diff --git a/epidermal_basset/README b/epidermal_basset/README new file mode 100644 index 000000000..675978281 --- /dev/null +++ b/epidermal_basset/README @@ -0,0 +1,14 @@ +to make models Kipoi compatible, first generated a file of the original +models with checkpoints - `models_orig.tsv`. + +this was the input to `run_kipoi_compatibility.py`, which inserted placeholder +ops into the metagraph. these new graphs were saved to a new dir, and +confirmed to run with kipoi test. + +these files were uploaded to zenodo with `upload_zenodo.py`. + +then the links were put into a new models.tsv table with update_w_zenodo_links.py. + +these links were then used with the files to generate a folder for each model. + +testing is done per folder, see script `run_tests.py`. \ No newline at end of file diff --git a/epidermal_basset/archive/model_OLD.yaml b/epidermal_basset/archive/model_OLD.yaml new file mode 100644 index 000000000..82e1954d5 --- /dev/null +++ b/epidermal_basset/archive/model_OLD.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: models.kipoi_compatible/encode-roadmap.basset.clf.testfold-0.model.meta + md5: 0bb5678c3651564e37fdad5c4b875cd2 + index: + url: models.kipoi_compatible/encode-roadmap.basset.clf.testfold-0.model.index + md5: cf33d8cd0a359a53381d66ee4c762542 + data: + url: models.kipoi_compatible/encode-roadmap.basset.clf.testfold-0.model.data-00000-of-00001 + md5: 2f1d71990db937fb83586dc191863e1a + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/archive/models_kipoi_local.tsv b/epidermal_basset/archive/models_kipoi_local.tsv new file mode 100644 index 000000000..da4b7281d --- /dev/null +++ b/epidermal_basset/archive/models_kipoi_local.tsv @@ -0,0 +1,31 @@ +model args_meta_url args_meta_md5 args_index_url args_index_md5 args_data_url args_data_md5 +encode-roadmap.basset.clf.testfold-0 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-0.model.meta 0bb5678c3651564e37fdad5c4b875cd2 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-0.model.index cf33d8cd0a359a53381d66ee4c762542 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-0.model.data-00000-of-00001 2f1d71990db937fb83586dc191863e1a +encode-roadmap.basset.clf.testfold-1 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-1.model.meta 3b24a69ecd953cb8e341fe134f46e3d9 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-1.model.index 8eaf8751af57976c35f770dc6498148f models.kipoi_compatible/encode-roadmap.basset.clf.testfold-1.model.data-00000-of-00001 499918cb6cd189200cdc8b8107634a4a +encode-roadmap.basset.clf.testfold-2 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-2.model.meta 708a9c8d20e2361242b3e63f0df84137 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-2.model.index 0489c73298dbb124ec45b02a2a6f1637 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-2.model.data-00000-of-00001 9127e729f247cb4e7b9973381a69b28e +encode-roadmap.basset.clf.testfold-3 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-3.model.meta 623f8a17121d98cf8f7993f25d330705 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-3.model.index cbb20d8327c3a0cdc0a708b05ca971cd models.kipoi_compatible/encode-roadmap.basset.clf.testfold-3.model.data-00000-of-00001 fd2997d78156594ea04196a18379e2fc +encode-roadmap.basset.clf.testfold-4 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-4.model.meta 3be06b38124f68700a423e43935bfe51 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-4.model.index b5c3978cce00de7a1a339a48b7697119 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-4.model.data-00000-of-00001 57508c46ffe33b5580fcd9772a0a7f95 +encode-roadmap.basset.clf.testfold-5 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-5.model.meta e5ff2b5579dd1ba1e7d1c0ea8a0c4be0 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-5.model.index 5c9451f8a316f6b9d44722b53d210010 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-5.model.data-00000-of-00001 79dde073c01580138d8b6cf50c1cae40 +encode-roadmap.basset.clf.testfold-6 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-6.model.meta cf060c2b1d4c9125ebec2accd74e298e models.kipoi_compatible/encode-roadmap.basset.clf.testfold-6.model.index fccf24f2e0846d5ea6f3d56f10d0cce2 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-6.model.data-00000-of-00001 e4bc8ce0c499759025499af7b06ec6ee +encode-roadmap.basset.clf.testfold-7 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-7.model.meta dd863fc91e1e46743efdd52315b84235 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-7.model.index ff03710e1c38f8561877fdc8df600c22 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-7.model.data-00000-of-00001 7b97299be0b9435360b24b3d526c5e60 +encode-roadmap.basset.clf.testfold-8 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-8.model.meta a7077d7359779958a43168aa8911a3af models.kipoi_compatible/encode-roadmap.basset.clf.testfold-8.model.index 28382fb71004441bfd891f8093272f9a models.kipoi_compatible/encode-roadmap.basset.clf.testfold-8.model.data-00000-of-00001 2e53513b09315728a17dda13997bc16d +encode-roadmap.basset.clf.testfold-9 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-9.model.meta 3e18c83fbb1cc29eabd5f8f17988d7c0 models.kipoi_compatible/encode-roadmap.basset.clf.testfold-9.model.index 20ddafde1077175ee9fba9fb04e8cacd models.kipoi_compatible/encode-roadmap.basset.clf.testfold-9.model.data-00000-of-00001 c87d589bbf7939c9731b67be8f414959 +ggr.basset.clf.pretrained.folds.testfold-0 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-0.model.meta 66dcdd18cd9e8e4560c79bf60c5c9160 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-0.model.index f7096b63b619eeb90f02ec28340d393b models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-0.model.data-00000-of-00001 af235f9a97e05d674cf9415d681396a0 +ggr.basset.clf.pretrained.folds.testfold-1 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-1.model.meta 794fac73c754d5fd88bbcf26b3439c77 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-1.model.index 49dbd1be34d3320eed34907120bbb5de models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-1.model.data-00000-of-00001 107c14caab90c3a53726d4f9bc9e5263 +ggr.basset.clf.pretrained.folds.testfold-2 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-2.model.meta 6018ae7864a8cdd27f2fd3fbcd605b05 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-2.model.index b97ea67092c63e8408246b8941f590d8 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-2.model.data-00000-of-00001 e567de2dfe2200fe3beaeb09f78a7f3f +ggr.basset.clf.pretrained.folds.testfold-3 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-3.model.meta 97267b947ea33270d7ddb92d5c8b657a models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-3.model.index 141afb11372af5c1995ecee9835c0aa0 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-3.model.data-00000-of-00001 81bab1510b2ba0b1d5d003dadf1dd3e0 +ggr.basset.clf.pretrained.folds.testfold-4 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-4.model.meta 9f39931ce83f26295f21f977e9fa0979 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-4.model.index e7d3e3bdda1efc774955293ebb66a741 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-4.model.data-00000-of-00001 12853a4eac1aff2e4c11717b6a597bb4 +ggr.basset.clf.pretrained.folds.testfold-5 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-5.model.meta 74ba6c7b7a505ade74e5b3ed6bf55f9f models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-5.model.index b451245b241bb982da4d849eb34b4c4f models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-5.model.data-00000-of-00001 e00b4bfbede3ddd1353711f5781627a2 +ggr.basset.clf.pretrained.folds.testfold-6 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-6.model.meta fd242cecdc33930243701f0389b2ba5e models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-6.model.index f05f29b0516d13570750e3115e3ec9de models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-6.model.data-00000-of-00001 1f8ad4b104f5cb8b745427c78cf8da75 +ggr.basset.clf.pretrained.folds.testfold-7 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-7.model.meta f77284d300cc441bfc47d8758cd1613c models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-7.model.index 7777626f6ef77b939798b1e1054f1417 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-7.model.data-00000-of-00001 7e09578aaaf13282b69e2b75eb8eacaf +ggr.basset.clf.pretrained.folds.testfold-8 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-8.model.meta bea6295aaa0308e534b81b3604b4a9b7 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-8.model.index 762199800bf1e069dcefc6dda8086d0a models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-8.model.data-00000-of-00001 6d79d9209b244f9e3128369c20469d33 +ggr.basset.clf.pretrained.folds.testfold-9 models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-9.model.meta 7a6552718b780b00dd9ab04877bccb3f models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-9.model.index 02c89e56a5a97ea38e76b05b07606edc models.kipoi_compatible/ggr.basset.clf.pretrained.folds.testfold-9.model.data-00000-of-00001 c86ae5ecb133c46cdcb1150d42691b87 +ggr.basset.regr.pretrained.folds.testfold-0 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-0.model.meta 9185274583bd67ca1bf3f376749dea0a models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-0.model.index 3ddad77aef022cf8246063da16984be4 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-0.model.data-00000-of-00001 dec73096e14c3f9ff424542a1ab11875 +ggr.basset.regr.pretrained.folds.testfold-1 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-1.model.meta 9eff2aa7236d73e24f4b09dc7ee4da35 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-1.model.index 9eee1096c7b06370fd83a47b8d6ad7f4 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-1.model.data-00000-of-00001 83431ab9129f960fd5435cb8f73d34bd +ggr.basset.regr.pretrained.folds.testfold-2 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-2.model.meta 000060b0ced8870536448e74267935cb models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-2.model.index 7085cac8d5487be035c0597b8dd0b8a0 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-2.model.data-00000-of-00001 d96ce0c3d876b0083b8693496c49d885 +ggr.basset.regr.pretrained.folds.testfold-3 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-3.model.meta 60a760c88e13aa0d0453e979c9a2b95e models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-3.model.index 960d7df63e2cdfaa336602ddc1d6dce2 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-3.model.data-00000-of-00001 9c6d827a455b42d1620bb32254e20025 +ggr.basset.regr.pretrained.folds.testfold-4 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-4.model.meta 05786276eb9044e0cc1b8645af24f7b9 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-4.model.index e3e69db7809d45768819db3a3972ce16 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-4.model.data-00000-of-00001 a48f6a1d6e47b7fa110a1c794ebd75dc +ggr.basset.regr.pretrained.folds.testfold-5 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-5.model.meta aecbc5165666647f2ad9d5113e056156 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-5.model.index dd56fde6d5dd0078c462c916119392dc models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-5.model.data-00000-of-00001 948789840faf196eff94ee0ecbfb3ad1 +ggr.basset.regr.pretrained.folds.testfold-6 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-6.model.meta 7c8c3b9c7f0116c5016c1f8733540c27 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-6.model.index 4d12f5815fb6b70c53becc57ceab44e7 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-6.model.data-00000-of-00001 8a7617a187769f1d23011fb24dfdf1a8 +ggr.basset.regr.pretrained.folds.testfold-7 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-7.model.meta 71ef1b06ca597efacaed20c56171495f models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-7.model.index ed884985af03ddc3ba3f55ae7112feca models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-7.model.data-00000-of-00001 1e07c75d137934c3f1b6259022b606a8 +ggr.basset.regr.pretrained.folds.testfold-8 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-8.model.meta 6789841151a1a99998e1e3850dc436c9 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-8.model.index 5de840becd3abb4f9a8f5f8d00a2c242 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-8.model.data-00000-of-00001 0b98d687db7b6c2d2247a053fd4074af +ggr.basset.regr.pretrained.folds.testfold-9 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-9.model.meta bbc8da8be311a4fa4104ff19332bad88 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-9.model.index e60798b9e9933bd5d06d50bb56094a45 models.kipoi_compatible/ggr.basset.regr.pretrained.folds.testfold-9.model.data-00000-of-00001 16463aeb88a068c146fe6eba660c1329 diff --git a/epidermal_basset/archive/models_orig.tsv b/epidermal_basset/archive/models_orig.tsv new file mode 100644 index 000000000..972bf4622 --- /dev/null +++ b/epidermal_basset/archive/models_orig.tsv @@ -0,0 +1,31 @@ +names args_meta_url args_meta_md5 args_index_url args_index_md5 args_data_url args_data_md5 +encode-roadmap.basset.clf.testfold-0 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-0/train/model.ckpt-122228.meta 7f338ea712a826c6ff8d977ae83e22f9 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-0/train/model.ckpt-122228.index cf33d8cd0a359a53381d66ee4c762542 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-0/train/model.ckpt-122228.data-00000-of-00001 2f1d71990db937fb83586dc191863e1a +encode-roadmap.basset.clf.testfold-1 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-1/train/model.ckpt-121488.meta aaed406ce2d4c62cc6e2eba3ca3f1d00 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-1/train/model.ckpt-121488.index 8eaf8751af57976c35f770dc6498148f /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-1/train/model.ckpt-121488.data-00000-of-00001 499918cb6cd189200cdc8b8107634a4a +encode-roadmap.basset.clf.testfold-2 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-2/train/model.ckpt-184050.meta f99e382724e70adbf970f02747eccbf6 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-2/train/model.ckpt-184050.index 0489c73298dbb124ec45b02a2a6f1637 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-2/train/model.ckpt-184050.data-00000-of-00001 9127e729f247cb4e7b9973381a69b28e +encode-roadmap.basset.clf.testfold-3 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-3/train/model.ckpt-124892.meta 083381102661114bdcdd74077eee79f9 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-3/train/model.ckpt-124892.index cbb20d8327c3a0cdc0a708b05ca971cd /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-3/train/model.ckpt-124892.data-00000-of-00001 fd2997d78156594ea04196a18379e2fc +encode-roadmap.basset.clf.testfold-4 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-4/train/model.ckpt-156555.meta 24a87507206a0f4c7f61b192c003871c /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-4/train/model.ckpt-156555.index b5c3978cce00de7a1a339a48b7697119 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-4/train/model.ckpt-156555.data-00000-of-00001 57508c46ffe33b5580fcd9772a0a7f95 +encode-roadmap.basset.clf.testfold-5 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-5/train/model.ckpt-93948.meta 8b56eb1dfbf91dea62a771dedec5108c /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-5/train/model.ckpt-93948.index 5c9451f8a316f6b9d44722b53d210010 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-5/train/model.ckpt-93948.data-00000-of-00001 79dde073c01580138d8b6cf50c1cae40 +encode-roadmap.basset.clf.testfold-6 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-6/train/model.ckpt-308940.meta 2eaa00429c40a003dd8a6e4fdd2de915 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-6/train/model.ckpt-308940.index fccf24f2e0846d5ea6f3d56f10d0cce2 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-6/train/model.ckpt-308940.data-00000-of-00001 e4bc8ce0c499759025499af7b06ec6ee +encode-roadmap.basset.clf.testfold-7 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-7/train/model.ckpt-91257.meta 6dfe2f0e56adf57f6bfd6362b2b0f227 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-7/train/model.ckpt-91257.index ff03710e1c38f8561877fdc8df600c22 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-7/train/model.ckpt-91257.data-00000-of-00001 7b97299be0b9435360b24b3d526c5e60 +encode-roadmap.basset.clf.testfold-8 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-8/train/model.ckpt-151955.meta 46a2ffce5158c962d72be3b3cf12a15a /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-8/train/model.ckpt-151955.index 28382fb71004441bfd891f8093272f9a /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-8/train/model.ckpt-151955.data-00000-of-00001 2e53513b09315728a17dda13997bc16d +encode-roadmap.basset.clf.testfold-9 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-9/train/model.ckpt-60996.meta edbf9d3e92b9bffb4da3e2828f5f39a0 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-9/train/model.ckpt-60996.index 20ddafde1077175ee9fba9fb04e8cacd /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/encode-roadmap.basset.clf.testfold-9/train/model.ckpt-60996.data-00000-of-00001 c87d589bbf7939c9731b67be8f414959 +ggr.basset.clf.pretrained.folds.testfold-0 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-0/train/model.ckpt-132614.meta 2ab642e0e5e658c7631e9ff19c71fbd2 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-0/train/model.ckpt-132614.index f7096b63b619eeb90f02ec28340d393b /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-0/train/model.ckpt-132614.data-00000-of-00001 af235f9a97e05d674cf9415d681396a0 +ggr.basset.clf.pretrained.folds.testfold-1 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-1/train/model.ckpt-131782.meta 65e4efdbb1b46f1d40895636ff3ce6d8 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-1/train/model.ckpt-131782.index 49dbd1be34d3320eed34907120bbb5de /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-1/train/model.ckpt-131782.data-00000-of-00001 107c14caab90c3a53726d4f9bc9e5263 +ggr.basset.clf.pretrained.folds.testfold-2 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-2/train/model.ckpt-189266.meta 782261a3527e5701db729b60798402f2 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-2/train/model.ckpt-189266.index b97ea67092c63e8408246b8941f590d8 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-2/train/model.ckpt-189266.data-00000-of-00001 e567de2dfe2200fe3beaeb09f78a7f3f +ggr.basset.clf.pretrained.folds.testfold-3 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-3/train/model.ckpt-130241.meta ed2e7b5864a4e27c1bcdf5bd39a3d124 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-3/train/model.ckpt-130241.index 141afb11372af5c1995ecee9835c0aa0 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-3/train/model.ckpt-130241.data-00000-of-00001 81bab1510b2ba0b1d5d003dadf1dd3e0 +ggr.basset.clf.pretrained.folds.testfold-4 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-4/train/model.ckpt-161932.meta 47d6bd1765232eca518f952aa390c831 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-4/train/model.ckpt-161932.index e7d3e3bdda1efc774955293ebb66a741 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-4/train/model.ckpt-161932.data-00000-of-00001 12853a4eac1aff2e4c11717b6a597bb4 +ggr.basset.clf.pretrained.folds.testfold-5 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-5/train/model.ckpt-104680.meta 836796deab0115c478799046ffc259e4 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-5/train/model.ckpt-104680.index b451245b241bb982da4d849eb34b4c4f /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-5/train/model.ckpt-104680.data-00000-of-00001 e00b4bfbede3ddd1353711f5781627a2 +ggr.basset.clf.pretrained.folds.testfold-6 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-6/train/model.ckpt-319382.meta 30263fabf0486566690976bbae01c76a /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-6/train/model.ckpt-319382.index f05f29b0516d13570750e3115e3ec9de /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-6/train/model.ckpt-319382.data-00000-of-00001 1f8ad4b104f5cb8b745427c78cf8da75 +ggr.basset.clf.pretrained.folds.testfold-7 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-7/train/model.ckpt-112201.meta 192af796388a4e81196b20aefd7bc80f /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-7/train/model.ckpt-112201.index 7777626f6ef77b939798b1e1054f1417 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-7/train/model.ckpt-112201.data-00000-of-00001 7e09578aaaf13282b69e2b75eb8eacaf +ggr.basset.clf.pretrained.folds.testfold-8 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-8/train/model.ckpt-157256.meta cd748e039896d7b8479e732a7c939a6a /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-8/train/model.ckpt-157256.index 762199800bf1e069dcefc6dda8086d0a /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-8/train/model.ckpt-157256.data-00000-of-00001 6d79d9209b244f9e3128369c20469d33 +ggr.basset.clf.pretrained.folds.testfold-9 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-9/train/model.ckpt-87046.meta 157b947c1fe776de59389720551d43a6 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-9/train/model.ckpt-87046.index 02c89e56a5a97ea38e76b05b07606edc /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.clf.pretrained.folds.testfold-9/train/model.ckpt-87046.data-00000-of-00001 c86ae5ecb133c46cdcb1150d42691b87 +ggr.basset.regr.pretrained.folds.testfold-0 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-0/train/model.ckpt-145519.meta 166fbec0c8151a45574cc7f08dd38a5c /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-0/train/model.ckpt-145519.index 3ddad77aef022cf8246063da16984be4 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-0/train/model.ckpt-145519.data-00000-of-00001 dec73096e14c3f9ff424542a1ab11875 +ggr.basset.regr.pretrained.folds.testfold-1 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-1/train/model.ckpt-139390.meta 959af5c6457f401579ee854d2327c7e7 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-1/train/model.ckpt-139390.index 9eee1096c7b06370fd83a47b8d6ad7f4 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-1/train/model.ckpt-139390.data-00000-of-00001 83431ab9129f960fd5435cb8f73d34bd +ggr.basset.regr.pretrained.folds.testfold-2 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-2/train/model.ckpt-194434.meta de7b3daafac97c860d5d8b6977430a3c /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-2/train/model.ckpt-194434.index 7085cac8d5487be035c0597b8dd0b8a0 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-2/train/model.ckpt-194434.data-00000-of-00001 d96ce0c3d876b0083b8693496c49d885 +ggr.basset.regr.pretrained.folds.testfold-3 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-3/train/model.ckpt-151657.meta aaef813a49a0b48072c89228e876d065 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-3/train/model.ckpt-151657.index 960d7df63e2cdfaa336602ddc1d6dce2 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-3/train/model.ckpt-151657.data-00000-of-00001 9c6d827a455b42d1620bb32254e20025 +ggr.basset.regr.pretrained.folds.testfold-4 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-4/train/model.ckpt-167318.meta 7d44122a482db6fba404bdb410d4c6f4 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-4/train/model.ckpt-167318.index e3e69db7809d45768819db3a3972ce16 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-4/train/model.ckpt-167318.data-00000-of-00001 a48f6a1d6e47b7fa110a1c794ebd75dc +ggr.basset.regr.pretrained.folds.testfold-5 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-5/train/model.ckpt-110042.meta f0be28d08e89cf5b25cb38ce2d6d5a39 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-5/train/model.ckpt-110042.index dd56fde6d5dd0078c462c916119392dc /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-5/train/model.ckpt-110042.data-00000-of-00001 948789840faf196eff94ee0ecbfb3ad1 +ggr.basset.regr.pretrained.folds.testfold-6 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-6/train/model.ckpt-332382.meta bdb67458c195ada1eb4d73af5addd84b /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-6/train/model.ckpt-332382.index 4d12f5815fb6b70c53becc57ceab44e7 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-6/train/model.ckpt-332382.data-00000-of-00001 8a7617a187769f1d23011fb24dfdf1a8 +ggr.basset.regr.pretrained.folds.testfold-7 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-7/train/model.ckpt-122745.meta a9a0c9e3fb344546687297fca335a193 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-7/train/model.ckpt-122745.index ed884985af03ddc3ba3f55ae7112feca /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-7/train/model.ckpt-122745.data-00000-of-00001 1e07c75d137934c3f1b6259022b606a8 +ggr.basset.regr.pretrained.folds.testfold-8 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-8/train/model.ckpt-165266.meta 54287421e562c7090fe22cf3995e1188 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-8/train/model.ckpt-165266.index 5de840becd3abb4f9a8f5f8d00a2c242 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-8/train/model.ckpt-165266.data-00000-of-00001 0b98d687db7b6c2d2247a053fd4074af +ggr.basset.regr.pretrained.folds.testfold-9 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-9/train/model.ckpt-92242.meta a84c58dbdaaaa3ee5cddacddac7e3d5e /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-9/train/model.ckpt-92242.index e60798b9e9933bd5d06d50bb56094a45 /mnt/lab_data/kundaje/users/dskim89/ggr/nn/models.2018-12-03/ggr.basset.regr.pretrained.folds.testfold-9/train/model.ckpt-92242.data-00000-of-00001 16463aeb88a068c146fe6eba660c1329 diff --git a/epidermal_basset/model.yaml b/epidermal_basset/model.yaml new file mode 100644 index 000000000..db7482398 --- /dev/null +++ b/epidermal_basset/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: {{ args_meta_url }} + md5: {{ args_meta_md5 }} + index: + url: {{ args_index_url }} + md5: {{ args_index_md5 }} + data: + url: {{ args_data_url }} + md5: {{ args_data_md5 }} + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org/10.1101/2020.10.16.342857 + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models.tsv b/epidermal_basset/models.tsv new file mode 100644 index 000000000..2f38770c3 --- /dev/null +++ b/epidermal_basset/models.tsv @@ -0,0 +1,31 @@ +model args_meta_url args_meta_md5 args_index_url args_index_md5 args_data_url args_data_md5 +encode-roadmap.basset.clf.testfold-0 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-0.model.meta/?download=1 0bb5678c3651564e37fdad5c4b875cd2 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-0.model.index/?download=1 cf33d8cd0a359a53381d66ee4c762542 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-0.model.data-00000-of-00001/?download=1 2f1d71990db937fb83586dc191863e1a +encode-roadmap.basset.clf.testfold-1 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-1.model.meta/?download=1 3b24a69ecd953cb8e341fe134f46e3d9 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-1.model.index/?download=1 8eaf8751af57976c35f770dc6498148f https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-1.model.data-00000-of-00001/?download=1 499918cb6cd189200cdc8b8107634a4a +encode-roadmap.basset.clf.testfold-2 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-2.model.meta/?download=1 708a9c8d20e2361242b3e63f0df84137 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-2.model.index/?download=1 0489c73298dbb124ec45b02a2a6f1637 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-2.model.data-00000-of-00001/?download=1 9127e729f247cb4e7b9973381a69b28e +encode-roadmap.basset.clf.testfold-3 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-3.model.meta/?download=1 623f8a17121d98cf8f7993f25d330705 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-3.model.index/?download=1 cbb20d8327c3a0cdc0a708b05ca971cd https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-3.model.data-00000-of-00001/?download=1 fd2997d78156594ea04196a18379e2fc +encode-roadmap.basset.clf.testfold-4 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-4.model.meta/?download=1 3be06b38124f68700a423e43935bfe51 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-4.model.index/?download=1 b5c3978cce00de7a1a339a48b7697119 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-4.model.data-00000-of-00001/?download=1 57508c46ffe33b5580fcd9772a0a7f95 +encode-roadmap.basset.clf.testfold-5 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-5.model.meta/?download=1 e5ff2b5579dd1ba1e7d1c0ea8a0c4be0 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-5.model.index/?download=1 5c9451f8a316f6b9d44722b53d210010 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-5.model.data-00000-of-00001/?download=1 79dde073c01580138d8b6cf50c1cae40 +encode-roadmap.basset.clf.testfold-6 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-6.model.meta/?download=1 cf060c2b1d4c9125ebec2accd74e298e https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-6.model.index/?download=1 fccf24f2e0846d5ea6f3d56f10d0cce2 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-6.model.data-00000-of-00001/?download=1 e4bc8ce0c499759025499af7b06ec6ee +encode-roadmap.basset.clf.testfold-7 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-7.model.meta/?download=1 dd863fc91e1e46743efdd52315b84235 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-7.model.index/?download=1 ff03710e1c38f8561877fdc8df600c22 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-7.model.data-00000-of-00001/?download=1 7b97299be0b9435360b24b3d526c5e60 +encode-roadmap.basset.clf.testfold-8 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-8.model.meta/?download=1 a7077d7359779958a43168aa8911a3af https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-8.model.index/?download=1 28382fb71004441bfd891f8093272f9a https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-8.model.data-00000-of-00001/?download=1 2e53513b09315728a17dda13997bc16d +encode-roadmap.basset.clf.testfold-9 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-9.model.meta/?download=1 3e18c83fbb1cc29eabd5f8f17988d7c0 https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-9.model.index/?download=1 20ddafde1077175ee9fba9fb04e8cacd https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-9.model.data-00000-of-00001/?download=1 c87d589bbf7939c9731b67be8f414959 +ggr.basset.clf.pretrained.folds.testfold-0 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-0.model.meta/?download=1 66dcdd18cd9e8e4560c79bf60c5c9160 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-0.model.index/?download=1 f7096b63b619eeb90f02ec28340d393b https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-0.model.data-00000-of-00001/?download=1 af235f9a97e05d674cf9415d681396a0 +ggr.basset.clf.pretrained.folds.testfold-1 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-1.model.meta/?download=1 794fac73c754d5fd88bbcf26b3439c77 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-1.model.index/?download=1 49dbd1be34d3320eed34907120bbb5de https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-1.model.data-00000-of-00001/?download=1 107c14caab90c3a53726d4f9bc9e5263 +ggr.basset.clf.pretrained.folds.testfold-2 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-2.model.meta/?download=1 6018ae7864a8cdd27f2fd3fbcd605b05 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-2.model.index/?download=1 b97ea67092c63e8408246b8941f590d8 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-2.model.data-00000-of-00001/?download=1 e567de2dfe2200fe3beaeb09f78a7f3f +ggr.basset.clf.pretrained.folds.testfold-3 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-3.model.meta/?download=1 97267b947ea33270d7ddb92d5c8b657a https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-3.model.index/?download=1 141afb11372af5c1995ecee9835c0aa0 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-3.model.data-00000-of-00001/?download=1 81bab1510b2ba0b1d5d003dadf1dd3e0 +ggr.basset.clf.pretrained.folds.testfold-4 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-4.model.meta/?download=1 9f39931ce83f26295f21f977e9fa0979 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-4.model.index/?download=1 e7d3e3bdda1efc774955293ebb66a741 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-4.model.data-00000-of-00001/?download=1 12853a4eac1aff2e4c11717b6a597bb4 +ggr.basset.clf.pretrained.folds.testfold-5 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-5.model.meta/?download=1 74ba6c7b7a505ade74e5b3ed6bf55f9f https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-5.model.index/?download=1 b451245b241bb982da4d849eb34b4c4f https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-5.model.data-00000-of-00001/?download=1 e00b4bfbede3ddd1353711f5781627a2 +ggr.basset.clf.pretrained.folds.testfold-6 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-6.model.meta/?download=1 fd242cecdc33930243701f0389b2ba5e https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-6.model.index/?download=1 f05f29b0516d13570750e3115e3ec9de https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-6.model.data-00000-of-00001/?download=1 1f8ad4b104f5cb8b745427c78cf8da75 +ggr.basset.clf.pretrained.folds.testfold-7 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-7.model.meta/?download=1 f77284d300cc441bfc47d8758cd1613c https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-7.model.index/?download=1 7777626f6ef77b939798b1e1054f1417 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-7.model.data-00000-of-00001/?download=1 7e09578aaaf13282b69e2b75eb8eacaf +ggr.basset.clf.pretrained.folds.testfold-8 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-8.model.meta/?download=1 bea6295aaa0308e534b81b3604b4a9b7 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-8.model.index/?download=1 762199800bf1e069dcefc6dda8086d0a https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-8.model.data-00000-of-00001/?download=1 6d79d9209b244f9e3128369c20469d33 +ggr.basset.clf.pretrained.folds.testfold-9 https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-9.model.meta/?download=1 7a6552718b780b00dd9ab04877bccb3f https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-9.model.index/?download=1 02c89e56a5a97ea38e76b05b07606edc https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-9.model.data-00000-of-00001/?download=1 c86ae5ecb133c46cdcb1150d42691b87 +ggr.basset.regr.pretrained.folds.testfold-0 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-0.model.meta/?download=1 9185274583bd67ca1bf3f376749dea0a https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-0.model.index/?download=1 3ddad77aef022cf8246063da16984be4 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-0.model.data-00000-of-00001/?download=1 dec73096e14c3f9ff424542a1ab11875 +ggr.basset.regr.pretrained.folds.testfold-1 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-1.model.meta/?download=1 9eff2aa7236d73e24f4b09dc7ee4da35 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-1.model.index/?download=1 9eee1096c7b06370fd83a47b8d6ad7f4 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-1.model.data-00000-of-00001/?download=1 83431ab9129f960fd5435cb8f73d34bd +ggr.basset.regr.pretrained.folds.testfold-2 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-2.model.meta/?download=1 000060b0ced8870536448e74267935cb https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-2.model.index/?download=1 7085cac8d5487be035c0597b8dd0b8a0 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-2.model.data-00000-of-00001/?download=1 d96ce0c3d876b0083b8693496c49d885 +ggr.basset.regr.pretrained.folds.testfold-3 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-3.model.meta/?download=1 60a760c88e13aa0d0453e979c9a2b95e https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-3.model.index/?download=1 960d7df63e2cdfaa336602ddc1d6dce2 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-3.model.data-00000-of-00001/?download=1 9c6d827a455b42d1620bb32254e20025 +ggr.basset.regr.pretrained.folds.testfold-4 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-4.model.meta/?download=1 05786276eb9044e0cc1b8645af24f7b9 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-4.model.index/?download=1 e3e69db7809d45768819db3a3972ce16 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-4.model.data-00000-of-00001/?download=1 a48f6a1d6e47b7fa110a1c794ebd75dc +ggr.basset.regr.pretrained.folds.testfold-5 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-5.model.meta/?download=1 aecbc5165666647f2ad9d5113e056156 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-5.model.index/?download=1 dd56fde6d5dd0078c462c916119392dc https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-5.model.data-00000-of-00001/?download=1 948789840faf196eff94ee0ecbfb3ad1 +ggr.basset.regr.pretrained.folds.testfold-6 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-6.model.meta/?download=1 7c8c3b9c7f0116c5016c1f8733540c27 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-6.model.index/?download=1 4d12f5815fb6b70c53becc57ceab44e7 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-6.model.data-00000-of-00001/?download=1 8a7617a187769f1d23011fb24dfdf1a8 +ggr.basset.regr.pretrained.folds.testfold-7 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-7.model.meta/?download=1 71ef1b06ca597efacaed20c56171495f https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-7.model.index/?download=1 ed884985af03ddc3ba3f55ae7112feca https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-7.model.data-00000-of-00001/?download=1 1e07c75d137934c3f1b6259022b606a8 +ggr.basset.regr.pretrained.folds.testfold-8 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-8.model.meta/?download=1 6789841151a1a99998e1e3850dc436c9 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-8.model.index/?download=1 5de840becd3abb4f9a8f5f8d00a2c242 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-8.model.data-00000-of-00001/?download=1 0b98d687db7b6c2d2247a053fd4074af +ggr.basset.regr.pretrained.folds.testfold-9 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-9.model.meta/?download=1 bbc8da8be311a4fa4104ff19332bad88 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-9.model.index/?download=1 e60798b9e9933bd5d06d50bb56094a45 https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-9.model.data-00000-of-00001/?download=1 16463aeb88a068c146fe6eba660c1329 diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-0/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-0/model.yaml new file mode 100644 index 000000000..75754e945 --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-0/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-0.model.meta/?download=1 + md5: 0bb5678c3651564e37fdad5c4b875cd2 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-0.model.index/?download=1 + md5: cf33d8cd0a359a53381d66ee4c762542 + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-0.model.data-00000-of-00001/?download=1 + md5: 2f1d71990db937fb83586dc191863e1a + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-1/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-1/model.yaml new file mode 100644 index 000000000..b647200e7 --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-1/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-1.model.meta/?download=1 + md5: 3b24a69ecd953cb8e341fe134f46e3d9 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-1.model.index/?download=1 + md5: 8eaf8751af57976c35f770dc6498148f + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-1.model.data-00000-of-00001/?download=1 + md5: 499918cb6cd189200cdc8b8107634a4a + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-2/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-2/model.yaml new file mode 100644 index 000000000..b8816eff2 --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-2/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-2.model.meta/?download=1 + md5: 708a9c8d20e2361242b3e63f0df84137 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-2.model.index/?download=1 + md5: 0489c73298dbb124ec45b02a2a6f1637 + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-2.model.data-00000-of-00001/?download=1 + md5: 9127e729f247cb4e7b9973381a69b28e + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-3/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-3/model.yaml new file mode 100644 index 000000000..3cd7ea1bd --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-3/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-3.model.meta/?download=1 + md5: 623f8a17121d98cf8f7993f25d330705 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-3.model.index/?download=1 + md5: cbb20d8327c3a0cdc0a708b05ca971cd + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-3.model.data-00000-of-00001/?download=1 + md5: fd2997d78156594ea04196a18379e2fc + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-4/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-4/model.yaml new file mode 100644 index 000000000..a6c562d30 --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-4/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-4.model.meta/?download=1 + md5: 3be06b38124f68700a423e43935bfe51 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-4.model.index/?download=1 + md5: b5c3978cce00de7a1a339a48b7697119 + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-4.model.data-00000-of-00001/?download=1 + md5: 57508c46ffe33b5580fcd9772a0a7f95 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-5/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-5/model.yaml new file mode 100644 index 000000000..b817a0b0b --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-5/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-5.model.meta/?download=1 + md5: e5ff2b5579dd1ba1e7d1c0ea8a0c4be0 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-5.model.index/?download=1 + md5: 5c9451f8a316f6b9d44722b53d210010 + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-5.model.data-00000-of-00001/?download=1 + md5: 79dde073c01580138d8b6cf50c1cae40 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-6/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-6/model.yaml new file mode 100644 index 000000000..512718333 --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-6/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-6.model.meta/?download=1 + md5: cf060c2b1d4c9125ebec2accd74e298e + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-6.model.index/?download=1 + md5: fccf24f2e0846d5ea6f3d56f10d0cce2 + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-6.model.data-00000-of-00001/?download=1 + md5: e4bc8ce0c499759025499af7b06ec6ee + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-7/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-7/model.yaml new file mode 100644 index 000000000..d29007df0 --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-7/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-7.model.meta/?download=1 + md5: dd863fc91e1e46743efdd52315b84235 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-7.model.index/?download=1 + md5: ff03710e1c38f8561877fdc8df600c22 + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-7.model.data-00000-of-00001/?download=1 + md5: 7b97299be0b9435360b24b3d526c5e60 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-8/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-8/model.yaml new file mode 100644 index 000000000..ccb13a0e4 --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-8/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-8.model.meta/?download=1 + md5: a7077d7359779958a43168aa8911a3af + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-8.model.index/?download=1 + md5: 28382fb71004441bfd891f8093272f9a + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-8.model.data-00000-of-00001/?download=1 + md5: 2e53513b09315728a17dda13997bc16d + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/encode-roadmap.basset.clf.testfold-9/model.yaml b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-9/model.yaml new file mode 100644 index 000000000..45c0bed7a --- /dev/null +++ b/epidermal_basset/models/encode-roadmap.basset.clf.testfold-9/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-9.model.meta/?download=1 + md5: 3e18c83fbb1cc29eabd5f8f17988d7c0 + index: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-9.model.index/?download=1 + md5: 20ddafde1077175ee9fba9fb04e8cacd + data: + url: https://zenodo.org/record/4777310/files/encode-roadmap.basset.clf.testfold-9.model.data-00000-of-00001/?download=1 + md5: c87d589bbf7939c9731b67be8f414959 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-0/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-0/model.yaml new file mode 100644 index 000000000..ac3670325 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-0/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-0.model.meta/?download=1 + md5: 66dcdd18cd9e8e4560c79bf60c5c9160 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-0.model.index/?download=1 + md5: f7096b63b619eeb90f02ec28340d393b + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-0.model.data-00000-of-00001/?download=1 + md5: af235f9a97e05d674cf9415d681396a0 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-1/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-1/model.yaml new file mode 100644 index 000000000..d831c459d --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-1/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-1.model.meta/?download=1 + md5: 794fac73c754d5fd88bbcf26b3439c77 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-1.model.index/?download=1 + md5: 49dbd1be34d3320eed34907120bbb5de + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-1.model.data-00000-of-00001/?download=1 + md5: 107c14caab90c3a53726d4f9bc9e5263 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-2/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-2/model.yaml new file mode 100644 index 000000000..1874f02ca --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-2/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-2.model.meta/?download=1 + md5: 6018ae7864a8cdd27f2fd3fbcd605b05 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-2.model.index/?download=1 + md5: b97ea67092c63e8408246b8941f590d8 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-2.model.data-00000-of-00001/?download=1 + md5: e567de2dfe2200fe3beaeb09f78a7f3f + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-3/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-3/model.yaml new file mode 100644 index 000000000..d222ccc1f --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-3/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-3.model.meta/?download=1 + md5: 97267b947ea33270d7ddb92d5c8b657a + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-3.model.index/?download=1 + md5: 141afb11372af5c1995ecee9835c0aa0 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-3.model.data-00000-of-00001/?download=1 + md5: 81bab1510b2ba0b1d5d003dadf1dd3e0 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-4/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-4/model.yaml new file mode 100644 index 000000000..d633186b6 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-4/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-4.model.meta/?download=1 + md5: 9f39931ce83f26295f21f977e9fa0979 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-4.model.index/?download=1 + md5: e7d3e3bdda1efc774955293ebb66a741 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-4.model.data-00000-of-00001/?download=1 + md5: 12853a4eac1aff2e4c11717b6a597bb4 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-5/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-5/model.yaml new file mode 100644 index 000000000..2bfc2ab77 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-5/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-5.model.meta/?download=1 + md5: 74ba6c7b7a505ade74e5b3ed6bf55f9f + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-5.model.index/?download=1 + md5: b451245b241bb982da4d849eb34b4c4f + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-5.model.data-00000-of-00001/?download=1 + md5: e00b4bfbede3ddd1353711f5781627a2 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-6/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-6/model.yaml new file mode 100644 index 000000000..704c76c27 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-6/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-6.model.meta/?download=1 + md5: fd242cecdc33930243701f0389b2ba5e + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-6.model.index/?download=1 + md5: f05f29b0516d13570750e3115e3ec9de + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-6.model.data-00000-of-00001/?download=1 + md5: 1f8ad4b104f5cb8b745427c78cf8da75 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-7/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-7/model.yaml new file mode 100644 index 000000000..bf1475728 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-7/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-7.model.meta/?download=1 + md5: f77284d300cc441bfc47d8758cd1613c + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-7.model.index/?download=1 + md5: 7777626f6ef77b939798b1e1054f1417 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-7.model.data-00000-of-00001/?download=1 + md5: 7e09578aaaf13282b69e2b75eb8eacaf + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-8/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-8/model.yaml new file mode 100644 index 000000000..9596b4529 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-8/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-8.model.meta/?download=1 + md5: bea6295aaa0308e534b81b3604b4a9b7 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-8.model.index/?download=1 + md5: 762199800bf1e069dcefc6dda8086d0a + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-8.model.data-00000-of-00001/?download=1 + md5: 6d79d9209b244f9e3128369c20469d33 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-9/model.yaml b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-9/model.yaml new file mode 100644 index 000000000..5320b5139 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.clf.pretrained.folds.testfold-9/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-9.model.meta/?download=1 + md5: 7a6552718b780b00dd9ab04877bccb3f + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-9.model.index/?download=1 + md5: 02c89e56a5a97ea38e76b05b07606edc + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.clf.pretrained.folds.testfold-9.model.data-00000-of-00001/?download=1 + md5: c86ae5ecb133c46cdcb1150d42691b87 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-0/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-0/model.yaml new file mode 100644 index 000000000..c4d0b5525 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-0/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-0.model.meta/?download=1 + md5: 9185274583bd67ca1bf3f376749dea0a + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-0.model.index/?download=1 + md5: 3ddad77aef022cf8246063da16984be4 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-0.model.data-00000-of-00001/?download=1 + md5: dec73096e14c3f9ff424542a1ab11875 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-1/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-1/model.yaml new file mode 100644 index 000000000..2c3982a51 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-1/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-1.model.meta/?download=1 + md5: 9eff2aa7236d73e24f4b09dc7ee4da35 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-1.model.index/?download=1 + md5: 9eee1096c7b06370fd83a47b8d6ad7f4 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-1.model.data-00000-of-00001/?download=1 + md5: 83431ab9129f960fd5435cb8f73d34bd + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-2/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-2/model.yaml new file mode 100644 index 000000000..fe193df9e --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-2/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-2.model.meta/?download=1 + md5: 000060b0ced8870536448e74267935cb + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-2.model.index/?download=1 + md5: 7085cac8d5487be035c0597b8dd0b8a0 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-2.model.data-00000-of-00001/?download=1 + md5: d96ce0c3d876b0083b8693496c49d885 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-3/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-3/model.yaml new file mode 100644 index 000000000..64f1c2eb2 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-3/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-3.model.meta/?download=1 + md5: 60a760c88e13aa0d0453e979c9a2b95e + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-3.model.index/?download=1 + md5: 960d7df63e2cdfaa336602ddc1d6dce2 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-3.model.data-00000-of-00001/?download=1 + md5: 9c6d827a455b42d1620bb32254e20025 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-4/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-4/model.yaml new file mode 100644 index 000000000..04a7ea4a3 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-4/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-4.model.meta/?download=1 + md5: 05786276eb9044e0cc1b8645af24f7b9 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-4.model.index/?download=1 + md5: e3e69db7809d45768819db3a3972ce16 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-4.model.data-00000-of-00001/?download=1 + md5: a48f6a1d6e47b7fa110a1c794ebd75dc + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-5/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-5/model.yaml new file mode 100644 index 000000000..c1e5307a7 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-5/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-5.model.meta/?download=1 + md5: aecbc5165666647f2ad9d5113e056156 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-5.model.index/?download=1 + md5: dd56fde6d5dd0078c462c916119392dc + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-5.model.data-00000-of-00001/?download=1 + md5: 948789840faf196eff94ee0ecbfb3ad1 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-6/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-6/model.yaml new file mode 100644 index 000000000..84f9df725 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-6/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-6.model.meta/?download=1 + md5: 7c8c3b9c7f0116c5016c1f8733540c27 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-6.model.index/?download=1 + md5: 4d12f5815fb6b70c53becc57ceab44e7 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-6.model.data-00000-of-00001/?download=1 + md5: 8a7617a187769f1d23011fb24dfdf1a8 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-7/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-7/model.yaml new file mode 100644 index 000000000..a40a7ceef --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-7/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-7.model.meta/?download=1 + md5: 71ef1b06ca597efacaed20c56171495f + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-7.model.index/?download=1 + md5: ed884985af03ddc3ba3f55ae7112feca + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-7.model.data-00000-of-00001/?download=1 + md5: 1e07c75d137934c3f1b6259022b606a8 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-8/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-8/model.yaml new file mode 100644 index 000000000..b2c506fc6 --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-8/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-8.model.meta/?download=1 + md5: 6789841151a1a99998e1e3850dc436c9 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-8.model.index/?download=1 + md5: 5de840becd3abb4f9a8f5f8d00a2c242 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-8.model.data-00000-of-00001/?download=1 + md5: 0b98d687db7b6c2d2247a053fd4074af + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-9/model.yaml b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-9/model.yaml new file mode 100644 index 000000000..61f1abc3d --- /dev/null +++ b/epidermal_basset/models/ggr.basset.regr.pretrained.folds.testfold-9/model.yaml @@ -0,0 +1,60 @@ +defined_as: kipoi.model.TensorFlowModel +args: # arguments of kipoi.model.TensorFlowModel + input_nodes: "inputs" + target_nodes: "basset/logits/biases/RMSProp_1" + checkpoint_path: + meta: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-9.model.meta/?download=1 + md5: bbc8da8be311a4fa4104ff19332bad88 + index: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-9.model.index/?download=1 + md5: e60798b9e9933bd5d06d50bb56094a45 + data: + url: https://zenodo.org/record/4777310/files/ggr.basset.regr.pretrained.folds.testfold-9.model.data-00000-of-00001/?download=1 + md5: 16463aeb88a068c146fe6eba660c1329 + +default_dataloader: + defined_as: kipoiseq.dataloaders.SeqIntervalDl + + default_args: # Optional arguments to the SeqIntervalDl dataloader + # See also https://kipoi.org/kipoiseq/dataloaders/#seqintervaldl + auto_resize_len: 1000 # Automatically resize sequence intervals + alphabet_axis: 2 + dummy_axis: 0 # Add a dummy axis. Omit in order not to create dummy_axis. + alphabet: "ACGT" # Order of letters in 1-hot encoding + ignore_targets: False # if True, dont return any target variables + +info: # General information about the model + authors: + - name: Daniel Kim + github: vervacity + email: danielskim@stanford.edu + doc: Model predicting accessibility/chromatin marks from sequence + cite_as: https://doi.org:/... # preferably a doi url to the paper + trained_on: see README + license: MIT # Software License - if not set defaults to MIT + # You can also specify the license in the LICENSE file + +dependencies: + conda: # install via conda + - python=2.7 + - h5py + pip: # install via pip + #- keras>=2.0.4 + - tensorflow>=1.8 + +schema: # Model schema. The schema defintion is essential for kipoi plug-ins to work. + inputs: # input = single numpy array + shape: (1,1000,4) # array shape of a single sample (omitting the batch dimension) + doc: input feature description + + # inputs: # input = dictionary of fields + # seq: + # shape: (100,4) + # doc: input feature description + # other_track: + # shape: (50,) + # doc: input feature description + targets: + shape: (1996,) + doc: model prediction description \ No newline at end of file diff --git a/epidermal_basset/run_tests.py b/epidermal_basset/run_tests.py new file mode 100644 index 000000000..a3c880713 --- /dev/null +++ b/epidermal_basset/run_tests.py @@ -0,0 +1,26 @@ + +import os +import glob + + +def main(): + """test all models in their folders + """ + model_yamls = sorted(glob.glob("models/*/model.yaml")) + for model_yaml in model_yamls: + model_dir = os.path.dirname(model_yaml) + + # test + kipoi_cmd = "kipoi test {}/".format(model_dir) + print(kipoi_cmd) + os.system(kipoi_cmd) + + # clean up + clean_cmd = "rm -r {}/downloaded/".format(model_dir) + print(clean_cmd) + os.system(clean_cmd) + + return + + +main() diff --git a/epidermal_basset/scripts/make_model_dirs.py b/epidermal_basset/scripts/make_model_dirs.py new file mode 100644 index 000000000..1162b5bc2 --- /dev/null +++ b/epidermal_basset/scripts/make_model_dirs.py @@ -0,0 +1,54 @@ + +import os + +import pandas as pd + + +def main(): + """for whatever reason, testing on models.tsv doesn't work + so just generating model dirs for each model + """ + # set up + out_dir = "models" + os.system("mkdir -p {}".format(out_dir)) + + # read in models + models_file = "models.tsv" + models = pd.read_csv(models_file, sep="\t") + + for model_idx in range(models.shape[0]): + model_info = models.iloc[model_idx] + + # set up model dir + model_dir = "{}/{}".format(out_dir, model_info["model"]) + os.system("mkdir -p {}".format(model_dir)) + os.system("cp model.yaml {}/model.TMP.yaml".format(model_dir)) + + # adjust model yaml + old_yaml = "{}/model.TMP.yaml".format(model_dir) + new_yaml = "{}/model.yaml".format(model_dir) + with open(new_yaml, "w") as out: + with open(old_yaml, "r") as fp: + for line in fp: + if "{{ args_meta_url }}" in line: + line = " url: {}\n".format(model_info["args_meta_url"]) + if "{{ args_meta_md5 }}" in line: + line = " md5: {}\n".format(model_info["args_meta_md5"]) + if "{{ args_index_url }}" in line: + line = " url: {}\n".format(model_info["args_index_url"]) + if "{{ args_index_md5 }}" in line: + line = " md5: {}\n".format(model_info["args_index_md5"]) + if "{{ args_data_url }}" in line: + line = " url: {}\n".format(model_info["args_data_url"]) + if "{{ args_data_md5 }}" in line: + line = " md5: {}\n".format(model_info["args_data_md5"]) + + out.write(line) + + # clean up + os.system("rm {}".format(old_yaml)) + + return + + +main() diff --git a/epidermal_basset/scripts/run_kipoi_compatibility.py b/epidermal_basset/scripts/run_kipoi_compatibility.py new file mode 100644 index 000000000..335a41d40 --- /dev/null +++ b/epidermal_basset/scripts/run_kipoi_compatibility.py @@ -0,0 +1,90 @@ +#!/bin/env python + +import os +import subprocess + +import tensorflow as tf +import pandas as pd + + +def get_md5sum(filename): + """get md5sum + """ + md5sum_val = subprocess.check_output(["md5sum", filename]).decode("ascii") + md5sum_val = md5sum_val.split()[0] + + return md5sum_val + + +def adjust_meta_file(old_meta_file, new_meta_file): + """properly insert placeholder into the graph so that Kipoi + can utilize its own dataloader + """ + # set up graph with placeholders to feed into model + tf.reset_default_graph() + inputs = tf.placeholder(tf.float32, + shape=(None, 1, 1000, 4), + name="inputs") + + # load metagraph + loaded_meta = tf.train.import_meta_graph( + old_meta_file, + input_map={ + "map/TensorArrayStack/TensorArrayGatherV3": inputs}) + #ops = tf.get_default_graph().get_operations() + #print(ops) + + # save out with placeholders + out = tf.train.export_meta_graph(filename=new_meta_file, as_text=True) + + return + + +def main(): + """take TF models and make kipoi compatible + Notes: key thing is to get into the TF meta file and figure out + which ops are the ones to adjust. Can use import_meta_graph with + get_operations (see below) to take a look. + + only replaced the input side + """ + # work dir + new_model_dir = "models.kipoi_compatible" + os.system("mkdir -p {}".format(new_model_dir)) + + # models + orig_model_table = "models_orig.tsv" + models = pd.read_csv(orig_model_table, sep="\t") + + # go through each model to update + for orig_model_idx in range(models.shape[0]): + model_info = models.iloc[orig_model_idx] + + # copy over data + model_data_name = "{}/{}.model.data-00000-of-00001".format( + new_model_dir, model_info["names"]) + os.system("cp {} {}".format(model_info["args_data_url"], model_data_name)) + models.loc[orig_model_idx, "args_data_url"] = model_data_name + + # copy over index + model_index_name = "{}/{}.model.index".format( + new_model_dir, model_info["names"]) + os.system("cp {} {}".format(model_info["args_index_url"], model_index_name)) + models.loc[orig_model_idx, "args_index_url"] = model_index_name + + # adjust meta and get new md5 sum + model_meta_name = "{}/{}.model.meta".format( + new_model_dir, model_info["names"]) + adjust_meta_file(model_info["args_meta_url"], model_meta_name) + models.loc[orig_model_idx, "args_meta_url"] = model_meta_name + md5sum_val = get_md5sum(model_meta_name) + models.loc[orig_model_idx, "args_meta_md5"] = md5sum_val + + # save out to models.tsv + models.to_csv( + "models.tsv", sep="\t", index=False, header=True) + + return + + +main() diff --git a/epidermal_basset/scripts/update_w_zenodo_links.py b/epidermal_basset/scripts/update_w_zenodo_links.py new file mode 100644 index 000000000..7d22d5654 --- /dev/null +++ b/epidermal_basset/scripts/update_w_zenodo_links.py @@ -0,0 +1,39 @@ + +import re + +import pandas as pd + +def main(): + """update models.tsv with zenodo links + """ + # zenodo record + zenodo_url_prefix = "https://zenodo.org/record/4777310/files" + zenodo_url_suffix = "?download=1" + + # read in models + old_models_file = "models_kipoi_local.tsv" + models = pd.read_csv(old_models_file, sep="\t") + + models["args_meta_url"] = [ + "{}/{}".format( + re.sub("models.kipoi_compatible", zenodo_url_prefix, url), zenodo_url_suffix) + for url in models["args_meta_url"]] + + models["args_index_url"] = [ + "{}/{}".format( + re.sub("models.kipoi_compatible", zenodo_url_prefix, url), zenodo_url_suffix) + for url in models["args_index_url"]] + + models["args_data_url"] = [ + "{}/{}".format( + re.sub("models.kipoi_compatible", zenodo_url_prefix, url), zenodo_url_suffix) + for url in models["args_data_url"]] + + # save out + models.to_csv( + "models.tsv", sep="\t", header=True, index=False) + + return + + +main() diff --git a/epidermal_basset/scripts/upload_zenodo.py b/epidermal_basset/scripts/upload_zenodo.py new file mode 100644 index 000000000..3cc754f12 --- /dev/null +++ b/epidermal_basset/scripts/upload_zenodo.py @@ -0,0 +1,112 @@ + +import os +import sys +import json +import requests + +import pandas as pd + + +def _upload_file( + filename, + full_path_filename, + bucket_url, + params): + """ we pass the file object (fp) directly to the request as the data to be uploaded + the target URL is a combination of the buckets link with the desired filename separated by a slash + """ + with open(full_path_filename, "rb") as fp: + r = requests.put( + "{}/{}".format(bucket_url, filename), + data=fp, + # No headers included in the request, since it's a raw byte request + params=params) + + return + + + +def main(): + """upload script: to set up links for models + """ + # USER INPUT + ACCESS_TOKEN = sys.argv[1] # always read in access token, do not save to script!! + deposition_id = None # either none or a numerical ID + + # file(s) - give with full paths + UPLOAD_DIR = "/mnt/lab_data/kundaje/users/dskim89/ggr/nn/kipoi/models.kipoi_compatible" + filenames = [] + model_table = pd.read_csv("models.tsv", sep="\t") + for model_idx in range(model_table.shape[0]): + model_info = model_table.iloc[model_idx] + model_data = "{}/{}".format(UPLOAD_DIR, os.path.basename(model_info["args_data_url"])) + model_meta = "{}/{}".format(UPLOAD_DIR, os.path.basename(model_info["args_meta_url"])) + model_index = "{}/{}".format(UPLOAD_DIR, os.path.basename(model_info["args_index_url"])) + filenames += [model_data, model_meta, model_index] + + # deposition metadata - fill this out as needed + metadata = { + "metadata": { + "title": "Convolutional Neural Net (CNN) models for epigenomic landscapes in epithelial differentiation - KIPOI COMPATIBLE", + "upload_type": "dataset", + "description": "Deep learning models trained on epigenomic landscapes in keratinocyte differentiation", + "creators": [ + {"name": "Kim, Daniel Sunwook", + "affiliation": "Stanford School of Medicine"}, + {"name": "Kundaje, Anshul", + "affiliation": "Stanford School of Medicine"} + ] + } + } + + # END USER INPUT + + + # set up + if deposition_id is None: + new_deposition = True + else: + new_deposition = False + + # API set up + headers = {"Content-Type": "application/json"} + params = {'access_token': ACCESS_TOKEN} + + # get the deposition: either make new, or grab old one + if new_deposition: + r = requests.post( + 'https://zenodo.org/api/deposit/depositions', + params=params, + json={}, + # Headers are not necessary here since "requests" automatically + # adds "Content-Type: application/json", because we're using + # the "json=" keyword argument + # headers=headers, + headers=headers) + deposition_id = r.json()["id"] + + # add deposition metadata + r = requests.put( + "https://zenodo.org/api/deposit/depositions/{}".format(deposition_id), + params={'access_token': ACCESS_TOKEN}, + data=json.dumps(metadata), + headers=headers) + + else: + r = requests.get( + "https://zenodo.org/api/deposit/depositions/{}".format(deposition_id), + params=params) + #print r.status_code + print r.json() + bucket_url = r.json()["links"]["bucket"] + + # upload files + for filename in filenames: + full_path_filename = filename + base_filename = os.path.basename(filename) + _upload_file(base_filename, full_path_filename, bucket_url, params) + + + return + +main()