Finetune Sh

as donto believed on

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-05-02.)

Finetune Sh

Adapts8 Layers	8^[1]
Changes Dir to	$(dirname "$0")^[1]
Checks File Existence	Train Jsonl^[1]
Counts Test Examples	wc -l < "$DATA_DIR/test.jsonl"^[1]
Counts Train Examples	wc -l < "$DATA_DIR/train.jsonl"^[1]
Counts Val Examples	wc -l < "$DATA_DIR/valid.jsonl"^[1]
Defaults Iters to	600^[1]
Echoes Iters	$ITERS^[1]

ex:finetune-sh

Finetune Sh has 40 facts recorded in Dontopedia across 1 reference.

40 facts·40 predicates·1 sources

Mostly:adapts8 layers(1), changes dir to(1), checks file existence(1)

Adapts8 Layers`adapts8Layers`

8^[1]all time · Part 30

Checks File Existence`checksFileExistence`

Train Jsonl^[1]all time · Part 30

Counts Test Examples`countsTestExamples`

wc -l < "$DATA_DIR/test.jsonl"^[1]all time · Part 30

Counts Train Examples`countsTrainExamples`

wc -l < "$DATA_DIR/train.jsonl"^[1]all time · Part 30

Counts Val Examples`countsValExamples`

wc -l < "$DATA_DIR/valid.jsonl"^[1]all time · Part 30

Defaults Iters to`defaultsItersTo`

600^[1]all time · Part 30

Echoes Iters`echoesIters`

$ITERS^[1]all time · Part 30

Echoes Model`echoesModel`

Smollm2 360m Instruct^[1]all time · Part 30

Has Purpose`hasPurpose`

Fine-tune SmolLM2-360M on SafierSemantics codebase using MLX LoRA^[1]all time · Part 30

Has Usage Instruction`hasUsageInstruction`

./finetune.sh [iters]^[1]all time · Part 30

Inbound mentions (1)

Other subjects in dontopedia point AT this entity as a value. These are inverse relationships — e.g. “X motherOf this subject” — and answer questions the forward facts can't. Grouped by predicate.

`postedScript`Posted Script(1)

Xenonfunex:xenonfun

Other facts (28)

The long tail: predicates that appear too rarely to warrant their own section. Filter or scroll to find a specific one. Each row links to its source.

28 facts

Predicate	Value	Ref
Instructs Save Adapters to	Adapters Dir	[1]
Invokes Command	uvx --from mlx-lm mlx_lm.lora	[1]
Is Bash Script	#/bin/bash	[1]
Passes Data Arg	Data Dir	[1]
Passes Model Arg	Smollm2 360m Instruct	[1]
Prints Completion	=== Training complete ===	[1]
Prints Data Stats Header	=== Training Data ===	[1]
Prints Header	=== Preparing training data ===	[1]
Prints Lo Ra Header	=== LoRA Fine-tuning ===	[1]
Recommends Fuse Script	Fuse Sh	[1]
Recommends Test Script	Test Generate Sh	[1]
Runs Script If Needed	Prepare Data Py	[1]
Sets Adapter Dir	./adapters	[1]
Sets Adapter Path	Adapters Dir	[1]
Sets Batch Size	2	[1]
Sets Data Dir	./data	[1]
Sets Iters Arg	$ITERS	[1]
Sets Learning Rate	0.00001	[1]
Sets Max Seq Length	2048	[1]
Sets Num Layers	8	[1]
Sets Steps Per Eval	50	[1]
Sets Steps Per Report	10	[1]
Sets Train Flag	true	[1]
Sets Variable	ITERS	[1]
Targets Instruct Model	Smollm2 360m Instruct	[1]
Uses Batch Size2	2	[1]
Uses Model	Smollm2 360m Instruct	[1]
Assumes Python3	true	[1]

Timeline axis is valid_time — when each source says the fact was true in the world, not when Dontopedia learned about it. Retracted rows are kept for provenance; coloured stripes indicate the context kind.

40 undated

—

adapts8Layersblah/random/part-30

—

assumesPython3blah/random/part-30

true

—

changesDirToblah/random/part-30

$(dirname "$0")

—

checksFileExistenceblah/random/part-30

ex:train-jsonl

—

countsTestExamplesblah/random/part-30

wc -l < "$DATA_DIR/test.jsonl"

—

countsTrainExamplesblah/random/part-30

wc -l < "$DATA_DIR/train.jsonl"

—

countsValExamplesblah/random/part-30

wc -l < "$DATA_DIR/valid.jsonl"

—

defaultsItersToblah/random/part-30

600

—

echoesItersblah/random/part-30

$ITERS

—

echoesModelblah/random/part-30

ex:smollm2-360m-instruct

—

exitsOnErrorblah/random/part-30

true

—

hasPurposeblah/random/part-30

Fine-tune SmolLM2-360M on SafierSemantics codebase using MLX LoRA

—

hasUsageInstructionblah/random/part-30

./finetune.sh [iters]

—

instructsSaveAdaptersToblah/random/part-30

ex:adapters-dir

—

invokesCommandblah/random/part-30

uvx --from mlx-lm mlx_lm.lora

—

isBashScriptblah/random/part-30

#/bin/bash

—

passesDataArgblah/random/part-30

ex:data-dir

—

passesModelArgblah/random/part-30

ex:smollm2-360m-instruct

—

printsCompletionblah/random/part-30

=== Training complete ===

—

printsDataStatsHeaderblah/random/part-30

=== Training Data ===

—

printsHeaderblah/random/part-30

=== Preparing training data ===

—

printsLoRAHeaderblah/random/part-30

=== LoRA Fine-tuning ===

—

recommendsFuseScriptblah/random/part-30

ex:fuse-sh

—

recommendsTestScriptblah/random/part-30

ex:test-generate-sh

—

runsScriptIfNeededblah/random/part-30

ex:prepare-data-py

—

setsAdapterDirblah/random/part-30

./adapters

—

setsAdapterPathblah/random/part-30

ex:adapters-dir

—

setsBatchSizeblah/random/part-30

—

setsDataDirblah/random/part-30

./data

—

setsItersArgblah/random/part-30

$ITERS

—

setsLearningRateblah/random/part-30

0.00001

—

setsMaxSeqLengthblah/random/part-30

2048

—

setsNumLayersblah/random/part-30

—

setsStepsPerEvalblah/random/part-30

—

setsStepsPerReportblah/random/part-30

—

setsTrainFlagblah/random/part-30

true

—

setsVariableblah/random/part-30

ITERS

—

targetsInstructModelblah/random/part-30

ex:smollm2-360m-instruct

—

usesBatchSize2blah/random/part-30

—

usesModelblah/random/part-30

ex:smollm2-360m-instruct

References (1)

1 references

[1]Part 3040 facts
customctx:discord/blah/random/part-30

Keep researching

Dontopedia is in a read-only public launch. Follow the references and disputed branches now; contributions will open after durable identity and moderation are in place.

Existing claims and sources remain available while the prepaid research lane recharges.

Finetune Sh

as donto believed on

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-05-02.)

Finetune Sh has 40 facts recorded in Dontopedia across 1 reference.

40 facts·40 predicates·1 sources

Adapts8 Layers`adapts8Layers`

8^[1]all time · Part 30

Changes Dir to`changesDirTo`

$(dirname "$0")^[1]all time · Part 30

Checks File Existence`checksFileExistence`

Train Jsonl^[1]all time · Part 30

Counts Test Examples`countsTestExamples`

wc -l < "$DATA_DIR/test.jsonl"^[1]all time · Part 30

Counts Train Examples`countsTrainExamples`

wc -l < "$DATA_DIR/train.jsonl"^[1]all time · Part 30

Counts Val Examples`countsValExamples`

wc -l < "$DATA_DIR/valid.jsonl"^[1]all time · Part 30

Defaults Iters to`defaultsItersTo`

600^[1]all time · Part 30

Echoes Iters`echoesIters`

$ITERS^[1]all time · Part 30

Echoes Model`echoesModel`

Smollm2 360m Instruct^[1]all time · Part 30

Exits on Error`exitsOnError`

true^[1]all time · Part 30

Has Purpose`hasPurpose`

Fine-tune SmolLM2-360M on SafierSemantics codebase using MLX LoRA^[1]all time · Part 30

Has Usage Instruction`hasUsageInstruction`

./finetune.sh [iters]^[1]all time · Part 30

Inbound mentions (1)

`postedScript`Posted Script(1)

Xenonfunex:xenonfun

Other facts (28)

The long tail: predicates that appear too rarely to warrant their own section. Filter or scroll to find a specific one. Each row links to its source.

28 facts

Predicate	Value	Ref
Instructs Save Adapters to	Adapters Dir	[1]
Invokes Command	uvx --from mlx-lm mlx_lm.lora	[1]
Is Bash Script	#/bin/bash	[1]
Passes Data Arg	Data Dir	[1]
Passes Model Arg	Smollm2 360m Instruct	[1]
Prints Completion	=== Training complete ===	[1]
Prints Data Stats Header	=== Training Data ===	[1]
Prints Header	=== Preparing training data ===	[1]
Prints Lo Ra Header	=== LoRA Fine-tuning ===	[1]
Recommends Fuse Script	Fuse Sh	[1]
Recommends Test Script	Test Generate Sh	[1]
Runs Script If Needed	Prepare Data Py	[1]
Sets Adapter Dir	./adapters	[1]
Sets Adapter Path	Adapters Dir	[1]
Sets Batch Size	2	[1]
Sets Data Dir	./data	[1]
Sets Iters Arg	$ITERS	[1]
Sets Learning Rate	0.00001	[1]
Sets Max Seq Length	2048	[1]
Sets Num Layers	8	[1]
Sets Steps Per Eval	50	[1]
Sets Steps Per Report	10	[1]
Sets Train Flag	true	[1]
Sets Variable	ITERS	[1]
Targets Instruct Model	Smollm2 360m Instruct	[1]
Uses Batch Size2	2	[1]
Uses Model	Smollm2 360m Instruct	[1]
Assumes Python3	true	[1]

Timeline

40 undated

—

adapts8Layersblah/random/part-30

—

assumesPython3blah/random/part-30

true

—

changesDirToblah/random/part-30

$(dirname "$0")

—

checksFileExistenceblah/random/part-30

ex:train-jsonl

—

countsTestExamplesblah/random/part-30

wc -l < "$DATA_DIR/test.jsonl"

—

countsTrainExamplesblah/random/part-30

wc -l < "$DATA_DIR/train.jsonl"

—

countsValExamplesblah/random/part-30

wc -l < "$DATA_DIR/valid.jsonl"

—

defaultsItersToblah/random/part-30

600

—

echoesItersblah/random/part-30

$ITERS

—

echoesModelblah/random/part-30

ex:smollm2-360m-instruct

—

exitsOnErrorblah/random/part-30

true

—

hasPurposeblah/random/part-30

Fine-tune SmolLM2-360M on SafierSemantics codebase using MLX LoRA

—

hasUsageInstructionblah/random/part-30

./finetune.sh [iters]

—

instructsSaveAdaptersToblah/random/part-30

ex:adapters-dir

—

invokesCommandblah/random/part-30

uvx --from mlx-lm mlx_lm.lora

—

isBashScriptblah/random/part-30

#/bin/bash

—

passesDataArgblah/random/part-30

ex:data-dir

—

passesModelArgblah/random/part-30

ex:smollm2-360m-instruct

—

printsCompletionblah/random/part-30

=== Training complete ===

—

printsDataStatsHeaderblah/random/part-30

=== Training Data ===

—

printsHeaderblah/random/part-30

=== Preparing training data ===

—

printsLoRAHeaderblah/random/part-30

=== LoRA Fine-tuning ===

—

recommendsFuseScriptblah/random/part-30

ex:fuse-sh

—

recommendsTestScriptblah/random/part-30

ex:test-generate-sh

—

runsScriptIfNeededblah/random/part-30

ex:prepare-data-py

—

setsAdapterDirblah/random/part-30

./adapters

—

setsAdapterPathblah/random/part-30

ex:adapters-dir

—

setsBatchSizeblah/random/part-30

—

setsDataDirblah/random/part-30

./data

—

setsItersArgblah/random/part-30

$ITERS

—

setsLearningRateblah/random/part-30

0.00001

—

setsMaxSeqLengthblah/random/part-30

2048

—

setsNumLayersblah/random/part-30

—

setsStepsPerEvalblah/random/part-30

—

setsStepsPerReportblah/random/part-30

—

setsTrainFlagblah/random/part-30

true

—

setsVariableblah/random/part-30

ITERS

—

targetsInstructModelblah/random/part-30

ex:smollm2-360m-instruct

—

usesBatchSize2blah/random/part-30

—

usesModelblah/random/part-30

ex:smollm2-360m-instruct

References (1)

1 references

[1]Part 3040 facts
customctx:discord/blah/random/part-30

Keep researching

Dontopedia is in a read-only public launch. Follow the references and disputed branches now; contributions will open after durable identity and moderation are in place.

Existing claims and sources remain available while the prepaid research lane recharges.

Finetune Sh

Adapts8 Layers`adapts8Layers`

Changes Dir to`changesDirTo`

Checks File Existence`checksFileExistence`

Counts Test Examples`countsTestExamples`

Counts Train Examples`countsTrainExamples`

Counts Val Examples`countsValExamples`

Defaults Iters to`defaultsItersTo`

Echoes Iters`echoesIters`

Echoes Model`echoesModel`

Exits on Error`exitsOnError`

Has Purpose`hasPurpose`

Has Usage Instruction`hasUsageInstruction`

Inbound mentions (1)

`postedScript`Posted Script(1)

Other facts (28)

Timeline

References (1)

See also

Keep researching

Finetune Sh

Adapts8 Layers`adapts8Layers`

Changes Dir to`changesDirTo`

Checks File Existence`checksFileExistence`

Counts Test Examples`countsTestExamples`

Counts Train Examples`countsTrainExamples`

Counts Val Examples`countsValExamples`

Defaults Iters to`defaultsItersTo`

Echoes Iters`echoesIters`

Echoes Model`echoesModel`

Exits on Error`exitsOnError`

Has Purpose`hasPurpose`

Has Usage Instruction`hasUsageInstruction`

Inbound mentions (1)

`postedScript`Posted Script(1)

Other facts (28)

Timeline

References (1)

See also

Keep researching

Finetune Sh

Adapts8 Layersadapts8Layers

Changes Dir tochangesDirTo

Checks File ExistencechecksFileExistence

Counts Test ExamplescountsTestExamples

Counts Train ExamplescountsTrainExamples

Counts Val ExamplescountsValExamples

Defaults Iters todefaultsItersTo

Echoes ItersechoesIters

Echoes ModelechoesModel

Exits on ErrorexitsOnError

Has PurposehasPurpose

Has Usage InstructionhasUsageInstruction

postedScriptPosted Script(1)

Finetune Sh

Adapts8 Layersadapts8Layers

Changes Dir tochangesDirTo

Checks File ExistencechecksFileExistence

Counts Test ExamplescountsTestExamples

Counts Train ExamplescountsTrainExamples

Counts Val ExamplescountsValExamples

Defaults Iters todefaultsItersTo

Echoes ItersechoesIters

Echoes ModelechoesModel

Exits on ErrorexitsOnError

Has PurposehasPurpose

Has Usage InstructionhasUsageInstruction

postedScriptPosted Script(1)

Adapts8 Layers`adapts8Layers`

Changes Dir to`changesDirTo`

Checks File Existence`checksFileExistence`

Counts Test Examples`countsTestExamples`

Counts Train Examples`countsTrainExamples`

Counts Val Examples`countsValExamples`

Defaults Iters to`defaultsItersTo`

Echoes Iters`echoesIters`

Echoes Model`echoesModel`

Exits on Error`exitsOnError`

Has Purpose`hasPurpose`

Has Usage Instruction`hasUsageInstruction`

`postedScript`Posted Script(1)

Adapts8 Layers`adapts8Layers`

Changes Dir to`changesDirTo`

Checks File Existence`checksFileExistence`

Counts Test Examples`countsTestExamples`

Counts Train Examples`countsTrainExamples`

Counts Val Examples`countsValExamples`

Defaults Iters to`defaultsItersTo`

Echoes Iters`echoesIters`

Echoes Model`echoesModel`

Exits on Error`exitsOnError`

Has Purpose`hasPurpose`

Has Usage Instruction`hasUsageInstruction`

`postedScript`Posted Script(1)