fitst commit

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "THUDM/glm-4-9b-chat-1m",
+  "_name_or_path": "THUDM/glm4-9b-chat-1m",
   "model_type": "chatglm",
   "architectures": [
     "ChatGLMModel"
@@ -23,20 +23,27 @@
   "hidden_dropout": 0.0,
   "hidden_size": 4096,
   "kv_channels": 128,
-  "layernorm_epsilon": 0.00000015625,
+  "layernorm_epsilon": 1.5625e-07,
   "multi_query_attention": true,
   "multi_query_group_num": 2,
   "num_attention_heads": 32,
+  "num_hidden_layers": 40,
   "num_layers": 40,
+  "rope_ratio": 500,
   "original_rope": true,
   "padded_vocab_size": 151552,
   "post_layer_norm": true,
   "rmsnorm": true,
-  "seq_length": 8192,
+  "seq_length": 131072,
   "use_cache": true,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.30.2",
   "tie_word_embeddings": false,
-  "eos_token_id": [151329, 151336, 151338],
+  "eos_token_id": [
+    151329,
+    151336,
+    151338
+  ],
   "pad_token_id": 151329
-}
+}
+