From 2d0b95853d306a06ba12c2d913862554291e0d75 Mon Sep 17 00:00:00 2001
From: Christian <christian.geishauser@hhu.de>
Date: Thu, 13 Apr 2023 12:00:41 +0200
Subject: [PATCH] small bugfix when using sgd for mle since sgd has no booked
information and not the full belief state is given in the dataset
---
convlab/policy/mle/loader.py | 3 ++-
convlab/policy/vector/vector_binary.py | 26 +++++++++++++++++---------
2 files changed, 19 insertions(+), 10 deletions(-)
diff --git a/convlab/policy/mle/loader.py b/convlab/policy/mle/loader.py
index 9c10d2a7..5af5ddf3 100755
--- a/convlab/policy/mle/loader.py
+++ b/convlab/policy/mle/loader.py
@@ -105,7 +105,8 @@ class PolicyDataVectorizer:
state['terminated'] = data_point['terminated']
if self.dst is not None and state['terminated']:
self.dst.init_session()
- state['booked'] = data_point['booked']
+ if "booked" in data_point:
+ state['booked'] = data_point['booked']
dialogue_act = flatten_acts(data_point['dialogue_acts'])
vectorized_state, mask = self.vector.state_vectorize(state)
diff --git a/convlab/policy/vector/vector_binary.py b/convlab/policy/vector/vector_binary.py
index c6b02a11..e444dec2 100755
--- a/convlab/policy/vector/vector_binary.py
+++ b/convlab/policy/vector/vector_binary.py
@@ -47,8 +47,15 @@ class VectorBinary(VectorBase):
opp_act_vec = self.vectorize_user_act(state)
last_act_vec = self.vectorize_system_act(state)
belief_state, domain_active_dict = self.vectorize_belief_state(state, domain_active_dict)
- book = self.vectorize_booked(state)
- degree, number_entities_dict = self.pointer()
+ if "booked" in state:
+ book = self.vectorize_booked(state)
+ else:
+ book = []
+ if self.db is not None:
+ degree, number_entities_dict = self.pointer()
+ else:
+ degree = []
+ number_entities_dict = {}
final = 1. if state['terminated'] else 0.
state_vec = np.r_[opp_act_vec, last_act_vec,
@@ -82,13 +89,14 @@ class VectorBinary(VectorBase):
belief_state = np.zeros(self.belief_state_dim)
i = 0
for domain in self.belief_domains:
- for slot, value in state['belief_state'][domain].items():
- if value:
- belief_state[i] = 1.
- i += 1
-
- if [slot for slot, value in state['belief_state'][domain].items() if value]:
- domain_active_dict[domain] = True
+ if domain in state['belief_state']:
+ for slot, value in state['belief_state'][domain].items():
+ if value:
+ belief_state[i] = 1.
+ i += 1
+
+ if [slot for slot, value in state['belief_state'][domain].items() if value]:
+ domain_active_dict[domain] = True
return belief_state, domain_active_dict
def vectorize_system_act(self, state):
--
GitLab