@@ -506,17 +506,18 @@ class UITARSAgent:
|
||||
if last_action_after_obs is not None and self.infer_mode == "double_image":
|
||||
self.history_images.append(last_action_after_obs["screenshot"])
|
||||
|
||||
self.history_images.append(obs["screenshot"])
|
||||
|
||||
if self.observation_type in ["screenshot", "screenshot_a11y_tree"]:
|
||||
base64_image = obs["screenshot"]
|
||||
self.history_images.append(obs["screenshot"])
|
||||
|
||||
if self.observation_type in ["screenshot", "screenshot_a11y_tree", "a11y_tree"]:
|
||||
base64_image = obs["screenshot"] if self.observation_type in ["screenshot", "screenshot_a11y_tree"] else None
|
||||
try:
|
||||
linearized_accessibility_tree = (
|
||||
linearize_accessibility_tree(
|
||||
accessibility_tree=obs["accessibility_tree"],
|
||||
platform=self.platform,
|
||||
)
|
||||
if self.observation_type == "screenshot_a11y_tree"
|
||||
if self.observation_type in ["screenshot_a11y_tree", "a11y_tree"]
|
||||
else None
|
||||
)
|
||||
except:
|
||||
@@ -535,7 +536,14 @@ class UITARSAgent:
|
||||
"accessibility_tree": linearized_accessibility_tree,
|
||||
}
|
||||
)
|
||||
else:
|
||||
elif self.observation_type == "a11y_tree":
|
||||
self.observations.append(
|
||||
{
|
||||
"screenshot": None,
|
||||
"accessibility_tree": linearized_accessibility_tree,
|
||||
}
|
||||
)
|
||||
else: # screenshot
|
||||
self.observations.append(
|
||||
{"screenshot": base64_image, "accessibility_tree": None}
|
||||
)
|
||||
@@ -760,4 +768,4 @@ class UITARSAgent:
|
||||
self.actions = []
|
||||
self.observations = []
|
||||
self.history_images = []
|
||||
self.history_responses = []
|
||||
self.history_responses = []
|
||||
|
||||
Reference in New Issue
Block a user