Update Readme

README.md

@@ -9,8 +9,7 @@ pipeline_tag: text-generation
 inference: false
 ---
 # BatGPT-15B-sirius
-
-<!-- Provide a quick summary of what the model is/does. -->
+Bidirectional Autoregressive Talker from Generative Pre-trained Transformer
 
 ## 介绍 (Introduction)
 

modeling_batgpt.py

@@ -940,56 +940,6 @@ class BatGPTForCausalLM(BatGPTPreTrainedModel):
             for layer_past in past
         )
 
-
-    def quantize(self, bits: int):
-        try:
-            # from .quantizer import QLinear
-            from quantizer import QLinear
-        except ImportError:
-            raise ImportError(
-                f"Needs QLinear to run quantize."
-            )
-
-        for layer in self.model.encoder.layers:
-            layer.self_attention.query_proj = QLinear(
-                bits=bits,
-                weight=layer.self_attention.query_proj.weight,
-                bias = layer.self_attention.query_proj.bias if self.config.qkv_bias else None,
-            )
-            layer.self_attention.key_proj = QLinear(
-                bits=bits,
-                weight=layer.self_attention.key_proj.weight,
-                bias = layer.self_attention.key_proj.bias if self.config.qkv_bias else None,
-            )
-            layer.self_attention.value_proj = QLinear(
-                bits=bits,
-                weight=layer.self_attention.value_proj.weight,
-                bias = layer.self_attention.value_proj.bias if self.config.qkv_bias else None,
-            )
-            layer.self_attention.dense = QLinear(
-                bits=bits,
-                weight=layer.self_attention.dense.weight,
-                bias = None,
-            )
-            layer.mlp.dense_h_to_4h = QLinear(
-                bits=bits,
-                weight=layer.mlp.dense_h_to_4h.weight,
-                bias = None,
-            )
-            layer.mlp.dense_4h_to_h = QLinear(
-                bits=bits,
-                weight=layer.mlp.dense_4h_to_h.weight,
-                bias = None,
-            )
-            if self.config.mlp_activation == "silu":
-                layer.mlp.gate_proj = QLinear(
-                    bits=bits,
-                    weight=layer.mlp.gate_proj.weight,
-                    bias = None,
-                )
-        return self
-
-
     def process_response(self, response):
         response = response.strip()
         return response