Spaces:
Running
Running
Commit
·
7e51da8
1
Parent(s):
7d69384
Updated costs
Browse files
src/manager/budget_manager.py
CHANGED
@@ -46,7 +46,7 @@ class BudgetManager():
|
|
46 |
ram_mem = mem.total/ 1024 ** 3
|
47 |
print("No GPU detected. Using CPU.")
|
48 |
print(f"Total RAM: {ram_mem:.2f} GB")
|
49 |
-
total_mem = gpu_mem + ram_mem
|
50 |
return round((total_mem / 16) * 100)
|
51 |
|
52 |
def get_total_resource_budget(self):
|
|
|
46 |
ram_mem = mem.total/ 1024 ** 3
|
47 |
print("No GPU detected. Using CPU.")
|
48 |
print(f"Total RAM: {ram_mem:.2f} GB")
|
49 |
+
total_mem = gpu_mem + ram_mem
|
50 |
return round((total_mem / 16) * 100)
|
51 |
|
52 |
def get_total_resource_budget(self):
|
src/tools/default_tools/agent_cost_manager.py
CHANGED
@@ -17,18 +17,18 @@ class AgentCostManager():
|
|
17 |
costs = {
|
18 |
"llama3.2": {
|
19 |
"description": "Avg Accuracy: 49.75%, Latency 0.9s, 63.4% on multi-task understanding, 40.8% on rewriting, 78.6% on reasoning.",
|
20 |
-
"create_resource_cost":
|
21 |
-
"invoke_resource_cost":
|
22 |
},
|
23 |
"mistral": {
|
24 |
"description": "Avg Accuracy: 51.3%, Latency 9.7s, 51% on LegalBench, 60.1% on multi-task understanding, 69.9% on TriviaQA, 67.9% on reasoning",
|
25 |
-
"create_resource_cost":
|
26 |
-
"invoke_resource_cost":
|
27 |
},
|
28 |
"deepseek-r1": {
|
29 |
"description": "Avg Accuracy: 77.3%, Latency: 120s, 69.9% on LegalBench, 71.1% on multi-task understanding, 92.2% on Math",
|
30 |
-
"create_resource_cost":
|
31 |
-
"invoke_resource_cost":
|
32 |
},
|
33 |
"gemini-2.5-flash-preview-05-20": {
|
34 |
"description": "Avg Accuracy: 75.8%, 82.8% on LegalBench, 81.6% on multi-task understanding, 91.6% on Math",
|
|
|
17 |
costs = {
|
18 |
"llama3.2": {
|
19 |
"description": "Avg Accuracy: 49.75%, Latency 0.9s, 63.4% on multi-task understanding, 40.8% on rewriting, 78.6% on reasoning.",
|
20 |
+
"create_resource_cost": 0,
|
21 |
+
"invoke_resource_cost": 40,
|
22 |
},
|
23 |
"mistral": {
|
24 |
"description": "Avg Accuracy: 51.3%, Latency 9.7s, 51% on LegalBench, 60.1% on multi-task understanding, 69.9% on TriviaQA, 67.9% on reasoning",
|
25 |
+
"create_resource_cost": 0,
|
26 |
+
"invoke_resource_cost": 100,
|
27 |
},
|
28 |
"deepseek-r1": {
|
29 |
"description": "Avg Accuracy: 77.3%, Latency: 120s, 69.9% on LegalBench, 71.1% on multi-task understanding, 92.2% on Math",
|
30 |
+
"create_resource_cost": 0,
|
31 |
+
"invoke_resource_cost": 150,
|
32 |
},
|
33 |
"gemini-2.5-flash-preview-05-20": {
|
34 |
"description": "Avg Accuracy: 75.8%, 82.8% on LegalBench, 81.6% on multi-task understanding, 91.6% on Math",
|