MCPcopy Create free account
hub / github.com/microsoft/JARVIS / formulate

Function formulate

taskbench/format_data.py:60–118  ·  view source on GitHub ↗
(data_dir, dependency_type)

Source from the content-addressed store, hash-verified

58@click.option('--data_dir', default='data_huggingface', help='Path to the data directory')
59@click.option('--dependency_type', default='resource')
60def formulate(data_dir, dependency_type):
61 rf = open(f"{data_dir}/data_raw.json", "r")
62 wf_format = open(f"{data_dir}/data.json", "w")
63 wf_error = open(f"{data_dir}/data_error.json", "w")
64 wf_ur = open(f"{data_dir}/user_requests.json", "w")
65
66 all = 0
67 format = 0
68 for line in rf:
69 all += 1
70 data = json.loads(line)
71 method = data["method"]
72 n_tools = data["number_of_tools"]
73 seed = data["seed"]
74 _id = data["id"]
75 sampled_nodes = data["sampled_nodes"]
76 sampled_links = data["sampled_links"]
77 for sampled_node in sampled_nodes:
78 sampled_node["task"] = sampled_node["id"]
79 sampled_node.pop("id")
80 if dependency_type == "resource":
81 sampled_node["task"] = sampled_node["task"].replace("_", " ")
82 if "input" in sampled_node:
83 sampled_node["input-type"] = sampled_node["input"]
84 sampled_node.pop("input")
85 sampled_node["output-type"] = sampled_node["output"]
86 sampled_node.pop("output")
87 else:
88 sampled_node["arguments"] = sampled_node["parameters"]
89 sampled_node.pop("parameters")
90 user_request, task_steps, links, nodes, check_by_teacher = formulate_sample(data, dependency_type)
91 if user_request is None:
92 wf_error.write(line)
93 continue
94 format += 1
95 result = {
96 "id": _id,
97 "seed": seed,
98 "type": method,
99 "n_tools": n_tools,
100 "sampled_nodes": sampled_nodes,
101 "sampled_links": sampled_links,
102 "user_request": user_request,
103 "task_steps": task_steps,
104 "task_nodes": nodes,
105 "task_links": links,
106 "check_by_teacher": check_by_teacher,
107 }
108 wf_format.write(json.dumps(result)+"\n")
109 ur_result = {
110 "id": _id,
111 "user_request": user_request,
112 }
113 wf_ur.write(json.dumps(ur_result)+"\n")
114 wf_format.close()
115 wf_error.close()
116 wf_ur.close()
117 rf.close()

Callers 1

format_data.pyFile · 0.85

Calls 1

formulate_sampleFunction · 0.85

Tested by

no test coverage detected