Upload best checkpoint from config

config.json

@@ -0,0 +1,36 @@
+{
+  "architectures": [
+    "SharedSpaceEncoderForMaskedLM"
+  ],
+  "attention_backend": "sdpa",
+  "attention_bias": false,
+  "attention_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "ffn_decompose": false,
+  "ffn_rank": null,
+  "head_dim": 32,
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 256,
+  "initializer_range": 0.02,
+  "intermediate_size": 1024,
+  "kv_latent_dim": 32,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 128,
+  "model_type": "shared_subspace_encoder",
+  "num_attention_heads": 8,
+  "num_dense_layers": 0,
+  "num_hidden_layers": 6,
+  "o_latent_dim": 64,
+  "output_subspace": true,
+  "pad_token_id": 0,
+  "q_latent_dim": 64,
+  "rms_norm_eps": 1e-06,
+  "rope_dims": 32,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "transformers_version": "4.56.1",
+  "vocab_rank": 128,
+  "vocab_size": 30522,
+  "vocab_subspace": false
+}

full_config.json

@@ -0,0 +1,78 @@
+{
+  "shorthand": "rd.32 - 6.mla-o.64.32.64 - mlp.1024 - model.256.lyr.6 - ah.8.32",
+  "notes": "Current top-performing MLA-o configuration",
+  "model": {
+    "hidden_size": 256,
+    "num_hidden_layers": 6,
+    "intermediate_size": 1024,
+    "hidden_dropout_prob": 0.1,
+    "attention_dropout_prob": 0.1,
+    "classifier_dropout": null,
+    "initializer_range": 0.02,
+    "layer_norm_eps": 1e-12,
+    "rms_norm_eps": 1e-06,
+    "vocab_size": 30522,
+    "rope_theta": 10000.0,
+    "rope_scaling": null,
+    "max_position_embeddings": 128,
+    "num_dense_layers": 0,
+    "q_latent_dim": 64,
+    "kv_latent_dim": 32,
+    "num_attention_heads": 8,
+    "head_dim": 32,
+    "rope_dims": 32,
+    "attention_bias": false,
+    "output_subspace": true,
+    "o_latent_dim": 64,
+    "attention_backend": "sdpa",
+    "ffn_decompose": false,
+    "ffn_rank": null,
+    "vocab_subspace": false,
+    "vocab_rank": 128
+  },
+  "pre_train": {
+    "output_dir": "checkpoints/mla-o_baseline",
+    "seed": 42,
+    "train_batch_size": 256,
+    "learning_rate": 0.0005,
+    "num_train_steps": 50000,
+    "eval_steps": 2000,
+    "weight_decay": 0.01,
+    "mlm_probability": 0.15,
+    "dataset_name": "wikitext",
+    "dataset_config": "wikitext-103-raw-v1",
+    "max_seq_length": 128,
+    "eval_batch_size": 64,
+    "fp16": false,
+    "bf16": true,
+    "torch_compile": true,
+    "torch_compile_backend": "inductor",
+    "torch_compile_mode": "default",
+    "best_checkpoint": "checkpoints/mla-o_baseline/checkpoint-50000",
+    "run_name": "mla-o_baseline",
+    "run_id": "koko",
+    "push_to_hub": true,
+    "hub_model_id": "kokolamba/encoder-mla-o-baseline-perf",
+    "hub_strategy": "end"
+  },
+  "fine_tune": {
+    "task": "sst2",
+    "batch_size": 16,
+    "lr": 2e-05,
+    "epochs": 3,
+    "seed": 42,
+    "max_length": 128,
+    "weight_decay": 0,
+    "warmup_ratio": 0,
+    "fp16": false,
+    "bf16": true,
+    "torch_compile": true,
+    "torch_compile_backend": "inductor",
+    "torch_compile_mode": "default",
+    "run_name": "configperf - mla-o_baseline - pt_id.koko - sst2",
+    "model_path": "checkpoints/mla-o_baseline/checkpoint-50000",
+    "tuned_from_id": "koko",
+    "run_id": "ruq0b2kl",
+    "run_url": "https://wandb.ai/abdulhakeemadefioye-personal/subspace-encoder-sst2/runs/ruq0b2kl"
+  }
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9315e89146c537272be004ba63444d8a3f348cea8bf01b125eca07d95cafc1e9
+size 52371776

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ff3a589a9664b6a7d2d111dae847708f0050e50f5a0425406eb8f1417f9dbab
+size 104799354

rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b83432253e3441b44b0af69b1d488725f0f394ab1e7b933b387b9e291b3e42de
+size 14244

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c41b24d556b463967f19cb66c03824bf85c07c05fbc80ebaac7fd319789f6b1
+size 1064

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,56 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "101": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "102": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "103": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "extra_special_tokens": {},
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b7ec0750c12d4bc7b33106f57f11ceb74562b0c2a3bae33fde49c19137ba6c6
+size 5560