init commit

config.json

@@ -0,0 +1,40 @@
+{
+  "_name_or_path": "/usr/local/google/home/vishvak/Projects/datamux-pretraining-root/checkpoints/release_checkpoints/bert/base/gaussian_attention_v2_index_pos/pt/10",
+  "architectures": [
+    "MuxedBertForMaskedLM"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "binary_hadamard_epsilon": 0.0,
+  "demuxing_variant": "index_pos",
+  "gaussian_hadamard_norm": 1.0,
+  "gradient_checkpointing": false,
+  "head_temperature": 100.0,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "hierarchical_softmax_buckets": 100,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "learn_muxing": 0,
+  "legacy_demuxing": false,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "muxing_variant": "gaussian_attention_v2",
+  "num_attention_heads": 12,
+  "num_hidden_demux_layers": 3,
+  "num_hidden_layers": 12,
+  "num_instances": 10,
+  "output_attentions": true,
+  "output_hidden_states": true,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "retrieval_loss_coeff": 0.0,
+  "retrieval_percentage": 1.0,
+  "task_loss_coeff": 1.0,
+  "transformers_version": "4.4.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "use_hierarchical_softmax": false,
+  "vocab_size": 30522
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28b06ff3ffc83a3277ff93bea0c61da5d7811dc918d1ebda6da2c78f768a1591
+size 506759303