added config.json, todo: work on tokenizer re-implementation

config.json

@@ -0,0 +1,40 @@
+
+{
+  "dataset": {
+    "remote_path": "flpelerin/tinystories-10k",
+    "batch_size": 4,
+    "seq_length": 512
+  },
+
+  "tokenizer": {
+    "max_length": 32000
+  },
+
+  "wandb": {
+    "entity": "florianpelerin110304",
+    "project": "mamba-trainer-test",
+    "name": "",
+    "api_key": "860f8753998c6e6dc356914de07e8855aa2f9642"
+  },
+
+  "model": {
+    "params": {
+      "d_model": 768,
+      "n_layer": 24,
+
+      "tie_embeddings": "false"
+    }
+  },
+
+  "trainer": {
+    "learning_rate": 1e-3,
+    "num_epochs": 4,
+
+    "inference": {
+      "frequency": 0,
+      "n_predict": 256,
+      "seed_text": "One day, a little girl"
+    }
+  }
+}
+