Upload 11 files

README.md

@@ -1,3 +1,9 @@
 ---
 license: mit
----
+datasets:
+- EleutherAI/pile
+language:
+- en
+metrics:
+- glue
+---

arch_budget_hours_24.json

@@ -0,0 +1,57 @@
+{
+    "architectures": [
+        "ScriptableCrammedBERT"
+    ],
+    "num_transformer_layers": 16,
+    "intermed_type": "fff",
+    "hidden_size": 768,
+    "intermed_size": 1,
+    "intermed_depth": 11,
+    "hidden_dropout_prob": 0.1,
+    "norm": "LayerNorm",
+    "norm_eps": 1e-12,
+    "norm_scheme": "pre",
+    "nonlin": "GELU",
+    "tie_weights": true,
+    "decoder_bias": false,
+    "sparse_prediction": 0.15,
+    "loss": "cross-entropy",
+    "objective_layout": "MLM",
+    "embedding": {
+        "vocab_size": null,
+        "pos_embedding": "scaled-sinusoidal",
+        "dropout_prob": 0.1,
+        "pad_token_id": 0,
+        "max_seq_length": 128,
+        "embedding_dim": 768,
+        "normalization": true,
+        "stable_low_precision": false
+    },
+    "attention": {
+        "type": "self-attention",
+        "causal_attention": false,
+        "num_attention_heads": 12,
+        "dropout_prob": 0.1,
+        "skip_output_projection": false,
+        "qkv_bias": false,
+        "rotary_embedding": false,
+        "seq_op_in_fp32": false,
+        "sequence_op": "torch-softmax"
+    },
+    "init": {
+        "type": "normal",
+        "std": 0.02
+    },
+    "ffn_layer_frequency": 1,
+    "skip_head_transform": true,
+    "use_bias": false,
+    "final_norm": true,
+    "num_labels": null,
+    "classification_head": {
+        "pooler": "avg",
+        "include_ff_layer": true,
+        "head_dim": 1024,
+        "nonlin": "Tanh",
+        "classifier_dropout": 0.1
+    }
+}

config.json

@@ -0,0 +1,65 @@
+{
+  "arch": {
+    "architectures": [
+      "ScriptableCrammedBERT"
+    ],
+    "attention": {
+      "causal_attention": false,
+      "dropout_prob": 0.1,
+      "num_attention_heads": 12,
+      "qkv_bias": false,
+      "rotary_embedding": false,
+      "seq_op_in_fp32": false,
+      "sequence_op": "torch-softmax",
+      "skip_output_projection": false,
+      "type": "self-attention"
+    },
+    "classification_head": {
+      "classifier_dropout": 0.1,
+      "head_dim": 1024,
+      "include_ff_layer": true,
+      "nonlin": "Tanh",
+      "pooler": "zero_index"
+    },
+    "decoder_bias": false,
+    "embedding": {
+      "dropout_prob": 0.1,
+      "embedding_dim": 768,
+      "max_seq_length": 128,
+      "normalization": true,
+      "pad_token_id": 0,
+      "pos_embedding": "scaled-sinusoidal",
+      "stable_low_precision": false,
+      "vocab_size": 32768
+    },
+    "ffn_layer_frequency": 1,
+    "final_norm": true,
+    "hidden_dropout_prob": 0.1,
+    "hidden_size": 768,
+    "init": {
+      "std": 0.02,
+      "type": "normal"
+    },
+    "intermed_depth": 11,
+    "intermed_size": 1,
+    "intermed_type": "fff",
+    "loss": "cross-entropy",
+    "nonlin": "GELU",
+    "norm": "LayerNorm",
+    "norm_eps": 1e-12,
+    "norm_scheme": "pre",
+    "num_labels": null,
+    "num_transformer_layers": 16,
+    "objective_layout": "MLM",
+    "skip_head_transform": true,
+    "sparse_prediction": 0.25,
+    "tie_weights": true,
+    "use_bias": false
+  },
+  "architectures": [
+    "ScriptableLMForPreTraining"
+  ],
+  "model_type": "crammedBERT",
+  "torch_dtype": "float32",
+  "transformers_version": "4.34.0"
+}

data_budget_hours_24.json

@@ -0,0 +1,12 @@
+{
+    "name": "pile-readymade",
+    "sources": {
+        "hub": {
+            "provider": "hub"
+        }
+    },
+    "hf_location": "JonasGeiping/the_pile_WordPiecex32768_2efdb9d060d1ae95faf952ec1a50f020",
+    "streaming": false,
+    "vocab_size": 32768,
+    "seq_length": 128
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd3070a6cdb36cefb0024fcbb6929548e99cbf193fede44ccc1f73ac8d5a0b5b
+size 755365540

model_config.json

@@ -0,0 +1 @@
+{"architectures": ["ScriptableCrammedBERT"], "num_transformer_layers": 16, "intermed_type": "fff", "hidden_size": 768, "intermed_size": 1, "intermed_depth": 11, "hidden_dropout_prob": 0.1, "norm": "LayerNorm", "norm_eps": 1e-12, "norm_scheme": "pre", "nonlin": "GELU", "tie_weights": true, "decoder_bias": false, "sparse_prediction": 0.25, "loss": "cross-entropy", "objective_layout": "MLM", "embedding": {"vocab_size": null, "pos_embedding": "scaled-sinusoidal", "dropout_prob": 0.1, "pad_token_id": 0, "max_seq_length": 128, "embedding_dim": 768, "normalization": true, "stable_low_precision": false}, "attention": {"type": "self-attention", "causal_attention": false, "num_attention_heads": 12, "dropout_prob": 0.1, "skip_output_projection": false, "qkv_bias": false, "rotary_embedding": false, "seq_op_in_fp32": false, "sequence_op": "torch-softmax"}, "init": {"type": "normal", "std": 0.02}, "ffn_layer_frequency": 1, "skip_head_transform": true, "use_bias": false, "final_norm": true, "num_labels": null, "classification_head": {"pooler": "avg", "include_ff_layer": true, "head_dim": 1024, "nonlin": "Tanh", "classifier_dropout": 0.1}}

special_tokens_map.json

@@ -0,0 +1,18 @@
+{
+  "additional_special_tokens": [
+    "<unk>",
+    "<cls>",
+    "<s>",
+    "</s>",
+    "<mask>",
+    "<sep>",
+    "<pad>"
+  ],
+  "bos_token": "<s>",
+  "cls_token": "<cls>",
+  "eos_token": "</s>",
+  "mask_token": "<mask>",
+  "pad_token": "<pad>",
+  "sep_token": "<sep>",
+  "unk_token": "<unk>"
+}

tokenizer_config.json

@@ -0,0 +1,79 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<cls>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<mask>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<sep>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<unk>",
+    "<cls>",
+    "<s>",
+    "</s>",
+    "<mask>",
+    "<sep>",
+    "<pad>"
+  ],
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<cls>",
+  "eos_token": "</s>",
+  "mask_token": "<mask>",
+  "model_max_length": 128,
+  "pad_token": "<pad>",
+  "sep_token": "<sep>",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "unk_token": "<unk>"
+}

train_budget_hours_24.json

@@ -0,0 +1,41 @@
+{
+    "optim": {
+        "type": "AdamW",
+        "lr": 0.0005,
+        "betas": [
+            0.9,
+            0.98
+        ],
+        "eps": 1e-12,
+        "weight_decay": 0.01,
+        "amsgrad": false,
+        "fused": null
+    },
+    "optim_mod": {
+        "name": "none"
+    },
+    "name": "bert-base",
+    "limited_decay_keys": [
+        "bias",
+        "LayerNorm.bias",
+        "LayerNorm.weight",
+        "norm"
+    ],
+    "warmup_steps": 30000,
+    "cooldown_steps": 0,
+    "steps": 900000,
+    "scheduler": "budget-cosine-decay",
+    "batch_size": 1536,
+    "batch_size_ramp": 0,
+    "gradient_clipping": null,
+    "pretrain_in_train_mode": false,
+    "objective": {
+        "name": "masked-lm",
+        "mlm_probability": 0.15,
+        "use_80_20_rule": true,
+        "disable_mlm": false,
+        "token_drop": 0.0
+    },
+    "reverse_dataset_order": false,
+    "budget": 24
+}