when-debugging-ml-training-use-ml-training-debugger

Name: when-debugging-ml-training-use-ml-training-debugger
Author: DNYoussef

32stars6forks

How to install when-debugging-ml-training-use-ml-training-debugger?

dnyoussef/context-cascade/when-debugging-ml-training-use-ml-training-debugger

$npx -y skills add DNYoussef/context-cascade --skill when-debugging-ml-training-use-ml-training-debugger

Installs into the current project.

Summary

/*============================================================================*/ /* WHEN-DEBUGGING-ML-TRAINING-USE-ML-TRAINING-DEBUGGER SKILL :: VERILINGUA x VERIX EDITION */ /*============================================================================*/

Files · 1

View on GitHub

SKILL.md

1	/============================================================================/
2	/* WHEN-DEBUGGING-ML-TRAINING-USE-ML-TRAINING-DEBUGGER SKILL :: VERILINGUA x VERIX EDITION */
3	/============================================================================/
4
5	---
6	name: when-debugging-ml-training-use-ml-training-debugger
7	version: 1.0.0
8	description: \|
9	[assert\|neutral] Debug ML training issues and optimize performance including loss divergence, overfitting, and slow convergence [ground:given] [conf:0.95] [state:confirmed]
10	category: machine-learning
11	tags:
12	- debugging
13	- ml
14	- training
15	- optimization
16	- troubleshooting
17	author: ruv
18	cognitive_frame:
19	primary: evidential
20	goal_analysis:
21	first_order: "Execute when-debugging-ml-training-use-ml-training-debugger workflow"
22	second_order: "Ensure quality and consistency"
23	third_order: "Enable systematic machine-learning processes"
24	---
25
26	/----------------------------------------------------------------------------/
27	/* S0 META-IDENTITY */
28	/----------------------------------------------------------------------------/
29
30	[define\|neutral] SKILL := {
31	name: "when-debugging-ml-training-use-ml-training-debugger",
32	category: "machine-learning",
33	version: "1.0.0",
34	layer: L1
35	} [ground:given] [conf:1.0] [state:confirmed]
36
37	/----------------------------------------------------------------------------/
38	/* S1 COGNITIVE FRAME */
39	/----------------------------------------------------------------------------/
40
41	[define\|neutral] COGNITIVE_FRAME := {
42	frame: "Evidential",
43	source: "Turkish",
44	force: "How do you know?"
45	} [ground:cognitive-science] [conf:0.92] [state:confirmed]
46
47	## Kanitsal Cerceve (Evidential Frame Activation)
48	Kaynak dogrulama modu etkin.
49
50	/----------------------------------------------------------------------------/
51	/* S2 TRIGGER CONDITIONS */
52	/----------------------------------------------------------------------------/
53
54	[define\|neutral] TRIGGER_POSITIVE := {
55	keywords: ["when-debugging-ml-training-use-ml-training-debugger", "machine-learning", "workflow"],
56	context: "user needs when-debugging-ml-training-use-ml-training-debugger capability"
57	} [ground:given] [conf:1.0] [state:confirmed]
58
59	/----------------------------------------------------------------------------/
60	/* S3 CORE CONTENT */
61	/----------------------------------------------------------------------------/
62
63	## When NOT to Use This Skill
64
65	- Simple data preprocessing without model training
66	- Statistical analysis that does not require ML models
67	- Rule-based systems without learning components
68	- Operations that do not involve model training or inference
69
70	## Success Criteria
71	- [assert\|neutral] Model training convergence: Loss decreasing consistently [ground:acceptance-criteria] [conf:0.90] [state:provisional]
72	- [assert\|neutral] Validation accuracy: Meeting or exceeding baseline targets [ground:acceptance-criteria] [conf:0.90] [state:provisional]
73	- [assert\|neutral] Training time: Within expected bounds for dataset size [ground:acceptance-criteria] [conf:0.90] [state:provisional]
74	- [assert\|neutral] GPU utilization: >80% during training [ground:acceptance-criteria] [conf:0.90] [state:provisional]
75	- [assert\|neutral] Model export success: 100% successful saves [ground:acceptance-criteria] [conf:0.90] [state:provisional]
76	- [assert\|neutral] Inference latency: <100ms for real-time applications [ground:acceptance-criteria] [conf:0.90] [state:provisional]
77
78	## Edge Cases & Error Handling
79
80	- GPU Memory Overflow: Reduce batch size, use gradient accumulation, or mixed precision
81	- Divergent Training: Implement learning rate scheduling, gradient clipping
82	- Data Pipeline Failures: Validate data integrity, handle missing/corrupted files
83	- Version Mismatches: Lock dependency versions, use containerization
84	- Checkpoint Corruption: Save multiple checkpoints, validate before loading
85	- Distributed Training Failures: Handle node failures, implement fault tolerance
86
87	## Guardrails & Safety
88	- [assert\|emphatic] NEVER: train on unvalidated or uncleaned data [ground:policy] [conf:0.98] [state:confirmed]
89	- [assert\|neutral] ALWAYS: validate model outputs before deployment [ground:policy] [conf:0.98] [state:confirmed]
90	- [assert\|neutral] ALWAYS: implement reproducibility (random seeds, version pinning) [ground:policy] [conf:0.98] [state:confirmed]
91	- [assert\|emphatic] NEVER: expose training data in model artifacts or logs [ground:policy] [conf:0.98] [state:confirmed]
92	- [assert\|neutral] ALWAYS: monitor for bias and fairness issues [ground:policy] [conf:0.98] [state:confirmed]
93	- [assert\|neutral] ALWAYS: implement model versioning and rollback capabilities [ground:policy] [conf:0.98] [state:confirmed]
94
95	## Evidence-Based Validation
96
97	- Verify hardware availabil

People who liked this also viewed

6 picks

Type

when-debugging-ml-training-use-ml-training-debugger

32stars6forks

View on GitHub

How to install when-debugging-ml-training-use-ml-training-debugger?

dnyoussef/context-cascade/when-debugging-ml-training-use-ml-training-debugger

$npx -y skills add DNYoussef/context-cascade --skill when-debugging-ml-training-use-ml-training-debugger

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	/============================================================================/
2	/* WHEN-DEBUGGING-ML-TRAINING-USE-ML-TRAINING-DEBUGGER SKILL :: VERILINGUA x VERIX EDITION */
3	/============================================================================/
4
5	---
6	name: when-debugging-ml-training-use-ml-training-debugger
7	version: 1.0.0
8	description: \|
9	[assert\|neutral] Debug ML training issues and optimize performance including loss divergence, overfitting, and slow convergence [ground:given] [conf:0.95] [state:confirmed]
10	category: machine-learning
11	tags:
12	- debugging
13	- ml
14	- training
15	- optimization
16	- troubleshooting
17	author: ruv
18	cognitive_frame:
19	primary: evidential
20	goal_analysis:
21	first_order: "Execute when-debugging-ml-training-use-ml-training-debugger workflow"
22	second_order: "Ensure quality and consistency"
23	third_order: "Enable systematic machine-learning processes"
24	---
25
26	/----------------------------------------------------------------------------/
27	/* S0 META-IDENTITY */
28	/----------------------------------------------------------------------------/
29
30	[define\|neutral] SKILL := {
31	name: "when-debugging-ml-training-use-ml-training-debugger",
32	category: "machine-learning",
33	version: "1.0.0",
34	layer: L1
35	} [ground:given] [conf:1.0] [state:confirmed]
36
37	/----------------------------------------------------------------------------/
38	/* S1 COGNITIVE FRAME */
39	/----------------------------------------------------------------------------/
40
41	[define\|neutral] COGNITIVE_FRAME := {
42	frame: "Evidential",
43	source: "Turkish",
44	force: "How do you know?"
45	} [ground:cognitive-science] [conf:0.92] [state:confirmed]
46
47	## Kanitsal Cerceve (Evidential Frame Activation)
48	Kaynak dogrulama modu etkin.
49
50	/----------------------------------------------------------------------------/
51	/* S2 TRIGGER CONDITIONS */
52	/----------------------------------------------------------------------------/
53
54	[define\|neutral] TRIGGER_POSITIVE := {
55	keywords: ["when-debugging-ml-training-use-ml-training-debugger", "machine-learning", "workflow"],
56	context: "user needs when-debugging-ml-training-use-ml-training-debugger capability"
57	} [ground:given] [conf:1.0] [state:confirmed]
58
59	/----------------------------------------------------------------------------/
60	/* S3 CORE CONTENT */
61	/----------------------------------------------------------------------------/
62
63	## When NOT to Use This Skill
64
65	- Simple data preprocessing without model training
66	- Statistical analysis that does not require ML models
67	- Rule-based systems without learning components
68	- Operations that do not involve model training or inference
69
70	## Success Criteria
71	- [assert\|neutral] Model training convergence: Loss decreasing consistently [ground:acceptance-criteria] [conf:0.90] [state:provisional]
72	- [assert\|neutral] Validation accuracy: Meeting or exceeding baseline targets [ground:acceptance-criteria] [conf:0.90] [state:provisional]
73	- [assert\|neutral] Training time: Within expected bounds for dataset size [ground:acceptance-criteria] [conf:0.90] [state:provisional]
74	- [assert\|neutral] GPU utilization: >80% during training [ground:acceptance-criteria] [conf:0.90] [state:provisional]
75	- [assert\|neutral] Model export success: 100% successful saves [ground:acceptance-criteria] [conf:0.90] [state:provisional]
76	- [assert\|neutral] Inference latency: <100ms for real-time applications [ground:acceptance-criteria] [conf:0.90] [state:provisional]
77
78	## Edge Cases & Error Handling
79
80	- GPU Memory Overflow: Reduce batch size, use gradient accumulation, or mixed precision
81	- Divergent Training: Implement learning rate scheduling, gradient clipping
82	- Data Pipeline Failures: Validate data integrity, handle missing/corrupted files
83	- Version Mismatches: Lock dependency versions, use containerization
84	- Checkpoint Corruption: Save multiple checkpoints, validate before loading
85	- Distributed Training Failures: Handle node failures, implement fault tolerance
86
87	## Guardrails & Safety
88	- [assert\|emphatic] NEVER: train on unvalidated or uncleaned data [ground:policy] [conf:0.98] [state:confirmed]
89	- [assert\|neutral] ALWAYS: validate model outputs before deployment [ground:policy] [conf:0.98] [state:confirmed]
90	- [assert\|neutral] ALWAYS: implement reproducibility (random seeds, version pinning) [ground:policy] [conf:0.98] [state:confirmed]
91	- [assert\|emphatic] NEVER: expose training data in model artifacts or logs [ground:policy] [conf:0.98] [state:confirmed]
92	- [assert\|neutral] ALWAYS: monitor for bias and fairness issues [ground:policy] [conf:0.98] [state:confirmed]
93	- [assert\|neutral] ALWAYS: implement model versioning and rollback capabilities [ground:policy] [conf:0.98] [state:confirmed]
94
95	## Evidence-Based Validation
96
97	- Verify hardware availabil

People who liked this also viewed

6 picks

Type