Add config.json to each of the models

llama-3.2-1B/QLoRA/config.json

@@ -0,0 +1,19 @@
+{
+    "dim": 2048,
+    "n_layers": 16,
+    "n_heads": 32,
+    "n_kv_heads": 8,
+    "vocab_size": 128256,
+    "ffn_dim_multiplier": 1.5,
+    "multiple_of": 256,
+    "norm_eps": 1e-05,
+    "rope_theta": 500000.0,
+    "use_scaled_rope": true,
+    "quantization_args": {
+        "group_size": 32
+    },
+    "lora_args": {
+        "rank": 16,
+        "scale": 2.0
+    }
+}

llama-3.2-1B/original/config.json

@@ -0,0 +1,12 @@
+{
+    "dim": 3072,
+    "n_layers": 28,
+    "n_heads": 24,
+    "n_kv_heads": 8,
+    "vocab_size": 128256,
+    "ffn_dim_multiplier": 1.0,
+    "multiple_of": 256,
+    "norm_eps": 1e-05,
+    "rope_theta": 500000.0,
+    "use_scaled_rope": true
+}

llama-3.2-1B/spinquant/config.json

@@ -0,0 +1,15 @@
+{
+    "dim": 2048,
+    "n_layers": 16,
+    "n_heads": 32,
+    "n_kv_heads": 8,
+    "vocab_size": 128256,
+    "ffn_dim_multiplier": 1.5,
+    "multiple_of": 256,
+    "norm_eps": 1e-05,
+    "rope_theta": 500000.0,
+    "use_scaled_rope": true,
+    "quantization_args": {
+        "group_size": 32
+    }
+}

llama-3.2-3B/QLoRA/config.json

@@ -0,0 +1,19 @@
+{
+    "dim": 3072,
+    "n_layers": 28,
+    "n_heads": 24,
+    "n_kv_heads": 8,
+    "vocab_size": 128256,
+    "ffn_dim_multiplier": 1.0,
+    "multiple_of": 256,
+    "norm_eps": 1e-05,
+    "rope_theta": 500000.0,
+    "use_scaled_rope": true,
+    "quantization_args": {
+        "group_size": 32
+    },
+    "lora_args": {
+        "rank": 16,
+        "scale": 2.0
+    }
+}

llama-3.2-3B/original/config.json

@@ -0,0 +1,12 @@
+{
+    "dim": 3072,
+    "n_layers": 28,
+    "n_heads": 24,
+    "n_kv_heads": 8,
+    "vocab_size": 128256,
+    "ffn_dim_multiplier": 1.0,
+    "multiple_of": 256,
+    "norm_eps": 1e-05,
+    "rope_theta": 500000.0,
+    "use_scaled_rope": true
+}

llama-3.2-3B/spinquant/config.json

@@ -0,0 +1,15 @@
+{
+    "dim": 3072,
+    "ffn_dim_multiplier": 1.0,
+    "multiple_of": 256,
+    "n_heads": 24,
+    "n_kv_heads": 8,
+    "n_layers": 28,
+    "norm_eps": 1e-05,
+    "rope_theta": 500000.0,
+    "use_scaled_rope": true,
+    "vocab_size": 128256,
+    "quantization_args": {
+        "group_size": 32
+    }
+}