MCPcopy Create free account
hub / github.com/RosettaCommons/RFdiffusion / __init__

Method __init__

rfdiffusion/Track_module.py:202–223  ·  view source on GitHub ↗
(self, d_msa=256, d_pair=128, d_state=16, 
            SE3_param={'l0_in_features':32, 'l0_out_features':16, 'num_edge_features':32}, p_drop=0.1)

Source from the content-addressed store, hash-verified

200
201class Str2Str(nn.Module):
202 def __init__(self, d_msa=256, d_pair=128, d_state=16,
203 SE3_param={'l0_in_features':32, 'l0_out_features':16, 'num_edge_features':32}, p_drop=0.1):
204 super(Str2Str, self).__init__()
205
206 # initial node & pair feature process
207 self.norm_msa = nn.LayerNorm(d_msa)
208 self.norm_pair = nn.LayerNorm(d_pair)
209 self.norm_state = nn.LayerNorm(d_state)
210
211 self.embed_x = nn.Linear(d_msa+d_state, SE3_param['l0_in_features'])
212 self.embed_e1 = nn.Linear(d_pair, SE3_param['num_edge_features'])
213 self.embed_e2 = nn.Linear(SE3_param['num_edge_features']+36+1, SE3_param['num_edge_features'])
214
215 self.norm_node = nn.LayerNorm(SE3_param['l0_in_features'])
216 self.norm_edge1 = nn.LayerNorm(SE3_param['num_edge_features'])
217 self.norm_edge2 = nn.LayerNorm(SE3_param['num_edge_features'])
218
219 self.se3 = SE3TransformerWrapper(**SE3_param)
220 self.sc_predictor = SCPred(d_msa=d_msa, d_state=SE3_param['l0_out_features'],
221 p_drop=p_drop)
222
223 self.reset_parameter()
224
225 def reset_parameter(self):
226 # initialize weights to normal distribution

Callers

nothing calls this directly

Calls 4

reset_parameterMethod · 0.95
SCPredClass · 0.85
__init__Method · 0.45

Tested by

no test coverage detected