Upload configuration_picolm_v3.py with huggingface_hub

configuration_picolm_v3.py

@@ -2,6 +2,12 @@ from transformers import PretrainedConfig
 
 class PicoLMV3Config(PretrainedConfig):
     model_type = "picolm_v3"
+    attribute_map = {
+        "hidden_size": "dim",
+        "num_hidden_layers": "n_layers",
+        "num_attention_heads": "n_heads",
+        "num_key_value_heads": "n_kv_heads",
+    }
 
     def __init__(
         self,
@@ -22,25 +28,32 @@ class PicoLMV3Config(PretrainedConfig):
         eos_token_id: int = 2,
         pad_token_id: int = 3,
         tie_word_embeddings: bool = True,
+        use_cache: bool = False,
         **kwargs,
     ):
         self.vocab_size = vocab_size
         self.emb_dim = emb_dim
         self.dim = dim
+        self.hidden_size = dim
         self.n_layers = n_layers
+        self.num_hidden_layers = n_layers
         self.layer_repeat = layer_repeat
         self.n_heads = n_heads
+        self.num_attention_heads = n_heads
         self.n_kv_heads = n_kv_heads
+        self.num_key_value_heads = n_kv_heads
         self.head_dim = head_dim
         self.intermediate_dim = intermediate_dim
         self.max_seq_len = max_seq_len
         self.orig_seq_len = orig_seq_len
         self.norm_eps = norm_eps
         self.rope_theta = rope_theta
+        self.use_cache = use_cache
         super().__init__(
             bos_token_id=bos_token_id,
             eos_token_id=eos_token_id,
             pad_token_id=pad_token_id,
             tie_word_embeddings=tie_word_embeddings,
+            use_cache=use_cache,
             **kwargs,
         )