Push model using huggingface_hub.

README.md

@@ -1,15 +1,10 @@
 ---
-language: en
-license: mit
 tags:
-- pytorch
-- pertTF
+- model_hub_mixin
+- pytorch_model_hub_mixin
 ---
 
-# pertTF-large
-
-This is model files for pertTF-large
-
-## Usage
-```python
-model = HFPerturbationTFModel.from_pretrained('weililab/pertTF-large')
+This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
+- Code: [More Information Needed]
+- Paper: [More Information Needed]
+- Docs: [More Information Needed]

config.json

@@ -1,9 +1,9 @@
 {
   "cell_emb_style": "cls",
-  "d_hid": null,
-  "d_model": 32,
+  "d_hid": 512,
+  "d_model": 512,
   "do_dab": false,
-  "do_mvc": false,
+  "do_mvc": true,
   "domain_spec_batchnorm": false,
   "dropout": 0.1,
   "ecs_threshold": 0.7,
@@ -12,8 +12,8 @@
   "input_emb_style": "continuous",
   "mvc_decoder_style": "inner product",
   "n_bins": 51,
-  "n_cls": 1,
-  "n_pert": 1,
+  "n_cls": 20,
+  "n_pert": 40,
   "n_ps": 1,
   "nhead": 8,
   "nlayers": 12,
@@ -29,6 +29,6 @@
   "pred_lochness_next": false,
   "ps_decoder2_nlayer": 3,
   "use_batch_labels": false,
-  "use_fast_transformer": false,
+  "use_fast_transformer": true,
   "vocab": null
 }

training_config.json

@@ -54,5 +54,36 @@
   "hvg_col": "highly_variable",
   "mvc_masked_train": true,
   "max_seq_len": 3001,
-  "pad_value": -2
+  "n_pert": 40,
+  "nlayers_pert": 4,
+  "n_ps": 1,
+  "ntoken": 36604,
+  "d_model": 512,
+  "nhead": 8,
+  "d_hid": 512,
+  "nlayers": 12,
+  "nlayers_cls": 3,
+  "n_cls": 20,
+  "vocab": null,
+  "dropout": 0.1,
+  "pad_token": "<pad>",
+  "pad_value": -2,
+  "do_mvc": true,
+  "do_dab": false,
+  "use_batch_labels": false,
+  "num_batch_labels": null,
+  "domain_spec_batchnorm": false,
+  "input_emb_style": "continuous",
+  "n_bins": 51,
+  "cell_emb_style": "cls",
+  "mvc_decoder_style": "inner product",
+  "ecs_threshold": 0.7,
+  "explicit_zero_prob": true,
+  "use_fast_transformer": true,
+  "fast_transformer_backend": "flash",
+  "pre_norm": false,
+  "pred_lochness_next": false,
+  "ps_decoder2_nlayer": 3,
+  "pert_pad_id": null,
+  "pert_dim": null
 }