Create dataset.py

dataset.py

@@ -0,0 +1,82 @@
+import datasets
+import os
+import pandas as pd
+
+class FeatureCache(datasets.GeneratorBasedBuilder):
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            features=datasets.Features({
+
+                "timestamp": datasets.Timestamp("us"),
+                "symbol": datasets.Value("string"),
+                "isin": datasets.Value("string"),
+                "series": datasets.Value("string"),
+
+                "open": datasets.Value("float64"),
+                "high": datasets.Value("float64"),
+                "low": datasets.Value("float64"),
+                "close": datasets.Value("float64"),
+                "volume": datasets.Value("int64"),
+
+                "interval_minutes": datasets.Value("int32"),
+                "segment": datasets.Value("string"),
+                "exchange": datasets.Value("string"),
+                "year": datasets.Value("int64"),
+                "month": datasets.Value("string"),
+
+                # === ALL TA FIELDS ===
+
+                **{col: datasets.Value("float64") for col in [
+                    "HT_DCPERIOD","HT_DCPHASE","ADD","DIV","MAX","MIN","MULT","SUB","SUM",
+                    "ACOS","ASIN","ATAN","CEIL","COS","COSH","EXP","FLOOR","LN","LOG10",
+                    "SIN","SINH","SQRT","TAN","TANH","ADX","ADXR","APO","AROONOSC","BOP",
+                    "CCI","CMO","DX","MFI","MINUS_DI","MINUS_DM","MOM","PLUS_DI","PLUS_DM",
+                    "PPO","ROC","ROCP","ROCR","ROCR100","RSI","TRIX","ULTOSC","WILLR",
+                    "DEMA","EMA","HT_TRENDLINE","KAMA","MA","MIDPOINT","MIDPRICE","SAR",
+                    "SAREXT","SMA","T3","TEMA","TRIMA","WMA","AVGPRICE","MEDPRICE",
+                    "TYPPRICE","WCLPRICE","BETA","CORREL","LINEARREG",
+                    "LINEARREG_ANGLE","LINEARREG_INTERCEPT","LINEARREG_SLOPE",
+                    "STDDEV","TSF","VAR","ATR","NATR","TRANGE","AD","ADOSC","OBV"
+                ]},
+
+                **{col: datasets.Value("int32") for col in [
+                    "HT_TRENDMODE","MAXINDEX","MININDEX",
+                    "CDL2CROWS","CDL3BLACKCROWS","CDL3INSIDE","CDL3LINESTRIKE",
+                    "CDL3OUTSIDE","CDL3STARSINSOUTH","CDL3WHITESOLDIERS",
+                    "CDLABANDONEDBABY","CDLADVANCEBLOCK","CDLBELTHOLD",
+                    "CDLBREAKAWAY","CDLCLOSINGMARUBOZU","CDLCONCEALBABYSWALL",
+                    "CDLCOUNTERATTACK","CDLDARKCLOUDCOVER","CDLDOJI","CDLDOJISTAR",
+                    "CDLDRAGONFLYDOJI","CDLENGULFING","CDLEVENINGDOJISTAR",
+                    "CDLEVENINGSTAR","CDLGAPSIDESIDEWHITE","CDLGRAVESTONEDOJI",
+                    "CDLHAMMER","CDLHANGINGMAN","CDLHARAMI","CDLHARAMICROSS",
+                    "CDLHIGHWAVE","CDLHIKKAKE","CDLHIKKAKEMOD","CDLHOMINGPIGEON",
+                    "CDLIDENTICAL3CROWS","CDLINNECK","CDLINVERTEDHAMMER",
+                    "CDLKICKING","CDLKICKINGBYLENGTH","CDLLADDERBOTTOM",
+                    "CDLLONGLEGGEDDOJI","CDLLONGLINE","CDLMARUBOZU",
+                    "CDLMATCHINGLOW","CDLMATHOLD","CDLMORNINGDOJISTAR",
+                    "CDLMORNINGSTAR","CDLONNECK","CDLPIERCING","CDLRICKSHAWMAN",
+                    "CDLRISEFALL3METHODS","CDLSEPARATINGLINES","CDLSHOOTINGSTAR",
+                    "CDLSHORTLINE","CDLSPINNINGTOP","CDLSTALLEDPATTERN",
+                    "CDLSTICKSANDWICH","CDLTAKURI","CDLTASUKIGAP",
+                    "CDLTHRUSTING","CDLTRISTAR","CDLUNIQUE3RIVER",
+                    "CDLUPSIDEGAP2CROWS","CDLXSIDEGAP3METHODS"
+                ]},
+
+            })
+        )
+
+    def _split_generators(self, dl_manager):
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"data_dir": "feature_cache"}
+            )
+        ]
+
+    def _generate_examples(self, data_dir):
+        for file in os.listdir(data_dir):
+            if file.endswith(".parquet"):
+                df = pd.read_parquet(os.path.join(data_dir, file))
+                for i, row in df.iterrows():
+                    yield f"{file}_{i}", row.to_dict()