Skip to content

Use fitted linear model for constraint 4. #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
curiousg23 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Use fitted linear model for constraint 4. #2

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 10 additions & 21 deletions exp/dynamic_batch_size/scheuler/dynamic_scheduler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,30 +220,19 @@ def solve_ilp(N, B, d, batch_runtimes):
# This is equivalent to: s[j_prime] <= M * s[j] where M is a large constant
model.addConstr(s[j_prime] <= B * s[j], name=f"sequential_{j}_{j_prime}")

# Constraint 4: e_j = sum_{k=1}^j f(s_k), forall j
# Model constraint 4 with fitted linear model.
for j in range(N):
# Create indicator variables for each possible batch size value for position j
indicators = {}
for batch_size in range(B + 1):
indicators[batch_size] = model.addVar(vtype=GRB.BINARY, name=f"ind_{j}_{batch_size}")

# Constraint: sum of indicators for position j must equal 1
model.addConstr(gp.quicksum(indicators[batch_size] for batch_size in range(B + 1)) == 1,
name=f"ind_sum_{j}")

# Constraint: s[j] = sum of batch_size * indicator[j, batch_size]
model.addConstr(s[j] == gp.quicksum(batch_size * indicators[batch_size] for batch_size in range(B + 1)),
name=f"s_value_{j}")

# e_j = sum_{k=1}^j f(s_k) = cumulative sum of batch runtimes
if j == 0:
# e[0] = f(s[0])
model.addConstr(e[j] == gp.quicksum(batch_runtimes.get(batch_size,0) * indicators[batch_size] for batch_size in range(B + 1)),
name=f"e_{j}")
model.addConstr(
e[j] == 21.576762387982544 * s[j] + 14.724000781249913
)
else:
# e[j] = e[j-1] + f(s[j])
model.addConstr(e[j] == e[j-1] + gp.quicksum(batch_runtimes.get(batch_size, 0) * indicators[batch_size] for batch_size in range(B + 1)),
name=f"e_{j}")
model.addConstr(
e[j]
== e[j - 1]
+ (21.576762387982544 * s[j] + 14.724000781249913),
name=f"e_{j}",
)

# Constraint 5: t_i = sum_j x_{i,j} e_j, forall i (if scheduled), otherwise t_i = large_value
for i in range(N):
Expand Down