Add HF dataset loading script for TXT telemetry files

metis-ol_mo_e-latent-telemetry.py

@@ -0,0 +1,69 @@
+import os
+import re
+import datasets
+from datasets import Dataset, Features, Value
+
+logger = datasets.logging.get_logger(__name__)
+
+
+class MetisOlmoeLatentTelemetry(datasets.GeneratorBasedBuilder):
+    """Metis OLMoE Latent Telemetry dataset for SNN routing research."""
+
+    VERSION = datasets.Version("1.0.0")
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description="Raw bare-metal telemetry logs from Metis SNN-quantized OLMoE model",
+            features=Features(
+                {
+                    "tick": Value("int64"),
+                    "best_walker": Value("int64"),
+                    "elapsed_us": Value("int64"),
+                    "source_file": Value("string"),
+                }
+            ),
+            homepage="https://huggingface.co/datasets/rmems/Metis-OLMoE-Latent-Telemetry",
+        )
+
+    def _split_generators(self, dl_manager):
+        """Returns SplitGenerators."""
+        data_dir = os.path.abspath(os.path.join(os.path.dirname(__file__)))
+        
+        # Find all .txt telemetry files
+        txt_files = []
+        for root, dirs, files in os.walk(data_dir):
+            for file in files:
+                if file.endswith(".txt") and ("telemetry" in file or "map_olmoe" in file):
+                    txt_files.append(os.path.join(root, file))
+        
+        logger.info(f"Found {len(txt_files)} telemetry files: {txt_files}")
+        
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={"filepaths": txt_files},
+            ),
+        ]
+
+    def _generate_examples(self, filepaths):
+        """Yields examples from the telemetry text files."""
+        pattern = re.compile(r"tick=(\d+)\s+best_walker=(\d+)\s+elapsed_us=(\d+)")
+        
+        idx = 0
+        for filepath in filepaths:
+            filename = os.path.basename(filepath)
+            with open(filepath, encoding="utf-8") as f:
+                for line in f:
+                    line = line.strip()
+                    if not line:
+                        continue
+                    match = pattern.match(line)
+                    if match:
+                        tick, best_walker, elapsed_us = match.groups()
+                        yield idx, {
+                            "tick": int(tick),
+                            "best_walker": int(best_walker),
+                            "elapsed_us": int(elapsed_us),
+                            "source_file": filename,
+                        }
+                        idx += 1