experiment-configuration-agent 0.1.6__py3-none-any.whl → 0.1.9__py3-none-any.whl

experiment_config_agent/config.py

@@ -19,8 +19,8 @@ class GluonConfig(BaseSettings):
 
 
     # This will now read from LM_PROVIDER environment variable
-    provider: str = Field(default=os.getenv('LM_SERVICE_PROVIDER', 'openai'), description="AI provider to use")
-    model: str = Field(default=os.getenv('LLM_MODEL_NAME', 'gpt-5-mini'), description="AI model to use")
+    provider: str = Field(default=os.getenv('LLM_SERVICE_PROVIDER', 'openai'), description="AI provider to use")
+    model: str = Field(default=os.getenv('LLM_MODEL_NAME', 'gpt-4.1-mini'), description="AI model to use")
     temperature: float = Field(default=0.3, ge=0.0, le=0.5, description="AI model temperature")
     max_tokens: int = Field(default=4000, ge=0, le=8000, description="Maximum tokens for AI response")
 

experiment_config_agent/constants.py

@@ -27,6 +27,121 @@ CORE MODEL CONCEPTS:
 - If bagging is enabled (num_bag_folds > 0), 'split_test_size' is ignored as CV is used.
 - If bagging is 0, 'split_test_size' (e.g., 0.1 to 0.2) is mandatory to monitor overfitting.
 
+
+5. TRAINING TIME LIMIT CALCULATION — MANDATORY EXECUTION
+
+You MUST calculate time_limit exactly using the following steps:
+
+STEP 1: Calculate BASE TIME from row count
+------------------------------------------
+Read rows from dataset_insights["rows"]
+
+    IF 0 < rows ≤ 500000:
+        base_time = 60 + (rows / 500000) * 540 → Range: 60-600 seconds
+
+    ELSE IF 500000 < rows ≤ 7000000:
+        base_time = 600 + (rows / 7000000) * 1200 → Range: 600-1800 seconds
+
+    ELSE (rows > 7000000):
+        base_time = 1800 + (rows / 7000000) * 1800 → Range: 1800-3600 seconds
+
+STEP 2: Apply COLUMN COUNT multiplier
+--------------------------------------
+Count columns from dataset_insights["feature_columns"]
+
+    IF 0 < columns ≤ 50:
+        column_multiplier = 1.0
+    ELSE IF 50 < columns ≤ 100:
+        column_multiplier = 1.2
+    ELSE IF 100 < columns ≤ 150:
+        column_multiplier = 1.4
+    ELSE IF 150 < columns ≤ 200:
+        column_multiplier = 1.6
+    ELSE (columns > 200):
+        column_multiplier = 1.8
+
+STEP 3: Apply MODEL COUNT multiplier
+------------------------------------
+Count the number of models selected in the 'models' list:
+
+    IF num_models = 1:
+        model_multiplier = 0.6
+    ELSE IF num_models = 2:
+        model_multiplier = 0.8
+    ELSE IF num_models = 3:
+        model_multiplier = 1.0
+    ELSE IF num_models = 4:
+        model_multiplier = 1.2
+    ELSE IF num_models = 5:
+        model_multiplier = 1.4
+    ELSE (num_models >= 6):
+        model_multiplier = 1.5
+
+STEP 4: Apply ENSEMBLE STRATEGY multiplier
+------------------------------------------
+Based on num_bag_folds, num_bag_sets, and num_stack_levels:
+
+    A. Bagging folds multiplier (from num_bag_folds):
+        IF num_bag_folds = 0:
+            bag_folds_multiplier = 1.0
+        ELSE IF num_bag_folds <= 5:
+            bag_folds_multiplier = 1.1
+        ELSE IF num_bag_folds <= 8:
+            bag_folds_multiplier = 1.3
+        ELSE (num_bag_folds > 8):
+            bag_folds_multiplier = 1.5
+
+    B. Bagging sets multiplier (from num_bag_sets):
+        IF num_bag_sets = 1:
+            bag_sets_multiplier = 1.0
+        ELSE IF num_bag_sets = 2:
+            bag_sets_multiplier = 1.2
+        ELSE (num_bag_sets >= 3):
+            bag_sets_multiplier = 1.4
+
+    C. Stacking multiplier (from num_stack_levels):
+        IF num_stack_levels = 0:
+            stack_multiplier = 1.0
+        ELSE IF num_stack_levels = 1:
+            stack_multiplier = 1.5
+        ELSE (num_stack_levels >= 2):
+            stack_multiplier = 2.0
+
+    ensemble_multiplier = bag_folds_multiplier * bag_sets_multiplier * stack_multiplier
+
+STEP 5: Calculate FINAL time_limit
+----------------------------------
+    time_limit = base_time * column_multiplier * model_multiplier * ensemble_multiplier
+
+STEP 6: Apply bounds and round
+------------------------------
+    MIN_TIME = 60
+    MAX_TIME = 3600
+
+    IF time_limit < MIN_TIME → time_limit = MIN_TIME
+    IF time_limit > MAX_TIME → time_limit = MAX_TIME
+
+    Round UP to nearest multiple of 5.
+
+EXAMPLE CALCULATION:
+-------------------
+Given: rows=250000, columns=45, models=['GBM','CAT','XGB'], num_bag_folds=5, num_bag_sets=2, num_stack_levels=1
+
+    base_time = 60 + (250000/500000) * 540 = 330 seconds
+    column_multiplier = 1.2 (60 columns)
+    model_multiplier = 1.0 (3 models)
+    bag_folds_multiplier = 1.3 (5 folds)
+    bag_sets_multiplier = 1.5 (2 sets)
+    stack_multiplier = 1.5 (1 level)
+    ensemble_multiplier = 1.3 * 1.5 * 1.5 = 2.925
+
+    time_limit = 330 * 1.2 * 1.0 * 2.925 = 1158.3 seconds
+    Rounded up to: 1160 seconds
+
+- You are NOT allowed to choose a different value than time_limit.
+- You are NOT allowed to approximate, proceed with the calculated value.
+- You MUST output the computed value exactly.
+
 PRESET SELECTION LOGIC (Ordered by Quality/Complexity):
 ======================================================
 
@@ -45,7 +160,9 @@ CONSTRAINTS:
 - You MUST only use the 6 allowed models ('GBM', 'CAT', 'XGB', 'RF', 'XT', 'KNN').
 - You MUST only use the 5 allowed presets ('extreme_quality', 'best_quality', 'high_quality', 'good_quality', 'medium_quality').
 - You MUST only use the 6 allowed metrics for both eval_metric and additional_metrics.
-
+- Do NOT choose mid-range values by default.
+- You MUST calculate time_limit using the formula in Section 5. The minimum is 60 seconds and maximum is 3600 seconds.
+- For each scenario, you MUST show the time_limit calculation steps clearly before stating the final value.
 Provide three distinct scenarios: Max Accuracy (Heavy), Production-Ready (Balanced), and Fast-Track (Speed)."""
 
 
@@ -201,7 +318,7 @@ Based on the above information, recommend an optimal AutoGluon configuration tha
 
 
 Consider multiple scenarios:
-- Scenario A: Maximum accuracy (accepting longer training time)
+- Scenario A: Maximum accuracy (accepting longer training time upto the relevant time limit for the given dataset rows)
 - Scenario B: Balanced accuracy and speed (production-ready)
 - Scenario C: Fast training and inference (prototyping/deployment constrained)
 

experiment_config_agent/models.py

@@ -1,5 +1,5 @@
 from pydantic import BaseModel, Field
-from typing import List, Literal
+from typing import List, Literal, Optional
 
 class AutoGluonConfig(BaseModel):
     eval_metric: str = Field(
@@ -7,11 +7,15 @@ class AutoGluonConfig(BaseModel):
         description="Primary metric to optimize. Allowed: 'accuracy', 'log_loss', 'f1', 'roc_auc', 'precision', 'recall'."
     )
     
-    preset: Literal[
+    preset: Optional[Literal[
         'best_quality', 'high_quality', 'good_quality', 'medium_quality'
-    ] = Field(
-        ..., 
-         description="Preset configurations. 'extreme_quality' and 'best_quality' enable bagging/stacking for maximum accuracy."
+    ]] = Field(
+        default='good_quality',
+        description=(
+            "Optional preset configuration. "
+            "'best_quality' enables bagging/stacking for maximum accuracy. "
+            "If not provided, AutoGluon default behavior will be used."
+        )
     )
     
     additional_metrics: List[str] = Field(
@@ -21,7 +25,7 @@ class AutoGluonConfig(BaseModel):
     
     time_limit: int = Field(
         ..., 
-        description="Total training time in seconds. AutoGluon will distribute this across models. Small datasets: 300, Medium: 3600, Large: 7200+."
+        description="Total training time in seconds. AutoGluon will distribute this across models. Small datasets (upto 500000 rows): Value between 60 seconds to 600 seconds, Medium (upto 7000000 rows): Value between 601 seconds to 1800 seconds, Large (7000000+ rows): Value between 1800 seconds to 3600 seconds."
     )
     
     num_bag_folds: int = Field(

{experiment_configuration_agent-0.1.6.dist-info → experiment_configuration_agent-0.1.9.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: experiment-configuration-agent
-Version: 0.1.6
+Version: 0.1.9
 Summary: Add your description here
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown

experiment_configuration_agent-0.1.9.dist-info/RECORD

@@ -0,0 +1,9 @@
+experiment_config_agent/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+experiment_config_agent/agent.py,sha256=bjFBjyR0_0hRUePfCDICitGfG7fVuZxBhPkwZZ0gsxM,3700
+experiment_config_agent/config.py,sha256=bzxTH9QQ5ds3SyvxELeDjZg8WKNe7DikOE5-H-CjjN8,1141
+experiment_config_agent/constants.py,sha256=9EYOuiDzaD4Pd-yxiC0pj1t67pxHf7EXZqQcqqhp2OQ,12271
+experiment_config_agent/models.py,sha256=eGftcqwXazqvArSifQNg2XlF1DW3Za8WZ0KCxDlHWRw,2572
+experiment_configuration_agent-0.1.9.dist-info/METADATA,sha256=o7AnkR64hwDtup9-Rk6G-V3Ot71fkyNZjVbBA2MrUw0,4453
+experiment_configuration_agent-0.1.9.dist-info/WHEEL,sha256=qELbo2s1Yzl39ZmrAibXA2jjPLUYfnVhUNTlyF1rq0Y,92
+experiment_configuration_agent-0.1.9.dist-info/top_level.txt,sha256=5c9CyVEjFUlvEf08vJIvi6BkzGuS4wdwtjdmCk2uL2U,24
+experiment_configuration_agent-0.1.9.dist-info/RECORD,,

{experiment_configuration_agent-0.1.6.dist-info → experiment_configuration_agent-0.1.9.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.9.0)
+Generator: setuptools (80.10.1)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

experiment_configuration_agent-0.1.6.dist-info/RECORD

@@ -1,9 +0,0 @@
-experiment_config_agent/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-experiment_config_agent/agent.py,sha256=bjFBjyR0_0hRUePfCDICitGfG7fVuZxBhPkwZZ0gsxM,3700
-experiment_config_agent/config.py,sha256=yAS1XWdOklcpuHwq5F3u-j2zQmX-ErLowj9IqcgnqH4,1138
-experiment_config_agent/constants.py,sha256=O7fsJQXVmt8Zs-A3sYxTafyNdpQv2H4tbL3E--rJ7Ug,8167
-experiment_config_agent/models.py,sha256=u8bANPWUUE2hlH8S7ZA2N9bRKuN6vp7vGRpVRwXn-aE,2271
-experiment_configuration_agent-0.1.6.dist-info/METADATA,sha256=y30snhqZgHUF5XXxG8DCPR9-1uQ2XplYdu-h6TEfu-I,4453
-experiment_configuration_agent-0.1.6.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-experiment_configuration_agent-0.1.6.dist-info/top_level.txt,sha256=5c9CyVEjFUlvEf08vJIvi6BkzGuS4wdwtjdmCk2uL2U,24
-experiment_configuration_agent-0.1.6.dist-info/RECORD,,