You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
constoptimizePerformancePrompt=`Optimize the following${lineOf}Python code:\n\n${context}\n\n# Start of code\n\n${code}\n\n# End of code\n\nRewrite the above Python code only from "Start of code" to "End of code", to make it more efficient WITHOUT CHANGING ITS RESULTS. Assume the code has already executed all these imports; do NOT include them in the optimized code:\n\n${imports}\n\nUse native libraries if that would make it faster than pure Python. Consider using the following other libraries, if appropriate:\n\n${libraries}\n\nYour output should only consist of valid Python code. Output the resulting Python with brief explanations only included as comments prefaced with #. Include a detailed explanatory comment before the code, starting with the text "# Proposed optimization:". Make the code as clear and simple as possible, while also making it as fast and memory-efficient as possible. Use vectorized operations${useGPUstring}whenever it would substantially increase performance, and quantify the speedup in terms of orders of magnitude. Eliminate as many for loops, while loops, and list or dict comprehensions as possible, replacing them with vectorized equivalents. If the performance is not likely to increase, leave the code unchanged. Fix any errors in the optimized code. Optimized${lineOf}code:`;
591
587
592
-
constcontext_ollama="";
593
-
constoptimizePerformancePrompt_ollama_prev=`Optimize the following${lineOf}Python code:\n\n${context_ollama}\n\n# Start of code\n\n${code}\n\n# End of code\n\nRewrite the above Python code only from "Start of code" to "End of code", to make it more efficient WITHOUT CHANGING ITS RESULTS. Only output your result in JSON, with the optimized code in "code". Optimized${lineOf}code:`;
594
-
595
-
constoptimizePerformancePrompt_ollama_pp=`Rewrite the following Python code to make it run faster. Use vectorization if possible, eliminating as many loops as possible. Try to reduce computational complexity of operations. Only output the optimized code in JSON with the key 'code'. Original code: ${code}. Optimized code:`;
596
-
597
-
// TODO parameterize based on CPU utilization, Python vs. C time, GPU choice, memory efficiency.
598
-
constoptimizePerformancePrompt_ollama=`# Original code\n${code}\n\n# This code is an optimized version of the original code that dramatically improves its performance. Whenever possible, the code has been changed to use native libraries and vectorization, and data structures like lists have been replaced by sets or dicts. Do not change any function names. Optimized code:\n`;
599
-
600
-
constpure_optimizePerformancePrompt=`Optimize the following${lineOf}Python code:\n\n${context}\n\n# Start of code\n\n${code}\n\n# End of code\n\nRewrite the above Python code only from "Start of code" to "End of code", to make it more efficient WITHOUT CHANGING ITS RESULTS. Assume the code has already executed all these imports; do NOT include them in the optimized code:\n\n${imports}\n\nONLY USE PURE PYTHON.\n\nYour output should only consist of valid Python code. Output the resulting Python with brief explanations only included as comments prefaced with #. Include a detailed explanatory comment before the code, starting with the text "# Proposed optimization:". Make the code as clear and simple as possible, while also making it as fast and memory-efficient as possible. If the performance is not likely to increase, leave the code unchanged. Fix any errors in the optimized code. Optimized${lineOf}code:`;
601
-
602
588
constmemoryEfficiencyPrompt=`Optimize the following${lineOf} Python code:\n\n${context}\n\n# Start of code\n\n${code}\n\n\n# End of code\n\nRewrite the above Python code only from "Start of code" to "End of code", to make it more memory-efficient WITHOUT CHANGING ITS RESULTS. Assume the code has already executed all these imports; do NOT include them in the optimized code:\n\n${imports}\n\nUse native libraries if that would make it more space efficient than pure Python. Consider using the following other libraries, if appropriate:\n\n${libraries}\n\nYour output should only consist of valid Python code. Output the resulting Python with brief explanations only included as comments prefaced with #. Include a detailed explanatory comment before the code, starting with the text "# Proposed optimization:". Make the code as clear and simple as possible, while also making it as fast and memory-efficient as possible. Use native libraries whenever possible to reduce memory consumption; invoke del on variables and array elements as soon as it is safe to do so. If the memory consumption is not likely to be reduced, leave the code unchanged. Fix any errors in the optimized code. Optimized${lineOf}code:`;
0 commit comments