add safetensors export

README.md

@@ -0,0 +1,10 @@
+# Apollo
+
+Decoder-only Transformer (RoPE, RMSNorm, SwiGLU, tied embeddings) that generates monophonic melodies as symbolic tokens (BAR / POS / PITCH / DUR) conditioned on a header (tempo, time signature, key, bars, density, register, ...).
+
+- `model.safetensors`: fp16 weights. `lm_head` is tied to `emb.weight`, so `head.weight` is not stored; build the model, load with `strict=False`, and the tie fills it.
+- `config.json`: architecture (`cfg`), context length, training step and validation loss.
+- `vocab.json`: token list in id order.
+- `best/` and `last/`: training checkpoints (`.pt`), used for resuming; load only from sources you trust.
+
+Trained on melodies extracted from the Lakh MIDI dataset. Check the licensing of the source data before any commercial use.

config.json

@@ -1 +1,17 @@
-{"cfg": {"vocab": 352, "max_len": 1024, "dropout": 0.1, "n_layer": 8, "d_model": 512, "n_head": 8, "d_ff": 1536}, "seq_len": 1024}
+{
+ "cfg": {
+  "vocab": 352,
+  "max_len": 1024,
+  "dropout": 0.1,
+  "n_layer": 8,
+  "d_model": 512,
+  "n_head": 8,
+  "d_ff": 1536
+ },
+ "seq_len": 1024,
+ "weights": "model.safetensors",
+ "dtype": "float16",
+ "tied_embeddings": true,
+ "step": 10000,
+ "val_loss": 0.923981524555829
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d9c05e1cdc348632f9c80cc5b042345336abb71606f0f0adf8154e3ff0de921
+size 54908976