64 per head

config.json

@@ -7,14 +7,14 @@
   "embd_pdrop": 0.0,
   "eos_token_id": 2,
   "hidden_act": "relu",
-  "hidden_size": 256,
+  "hidden_size": 64,
   "initializer_range": 0.02,
   "intermediate_size": 8,
   "max_position_embeddings": 4096,
   "model_type": "phi3",
-  "num_attention_heads": 32,
+  "num_attention_heads": 1,
   "num_hidden_layers": 2,
-  "num_key_value_heads": 32,
+  "num_key_value_heads": 1,
   "original_max_position_embeddings": 4096,
   "pad_token_id": 2,
   "resid_pdrop": 0.0,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de0378f8a2e4432d986b1d7d3285b2891eaf41c6966657443ed265c457075d5f
-size 2204184
+oid sha256:965e9fbb65406e8b504fc8680a04a696b75ac12ecbf327109f959825f7e5aa24
+size 158952