AaronTekle commited on
Commit
d989a5c
·
verified ·
1 Parent(s): 95911a9

Update config.py

Browse files
Files changed (1) hide show
  1. config.py +30 -4
config.py CHANGED
@@ -1,36 +1,62 @@
1
- # uses postponed evaluation of type annotations for forward-compatible typing behavior
2
  from __future__ import annotations
 
3
  # Provides access to environment variables used to configure the application at runtime
4
  import os
5
 
6
  # defines the display title used by the ML app
7
  APP_TITLE = "Agentic Machine Learning"
 
8
  # reads the Hugging Face authentication token from the environment and removes surrounding whitespace
9
  HF_TOKEN = os.getenv("HF_TOKEN", "").strip()
 
10
  # selects the HF model ID while providing a default model when none is configured
11
- HF_MODEL_ID = os.getenv("HF_MODEL_ID", "Qwen/Qwen3-Coder-30B-A3B-Instruct").strip()
 
 
 
 
12
  # selects the HF inference provider and falls back to automatic provider selection
13
  HF_PROVIDER = os.getenv("HF_PROVIDER", "auto").strip() or "auto"
14
 
15
  # sets the max allowed upload size in megabytes from an environment variable or default value
16
  MAX_UPLOAD_MB = int(os.getenv("MAX_UPLOAD_MB", "50"))
 
17
  # limits the number of rows used when profiling uploaded datasets
18
  MAX_PROFILE_ROWS = int(os.getenv("MAX_PROFILE_ROWS", "100000"))
 
19
  # limits the number of rows used during model training to control runtime and resource usage
20
  MAX_TRAIN_ROWS = int(os.getenv("MAX_TRAIN_ROWS", "25000"))
 
21
  # max number of tool-driven reasoning steps the agent may execute
22
  MAX_AGENT_STEPS = int(os.getenv("MAX_AGENT_STEPS", "4"))
 
23
  # default max number of tokens allowed in generated model responses
24
  DEFAULT_MAX_TOKENS = int(os.getenv("DEFAULT_MAX_TOKENS", "1800"))
 
25
  # deterministic random seed
26
  RANDOM_STATE = int(os.getenv("RANDOM_STATE", "42"))
27
 
 
 
 
 
 
28
  # lists the dataset file extensions accepted by the app
29
- SUPPORTED_DATA_EXTENSIONS = {".csv", ".parquet", ".json", ".jsonl", ".xlsx", ".xls"}
 
 
 
 
 
 
 
30
 
31
  # HF repo containing the bundled example dataset
32
  EXAMPLE_DATASET_REPO = "scikit-learn/adult-census-income"
 
33
  # filename of the example dataset within the repo
34
  EXAMPLE_DATASET_FILE = "adult.csv"
 
35
  # target column used for the example ML modeling task
36
- EXAMPLE_DATASET_TARGET = "income"
 
1
+ # Enables postponed evaluation of type annotations for forward-compatible typing behavior
2
  from __future__ import annotations
3
+
4
  # Provides access to environment variables used to configure the application at runtime
5
  import os
6
 
7
  # defines the display title used by the ML app
8
  APP_TITLE = "Agentic Machine Learning"
9
+
10
  # reads the Hugging Face authentication token from the environment and removes surrounding whitespace
11
  HF_TOKEN = os.getenv("HF_TOKEN", "").strip()
12
+
13
  # selects the HF model ID while providing a default model when none is configured
14
+ HF_MODEL_ID = os.getenv(
15
+ "HF_MODEL_ID",
16
+ "Qwen/Qwen3-Coder-30B-A3B-Instruct",
17
+ ).strip()
18
+
19
  # selects the HF inference provider and falls back to automatic provider selection
20
  HF_PROVIDER = os.getenv("HF_PROVIDER", "auto").strip() or "auto"
21
 
22
  # sets the max allowed upload size in megabytes from an environment variable or default value
23
  MAX_UPLOAD_MB = int(os.getenv("MAX_UPLOAD_MB", "50"))
24
+
25
  # limits the number of rows used when profiling uploaded datasets
26
  MAX_PROFILE_ROWS = int(os.getenv("MAX_PROFILE_ROWS", "100000"))
27
+
28
  # limits the number of rows used during model training to control runtime and resource usage
29
  MAX_TRAIN_ROWS = int(os.getenv("MAX_TRAIN_ROWS", "25000"))
30
+
31
  # max number of tool-driven reasoning steps the agent may execute
32
  MAX_AGENT_STEPS = int(os.getenv("MAX_AGENT_STEPS", "4"))
33
+
34
  # default max number of tokens allowed in generated model responses
35
  DEFAULT_MAX_TOKENS = int(os.getenv("DEFAULT_MAX_TOKENS", "1800"))
36
+
37
  # deterministic random seed
38
  RANDOM_STATE = int(os.getenv("RANDOM_STATE", "42"))
39
 
40
+ # Maximum ZeroGPU allocation for the decorated agent callback.
41
+ # Hugging Face ZeroGPU defaults to 60 seconds; this project allows more time
42
+ # because the agent can execute several bounded tool/inference steps.
43
+ ZERO_GPU_DURATION = int(os.getenv("ZERO_GPU_DURATION", "120"))
44
+
45
  # lists the dataset file extensions accepted by the app
46
+ SUPPORTED_DATA_EXTENSIONS = {
47
+ ".csv",
48
+ ".parquet",
49
+ ".json",
50
+ ".jsonl",
51
+ ".xlsx",
52
+ ".xls",
53
+ }
54
 
55
  # HF repo containing the bundled example dataset
56
  EXAMPLE_DATASET_REPO = "scikit-learn/adult-census-income"
57
+
58
  # filename of the example dataset within the repo
59
  EXAMPLE_DATASET_FILE = "adult.csv"
60
+
61
  # target column used for the example ML modeling task
62
+ EXAMPLE_DATASET_TARGET = "income"