easy_ml 0.2.0.pre.rc52 → 0.2.0.pre.rc56

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b0b6da194500895ff325b1408757ee328b355da448a5fcc53439d7b14ccea9c8
-  data.tar.gz: da221df414245aafd56c5a05367146b9aae131885a99fa19e11aaa0f3cd32a20
+  metadata.gz: a2eb7d933162cc05c64a1ea7c21c65f9c80283a1ae45f37226282c503607008f
+  data.tar.gz: e25a1b7b1970753ae6f65917943607102ad52ef9f90831de0660563441448975
 SHA512:
-  metadata.gz: 1d04422529423e09cff72496a29ff41038fdde5adfc0f267065981d04c24c9e9475d9ad0b2ab083e2d5e8cef22581180df97c93e960579cca8657b1771adc977
-  data.tar.gz: d9904acd55317dc68d774c26f392399990fdd52f45f394f8fdc6addf11456ea1361e2d972a438647c7dc0f679e3642cd8ae1f13cc9e0824c72865de8ecb0b6a9
+  metadata.gz: 6fc39e6b2838ab6242df1848411450764b0434b63bb7e4f1cb60151850e794f1d0a71a956b70b7ca78e159003efc6947d7dbaad9010c7c6899a1baeb8c7570b2
+  data.tar.gz: f50ceecc6935fea0c1f82e5b76beaef2e6ee329087d0f7b7739d4a88b4738d9ed37d13acd47bf12d118092c0b66774772dab9a30f60e12c3854360329caacfa9

data/app/controllers/easy_ml/application_controller.rb

@@ -52,7 +52,7 @@ module EasyML
 
       flash_messages << { type: "success", message: flash[:notice] } if flash[:notice]
 
-      flash_messages << { type: "error", message: flash[:alert] } if flash[:alert]
+      flash_messages << { type: "error", message: flash[:error] } if flash[:error]
 
       flash_messages << { type: "info", message: flash[:info] } if flash[:info]
 

data/app/controllers/easy_ml/deploys_controller.rb

@@ -13,7 +13,7 @@ module EasyML
       flash[:notice] = "Model deployment has started"
       redirect_to easy_ml_model_path(@deploy.model)
     rescue => e
-      flash[:alert] = "Trouble deploying model: #{e.message}"
+      flash[:error] = "Trouble deploying model: #{e.message}"
       redirect_to easy_ml_model_path(@deploy.model)
     end
   end

data/app/controllers/easy_ml/models_controller.rb

@@ -70,6 +70,9 @@ module EasyML
         flash[:notice] = "Model was successfully updated."
         redirect_to easy_ml_models_path
       else
+        errors = model.errors.to_hash(true)
+        values = errors.values.flatten
+        flash.now[:error] = values.join(", ")
         render inertia: "pages/EditModelPage", props: {
           model: model_to_json(model),
           datasets: EasyML::Dataset.all.map { |dataset| dataset_to_json(dataset) },
@@ -99,7 +102,7 @@ module EasyML
         flash[:notice] = "Model was successfully deleted."
         redirect_to easy_ml_models_path
       else
-        flash[:alert] = "Failed to delete the model."
+        flash[:error] = "Failed to delete the model."
         redirect_to easy_ml_models_path
       end
     end

data/app/frontend/components/AlertProvider.tsx

@@ -38,9 +38,12 @@ export function AlertProvider({ children }: { children: React.ReactNode }) {
     const id = Math.random().toString(36).substring(7);
     setAlerts(prev => [...prev, { id, type, message }]);
 
-    setTimeout(() => {
-      removeAlert(id);
-    }, numSeconds * 1000);
+    // Only auto-dismiss non-error alerts
+    if (type !== 'error') {
+      setTimeout(() => {
+        removeAlert(id);
+      }, numSeconds * 1000);
+    }
   }, [removeAlert]);
 
   return (

data/app/frontend/components/ModelForm.tsx

@@ -103,29 +103,6 @@ export function ModelForm({ initialData, datasets, constants, isEditing, errors:
   const objectives: { value: string; label: string; description?: string }[] = 
     constants.objectives[data.model.model_type]?.[data.model.task] || [];
 
-  useEffect(() => {
-    // Only set default metrics if none were provided from the backend
-    if (!initialData?.metrics) {
-      const availableMetrics = constants.metrics[data.model.task]?.map(metric => metric.value) || [];
-      setData({
-        ...data,
-        model: {
-          ...data.model,
-          objective: data.model.task === 'classification' ? 'binary:logistic' : 'reg:squarederror',
-          metrics: availableMetrics
-        }
-      });
-    } else {
-      setData({
-        ...data,
-        model: {
-          ...data.model,
-          objective: data.model.task === 'classification' ? 'binary:logistic' : 'reg:squarederror'
-        }
-      });
-    }
-  }, [data.model.task]);
-
   useEffect(() => {
     if (isDataSet) {
       save();
@@ -187,11 +164,21 @@ export function ModelForm({ initialData, datasets, constants, isEditing, errors:
     save();
   };
 
-  console.log(data.model)
   const selectedDataset = datasets.find(d => d.id === data.model.dataset_id);
 
   const filteredTunerJobConstants = constants.tuner_job_constants[data.model.model_type] || {};
 
+  const handleTaskChange = (value: string) => {
+    // First update the task
+    setData('model.task', value);
+    
+    // Then force reset metrics to empty array
+    setData('model.metrics', []);
+
+    // Update objective based on new task
+    setData('model.objective', value === 'classification' ? 'binary:logistic' : 'reg:squarederror');
+  };
+
   return (
     <form onSubmit={handleSubmit} className="space-y-8">
       <div className="flex justify-between items-center border-b pb-4">
@@ -266,8 +253,7 @@ export function ModelForm({ initialData, datasets, constants, isEditing, errors:
             <SearchableSelect
               options={constants.tasks}
               value={data.model.task}
-              onChange={(value) => setData('model.task', value as string)}
-              placeholder="Select task"
+              onChange={handleTaskChange}
             />
             <ErrorDisplay error={errors.task} />
           </div>
@@ -300,24 +286,21 @@ export function ModelForm({ initialData, datasets, constants, isEditing, errors:
                   type="checkbox"
                   checked={data.model.metrics.includes(metric.value)}
                   onChange={(e) => {
-                    const metrics = e.target.checked
+                    const newMetrics = e.target.checked
                       ? [...data.model.metrics, metric.value]
                       : data.model.metrics.filter(m => m !== metric.value);
-                    setData('model.metrics', metrics);
+                    setData('model.metrics', newMetrics);
                   }}
-                  className="h-4 w-4 rounded border-gray-300 text-blue-600 focus:ring-blue-500"
+                  className="h-4 w-4 text-blue-600 focus:ring-blue-500 border-gray-300 rounded"
                 />
                 <div className="ml-3">
-                  <span className="block text-sm font-medium text-gray-900">
-                    {metric.label}
-                  </span>
-                  <span className="block text-xs text-gray-500">
-                    {metric.direction === 'maximize' ? 'Higher is better' : 'Lower is better'}
-                  </span>
+                  <span className="block text-sm font-medium text-gray-900">{metric.label}</span>
+                  <span className="block text-xs text-gray-500">Direction: {metric.direction}</span>
                 </div>
               </label>
             ))}
           </div>
+          <ErrorDisplay error={errors.metrics} />
         </div>
       </div>
 
@@ -374,6 +357,7 @@ export function ModelForm({ initialData, datasets, constants, isEditing, errors:
           dataset: selectedDataset,
           retraining_job: data.model.retraining_job_attributes
         }}
+        metrics={constants.metrics}
         tunerJobConstants={filteredTunerJobConstants}
         timezone={constants.timezone}
         retrainingJobConstants={constants.retraining_job_constants}

data/app/frontend/components/ScheduleModal.tsx

@@ -37,27 +37,20 @@ interface ScheduleModalProps {
       tuning_enabled?: boolean;
     };
   };
+  metrics: {
+    [key: string]: Array<{
+      value: string;
+      label: string;
+      description: string;
+      direction: string;
+    }>;
+  };
   tunerJobConstants: any;
   timezone: string;
   retrainingJobConstants: any;
 }
 
-const METRICS = {
-  classification: [
-    { value: 'accuracy_score', label: 'Accuracy', description: 'Overall prediction accuracy', direction: 'maximize' },
-    { value: 'precision_score', label: 'Precision', description: 'Ratio of true positives to predicted positives', direction: 'maximize' },
-    { value: 'recall_score', label: 'Recall', description: 'Ratio of true positives to actual positives', direction: 'maximize' },
-    { value: 'f1_score', label: 'F1 Score', description: 'Harmonic mean of precision and recall', direction: 'maximize' }
-  ],
-  regression: [
-    { value: 'mean_absolute_error', label: 'Mean Absolute Error', description: 'Average absolute differences between predicted and actual values', direction: 'minimize' },
-    { value: 'mean_squared_error', label: 'Mean Squared Error', description: 'Average squared differences between predicted and actual values', direction: 'minimize' },
-    { value: 'root_mean_squared_error', label: 'Root Mean Squared Error', description: 'Square root of mean squared error', direction: 'minimize' },
-    { value: 'r2_score', label: 'R² Score', description: 'Proportion of variance in the target that is predictable', direction: 'maximize' }
-  ]
-};
-
-export function ScheduleModal({ isOpen, onClose, onSave, initialData, tunerJobConstants, timezone, retrainingJobConstants }: ScheduleModalProps) {
+export function ScheduleModal({ isOpen, onClose, onSave, initialData, metrics, tunerJobConstants, timezone, retrainingJobConstants }: ScheduleModalProps) {
   const [showBatchTrainingInfo, setShowBatchTrainingInfo] = useState(false);
   const [activeBatchPopover, setActiveBatchPopover] = useState<'size' | 'overlap' | null>(null);
 
@@ -97,8 +90,8 @@ export function ScheduleModal({ isOpen, onClose, onSave, initialData, tunerJobCo
         day_of_week: initialData.retraining_job?.at?.day_of_week ?? 1,
         day_of_month: initialData.retraining_job?.at?.day_of_month ?? 1
       },
-      metric: initialData.retraining_job?.metric || METRICS[initialData.task === 'classification' ? 'classification' : 'regression'][0].value,
-      threshold: initialData.retraining_job?.threshold || (initialData.task === 'classification' ? 0.85 : 0.1),
+      metric: initialData.retraining_job?.metric || (metrics[initialData.task]?.[0]?.value ?? ''),
+      threshold: initialData.retraining_job?.threshold ?? (initialData.task === 'classification' ? 0.85 : 0.1),
       tuner_config: initialData.retraining_job?.tuner_config ? {
         n_trials: initialData.retraining_job.tuner_config.n_trials || 10,
         config: {
@@ -336,9 +329,9 @@ export function ScheduleModal({ isOpen, onClose, onSave, initialData, tunerJobCo
   };
 
   return (
-    <div className="fixed inset-0 bg-black bg-opacity-50 flex items-center justify-center z-50">
-      <div className="bg-white rounded-lg w-full max-w-6xl max-h-[90vh] overflow-hidden">
-        <div className="flex justify-between items-center p-4 border-b">
+    <div className="fixed inset-0 bg-black bg-opacity-50 flex items-start justify-center pt-[5vh] z-50">
+      <div className="bg-white rounded-lg w-full max-w-6xl flex flex-col" style={{ maxHeight: '90vh' }}>
+        <div className="flex-none flex justify-between items-center p-4 border-b">
           <h2 className="text-lg font-semibold">Training Configuration</h2>
           <button
             onClick={onClose}
@@ -348,7 +341,7 @@ export function ScheduleModal({ isOpen, onClose, onSave, initialData, tunerJobCo
           </button>
         </div>
 
-        <div className="p-6 grid grid-cols-2 gap-8 max-h-[calc(90vh-8rem)] overflow-y-auto">
+        <div className="flex-1 p-6 grid grid-cols-2 gap-8 overflow-y-auto">
           {/* Left Column */}
           <div className="space-y-8">
             {/* Training Schedule */}
@@ -575,7 +568,7 @@ export function ScheduleModal({ isOpen, onClose, onSave, initialData, tunerJobCo
                           Metric
                         </label>
                         <SearchableSelect
-                          options={METRICS[initialData.task === 'classification' ? 'classification' : 'regression'].map((metric) => ({
+                          options={metrics[initialData.task].map((metric) => ({
                             value: metric.value,
                             label: metric.label,
                             description: metric.description
@@ -610,8 +603,7 @@ export function ScheduleModal({ isOpen, onClose, onSave, initialData, tunerJobCo
                             <h3 className="text-sm font-medium text-blue-800">Deployment Criteria</h3>
                             <p className="mt-2 text-sm text-blue-700">
                               {(() => {
-                                const metricsList = METRICS[initialData.task === 'classification' ? 'classification' : 'regression'];
-                                const selectedMetric = metricsList.find(m => m.value === formData.retraining_job_attributes.metric);
+                                const selectedMetric = metrics[initialData.task].find(m => m.value === formData.retraining_job_attributes.metric);
                                 const direction = selectedMetric?.direction === 'minimize' ? 'below' : 'above';
                                 
                                 return `The model will be automatically deployed when the ${selectedMetric?.label} is ${direction} ${formData.retraining_job_attributes.threshold}.`;
@@ -711,7 +703,7 @@ export function ScheduleModal({ isOpen, onClose, onSave, initialData, tunerJobCo
           </div>
         </div>
 
-        <div className="flex justify-end gap-4 p-4 border-t">
+        <div className="flex-none flex justify-end gap-4 p-4 border-t bg-white">
           <button
             onClick={onClose}
             className="px-4 py-2 text-sm font-medium text-gray-700 hover:text-gray-500"

data/app/frontend/components/SearchableSelect.tsx

@@ -1,5 +1,6 @@
 import React, { useState, useRef, useEffect, forwardRef } from 'react';
 import { Search, Check } from 'lucide-react';
+import { createPortal } from 'react-dom';
 
 interface Option {
   value: string | number;
@@ -20,6 +21,7 @@ export const SearchableSelect = forwardRef<HTMLButtonElement, SearchableSelectPr
   ({ options, value, onChange, placeholder = 'Search...', renderOption }, ref) => {
     const [isOpen, setIsOpen] = useState(false);
     const [searchQuery, setSearchQuery] = useState('');
+    const [dropdownPosition, setDropdownPosition] = useState({ top: 0, left: 0, width: 0 });
     const containerRef = useRef<HTMLDivElement>(null);
     const inputRef = useRef<HTMLInputElement>(null);
 
@@ -47,11 +49,98 @@ export const SearchableSelect = forwardRef<HTMLButtonElement, SearchableSelectPr
       }
     }, [isOpen]);
 
+    useEffect(() => {
+      if (isOpen && containerRef.current) {
+        const rect = containerRef.current.getBoundingClientRect();
+        setDropdownPosition({
+          top: rect.bottom + window.scrollY,
+          left: rect.left + window.scrollX,
+          width: rect.width
+        });
+      }
+    }, [isOpen]);
+
+    const handleOptionClick = (optionValue: Option['value'], e: React.MouseEvent) => {
+      debugger;
+      e.preventDefault();
+      e.stopPropagation();
+      onChange(optionValue);
+      setIsOpen(false);
+      setSearchQuery('');
+    };
+
+    const dropdown = isOpen && createPortal(
+      <div 
+        className="fixed bg-white shadow-lg rounded-md overflow-hidden border border-gray-200"
+        style={{
+          top: dropdownPosition.top,
+          left: dropdownPosition.left,
+          width: dropdownPosition.width,
+          zIndex: 9999
+        }}
+      >
+        <div className="p-2 border-b">
+          <div className="relative">
+            <Search className="absolute left-3 top-1/2 transform -translate-y-1/2 w-4 h-4 text-gray-400" />
+            <input
+              ref={inputRef}
+              type="text"
+              className="w-full pl-9 pr-4 py-2 border border-gray-300 rounded-md focus:outline-none focus:ring-1 focus:ring-blue-500 focus:border-blue-500"
+              placeholder="Search..."
+              value={searchQuery}
+              onChange={(e) => setSearchQuery(e.target.value)}
+              onMouseDown={(e) => e.stopPropagation()}
+            />
+          </div>
+        </div>
+
+        <div className="max-h-60 overflow-y-auto">
+          {filteredOptions.length === 0 ? (
+            <div className="text-center py-4 text-sm text-gray-500">
+              No results found
+            </div>
+          ) : (
+            <ul className="py-1">
+              {filteredOptions.map((option) => (
+                <li key={option.value}>
+                  <button
+                    type="button"
+                    className={`w-full text-left px-4 py-2 hover:bg-gray-100 ${
+                      option.value === value ? 'bg-blue-50' : ''
+                    }`}
+                    onMouseDown={(e) => handleOptionClick(option.value, e)}
+                  >
+                    <div className="flex items-center justify-between">
+                      <span className="block font-medium">
+                        {option.label}
+                      </span>
+                      {option.value === value && (
+                        <Check className="w-4 h-4 text-blue-600" />
+                      )}
+                    </div>
+                    {option.description && (
+                      <span className="block text-sm text-gray-500">
+                        {option.description}
+                      </span>
+                    )}
+                  </button>
+                </li>
+              ))}
+            </ul>
+          )}
+        </div>
+      </div>,
+      document.body
+    );
+
     return (
       <div className="relative" ref={containerRef}>
         <button
           type="button"
-          onClick={() => setIsOpen(!isOpen)}
+          onMouseDown={(e) => {
+            e.preventDefault();
+            setIsOpen(!isOpen);
+          }}
           className="w-full bg-white relative border border-gray-300 rounded-md shadow-sm pl-3 pr-10 py-2 text-left cursor-pointer focus:outline-none focus:ring-1 focus:ring-blue-500 focus:border-blue-500"
           ref={ref}
         >
@@ -61,72 +150,8 @@ export const SearchableSelect = forwardRef<HTMLButtonElement, SearchableSelectPr
             <span className="block truncate text-gray-500">{placeholder}</span>
           )}
         </button>
-
-        {isOpen && (
-          <div className="absolute z-10 mt-1 w-full bg-white shadow-lg max-h-96 rounded-md overflow-hidden">
-            <div className="p-2 border-b">
-              <div className="relative">
-                <Search className="absolute left-3 top-1/2 transform -translate-y-1/2 w-4 h-4 text-gray-400" />
-                <input
-                  ref={inputRef}
-                  type="text"
-                  className="w-full pl-9 pr-4 py-2 border border-gray-300 rounded-md focus:outline-none focus:ring-1 focus:ring-blue-500 focus:border-blue-500"
-                  placeholder="Search..."
-                  value={searchQuery}
-                  onChange={(e) => setSearchQuery(e.target.value)}
-                  onClick={(e) => e.stopPropagation()}
-                />
-              </div>
-            </div>
-
-            <div className="max-h-60 overflow-y-auto">
-              {filteredOptions.length === 0 ? (
-                <div className="text-center py-4 text-sm text-gray-500">
-                  No results found
-                </div>
-              ) : (
-                <ul className="py-1">
-                  {filteredOptions.map((option) => (
-                    <li key={option.value}>
-                      <button
-                        type="button"
-                        className={`w-full text-left px-4 py-2 hover:bg-gray-100 ${
-                          option.value === value ? 'bg-blue-50' : ''
-                        }`}
-                        onClick={() => {
-                          onChange(option.value);
-                          setIsOpen(false);
-                          setSearchQuery('');
-                        }}
-                      >
-                        {renderOption ? (
-                          renderOption(option)
-                        ) : (
-                          <div className="flex items-center justify-between">
-                            <div>
-                              <div className="font-medium">{option.label}</div>
-                              {option.description && (
-                                <div className="text-sm text-gray-500">
-                                  {option.description}
-                                </div>
-                              )}
-                            </div>
-                            {option.value === value && (
-                              <Check className="w-4 h-4 text-blue-600" />
-                            )}
-                          </div>
-                        )}
-                      </button>
-                    </li>
-                  ))}
-                </ul>
-              )}
-            </div>
-          </div>
-        )}
+        {dropdown}
       </div>
     );
   }
-);
-
-SearchableSelect.displayName = 'SearchableSelect';
+);

data/app/jobs/easy_ml/refresh_dataset_job.rb

@@ -3,6 +3,8 @@ module EasyML
     def perform(id)
       begin
         dataset = EasyML::Dataset.find(id)
+        return if dataset.workflow_status == :analyzing
+
         puts "Refreshing dataset #{dataset.name}"
         puts "Needs refresh? #{dataset.needs_refresh?}"
         unless dataset.needs_refresh?
@@ -12,6 +14,7 @@ module EasyML
         create_event(dataset, "started")
 
         puts "Prepare! #{dataset.name}"
+        dataset.unlock!
         dataset.prepare
         if dataset.features.needs_fit.any?
           dataset.fit_features(async: true)

data/app/models/easy_ml/column.rb

@@ -32,6 +32,7 @@ module EasyML
     before_save :ensure_valid_datatype
     after_create :set_date_column_if_date_splitter
     after_save :handle_date_column_change
+    before_save :set_defaults
 
     # Scopes
     scope :visible, -> { where(hidden: false) }
@@ -98,6 +99,84 @@ module EasyML
 
     private
 
+    def set_defaults
+      self.preprocessing_steps = set_preprocessing_steps_defaults
+    end
+
+    def set_preprocessing_steps_defaults
+      preprocessing_steps.inject({}) do |h, (type, config)|
+        h.tap do
+          h[type] = set_preprocessing_step_defaults(config)
+        end
+      end
+    end
+
+    ALLOWED_PARAMS = {
+      constant: [:constant],
+      categorical: %i[categorical_min one_hot ordinal_encoding],
+      most_frequent: %i[one_hot ordinal_encoding],
+      mean: [:clip],
+      median: [:clip],
+    }
+
+    REQUIRED_PARAMS = {
+      constant: [:constant],
+      categorical: %i[categorical_min one_hot ordinal_encoding],
+    }
+
+    DEFAULT_PARAMS = {
+      categorical_min: 1,
+      one_hot: true,
+      ordinal_encoding: false,
+      clip: { min: 0, max: 1_000_000_000 },
+      constant: nil,
+    }
+
+    XOR_PARAMS = [{
+      params: [:one_hot, :ordinal_encoding],
+      default: :one_hot,
+    }]
+
+    def set_preprocessing_step_defaults(config)
+      config.deep_symbolize_keys!
+      config[:params] ||= {}
+      params = config[:params].symbolize_keys
+
+      required = REQUIRED_PARAMS.fetch(config[:method].to_sym, [])
+      allowed = ALLOWED_PARAMS.fetch(config[:method].to_sym, [])
+
+      missing = required - params.keys
+      missing.reject! do |param|
+        XOR_PARAMS.any? do |rule|
+          if rule[:params].include?(param)
+            missing_param = rule[:params].find { |p| p != param }
+            params[missing_param] == true
+          else
+            false
+          end
+        end
+      end
+      extra = params.keys - allowed
+
+      missing.each do |key|
+        params[key] = DEFAULT_PARAMS.fetch(key)
+      end
+
+      extra.each do |key|
+        params.delete(key)
+      end
+
+      # Only set one of one_hot or ordinal_encoding to true,
+      # by default set one_hot to true
+      xor = XOR_PARAMS.find { |rule| rule[:params] & params.keys == rule[:params] }
+      if xor && xor[:params].all? { |param| params[param] }
+        xor[:params].each { |param| params[param] = false }
+        params[xor[:default]] = true
+      end
+
+      config.merge!(params: params)
+    end
+
     def handle_date_column_change
       return unless saved_change_to_is_date_column? && is_date_column?
 

data/app/models/easy_ml/column_list.rb

@@ -1,14 +1,18 @@
 module EasyML
   module ColumnList
-    def sync
+    def sync(only_new: false)
       return unless dataset.schema.present?
 
       EasyML::Column.transaction do
         col_names = syncable
         existing_columns = where(name: col_names)
         import_new(col_names, existing_columns)
-        update_existing(existing_columns)
-        delete_missing(existing_columns)
+
+        if !only_new
+          update_existing(existing_columns)
+          delete_missing(existing_columns)
+        end
+
         if existing_columns.none? # Totally new dataset
           dataset.after_create_columns
         end

data/app/models/easy_ml/dataset.rb

@@ -175,7 +175,6 @@ module EasyML
 
     def actually_refresh
       refreshing do
-        split_data
         process_data
         fully_reload
         learn
@@ -273,10 +272,10 @@ module EasyML
       raw.split_at.present? && raw.split_at < datasource.last_updated_at
     end
 
-    def learn
+    def learn(only_new: false)
       learn_schema
       learn_statistics
-      columns.sync
+      columns.sync(only_new: only_new)
     end
 
     def refreshing
@@ -399,7 +398,7 @@ module EasyML
 
       # Learn will update columns, so if any features have been added
       # since the last time columns were learned, we should re-learn the schema
-      learn if idx == 0 && needs_learn?(df)
+      learn(only_new: true) if idx == 1 && needs_learn?(df)
       df = apply_column_mask(df, inference: inference) unless all_columns
       raise_on_nulls(df) if inference
       df, = processed.split_features_targets(df, true, target) if split_ys

data/app/models/easy_ml/event.rb

@@ -56,14 +56,16 @@ module EasyML
       create_event(model, "failed", error)
     end
 
+    def self.easy_ml_context(stacktrace)
+      stacktrace.select { |loc| loc.match?(/easy_ml/) }
+    end
+
     def self.format_stacktrace(error)
       return nil if error.nil?
 
       topline = error.inspect
 
-      stacktrace = error.backtrace.select do |loc|
-        loc.match?(/easy_ml/)
-      end
+      stacktrace = easy_ml_context(error.backtrace)
 
       %(#{topline}