Upload model

config.json

@@ -11,7 +11,9 @@
   "ffn_num_input": 32,
   "key_size": 32,
   "model_type": "IQsignal_transformer",
-  "norm_shape": 32,
+  "norm_shape": [
+    32
+  ],
   "num_heads": 4,
   "num_hiddens": 32,
   "num_layers": 2,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f80269a38e99e9ac23a319b1a548a35ecd51ddb66aadc3b05b5ea85a32179498
+oid sha256:f0086aa0b6c4b67c2d818b2038c0232b0c8d86445867f1d73489eb2e5f4bf41d
 size 79108

modeling_IQtransformer.py

@@ -5,6 +5,8 @@ import math
 
 from transformers import PretrainedConfig
 
+# 把transformerConfig和transformerModel都放在一个文件中，避免类别不匹配引起的错误
+
 class transformerConfig(PretrainedConfig):
     model_type = "IQsignal_transformer"
 
@@ -15,7 +17,7 @@ class transformerConfig(PretrainedConfig):
         query_size      : int = 32,
         value_size      : int = 32,
         num_hiddens     : int = 32,
-        norm_shape      : int = 32,
+        norm_shape      : int = [32],
         ffn_num_input   : int = 32,
         ffn_num_hiddens : int = 64,
         num_heads       : int = 4,
@@ -211,7 +213,7 @@ class transformerModel(PreTrainedModel):
         self.Linear = nn.Linear(config.vocab_size, config.vocab_size)
         # self.embedding = nn.Embedding(vocab_size, num_hiddens)      # 将输入vocab_size的维度  转化为  想要的num_hiddens维度
         # self.pos_encoding = d2l.PositionalEncoding(num_hiddens, dropout)
-        self.ln = nn.LayerNorm(config.vocab_size)
+        self.ln = nn.LayerNorm(config.norm_shape)
         self.blks = nn.Sequential()
         for i in range(config.num_layers):
             self.blks.add_module("block" + str(i),