Add encoder–decoder checkpoint prefix_suffix_run_20251006_033115

README.md

@@ -0,0 +1,12 @@
+# Encoder–Decoder (Prefix→Suffix) — Names
+
+**Author:** @paph3285
+
+This model is a small Transformer encoder–decoder trained to generate name suffixes conditioned on prefixes (character-level baby-name dataset). It demonstrates Google Drive + Hugging Face integrations.
+
+Files:
+- `model_fix.pt` — weights
+- `config.json` — model hyperparameters + special tokens
+- `alphabet.json` — base vocabulary
+
+License: Apache-2.0

alphabet.json

@@ -0,0 +1,32 @@
+{
+  "alphabet": [
+    " ",
+    "a",
+    "b",
+    "c",
+    "d",
+    "e",
+    "f",
+    "g",
+    "h",
+    "i",
+    "j",
+    "k",
+    "l",
+    "m",
+    "n",
+    "o",
+    "p",
+    "q",
+    "r",
+    "s",
+    "t",
+    "u",
+    "v",
+    "w",
+    "x",
+    "y",
+    "z",
+    "."
+  ]
+}

config.json

@@ -0,0 +1,13 @@
+{
+  "model_type": "ImprovedEncoderDecoder",
+  "n_embd": 128,
+  "n_head": 4,
+  "n_layer": 3,
+  "max_len": 128,
+  "vocab_size": 30,
+  "special_tokens": {
+    "PAD_ID": 28,
+    "BOS_ID": 29,
+    "EOS_ID": 27
+  }
+}

model_fix.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a11bc5d4654c99d7d34a39bfc062a6b38ffe9606eb67674c470f7d30a590223
+size 18342027