Update NCCL report strategy and tests to enforce correct data types (#…

…273)

src/cloudai/report_generator/util.py

@@ -21,13 +21,24 @@
 
 bokeh_size_unit_js_tick_formatter = """
     function tick_formatter(tick) {
-        var units = ['B', 'KB', 'MB', 'GB', 'TB', 'PB', 'EB', 'ZB'];
-        var i = 0;
+        const units = ['B', 'KB', 'MB', 'GB', 'TB', 'PB', 'EB', 'ZB'];
+        let i = 0;
+
+        // Handle negative ticks and large values safely
+        if (tick < 0) {
+            return '0B';  // Handle negative numbers by returning 0B as a fallback
+        }
+
+        // Loop through units until tick is smaller than 1024 or max unit is reached
         while (tick >= 1024 && i < units.length - 1) {
             tick /= 1024;
             i++;
         }
-        return tick.toFixed(1) + units[i];
+
+        // Use Number.isInteger() to check if tick is an integer (ES6 feature)
+        return Number.isInteger(tick)
+            ? `${Math.floor(tick)}${units[i]}`  // If integer, no decimal
+            : `${tick.toFixed(1)}${units[i]}`;  // Else, one decimal point
     }
     return tick_formatter(tick);
     """
@@ -62,9 +73,12 @@ def bytes_to_human_readable(num_bytes: float) -> str:
     """
     for unit in ["B", "KB", "MB", "GB", "TB", "PB", "EB", "ZB"]:
         if abs(num_bytes) < 1024.0:
-            return f"{num_bytes:3.1f}{unit}"
+            if num_bytes == int(num_bytes):
+                return f"{int(num_bytes)}{unit}"
+            else:
+                return f"{num_bytes:3.1f}{unit}"
         num_bytes /= 1024.0
-    return f"{num_bytes:.1f}YB"
+    return f"{num_bytes}YB"
 
 
 def add_human_readable_sizes(

src/cloudai/schema/test_template/nccl_test/report_generation_strategy.py

@@ -69,12 +69,15 @@ def generate_report(self, test_name: str, directory_path: Path, sol: Optional[fl
                     "#Wrong In-place",
                 ],
             )
-            df["Size (B)"] = df["Size (B)"].astype(float)
+            df["Size (B)"] = df["Size (B)"].astype(int)
+            df["Time (us) Out-of-place"] = df["Time (us) Out-of-place"].astype(float).round(1)
+            df["Time (us) In-place"] = df["Time (us) In-place"].astype(float).round(1)
             df["Algbw (GB/s) Out-of-place"] = df["Algbw (GB/s) Out-of-place"].astype(float)
             df["Busbw (GB/s) Out-of-place"] = df["Busbw (GB/s) Out-of-place"].astype(float)
             df["Algbw (GB/s) In-place"] = df["Algbw (GB/s) In-place"].astype(float)
             df["Busbw (GB/s) In-place"] = df["Busbw (GB/s) In-place"].astype(float)
             df = add_human_readable_sizes(df, "Size (B)", "Size Human-readable")
+
             self._generate_bokeh_report(test_name, df, directory_path, sol)
             self._generate_csv_report(df, directory_path)
 

tests/test_slurm_report_generation_strategy.py

@@ -85,13 +85,21 @@ def test_nccl_report_generation(setup_test_environment):
     df = pd.read_csv(csv_report_path)
     assert not df.empty, "CSV report is empty."
 
-    # Validate specific values if needed
+    # Validate data types
+    assert df["Size (B)"].dtype == int, "Size (B) is not an integer."
+    assert df["Time (us) Out-of-place"].dtype == float, "Time (us) Out-of-place is not a float."
+    assert df["Time (us) In-place"].dtype == float, "Time (us) In-place is not a float."
+
+    # Validate human-readable sizes
+    assert df.iloc[0]["Size Human-readable"] == "976.6KB", "First row Size Human-readable does not match."
+    assert df.iloc[-1]["Size Human-readable"] == "11.4MB", "Last row Size Human-readable does not match."
+
     # Example: Checking that the first entry matches the expected value
-    assert df.iloc[0]["Size (B)"] == 1000000.0, "First row Size (B) does not match."
+    assert df.iloc[0]["Size (B)"] == 1000000, "First row Size (B) does not match."
     assert df.iloc[0]["Algbw (GB/s) Out-of-place"] == 10.10, "First row Algbw (GB/s) Out-of-place does not match."
     assert df.iloc[0]["Busbw (GB/s) Out-of-place"] == 20.20, "First row Busbw (GB/s) Out-of-place does not match."
 
     # Checking that the last entry matches the expected value
-    assert df.iloc[-1]["Size (B)"] == 12000000.0, "Last row Size (B) does not match."
+    assert df.iloc[-1]["Size (B)"] == 12000000, "Last row Size (B) does not match."
     assert df.iloc[-1]["Algbw (GB/s) Out-of-place"] == 120.30, "Last row Algbw (GB/s) Out-of-place does not match."
     assert df.iloc[-1]["Busbw (GB/s) Out-of-place"] == 130.40, "Last row Busbw (GB/s) Out-of-place does not match."