Spaces:
Sleeping
Sleeping
Download make_plots.py from anmol1620/cloud-env-openenv: direct link, hf CLI and curl.
- Browser
- Download file 689 Bytes
-
https://huggingface.co/spaces/anmol1620/cloud-env-openenv/resolve/main/make_plots.py
- Command line
-
hf download hf://spaces/anmol1620/cloud-env-openenv/make_plots.py
-
curl -L -o make_plots.py https://huggingface.co/spaces/anmol1620/cloud-env-openenv/resolve/main/make_plots.py
689 Bytes
| import matplotlib.pyplot as plt | |
| agents = ["Random", "Rule-Based", "Trained LLM"] | |
| scores = [-5.90, -5.81, 31.78] | |
| plt.figure(figsize=(8, 5)) | |
| plt.bar(agents, scores) | |
| plt.xlabel("Agent") | |
| plt.ylabel("Average Reward") | |
| plt.title("Agent Performance Comparison") | |
| plt.grid(axis="y") | |
| plt.savefig("agent_comparison.png", dpi=300, bbox_inches="tight") | |
| plt.close() | |
| losses = [1.9, 1.55, 1.28, 1.05, 0.91, 0.78, 0.69, 0.61, 0.56, 0.51] | |
| plt.figure(figsize=(8, 5)) | |
| plt.plot(losses, marker="o") | |
| plt.xlabel("Training Step") | |
| plt.ylabel("Loss") | |
| plt.title("LLM Training Loss Curve") | |
| plt.grid(True) | |
| plt.savefig("loss_curve.png", dpi=300, bbox_inches="tight") | |
| plt.close() | |
| print("Plots saved successfully!") |