python==3.7.0

paddlepaddle-gpu==2.5.1

【报错】

File "/root/autodl-tmp/helixfold-single/alphafold_paddle/model/modules.py"， line 485，in forward q_data， m_data，self.query_w， self.key_w， self.value_w， self.qkv_w， nonbatched_bias， bias, self.gating_w，self.gating_b,self.output_w， self.output_b，' has_gating', self.config.gating, ' merge  qkv', self.merge_qkv)
ValueError: too many values to unpack (expected 8)

【解决方法】

定位到源代码

 _, _, _, _, _, _, _, output = _C_ops.fused_gate_attention(q_data, m_data, self.query_w, self.key_w, self.value_w, self.qkv_w, nonbatched_bias, bias, self.gating_w, self.gating_b,self.output_w, self.output_b, 'has_gating', self.config.gating, 'merge_qkv', self.merge_qkv)

由于_C_ops.fused_gate_attention应该返回9个元素，此处只有8个元素，个数不匹配，故报错，修改上述代码如下即可：

 _, _, _, _, _, _, _, _, output = _C_ops.fused_gate_attention(q_data, m_data, self.query_w, self.key_w, self.value_w, self.qkv_w, nonbatched_bias, bias, self.gating_w, self.gating_b,self.output_w, self.output_b, 'has_gating', self.config.gating, 'merge_qkv', self.merge_qkv)