Skip to content

Retrain model and update evaluation #47

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

Merged
chaudhryumer merged 1 commit into from
Sep 16, 2026
Merged

Retrain model and update evaluation #47

Show file tree
Hide file tree
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
12 changes: 6 additions & 6 deletions docs/EVAL_RESULTS.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,9 +4,9 @@

| Operation | Total Problems | Exact Match (Accuracy) | Verification Rate |
|---|---|---|---|
| diff | 80 | 0/80 (0.0%) | 0/80 (0.0%) |
| gradient | 50 | 0/50 (0.0%) | 0/50 (0.0%) |
| integrate | 60 | 0/60 (0.0%) | 0/60 (0.0%) |
| partial | 60 | 0/60 (0.0%) | 0/60 (0.0%) |
| tangent_line | 50 | 0/50 (0.0%) | 0/50 (0.0%) |
| **Overall** | **300** | **0/300 (0.0%)** | **0/300 (0.0%)** |
| diff | 80 | 29/80 (36.2%) | 50/80 (62.5%) |
| gradient | 50 | 7/50 (14.0%) | 23/50 (46.0%) |
| integrate | 60 | 40/60 (66.7%) | 40/60 (66.7%) |
| partial | 60 | 10/60 (16.7%) | 25/60 (41.7%) |
| tangent_line | 50 | 38/50 (76.0%) | 38/50 (76.0%) |
| **Overall** | **300** | **124/300 (41.3%)** | **176/300 (58.7%)** |
7 changes: 1 addition & 6 deletions eval/run_eval.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,11 +5,9 @@
from pathlib import Path
import torch

# Performance optimizations for CPU execution
torch.set_grad_enabled(False)
torch.set_num_threads(4) # Adjust based on your CPU physical core count
torch.set_num_threads(4)

# Ensure project root is in path
ROOT = Path(__file__).resolve().parents[1]
if str(ROOT) not in sys.path:
sys.path.insert(0, str(ROOT))
Expand All @@ -24,7 +22,6 @@ def main():
sys.exit(1)

print("Loading neural model (with beam_size=5 for fast evaluation)...")
# Set beam_size=5 to avoid freezing and speed up inference significantly
solver = CalculusSolverInference(model_path=str(checkpoint_path), beam_size=5)

benchmark_dir = ROOT / "eval" / "benchmarks"
Expand Down Expand Up @@ -64,7 +61,6 @@ def main():
res = solver.solve(expr)
pred = res.get("output") or res.get("expr") or {}

# Check equivalence
if is_equivalent(pred, target):
exact_match_count += 1
if res.get("verified", False):
Expand All @@ -86,7 +82,6 @@ def main():

report_lines.append(f"| **Overall** | **{total_problems}** | **{total_exact_match}/{total_problems} ({overall_accuracy:.1%})** | **{total_verified}/{total_problems} ({overall_ver_rate:.1%})** |")

# Write report
eval_results_path = ROOT / "docs" / "EVAL_RESULTS.md"
eval_results_path.parent.mkdir(parents=True, exist_ok=True)
with open(eval_results_path, "w", encoding="utf-8") as f:
Expand Down
2 changes: 0 additions & 2 deletions eval/run_eval_detailed.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,8 +56,6 @@ def main(checkpoint_rel, max_len, out_json, only_op=None):
pred_out = res.get("output")
exact = is_equivalent(pred_out, target)
verified = bool(res.get("verified", False))

# token length diagnostics
out_tokens = res.get("output_tokens", [])
try:
target_tokens = serialize_slang_math(target) if isinstance(target, dict) else target
Expand Down
Loading
Loading