adjust to new readers for csv

Signed-off-by: dafnapension <[email protected]>

Author: dafnapension

performance/compare_benchmark_performance_results.py

@@ -37,22 +37,17 @@
 ratio_line4 = (
     pr_perf["load_time"] / main_perf["load_time"] if main_perf["load_time"] > 0 else 1
 )
-line4 = f" Load time          | {main_perf['load_time_no_initial_ms']:>11} | {pr_perf['load_time_no_initial_ms']:>11} | {ratio_line4:.2f}\n"
+line4 = f" Load time          | {main_perf['load_time']:>11} | {pr_perf['load_time']:>11} | {ratio_line4:.2f}\n"
 line5 = f" DS Gen. inc. Load  | {main_perf['generate_benchmark_dataset_time']:>11} | {pr_perf['generate_benchmark_dataset_time']:>11} | {pr_perf['generate_benchmark_dataset_time'] / main_perf['generate_benchmark_dataset_time']:.2f}\n"
-line6 = f" DS Gen. exc. Load  | {round(main_perf['generate_benchmark_dataset_time'] - main_perf['load_time_no_initial_ms'], 3):>11} | {round(pr_perf['generate_benchmark_dataset_time'] - pr_perf['load_time_no_initial_ms'], 3):>11} | {ratio1:.2f}\n"
-line7 = f" Inference time     | {main_perf['inference_time']:>11} | {pr_perf['inference_time']:>11} | {pr_perf['inference_time'] / main_perf['inference_time']:.2f}\n"
-line8 = f" Evaluate  time     | {main_perf['evaluation_time']:>11} | {pr_perf['evaluation_time']:>11} | {ratio2:.2f}\n"
-line9 = f" Benchmark Instant. | {main_perf['instantiate_benchmark_time']:>11} | {pr_perf['instantiate_benchmark_time']:>11} | {pr_perf['instantiate_benchmark_time'] / main_perf['instantiate_benchmark_time']:.2f}\n"
-line10 = f" Model Instantiation| {main_perf['instantiate_model_time']:>11} | {pr_perf['instantiate_model_time']:>11} | {pr_perf['instantiate_model_time'] / main_perf['instantiate_model_time']:.2f}\n"
+line6 = f" DS Gen. exc. Load  | {round(main_perf['generate_benchmark_dataset_time'] - main_perf['load_time'], 3):>11} | {round(pr_perf['generate_benchmark_dataset_time'] - pr_perf['load_time'], 3):>11} | {ratio1:.2f}\n"
+line7 = f" Benchmark Instant. | {main_perf['instantiate_benchmark_time']:>11} | {pr_perf['instantiate_benchmark_time']:>11} | {pr_perf['instantiate_benchmark_time'] / main_perf['instantiate_benchmark_time']:.2f}\n"
 
 print("### Performance Comparison Results, time expressed in seconds:\n")
 print(line1 + line2 + line3 + line4 + line5 + line6 + line7)
 print("\n\n")
 # Performance degradation check (5% threshold)
 if ratio1 > 1.05:
-    print(
-        "\n**Warning**: Performance degradation in Dataset Generation and/or Evaluation exceeds 5%!"
-    )
+    print("\n**Warning**: Performance degradation in Dataset Generation exceeds 5%!")
     print(
         "Explore branch performance via 'python performance/bluebench_profiler.py --output_file=<path to json file>',"
         "followed by 'snakeviz <the performance.prof file specified in the output json file>'."

src/unitxt/loaders.py

@@ -447,9 +447,13 @@ def load_iterables(self):
         return iterables
 
     def split_generator(self, split: str) -> Generator:
-        dataset = pd.read_csv(
-            self.files[split], nrows=self.get_limit(), sep=self.sep
-        ).to_dict("records")
+        dataset = self.__class__._loader_cache.get(str(self) + "_" + split, None)
+        if dataset is None:
+            reader = self.get_reader()
+            dataset = reader(self.files[split], **self.get_args()).to_dict("records")
+            self.__class__._loader_cache.max_size = settings.loader_cache_size
+            self.__class__._loader_cache[str(self) + "_" + split] = dataset
+
         yield from dataset
 
 

utils/.secrets.baseline

@@ -151,7 +151,7 @@
         "filename": "src/unitxt/loaders.py",
         "hashed_secret": "840268f77a57d5553add023cfa8a4d1535f49742",
         "is_verified": false,
-        "line_number": 595,
+        "line_number": 599,
         "is_secret": false
       }
     ],