Add files using upload-large-folder tool

README.md

@@ -21,12 +21,12 @@ gene vocabulary, the inference implementation, and runnable examples.
 
 ## Checkpoints
 
-| Checkpoint | Directory | Vocabulary size | Required paired tokenizer |
+| Checkpoint | Directory | Vocabulary size | Required paired assets |
 | --- | --- | ---: | --- |
-| 2026-06-25 / 30M | `checkpoints/scRep_20260625_30M/` | 19,240 | `checkpoints/scRep_20260625_30M/tokenizer/` |
-| 2026-07-17 / 3M | `checkpoints/scRep_20260717_3M/` | 19,239 | `checkpoints/scRep_20260717_3M/tokenizer/` |
+| 2026-06-25 / 30M | `checkpoints/scRep_20260625_30M/` | 19,240 | `checkpoints/scRep_20260625_30M/assets/` |
+| 2026-07-17 / 3M | `checkpoints/scRep_20260717_3M/` | 19,239 | `checkpoints/scRep_20260717_3M/assets/` |
 
-`model.safetensors`, `config.json`, and `tokenizer/gene_vocab.json` from the
+`model.safetensors`, `config.json`, and `assets/gene_vocab.json` from the
 same checkpoint directory must always be used together. The checkpoint names
 are retained as release identifiers; they do not by themselves specify a
 parameter count or a benchmark claim.
@@ -81,11 +81,11 @@ scRep_release/
     ├── scRep_20260625_30M/
     │   ├── config.json
     │   ├── model.safetensors
-    │   └── tokenizer/{gene_vocab.json,gene_names.txt}
+    │   └── assets/{gene_vocab.json,gene_names.txt,meta_vocab.json,manifest.json}
     └── scRep_20260717_3M/
         ├── config.json
         ├── model.safetensors
-        └── tokenizer/{gene_vocab.json,gene_names.txt}
+        └── assets/{gene_vocab.json,gene_names.txt,meta_vocab.json,manifest.json}
 ```
 
 ## Limitations and responsible use

checkpoints/scRep_20260625_30M/assets/meta_vocab.json

@@ -0,0 +1,4 @@
+{
+  "meta_vocab_sizes": {},
+  "meta_maps": {}
+}

checkpoints/scRep_20260625_30M/config.json

@@ -1,8 +1,40 @@
 {
-  "d_model": 768, "n_heads": 12, "dropout": 0.0, "encoder_layers": 12,
-  "gradient_checkpointing": false, "gene_vocab_size": 19240, "bin_vocab_size": 53,
-  "gene_pad_id": 0, "bin_pad_id": 0, "bin_mask_id": 1, "bin_offset": 3,
-  "gene_cls_id": 1, "expr_embedding_mode": "bin", "expr_rank_mlp_hidden_dim": 256,
-  "dino_out_dim": 4096, "dino_hidden_dim": 2048, "dino_bottleneck_dim": 256,
-  "inference_use_teacher": true
-}
+  "d_model": 768,
+  "n_heads": 12,
+  "dropout": 0.0,
+  "encoder_layers": 12,
+  "gene_vocab_size": 19240,
+  "bin_vocab_size": 53,
+  "gene_pad_id": 0,
+  "bin_pad_id": 0,
+  "bin_mask_id": 1,
+  "bin_offset": 3,
+  "gene_cls_id": 1,
+  "expr_embedding_mode": "bin",
+  "expr_rank_mlp_hidden_dim": 256,
+  "dino_out_dim": 4096,
+  "dino_hidden_dim": 2048,
+  "dino_bottleneck_dim": 256,
+  "student_temperature": 0.1,
+  "teacher_temperature": 0.04,
+  "teacher_momentum": 0.996,
+  "teacher_momentum_end": 1.0,
+  "center_momentum": 0.9,
+  "token_center_momentum": 0.9,
+  "cross_modal_dino_weight": 0.0,
+  "cross_modal_pairing_mode": "disabled",
+  "token_loss_weight": 1.0,
+  "token_unmasked_loss_weight": 0.0,
+  "inference_use_teacher": true,
+  "teacher_cls_decoder_loss_weight": 0.0,
+  "teacher_cls_decoder_hidden_dim": 1024,
+  "teacher_cls_decoder_condition_on_obs": false,
+  "teacher_cls_decoder_rank_loss_alpha": 0.05,
+  "teacher_cls_decoder_rank_num_pairs": 4096,
+  "teacher_student_decoder_distill_weight": 0.0,
+  "teacher_student_decoder_loss_mode": "distribution",
+  "teacher_student_decoder_rank_alpha": 0.05,
+  "teacher_student_decoder_raw_mmd_weight": 0.1,
+  "teacher_student_decoder_raw_rank_weight": 0.01,
+  "teacher_student_decoder_raw_mse_weight": 0.0
+}

checkpoints/scRep_20260717_3M/assets/manifest.json

@@ -0,0 +1,401 @@
+[
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0023_0_50000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0023_0_50000.h5ad",
+    "dataset_id": 0,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0024_50000_100000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0024_50000_100000.h5ad",
+    "dataset_id": 1,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0025_100000_150000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0025_100000_150000.h5ad",
+    "dataset_id": 2,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0026_150000_200000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0026_150000_200000.h5ad",
+    "dataset_id": 3,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0027_200000_250000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0027_200000_250000.h5ad",
+    "dataset_id": 4,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0028_250000_300000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0028_250000_300000.h5ad",
+    "dataset_id": 5,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0029_300000_350000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0029_300000_350000.h5ad",
+    "dataset_id": 6,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0030_350000_400000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0030_350000_400000.h5ad",
+    "dataset_id": 7,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0031_400000_450000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0031_400000_450000.h5ad",
+    "dataset_id": 8,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0032_450000_500000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0032_450000_500000.h5ad",
+    "dataset_id": 9,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0033_500000_550000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0033_500000_550000.h5ad",
+    "dataset_id": 10,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0034_550000_600000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0034_550000_600000.h5ad",
+    "dataset_id": 11,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0035_600000_650000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0035_600000_650000.h5ad",
+    "dataset_id": 12,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0036_650000_700000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0036_650000_700000.h5ad",
+    "dataset_id": 13,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0037_700000_750000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0037_700000_750000.h5ad",
+    "dataset_id": 14,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0038_750000_800000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0038_750000_800000.h5ad",
+    "dataset_id": 15,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0039_800000_850000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0039_800000_850000.h5ad",
+    "dataset_id": 16,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0040_850000_900000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0040_850000_900000.h5ad",
+    "dataset_id": 17,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0041_900000_950000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0041_900000_950000.h5ad",
+    "dataset_id": 18,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0042_950000_1000000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0042_950000_1000000.h5ad",
+    "dataset_id": 19,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0043_1000000_1050000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0043_1000000_1050000.h5ad",
+    "dataset_id": 20,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0044_1050000_1100000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0044_1050000_1100000.h5ad",
+    "dataset_id": 21,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0045_1100000_1150000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0045_1100000_1150000.h5ad",
+    "dataset_id": 22,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0046_1150000_1200000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0046_1150000_1200000.h5ad",
+    "dataset_id": 23,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0047_1200000_1250000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0047_1200000_1250000.h5ad",
+    "dataset_id": 24,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0048_1250000_1300000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0048_1250000_1300000.h5ad",
+    "dataset_id": 25,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0049_1300000_1350000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0049_1300000_1350000.h5ad",
+    "dataset_id": 26,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0050_1350000_1400000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0050_1350000_1400000.h5ad",
+    "dataset_id": 27,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0051_1400000_1450000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0051_1400000_1450000.h5ad",
+    "dataset_id": 28,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0052_1450000_1500000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0052_1450000_1500000.h5ad",
+    "dataset_id": 29,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0053_1500000_1550000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0053_1500000_1550000.h5ad",
+    "dataset_id": 30,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0054_1550000_1600000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0054_1550000_1600000.h5ad",
+    "dataset_id": 31,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0055_1600000_1650000.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0055_1600000_1650000.h5ad",
+    "dataset_id": 32,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/cellxgene_supplement_old19240_shard0056_1650000_1667317.h5ad",
+    "name": "cellxgene_supplement_old19240_shard0056_1650000_1667317.h5ad",
+    "dataset_id": 33,
+    "n_cells": 17317,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0000_0_50000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0000_0_50000.h5ad",
+    "dataset_id": 34,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0001_50000_100000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0001_50000_100000.h5ad",
+    "dataset_id": 35,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0002_100000_150000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0002_100000_150000.h5ad",
+    "dataset_id": 36,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0003_150000_200000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0003_150000_200000.h5ad",
+    "dataset_id": 37,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0004_200000_250000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0004_200000_250000.h5ad",
+    "dataset_id": 38,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0005_250000_300000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0005_250000_300000.h5ad",
+    "dataset_id": 39,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0006_300000_350000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0006_300000_350000.h5ad",
+    "dataset_id": 40,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0007_350000_400000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0007_350000_400000.h5ad",
+    "dataset_id": 41,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0008_400000_450000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0008_400000_450000.h5ad",
+    "dataset_id": 42,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0009_450000_500000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0009_450000_500000.h5ad",
+    "dataset_id": 43,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0010_500000_550000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0010_500000_550000.h5ad",
+    "dataset_id": 44,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0011_550000_600000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0011_550000_600000.h5ad",
+    "dataset_id": 45,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0012_600000_650000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0012_600000_650000.h5ad",
+    "dataset_id": 46,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0013_650000_700000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0013_650000_700000.h5ad",
+    "dataset_id": 47,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0014_700000_750000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0014_700000_750000.h5ad",
+    "dataset_id": 48,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0015_750000_800000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0015_750000_800000.h5ad",
+    "dataset_id": 49,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0016_800000_850000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0016_800000_850000.h5ad",
+    "dataset_id": 50,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0017_850000_900000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0017_850000_900000.h5ad",
+    "dataset_id": 51,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0018_900000_950000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0018_900000_950000.h5ad",
+    "dataset_id": 52,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0019_950000_1000000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0019_950000_1000000.h5ad",
+    "dataset_id": 53,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0020_1000000_1050000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0020_1000000_1050000.h5ad",
+    "dataset_id": 54,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0021_1050000_1100000.h5ad",
+    "name": "tabula_sapiens_old19240_shard0021_1050000_1100000.h5ad",
+    "dataset_id": 55,
+    "n_cells": 50000,
+    "obs_nunique": {}
+  },
+  {
+    "path": "/vepfs-mlp2/mlp-public/400092/data/Tabula_Sapiens_Plus_3M_old19240_shards/tabula_sapiens_old19240_shard0022_1100000_1136218.h5ad",
+    "name": "tabula_sapiens_old19240_shard0022_1100000_1136218.h5ad",
+    "dataset_id": 56,
+    "n_cells": 36218,
+    "obs_nunique": {}
+  }
+]

checkpoints/scRep_20260717_3M/assets/meta_vocab.json

@@ -0,0 +1,4 @@
+{
+  "meta_vocab_sizes": {},
+  "meta_maps": {}
+}

checkpoints/scRep_20260717_3M/config.json

@@ -1,8 +1,41 @@
 {
-  "d_model": 768, "n_heads": 12, "dropout": 0.0, "encoder_layers": 12,
-  "gradient_checkpointing": false, "gene_vocab_size": 19239, "bin_vocab_size": 53,
-  "gene_pad_id": 0, "bin_pad_id": 0, "bin_mask_id": 1, "bin_offset": 3,
-  "gene_cls_id": 1, "expr_embedding_mode": "bin", "expr_rank_mlp_hidden_dim": 256,
-  "dino_out_dim": 4096, "dino_hidden_dim": 2048, "dino_bottleneck_dim": 256,
-  "inference_use_teacher": true
-}
+  "d_model": 768,
+  "n_heads": 12,
+  "dropout": 0.0,
+  "encoder_layers": 12,
+  "gradient_checkpointing": true,
+  "gene_vocab_size": 19239,
+  "bin_vocab_size": 53,
+  "gene_pad_id": 0,
+  "bin_pad_id": 0,
+  "bin_mask_id": 1,
+  "bin_offset": 3,
+  "gene_cls_id": 1,
+  "expr_embedding_mode": "bin",
+  "expr_rank_mlp_hidden_dim": 256,
+  "dino_out_dim": 4096,
+  "dino_hidden_dim": 2048,
+  "dino_bottleneck_dim": 256,
+  "student_temperature": 0.1,
+  "teacher_temperature": 0.04,
+  "teacher_momentum": 0.996,
+  "teacher_momentum_end": 1.0,
+  "center_momentum": 0.9,
+  "token_center_momentum": 0.9,
+  "cross_modal_dino_weight": 0.0,
+  "cross_modal_pairing_mode": "disabled",
+  "token_loss_weight": 1.0,
+  "token_unmasked_loss_weight": 0.0,
+  "inference_use_teacher": true,
+  "teacher_cls_decoder_loss_weight": 0.0,
+  "teacher_cls_decoder_hidden_dim": 1024,
+  "teacher_cls_decoder_condition_on_obs": false,
+  "teacher_cls_decoder_rank_loss_alpha": 0.05,
+  "teacher_cls_decoder_rank_num_pairs": 4096,
+  "teacher_student_decoder_distill_weight": 0.0,
+  "teacher_student_decoder_loss_mode": "distribution",
+  "teacher_student_decoder_rank_alpha": 0.05,
+  "teacher_student_decoder_raw_mmd_weight": 0.1,
+  "teacher_student_decoder_raw_rank_weight": 0.01,
+  "teacher_student_decoder_raw_mse_weight": 0.0
+}

scRep_inference.py

@@ -135,8 +135,8 @@ def load_scRep_bundle(model_dir_or_args: str | Path | Any, args: Any | None = No
     weights_dir = Path(model_dir).expanduser().resolve()
     from safetensors.torch import load_file
     state = load_file(str(weights_dir / "model.safetensors"), device="cpu")
-    asset_arg = Path(str(getattr(args, "asset_dir", "") or ROOT / "assets")).expanduser()
-    candidates = [weights_dir, weights_dir / "tokenizer", weights_dir.parent, asset_arg, ROOT / "assets"]
+    asset_arg = Path(str(getattr(args, "asset_dir", "") or weights_dir / "assets")).expanduser()
+    candidates = [weights_dir, weights_dir / "assets", weights_dir / "tokenizer", asset_arg]
     def asset(name: str) -> Path:
         found = next((p / name for p in candidates if (p / name).is_file()), None)
         if found is None: raise FileNotFoundError(f"Missing {name}; pass --asset_dir")