Model: Fix max seq len handling

Previously, the max sequence length was overriden by the user's
config and never took the model's config.json into account.

Now, set the default to 4096, but include config.prepare when
selecting the max sequence length. The yaml and API request
now serve as overrides rather than parameters.

Signed-off-by: kingbri <bdashore3@proton.me>

model.py

@@ -79,13 +79,21 @@ class ModelContainer:
 
         self.config = ExLlamaV2Config()
         self.config.model_dir = str(model_directory.resolve())
+
+        # Make the max seq len 4096 before preparing the config
+        # This is a better default than 2038
+        self.config.max_seq_len = 4096
         self.config.prepare()
 
+        # Then override the max_seq_len if present
+        override_max_seq_len = kwargs.get("max_seq_len")
+        if override_max_seq_len:
+            self.config.max_seq_len = kwargs.get("max_seq_len")
+
         # Grab the base model's sequence length before overrides for rope calculations
         base_seq_len = self.config.max_seq_len
 
-        # Then override the max_seq_len if present
-        self.config.max_seq_len = unwrap(kwargs.get("max_seq_len"), 4096)
+        # Set the rope scale
         self.config.scale_pos_emb = unwrap(kwargs.get("rope_scale"), 1.0)
 
         # Automatically calculate rope alpha