saisha09 commited on
Commit
f76faba
·
1 Parent(s): 545222f

deleting cost_benefit.py

Browse files
Files changed (1) hide show
  1. src/cost_benefit.py +0 -48
src/cost_benefit.py DELETED
@@ -1,48 +0,0 @@
1
- import argparse
2
- import subprocess
3
- import time
4
- import requests
5
-
6
- def detect_available_budget(runtime_env: str) -> int:
7
- """
8
- Return an approximate VRAM‑based budget (MB) when running locally,
9
- else default to 100.
10
- """
11
- import torch
12
- if "local" in runtime_env and torch.cuda.is_available():
13
- total_vram_mb = torch.cuda.get_device_properties(0).total_memory // (1024 ** 2)
14
- return min(total_vram_mb, 100)
15
- return 100
16
-
17
- def get_best_model(runtime_env: str, *, use_local_only: bool = False, use_api_only: bool = False) -> dict:
18
- """
19
- Pick the fastest model that fits in the detected budget while
20
- respecting the locality filters.
21
- """
22
- static_costs = {
23
- "llama3.2": {"size": 20, "token_cost": 0.0001, "tokens_sec": 30, "type": "local"},
24
- "mistral": {"size": 40, "token_cost": 0.0002, "tokens_sec": 50, "type": "local"},
25
- "gemini-2.0-flash": {"size": 60, "token_cost": 0.0005, "tokens_sec": 60, "type": "api"},
26
- "gemini-2.5-pro-preview-03-25": {"size": 80, "token_cost": 0.002 , "tokens_sec": 45, "type": "api"},
27
- }
28
-
29
- budget = detect_available_budget(runtime_env)
30
- best_model, best_speed = None, -1
31
-
32
- for model, info in static_costs.items():
33
- if info["size"] > budget:
34
- continue
35
- if use_local_only and info["type"] != "local":
36
- continue
37
- if use_api_only and info["type"] != "api":
38
- continue
39
- if info["tokens_sec"] > best_speed:
40
- best_model, best_speed = model, info["tokens_sec"]
41
-
42
- chosen = best_model or "llama3.2" # sensible default
43
- return {
44
- "model": chosen,
45
- "token_cost": static_costs[chosen]["token_cost"],
46
- "tokens_sec": static_costs[chosen]["tokens_sec"],
47
- "note": None if best_model else "Defaulted because no model met the constraints",
48
- }