diff --git a/README.md b/README.md new file mode 100644 index 0000000000000000000000000000000000000000..97f7ed23e7feabb4612ac71e8cddd696edbedc22 --- /dev/null +++ b/README.md @@ -0,0 +1,57 @@ +--- +license: cc-by-nc-sa-4.0 +library_name: pytorch +tags: + - biology + - genomics + - drug-discovery + - target-prioritization + - variant-prioritization +--- + +# Varformer + +Varformer is a multimodal transformer for population-scale drug target and variant prioritization. It learns contextual representations of the missense variants observed in a population and integrates them with gene-level biological features through cross-modal attention. It is trained with a non-negative positive-unlabeled objective for the gene score, and an allele-frequency objective for a per-variant score. + +Code, documentation and installation: https://github.com/aaronwtr/Varformer + +## Models in this repository + +The 25 models of the paper. Each was trained with five seeds (13, 23, 42, 85, 482), with the variant head (`variant_loss_weight` 0.1, pooled binary cross-entropy) and 32-bit precision. + +| Folder | Cohort | Variant source | Gene inputs | Use in the paper | +|---|---|---|---|---| +| `afr/` | African / African American | gnomAD v4.1 | without `geneticConstraint` | gene prioritization | +| `amr/` | Admixed American | gnomAD v4.1 | without `geneticConstraint` | gene prioritization | +| `nfe/` | Non-Finnish European | gnomAD v4.1 | without `geneticConstraint` | gene prioritization | +| `sas/` | South Asian | Genes & Health | without `geneticConstraint` | gene prioritization | +| `sas-gnh/` | South Asian | Genes & Health | with `geneticConstraint` | variant scores in the Genes & Health analyses | + +Each model has two files: + +- `seed.safetensors`: the model weights only. The files contain no training data, no genotypes and no individual-level information. +- `seed.json`: the training settings, the input dimensions, the class prior, the decision threshold, the selected epoch, the validation Spearman correlation, and SHA-256 checksums of the weights and of the source checkpoint. + +`index.json` lists all 25 models. + +## Use + +```python +from varformer import Varformer + +model = Varformer.from_pretrained("nfe", seed="best") +predictions = model.predict(genes=["ENSG00000141510"]) +variant_scores = model.score_variants(genes=["ENSG00000141510"]) +``` + +`from_pretrained` checks the weights against their checksum. The model inputs are built from local copies of the input data (gnomAD or Genes & Health variants, AlphaMissense scores, Open Targets, Human Protein Atlas and GTEx features); see the data sources in the code repository. Genes & Health data are available only to approved researchers. + +The gene score ranks genes by target-prioritization evidence; it is not a calibrated probability of clinical success. The variant score ranks variants for research prioritization; it is not validated for clinical variant classification. + +## License + +The weights are released under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/): attribution, non-commercial use only, and the same license for derived works. The models take AlphaMissense scores as input, and AlphaMissense is released under the same license. The code is MIT-licensed. For commercial use, train a model with another pathogenicity predictor; the code repository describes how. + +## Citation + +The paper citation will be added when the preprint is available. Until then, please cite the software release described in the code repository. diff --git a/afr/seed13.json b/afr/seed13.json new file mode 100644 index 0000000000000000000000000000000000000000..2d3f95e1e71dddca4d669381a8cedd266121b60a --- /dev/null +++ b/afr/seed13.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08368786966898413, + "decision_threshold": 0.7535547614097595, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "afr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 13, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.7535547373130804, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "afr", + "seed": 13, + "selected_epoch": 23, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "cd70d456eee7b6da5aafd7243daeac5c72ad3afdac2025f458bda4e002280ccd", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "4c79b8b76c3d5b8cade53308e3547e0969e0f62d1c85b3d044fdce88b354385a" +} diff --git a/afr/seed13.safetensors b/afr/seed13.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..294835d0449ca8e4720257555287ae9a0bb66e79 --- /dev/null +++ b/afr/seed13.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c79b8b76c3d5b8cade53308e3547e0969e0f62d1c85b3d044fdce88b354385a +size 33527716 diff --git a/afr/seed23.json b/afr/seed23.json new file mode 100644 index 0000000000000000000000000000000000000000..3120c51815bc2d9f50d000878c7afd009f4a70fd --- /dev/null +++ b/afr/seed23.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08368786966898413, + "decision_threshold": 0.5682888627052307, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "afr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 23, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.5682888836537554, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "afr", + "seed": 23, + "selected_epoch": 16, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "3ba8ce07d1ddc7b2c94dc4624294cae47c0fb636f05fc91d66887e224d084df5", + "validation_spearman": 0.38, + "variant_head": true, + "weights_sha256": "a0822954a24ce0da917436847a7531fbac10cf3a5b4caf47112b96e194ec0346" +} diff --git a/afr/seed23.safetensors b/afr/seed23.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..760dd4f61a8311744f0f9ea017a4a8b43b3f821c --- /dev/null +++ b/afr/seed23.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0822954a24ce0da917436847a7531fbac10cf3a5b4caf47112b96e194ec0346 +size 33527716 diff --git a/afr/seed42.json b/afr/seed42.json new file mode 100644 index 0000000000000000000000000000000000000000..39694f0123dfee0f4311565e3bdf042c847e4410 --- /dev/null +++ b/afr/seed42.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08368786966898413, + "decision_threshold": 0.5650092959403992, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "afr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 42, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.5650092758701416, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "afr", + "seed": 42, + "selected_epoch": 28, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "6b420977f96351c6fe627ed22c738276960f20025ce4f80fe4afaf40e66cf41f", + "validation_spearman": 0.39, + "variant_head": true, + "weights_sha256": "6be35ab0c05403c85408a61cf381d08a302fafb0288ee4c11cd8b3a8fa1d643b" +} diff --git a/afr/seed42.safetensors b/afr/seed42.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bfb61f9d5fc4030e1c903d5c0d865a47a3b5570c --- /dev/null +++ b/afr/seed42.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6be35ab0c05403c85408a61cf381d08a302fafb0288ee4c11cd8b3a8fa1d643b +size 33527716 diff --git a/afr/seed482.json b/afr/seed482.json new file mode 100644 index 0000000000000000000000000000000000000000..01bfcd72a20bb0713c7026781fb85446f3f09f72 --- /dev/null +++ b/afr/seed482.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08368786966898413, + "decision_threshold": 0.531441330909729, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "afr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 482, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.5314413357773077, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "afr", + "seed": 482, + "selected_epoch": 30, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "a5e14abe23f9e43c9edb131eaaaf794d26f7143d13570a35df9de67a312401e3", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "efe418a39c330fd0887f1024692a793dac9b150e5535b484a895e21bc77073f3" +} diff --git a/afr/seed482.safetensors b/afr/seed482.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..58f109adb1cf3a7ea8f4ce24b9651b2173717d17 --- /dev/null +++ b/afr/seed482.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:efe418a39c330fd0887f1024692a793dac9b150e5535b484a895e21bc77073f3 +size 33527716 diff --git a/afr/seed85.json b/afr/seed85.json new file mode 100644 index 0000000000000000000000000000000000000000..2de439c601dd25165da9dfc0ed3f19d2a15cc86b --- /dev/null +++ b/afr/seed85.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08368786966898413, + "decision_threshold": 0.45899805426597595, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "afr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 85, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.45899805593901283, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "afr", + "seed": 85, + "selected_epoch": 20, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "75eef825c2597af5aa88fa317cf8957c7e9aecb391aab086a9bf2f24a2f0d46a", + "validation_spearman": 0.38, + "variant_head": true, + "weights_sha256": "2936c16df39cf65822ea29deaea43078f9e4bec601ca9b0662ba2a6c1640fd9d" +} diff --git a/afr/seed85.safetensors b/afr/seed85.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9ebfb5880fd5a03dd0399d7afdd419d1b5d2372b --- /dev/null +++ b/afr/seed85.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2936c16df39cf65822ea29deaea43078f9e4bec601ca9b0662ba2a6c1640fd9d +size 33527716 diff --git a/amr/seed13.json b/amr/seed13.json new file mode 100644 index 0000000000000000000000000000000000000000..335e154c244b2d603dfc0fe0dda0182fbd5b712a --- /dev/null +++ b/amr/seed13.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08360128617363344, + "decision_threshold": 0.5641329884529114, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "amr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 13, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.5641329796367903, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "amr", + "seed": 13, + "selected_epoch": 26, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "956a348cb79f0a3534a0a7f526d4e0367f988315017649e3e99a4a8ab75d027a", + "validation_spearman": 0.4, + "variant_head": true, + "weights_sha256": "dcecf4c1f6bbd20e984e9484a8c58daef36814517f6fd59a2597504e99e7b239" +} diff --git a/amr/seed13.safetensors b/amr/seed13.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b5d5da70619b91b65658085436b9e468d3dc68c4 --- /dev/null +++ b/amr/seed13.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcecf4c1f6bbd20e984e9484a8c58daef36814517f6fd59a2597504e99e7b239 +size 33527716 diff --git a/amr/seed23.json b/amr/seed23.json new file mode 100644 index 0000000000000000000000000000000000000000..b141b5565ef5cc7d772aeec8a9d5f60ed09e92a8 --- /dev/null +++ b/amr/seed23.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08360128617363344, + "decision_threshold": 0.5932214856147766, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "amr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 23, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.5932215043969471, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "amr", + "seed": 23, + "selected_epoch": 29, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "90b9174033575013f3376e752b267513006e76cb26ba652a27a9321618a0f924", + "validation_spearman": 0.39, + "variant_head": true, + "weights_sha256": "6445517f741d337a2b6fb1818a945db28973268a41d77cf4fa445a1806b51caf" +} diff --git a/amr/seed23.safetensors b/amr/seed23.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..421e041c392e3e895583b8c7b5ff47d74f355065 --- /dev/null +++ b/amr/seed23.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6445517f741d337a2b6fb1818a945db28973268a41d77cf4fa445a1806b51caf +size 33527716 diff --git a/amr/seed42.json b/amr/seed42.json new file mode 100644 index 0000000000000000000000000000000000000000..9ca35150a25d0ff156a85914b3a267413a07f6c3 --- /dev/null +++ b/amr/seed42.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08360128617363344, + "decision_threshold": 0.6023826003074646, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "amr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 42, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.6023825853583891, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "amr", + "seed": 42, + "selected_epoch": 40, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "eb871dbf93b71b98fb357facd44ab82e761266a8a0900123759098a2dcf2dcbc", + "validation_spearman": 0.39, + "variant_head": true, + "weights_sha256": "f0d290cb431bbe377bbd5144f382a6ff5bab47b559c730bc8a97e25aa678ae6a" +} diff --git a/amr/seed42.safetensors b/amr/seed42.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..439cd007c27924f369a2c62d13825092b5b66a77 --- /dev/null +++ b/amr/seed42.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0d290cb431bbe377bbd5144f382a6ff5bab47b559c730bc8a97e25aa678ae6a +size 33527716 diff --git a/amr/seed482.json b/amr/seed482.json new file mode 100644 index 0000000000000000000000000000000000000000..249349c89c63e0a73ac0f0871055f03f28634def --- /dev/null +++ b/amr/seed482.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08360128617363344, + "decision_threshold": 0.8855880498886108, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "amr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 482, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.8855880675208528, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "amr", + "seed": 482, + "selected_epoch": 37, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "b4a6e256e98a67436fd5b8a91d617822dfb44e446976750c1987d94155a235eb", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "f5ab4e1eb2ffe5125628709db8a43317ec665abbbbcb3d68a2be99c465f89f68" +} diff --git a/amr/seed482.safetensors b/amr/seed482.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..80f235db85e2134ff0cc17f997bd196182133eb2 --- /dev/null +++ b/amr/seed482.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f5ab4e1eb2ffe5125628709db8a43317ec665abbbbcb3d68a2be99c465f89f68 +size 33527716 diff --git a/amr/seed85.json b/amr/seed85.json new file mode 100644 index 0000000000000000000000000000000000000000..b9aa4219cd21bed71755691098d48c5eded8ad7d --- /dev/null +++ b/amr/seed85.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08360128617363344, + "decision_threshold": 0.6900997757911682, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "amr", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 85, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.6900997972565062, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "amr", + "seed": 85, + "selected_epoch": 30, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "31237492810c1b61e9fce7b88abd456d1170733a8e86838335c749c019f786a7", + "validation_spearman": 0.4, + "variant_head": true, + "weights_sha256": "24220f19f9c837613afd9c4b9bc493ed51715950f22f308bb09aebb4cc8c32b0" +} diff --git a/amr/seed85.safetensors b/amr/seed85.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f6144a020920cfa83c35487c8efc72fb358b7b52 --- /dev/null +++ b/amr/seed85.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24220f19f9c837613afd9c4b9bc493ed51715950f22f308bb09aebb4cc8c32b0 +size 33527716 diff --git a/index.json b/index.json new file mode 100644 index 0000000000000000000000000000000000000000..86d61848873db4eeca26aa1865ac9ba8cffc967c --- /dev/null +++ b/index.json @@ -0,0 +1,317 @@ +[ + { + "file": "afr/seed13.safetensors", + "population": "afr", + "seed": 13, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.41, + "selected_epoch": 23, + "weights_sha256": "4c79b8b76c3d5b8cade53308e3547e0969e0f62d1c85b3d044fdce88b354385a", + "source_checkpoint_sha256": "cd70d456eee7b6da5aafd7243daeac5c72ad3afdac2025f458bda4e002280ccd", + "source_check": "evaluation_manifest" + }, + { + "file": "afr/seed23.safetensors", + "population": "afr", + "seed": 23, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.38, + "selected_epoch": 16, + "weights_sha256": "a0822954a24ce0da917436847a7531fbac10cf3a5b4caf47112b96e194ec0346", + "source_checkpoint_sha256": "3ba8ce07d1ddc7b2c94dc4624294cae47c0fb636f05fc91d66887e224d084df5", + "source_check": "evaluation_manifest" + }, + { + "file": "afr/seed42.safetensors", + "population": "afr", + "seed": 42, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.39, + "selected_epoch": 28, + "weights_sha256": "6be35ab0c05403c85408a61cf381d08a302fafb0288ee4c11cd8b3a8fa1d643b", + "source_checkpoint_sha256": "6b420977f96351c6fe627ed22c738276960f20025ce4f80fe4afaf40e66cf41f", + "source_check": "evaluation_manifest" + }, + { + "file": "afr/seed85.safetensors", + "population": "afr", + "seed": 85, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.38, + "selected_epoch": 20, + "weights_sha256": "2936c16df39cf65822ea29deaea43078f9e4bec601ca9b0662ba2a6c1640fd9d", + "source_checkpoint_sha256": "75eef825c2597af5aa88fa317cf8957c7e9aecb391aab086a9bf2f24a2f0d46a", + "source_check": "evaluation_manifest" + }, + { + "file": "afr/seed482.safetensors", + "population": "afr", + "seed": 482, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.41, + "selected_epoch": 30, + "weights_sha256": "efe418a39c330fd0887f1024692a793dac9b150e5535b484a895e21bc77073f3", + "source_checkpoint_sha256": "a5e14abe23f9e43c9edb131eaaaf794d26f7143d13570a35df9de67a312401e3", + "source_check": "evaluation_manifest" + }, + { + "file": "amr/seed13.safetensors", + "population": "amr", + "seed": 13, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.4, + "selected_epoch": 26, + "weights_sha256": "dcecf4c1f6bbd20e984e9484a8c58daef36814517f6fd59a2597504e99e7b239", + "source_checkpoint_sha256": "956a348cb79f0a3534a0a7f526d4e0367f988315017649e3e99a4a8ab75d027a", + "source_check": "evaluation_manifest" + }, + { + "file": "amr/seed23.safetensors", + "population": "amr", + "seed": 23, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.39, + "selected_epoch": 29, + "weights_sha256": "6445517f741d337a2b6fb1818a945db28973268a41d77cf4fa445a1806b51caf", + "source_checkpoint_sha256": "90b9174033575013f3376e752b267513006e76cb26ba652a27a9321618a0f924", + "source_check": "evaluation_manifest" + }, + { + "file": "amr/seed42.safetensors", + "population": "amr", + "seed": 42, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.39, + "selected_epoch": 40, + "weights_sha256": "f0d290cb431bbe377bbd5144f382a6ff5bab47b559c730bc8a97e25aa678ae6a", + "source_checkpoint_sha256": "eb871dbf93b71b98fb357facd44ab82e761266a8a0900123759098a2dcf2dcbc", + "source_check": "evaluation_manifest" + }, + { + "file": "amr/seed85.safetensors", + "population": "amr", + "seed": 85, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.4, + "selected_epoch": 30, + "weights_sha256": "24220f19f9c837613afd9c4b9bc493ed51715950f22f308bb09aebb4cc8c32b0", + "source_checkpoint_sha256": "31237492810c1b61e9fce7b88abd456d1170733a8e86838335c749c019f786a7", + "source_check": "evaluation_manifest" + }, + { + "file": "amr/seed482.safetensors", + "population": "amr", + "seed": 482, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.41, + "selected_epoch": 37, + "weights_sha256": "f5ab4e1eb2ffe5125628709db8a43317ec665abbbbcb3d68a2be99c465f89f68", + "source_checkpoint_sha256": "b4a6e256e98a67436fd5b8a91d617822dfb44e446976750c1987d94155a235eb", + "source_check": "evaluation_manifest" + }, + { + "file": "nfe/seed13.safetensors", + "population": "nfe", + "seed": 13, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.4, + "selected_epoch": 34, + "weights_sha256": "0c5de915aae59ad0b4ade78eeeff0039fb7666ced4840f5347005c474e29dd30", + "source_checkpoint_sha256": "a10dda732716d55944ce5402e6065cd3217d361448d8e4cb849ae6910c905d68", + "source_check": "evaluation_manifest" + }, + { + "file": "nfe/seed23.safetensors", + "population": "nfe", + "seed": 23, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.39, + "selected_epoch": 19, + "weights_sha256": "dd3fb603b602c686b3ad68908695c12a5bbe3d03efb54dc7301a036094434588", + "source_checkpoint_sha256": "1bf2dbc21482cace72b2f636d890f976f12a6e8f4ebc8a83d275cf8ab77233bd", + "source_check": "evaluation_manifest" + }, + { + "file": "nfe/seed42.safetensors", + "population": "nfe", + "seed": 42, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.4, + "selected_epoch": 40, + "weights_sha256": "98b7f98162a238e3525c2af28c4cb8aa54a4d6ce7f9e387d67235a5f89366a44", + "source_checkpoint_sha256": "734bc122f8f980211f7788d94c261828cae06c6c4ec869655b85bade0ef6a98f", + "source_check": "evaluation_manifest" + }, + { + "file": "nfe/seed85.safetensors", + "population": "nfe", + "seed": 85, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.39, + "selected_epoch": 29, + "weights_sha256": "acf777405901658bc1b7e9d071473d87ae49a21720bd3dcd20bb6f945049f371", + "source_checkpoint_sha256": "f588461020fc7b6a39ffe34575f493e3223f12b15d6caffc9498bc1aacc7175f", + "source_check": "evaluation_manifest" + }, + { + "file": "nfe/seed482.safetensors", + "population": "nfe", + "seed": 482, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.38, + "selected_epoch": 15, + "weights_sha256": "dd2b78a9c8bf08f08127ec5e8ca606ef5a460196686e17cd16ae4cda01b5283d", + "source_checkpoint_sha256": "fb5358431175a362baeee1ac49ef1e69c51c06cef9c4ddcedd945c101547f2f4", + "source_check": "evaluation_manifest" + }, + { + "file": "sas/seed13.safetensors", + "population": "sas", + "seed": 13, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.44, + "selected_epoch": 48, + "weights_sha256": "f3ed37f7c609b6ead69b442b0e0761a18c74336abf56788ebc0058953c158099", + "source_checkpoint_sha256": "e5cecb69c2ccb5bcd80c56aae1b4a84e8a379a69fb41f96a8d6b892ee46a9ce2", + "source_check": "evaluation_manifest" + }, + { + "file": "sas/seed23.safetensors", + "population": "sas", + "seed": 23, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.41, + "selected_epoch": 16, + "weights_sha256": "3b297d849b09f3e6d025351a62db91bae7d150ddcc6ffcf3bd0c2699b9fe1554", + "source_checkpoint_sha256": "88a42931d355344ed42d80cde5b02ff0ceb3787a5de56bfd212d522acd785602", + "source_check": "evaluation_manifest" + }, + { + "file": "sas/seed42.safetensors", + "population": "sas", + "seed": 42, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.41, + "selected_epoch": 38, + "weights_sha256": "ad938cb10f958bece96a881a1db4c00b6a07a16c5e1097d211d539ee6cda99f6", + "source_checkpoint_sha256": "f3e4d7f90f82ce87f7cb89d323fe41e4f7bb938879ec276aaf7d3e5730d2c880", + "source_check": "evaluation_manifest" + }, + { + "file": "sas/seed85.safetensors", + "population": "sas", + "seed": 85, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.42, + "selected_epoch": 28, + "weights_sha256": "59e6f1f76b2c95457863a71d5e2e93f5b2a8dcdc23e152b207df0630da13c2ec", + "source_checkpoint_sha256": "f006488787655ba28636e87050c74acccaef9a12e8da296ccb8d23c84a325593", + "source_check": "evaluation_manifest" + }, + { + "file": "sas/seed482.safetensors", + "population": "sas", + "seed": 482, + "gc_exclude_features": [ + "geneticConstraint" + ], + "validation_spearman": 0.41, + "selected_epoch": 21, + "weights_sha256": "ebf099391138a2d98189cd6fcb7d6117d9d274674f8eeeef18aa5f3907d83f57", + "source_checkpoint_sha256": "5611864f162ff430d278fba27b80e852995b28be4e91843e8c503276f75d7c21", + "source_check": "evaluation_manifest" + }, + { + "file": "sas-gnh/seed13.safetensors", + "population": "sas", + "seed": 13, + "gc_exclude_features": [], + "validation_spearman": 0.43, + "selected_epoch": 36, + "weights_sha256": "eda3c7c410005d6eea833aa36755176878ec4615390a7546f844d687b1c86062", + "source_checkpoint_sha256": "ff2adb781074ed57e02943fdc8ca774376271a638dccedffd99b61a4519e55eb", + "source_check": "run_manifest" + }, + { + "file": "sas-gnh/seed23.safetensors", + "population": "sas", + "seed": 23, + "gc_exclude_features": [], + "validation_spearman": 0.41, + "selected_epoch": 37, + "weights_sha256": "9ff25a0fe7c7ecd54e00c1c66a654ff5161ef0330ebe6144d64a9374c0df3b66", + "source_checkpoint_sha256": "729944b5fde327e3ab64935d5e129385fb96ae0d711a428cd847941b5ff848ca", + "source_check": "run_manifest" + }, + { + "file": "sas-gnh/seed42.safetensors", + "population": "sas", + "seed": 42, + "gc_exclude_features": [], + "validation_spearman": 0.4, + "selected_epoch": 29, + "weights_sha256": "92022a8a561fb8dba521ffd88ab010df41fce23ec848472bb68720e3db311b77", + "source_checkpoint_sha256": "e535bef88980c2303ffa2e95925bbb4d399585020e2c190f7281061665fdb89d", + "source_check": "run_manifest" + }, + { + "file": "sas-gnh/seed85.safetensors", + "population": "sas", + "seed": 85, + "gc_exclude_features": [], + "validation_spearman": 0.41, + "selected_epoch": 28, + "weights_sha256": "55c37906838a195769e3a010a6adf3732deb9dc3a91ca0de4d7f0d428a62bc6a", + "source_checkpoint_sha256": "31fe3d9549eab5ae3041ca244fc3865b7b3351f1f7b196bf59811e76f020b011", + "source_check": "run_manifest" + }, + { + "file": "sas-gnh/seed482.safetensors", + "population": "sas", + "seed": 482, + "gc_exclude_features": [], + "validation_spearman": 0.41, + "selected_epoch": 30, + "weights_sha256": "75c1187f3e93af73579cd327f044276af41adeca9b209de005b0c2b267a21c7b", + "source_checkpoint_sha256": "2ae1e885532c107625270a2b9bcf85d4ca4d22dc17e37b02c9a4021a604c5d65", + "source_check": "run_manifest" + } +] diff --git a/nfe/seed13.json b/nfe/seed13.json new file mode 100644 index 0000000000000000000000000000000000000000..b2019a3544b6bfdd6ff78578b95b02b568feea93 --- /dev/null +++ b/nfe/seed13.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.0814781160078752, + "decision_threshold": 0.6301044821739197, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "nfe", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 13, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.6301045029719057, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "nfe", + "seed": 13, + "selected_epoch": 34, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "a10dda732716d55944ce5402e6065cd3217d361448d8e4cb849ae6910c905d68", + "validation_spearman": 0.4, + "variant_head": true, + "weights_sha256": "0c5de915aae59ad0b4ade78eeeff0039fb7666ced4840f5347005c474e29dd30" +} diff --git a/nfe/seed13.safetensors b/nfe/seed13.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..25a7b94468687f43f42d2383e2b1cdc5d76f1b89 --- /dev/null +++ b/nfe/seed13.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c5de915aae59ad0b4ade78eeeff0039fb7666ced4840f5347005c474e29dd30 +size 33527716 diff --git a/nfe/seed23.json b/nfe/seed23.json new file mode 100644 index 0000000000000000000000000000000000000000..6e47191072bfecabf8ec04eced7e2918902030a5 --- /dev/null +++ b/nfe/seed23.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.0814781160078752, + "decision_threshold": 0.6867796182632446, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "nfe", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 23, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.6867796320744065, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "nfe", + "seed": 23, + "selected_epoch": 19, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "1bf2dbc21482cace72b2f636d890f976f12a6e8f4ebc8a83d275cf8ab77233bd", + "validation_spearman": 0.39, + "variant_head": true, + "weights_sha256": "dd3fb603b602c686b3ad68908695c12a5bbe3d03efb54dc7301a036094434588" +} diff --git a/nfe/seed23.safetensors b/nfe/seed23.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1bd3b4e21acc28e03ecb63d981f0bb7fab5be77e --- /dev/null +++ b/nfe/seed23.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dd3fb603b602c686b3ad68908695c12a5bbe3d03efb54dc7301a036094434588 +size 33527716 diff --git a/nfe/seed42.json b/nfe/seed42.json new file mode 100644 index 0000000000000000000000000000000000000000..0dcc442ddc6dd89ca323ab1a1dd005575b2a18bd --- /dev/null +++ b/nfe/seed42.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.0814781160078752, + "decision_threshold": 0.8719609975814819, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "nfe", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 42, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.8719609837432385, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "nfe", + "seed": 42, + "selected_epoch": 40, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "734bc122f8f980211f7788d94c261828cae06c6c4ec869655b85bade0ef6a98f", + "validation_spearman": 0.4, + "variant_head": true, + "weights_sha256": "98b7f98162a238e3525c2af28c4cb8aa54a4d6ce7f9e387d67235a5f89366a44" +} diff --git a/nfe/seed42.safetensors b/nfe/seed42.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0fbad2246368b05d0c3747acf4d3ab8bbaf2af6f --- /dev/null +++ b/nfe/seed42.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:98b7f98162a238e3525c2af28c4cb8aa54a4d6ce7f9e387d67235a5f89366a44 +size 33527716 diff --git a/nfe/seed482.json b/nfe/seed482.json new file mode 100644 index 0000000000000000000000000000000000000000..c23d692cfb5570a148b0a35e1a634f930ab5cf40 --- /dev/null +++ b/nfe/seed482.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.0814781160078752, + "decision_threshold": 0.3766169250011444, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "nfe", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 482, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.37661692854871753, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "nfe", + "seed": 482, + "selected_epoch": 15, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "fb5358431175a362baeee1ac49ef1e69c51c06cef9c4ddcedd945c101547f2f4", + "validation_spearman": 0.38, + "variant_head": true, + "weights_sha256": "dd2b78a9c8bf08f08127ec5e8ca606ef5a460196686e17cd16ae4cda01b5283d" +} diff --git a/nfe/seed482.safetensors b/nfe/seed482.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..781adef3324fa6de4319d049b94fc710d7594def --- /dev/null +++ b/nfe/seed482.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dd2b78a9c8bf08f08127ec5e8ca606ef5a460196686e17cd16ae4cda01b5283d +size 33527716 diff --git a/nfe/seed85.json b/nfe/seed85.json new file mode 100644 index 0000000000000000000000000000000000000000..a85325cd207cd37996a5d5d598a868c975605a5f --- /dev/null +++ b/nfe/seed85.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.0814781160078752, + "decision_threshold": 0.8027001619338989, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 527, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "nfe", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 85, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.8027001751492858, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8379686, + "population": "nfe", + "seed": 85, + "selected_epoch": 29, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "f588461020fc7b6a39ffe34575f493e3223f12b15d6caffc9498bc1aacc7175f", + "validation_spearman": 0.39, + "variant_head": true, + "weights_sha256": "acf777405901658bc1b7e9d071473d87ae49a21720bd3dcd20bb6f945049f371" +} diff --git a/nfe/seed85.safetensors b/nfe/seed85.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c4f995ac1704343c474069b783cb9a44e1c817e1 --- /dev/null +++ b/nfe/seed85.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acf777405901658bc1b7e9d071473d87ae49a21720bd3dcd20bb6f945049f371 +size 33527716 diff --git a/sas-gnh/seed13.json b/sas-gnh/seed13.json new file mode 100644 index 0000000000000000000000000000000000000000..e9010830aae7aafc2020b6c371b5764edd707864 --- /dev/null +++ b/sas-gnh/seed13.json @@ -0,0 +1,72 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.4418480694293976, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 10, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 13, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.4418480829839327, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378182, + "population": "sas", + "seed": 13, + "selected_epoch": 36, + "source_check": "run_manifest", + "source_checkpoint_sha256": "ff2adb781074ed57e02943fdc8ca774376271a638dccedffd99b61a4519e55eb", + "validation_spearman": 0.43, + "variant_head": true, + "weights_sha256": "eda3c7c410005d6eea833aa36755176878ec4615390a7546f844d687b1c86062" +} diff --git a/sas-gnh/seed13.safetensors b/sas-gnh/seed13.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..458facff1a3e5b5caa0b9589bf2d587c2c3e7a03 --- /dev/null +++ b/sas-gnh/seed13.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eda3c7c410005d6eea833aa36755176878ec4615390a7546f844d687b1c86062 +size 33521700 diff --git a/sas-gnh/seed23.json b/sas-gnh/seed23.json new file mode 100644 index 0000000000000000000000000000000000000000..f9137faaa1ec5b2ac04e9194a210a046953f3c0c --- /dev/null +++ b/sas-gnh/seed23.json @@ -0,0 +1,72 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.7796894907951355, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 10, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 23, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.7796894906005755, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378182, + "population": "sas", + "seed": 23, + "selected_epoch": 37, + "source_check": "run_manifest", + "source_checkpoint_sha256": "729944b5fde327e3ab64935d5e129385fb96ae0d711a428cd847941b5ff848ca", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "9ff25a0fe7c7ecd54e00c1c66a654ff5161ef0330ebe6144d64a9374c0df3b66" +} diff --git a/sas-gnh/seed23.safetensors b/sas-gnh/seed23.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..571fd629801d3b5e03ea1c21bb261ab1e7d3489a --- /dev/null +++ b/sas-gnh/seed23.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9ff25a0fe7c7ecd54e00c1c66a654ff5161ef0330ebe6144d64a9374c0df3b66 +size 33521700 diff --git a/sas-gnh/seed42.json b/sas-gnh/seed42.json new file mode 100644 index 0000000000000000000000000000000000000000..9fdeb3be9bd7ee147d3b801bfef47a513f8310a9 --- /dev/null +++ b/sas-gnh/seed42.json @@ -0,0 +1,72 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.6333062052726746, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 10, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 42, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.6333061968236093, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378182, + "population": "sas", + "seed": 42, + "selected_epoch": 29, + "source_check": "run_manifest", + "source_checkpoint_sha256": "e535bef88980c2303ffa2e95925bbb4d399585020e2c190f7281061665fdb89d", + "validation_spearman": 0.4, + "variant_head": true, + "weights_sha256": "92022a8a561fb8dba521ffd88ab010df41fce23ec848472bb68720e3db311b77" +} diff --git a/sas-gnh/seed42.safetensors b/sas-gnh/seed42.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7ac56e49de7758b5720cb808fc6fe22470005290 --- /dev/null +++ b/sas-gnh/seed42.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:92022a8a561fb8dba521ffd88ab010df41fce23ec848472bb68720e3db311b77 +size 33521700 diff --git a/sas-gnh/seed482.json b/sas-gnh/seed482.json new file mode 100644 index 0000000000000000000000000000000000000000..ab3022f470c27491219368808281e7f92d07c4db --- /dev/null +++ b/sas-gnh/seed482.json @@ -0,0 +1,72 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.7612968683242798, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 10, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 482, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.7612968543445895, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378182, + "population": "sas", + "seed": 482, + "selected_epoch": 30, + "source_check": "run_manifest", + "source_checkpoint_sha256": "2ae1e885532c107625270a2b9bcf85d4ca4d22dc17e37b02c9a4021a604c5d65", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "75c1187f3e93af73579cd327f044276af41adeca9b209de005b0c2b267a21c7b" +} diff --git a/sas-gnh/seed482.safetensors b/sas-gnh/seed482.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3368cf79d8fc017c961909913b720be1481c1519 --- /dev/null +++ b/sas-gnh/seed482.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75c1187f3e93af73579cd327f044276af41adeca9b209de005b0c2b267a21c7b +size 33521700 diff --git a/sas-gnh/seed85.json b/sas-gnh/seed85.json new file mode 100644 index 0000000000000000000000000000000000000000..31e40adb48f4675dbe42da91855009bfa454ee39 --- /dev/null +++ b/sas-gnh/seed85.json @@ -0,0 +1,72 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.2809520363807678, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 10, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 85, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.2809520276975065, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378182, + "population": "sas", + "seed": 85, + "selected_epoch": 28, + "source_check": "run_manifest", + "source_checkpoint_sha256": "31fe3d9549eab5ae3041ca244fc3865b7b3351f1f7b196bf59811e76f020b011", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "55c37906838a195769e3a010a6adf3732deb9dc3a91ca0de4d7f0d428a62bc6a" +} diff --git a/sas-gnh/seed85.safetensors b/sas-gnh/seed85.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b105b67b3af6a51a1cdf0c25a186fd0f95f46f4d --- /dev/null +++ b/sas-gnh/seed85.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55c37906838a195769e3a010a6adf3732deb9dc3a91ca0de4d7f0d428a62bc6a +size 33521700 diff --git a/sas/seed13.json b/sas/seed13.json new file mode 100644 index 0000000000000000000000000000000000000000..639745d2053807efae7450ad81946eab2bd49cba --- /dev/null +++ b/sas/seed13.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.6496922969818115, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 13, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.6496922880823696, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378150, + "population": "sas", + "seed": 13, + "selected_epoch": 48, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "e5cecb69c2ccb5bcd80c56aae1b4a84e8a379a69fb41f96a8d6b892ee46a9ce2", + "validation_spearman": 0.44, + "variant_head": true, + "weights_sha256": "f3ed37f7c609b6ead69b442b0e0761a18c74336abf56788ebc0058953c158099" +} diff --git a/sas/seed13.safetensors b/sas/seed13.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1af1e28ae7525bcf2f73ce9f25b4bd4b75954e9c --- /dev/null +++ b/sas/seed13.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3ed37f7c609b6ead69b442b0e0761a18c74336abf56788ebc0058953c158099 +size 33521572 diff --git a/sas/seed23.json b/sas/seed23.json new file mode 100644 index 0000000000000000000000000000000000000000..c9f4f6c92efca00ea8da4749fe57ed333bcc6761 --- /dev/null +++ b/sas/seed23.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.8452841639518738, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 23, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.8452841911906526, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378150, + "population": "sas", + "seed": 23, + "selected_epoch": 16, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "88a42931d355344ed42d80cde5b02ff0ceb3787a5de56bfd212d522acd785602", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "3b297d849b09f3e6d025351a62db91bae7d150ddcc6ffcf3bd0c2699b9fe1554" +} diff --git a/sas/seed23.safetensors b/sas/seed23.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ddbdfae8489b1469cbb53ad5127009090ad0b1d2 --- /dev/null +++ b/sas/seed23.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b297d849b09f3e6d025351a62db91bae7d150ddcc6ffcf3bd0c2699b9fe1554 +size 33521572 diff --git a/sas/seed42.json b/sas/seed42.json new file mode 100644 index 0000000000000000000000000000000000000000..84dad28432951ab212bcaba888f1f43a27f56368 --- /dev/null +++ b/sas/seed42.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.9446958899497986, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 42, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.9446958622065983, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378150, + "population": "sas", + "seed": 42, + "selected_epoch": 38, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "f3e4d7f90f82ce87f7cb89d323fe41e4f7bb938879ec276aaf7d3e5730d2c880", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "ad938cb10f958bece96a881a1db4c00b6a07a16c5e1097d211d539ee6cda99f6" +} diff --git a/sas/seed42.safetensors b/sas/seed42.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..30f73ec31756eef4e53c8349b28557e6d7ca7283 --- /dev/null +++ b/sas/seed42.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad938cb10f958bece96a881a1db4c00b6a07a16c5e1097d211d539ee6cda99f6 +size 33521572 diff --git a/sas/seed482.json b/sas/seed482.json new file mode 100644 index 0000000000000000000000000000000000000000..f774902c42d31751bc89132a261bd790dd3d6f70 --- /dev/null +++ b/sas/seed482.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.630003809928894, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 482, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.6300037906275539, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378150, + "population": "sas", + "seed": 482, + "selected_epoch": 21, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "5611864f162ff430d278fba27b80e852995b28be4e91843e8c503276f75d7c21", + "validation_spearman": 0.41, + "variant_head": true, + "weights_sha256": "ebf099391138a2d98189cd6fcb7d6117d9d274674f8eeeef18aa5f3907d83f57" +} diff --git a/sas/seed482.safetensors b/sas/seed482.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..aaf78c7cc57878f7b3f655d746fc74bb9f7b517d --- /dev/null +++ b/sas/seed482.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ebf099391138a2d98189cd6fcb7d6117d9d274674f8eeeef18aa5f3907d83f57 +size 33521572 diff --git a/sas/seed85.json b/sas/seed85.json new file mode 100644 index 0000000000000000000000000000000000000000..6ced74b7b9b744632400ebc36938c363d40b36ab --- /dev/null +++ b/sas/seed85.json @@ -0,0 +1,76 @@ +{ + "class_prior": 0.08849664510669165, + "decision_threshold": 0.49799177050590515, + "dimensions": { + "max_seq_len": 1024, + "num_features_gc": 9, + "num_features_go": 524, + "num_genes": 1024, + "num_mutations": 381 + }, + "gc_exclude_features": [ + "geneticConstraint" + ], + "hyperparameters": { + "T0": 200, + "af_log10_floor": -6.0, + "batch_size": 128, + "d_model": 256, + "depth_cls_head": 4, + "dim_feedforward": 4096, + "dropout": 0.3, + "epochs": 100, + "gc_exclude_features": [ + "geneticConstraint" + ], + "gc_width": 32, + "go_width": 512, + "grad_accum": null, + "gradient_clip_val": 1.0, + "gv_attn_dim": 256, + "include_open_targets_labels": true, + "lr_end": 1e-05, + "lr_start": 0.0001, + "mask_cross_attention_padding": true, + "max_seq_len": 1024, + "min_callability_fraction": 0.8, + "mutation_embedding_max_norm": 20.0, + "nhead": 8, + "num_encoder_layers": 3, + "optimizer": "AdamW", + "optimizer_foreach": null, + "population": "sas", + "position_encoding": "modulo", + "precision": "32-true", + "pu_class_prior": null, + "pusb": true, + "return_attn": true, + "scheduler": "CosineAnnealingLR", + "seed": 85, + "shuffle_pvc_across_genes": false, + "stop_on_nan": true, + "threshold": 0.4979917558848786, + "threshold_positive_fraction": null, + "use_gc": true, + "use_go": true, + "use_pvc": true, + "variant_frequency_qc": false, + "variant_head": true, + "variant_loss_weight": 0.1, + "variant_objective": "pooled_bce", + "variant_pairs_per_gene": 256, + "variant_pairs_per_side": 96, + "variant_token_ablate": [], + "variant_token_schema": "original", + "weight_decay": 0.0003 + }, + "parameters": 8378150, + "population": "sas", + "seed": 85, + "selected_epoch": 28, + "source_check": "evaluation_manifest", + "source_checkpoint_sha256": "f006488787655ba28636e87050c74acccaef9a12e8da296ccb8d23c84a325593", + "validation_spearman": 0.42, + "variant_head": true, + "weights_sha256": "59e6f1f76b2c95457863a71d5e2e93f5b2a8dcdc23e152b207df0630da13c2ec" +} diff --git a/sas/seed85.safetensors b/sas/seed85.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..beacaec82a8bbdf035def0242c97ffaab6a7511c --- /dev/null +++ b/sas/seed85.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59e6f1f76b2c95457863a71d5e2e93f5b2a8dcdc23e152b207df0630da13c2ec +size 33521572