Upload LSTM model artifacts (2025-10-04)

README.md

@@ -1,3 +1,29 @@
----
-license: mit
----
+---
+language: en
+library_name: pytorch
+model_name: DataSynthis_ML_JobTask
+tags:
+- time-series
+- forecasting
+- lstm
+- tesla
+task: time-series-forecasting
+license: mit
+---
+
+# DataSynthis_ML_JobTask — LSTM (Multivariate)
+
+This repo hosts a PyTorch LSTM with attention for multivariate stock forecasting (OHLCV + technicals).
+
+Included artifacts:
+- pytorch_model.bin
+- config.json
+- training_stats.json
+- preprocessing.json
+
+Quick usage:
+1) Recreate the LSTM architecture from the notebook and load `pytorch_model.bin` into it
+2) Use `config.json` for feature list and `seq_length`
+3) Apply scaling with the parameters in `preprocessing.json`:
+   - Transform: X_scaled = X * scale_ + min_
+   - Inverse: X = (X_scaled - min_) / scale_

config.json

@@ -0,0 +1,20 @@
+{
+  "model_name": "DataSynthis_ML_JobTask",
+  "model_type": "lstm_forecaster",
+  "framework": "pytorch",
+  "input_size": 10,
+  "hidden_size": 128,
+  "num_layers": 2,
+  "dropout": 0.2,
+  "seq_length": 60,
+  "feature_columns": [
+    "Open",
+    "High",
+    "Low",
+    "Close",
+    "Volume"
+  ],
+  "target_column": "Close",
+  "test_size": 60,
+  "trained_device": "cuda:0"
+}

preprocessing.json

@@ -0,0 +1,80 @@
+{
+  "type": "MinMaxScaler",
+  "feature_range": [
+    0,
+    1
+  ],
+  "n_features_in_": 10,
+  "scale_": [
+    0.002487673569350158,
+    0.0024742349231429736,
+    0.002523574451582122,
+    0.0024975523914631745,
+    1.1068534149748413e-09,
+    2.4415443202382026,
+    0.0025431634113237526,
+    0.0027610656809249445,
+    11.024082474601364,
+    4.755031972326864
+  ],
+  "min_": [
+    -0.023603046617220423,
+    -0.025562145308087716,
+    -0.023730010187354034,
+    -0.023921556976927598,
+    -0.011754783267032814,
+    0.5142581908762522,
+    -0.02514069596286517,
+    -0.030808891003557718,
+    -0.105367594961265,
+    -0.040093783553332325
+  ],
+  "data_min_": [
+    9.48799991607666,
+    10.33133316040039,
+    9.403332710266112,
+    9.57800006866455,
+    10620000.0,
+    -0.21062824320390794,
+    9.885599899291993,
+    11.158333253860473,
+    0.009557946904336374,
+    0.008431864136070681
+  ],
+  "data_max_": [
+    411.4700012207031,
+    414.4966735839844,
+    405.6666564941406,
+    409.9700012207031,
+    914082000.0,
+    0.19894859376394924,
+    403.09666748046874,
+    373.3373306274414,
+    0.10026844388255864,
+    0.21873539223425345
+  ],
+  "data_range_": [
+    401.98200130462646,
+    404.165340423584,
+    396.2633237838745,
+    400.3920011520386,
+    903462000.0,
+    0.4095768369678572,
+    393.21106758117674,
+    362.17899737358096,
+    0.09071049697822225,
+    0.21030352809818276
+  ],
+  "feature_names_in_": [
+    "Open",
+    "High",
+    "Low",
+    "Close",
+    "Volume",
+    "Returns",
+    "MA_5",
+    "MA_20",
+    "Volatility",
+    "Price_Range"
+  ]
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11758aa85bc9f15771a1ec3cc5137e3147a69a40cd5bf1ac8269a9e29651a790
+size 2690907

training_stats.json

@@ -0,0 +1,86 @@
+{
+  "train_losses": [
+    0.09857238518695037,
+    0.03774987465391556,
+    0.02757480355600516,
+    0.022716590855270623,
+    0.017580508440732955,
+    0.012939690332859754,
+    0.009969860082492232,
+    0.010116041187817853,
+    0.007874186274905999,
+    0.007968392308490972,
+    0.005510548284898202,
+    0.006263008837898572,
+    0.005881796474568546,
+    0.004392406235759457,
+    0.005273079398709039,
+    0.0043719599954783915,
+    0.003990105966416498,
+    0.0037319724758466083,
+    0.0035608692715565363,
+    0.0034520207981889447,
+    0.0034128626730913917,
+    0.0034679228672757747,
+    0.0026424937532283366,
+    0.0034531054436229168,
+    0.002834749217921247,
+    0.003169918009856095,
+    0.0029859959652336937,
+    0.0024479756112365672,
+    0.002451325758981208,
+    0.002955105664053311,
+    0.0021894385456107555,
+    0.002697460786051427,
+    0.0025778446812182664,
+    0.002681724181942021,
+    0.0021836976093860963,
+    0.0024333174786685657,
+    0.0017869161946388581,
+    0.00232065737945959,
+    0.0022055609753200162,
+    0.0027215502535303435
+  ],
+  "val_losses": [
+    0.09563604276627302,
+    0.010005674014488855,
+    0.011327980008597175,
+    0.009179237803133825,
+    0.010561205369109908,
+    0.009789475317423543,
+    0.012465575011447072,
+    0.010488318589826425,
+    0.032734413631260395,
+    0.0059351618789757294,
+    0.003052401317593952,
+    0.005954862145396571,
+    0.003063126699998975,
+    0.006719038724744071,
+    0.010634368130316338,
+    0.0019871961315705753,
+    0.0031255605863407254,
+    0.002428303414490074,
+    0.0022362192588237426,
+    0.0011847723993317534,
+    0.0034753913253856203,
+    0.002814347312475244,
+    0.0028507958243911466,
+    0.0017110408031536888,
+    0.0015859827399253845,
+    0.0015160128435430427,
+    0.0013432229170575738,
+    0.002925180442010363,
+    0.002052574942354113,
+    0.0013953568510866414,
+    0.0017695200610129784,
+    0.001256332907360047,
+    0.0013786422884246956,
+    0.0011955285299336538,
+    0.0012340102014907945,
+    0.001634793191139276,
+    0.0013133611452455323,
+    0.0014318732913428296,
+    0.0019119653152301908,
+    0.0025325697109413645
+  ]
+}