 4,
+    "order": [
+      "hilbert",
+      "hilbert-trans",
+      "z",
+      "z-trans"
+    ],
+    "stride": [
+      2,
+      2,
+      2,
+      2
+    ],
+    "enc_depths": [
+      3,
+      3,
+      3,
+      9,
+      3
+    ],
+    "enc_channels": [
+      48,
+      96,
+      192,
+      384,
+      512
+    ],
+    "enc_num_head": [
+      3,
+      6,
+      12,
+      24,
+      32
+    ],
+    "enc_patch_size": [
+      256,
+      256,
+      256,
+      256,
+      256
+    ],
+    "mlp_ratio": 4,
+    "qkv_bias": true,
+    "qk_scale": null,
+    "layer_scale": 1e-05,
+    "attn_drop": 0.0,
+    "proj_drop": 0.0,
+    "drop_path": 0.3,
+    "shuffle_orders": true,
+    "pre_norm": true,
+    "enable_rpe": false,
+    "enable_flash": true,
+    "upcast_attention": false,
+    "upcast_softmax": false,
+    "traceable": true,
+    "enc_mode": true,
+    "mask_token": true,
+    "cpe_first_layer_only": false,
+    "cpe_shared_weight": false
+  }
+}