Orb

Dokumentacija

Tenzorji, nevronske mreže in jezikovni modeli

Vsak operator, ki ga program .lolo lahko pokliče, iz kanoničnega registra, ki ga prevajalnik vgradi.

Na tej strani

Imenski prostori na tej strani: llm/, nn/, tensor/, train/.

llm/

`llm/call-tools`

Call the LLM with tool definitions. Returns the assistant response with optional tool calls and token usage.

lolo

(llm/call-tools prompt tools)

Parametri

Ime	Tip	Opis
`prompt`	`string`	User prompt
`tools`	`array`	Tool definitions

Vrne: LlmCallToolsResult · Teče na: any

Primer

lolo

(llm/call-tools [{"role":"user","content":"What time is it?"}] [{"name":"getTime","description":"Get current time","parameters":{}}])

`llm/compact`

Compact the context window. Returns before/after token counts.

lolo

(llm/compact strategy)

Parametri

Ime	Tip	Opis
`strategy`	`{'kind': 'union', 'of': [{'kind': 'literal', 'value': 'hybrid'}, {'kind': 'literal', 'value': 'summarize'}, {'kind': 'literal', 'value': 'truncate'}, {'kind': 'literal', 'value': 'extract'}]}`	Compaction strategy

Vrne: object · Teče na: any

Primer

lolo

(llm/compact) // => {before: 10000 after: 8000}

`llm/embed`

Generate embeddings for an array of texts. Returns a 2D array of floats.

lolo

(llm/embed texts)

Parametri

Ime	Tip	Opis
`texts`	`array`	Texts to embed

Vrne: array · Teče na: any

Primer

lolo

(llm/embed ["hello" "world"])

`llm/generate`

Generate text from an LLM. Returns the generated string.

lolo

(llm/generate prompt)

Parametri

Ime	Tip	Opis
`prompt`	`string`	Generation prompt

Vrne: string · Teče na: any

Primer

lolo

(llm/generate "Write a haiku") // => "Spring rain falls gently"

`llm/switch`

Switch the active LLM provider and optionally the model.

lolo

(llm/switch provider)

Parametri

Ime	Tip	Opis
`provider`	`string`	LLM provider name

Vrne: void · Teče na: any

Primer

lolo

(llm/switch "claude")

`llm/token-count`

Get the current token count in the context window.

lolo

(llm/token-count)

Vrne: number

Primer

lolo

(llm/token-count) // => 5432

nn/

`nn/batchnorm`

Batch normalization layer

lolo

(nn/batchnorm input)

Parametri

Ime	Tip	Opis
`input`	`any`	Input layer or config

Vrne: nn/layer

Primer

lolo

(nn/batchnorm 64)

`nn/clone`

Create a deep copy of the network with same weights

lolo

(nn/clone module)

Parametri

Ime	Tip	Opis
`module`	`any`	Neural network module

Vrne: nn/module

Primer

lolo

(nn/clone {"layers":[{"type":"dense","units":2}]})

`nn/dropout`

Dropout layer for regularization (randomly zeros elements during training)

lolo

(nn/dropout p)

Parametri

Ime	Tip	Opis
`p`	`number`	Dropout probability

Vrne: nn/layer

Primer

lolo

(nn/dropout)

`nn/forward`

Execute forward pass through network

lolo

(nn/forward module input)

Parametri

Ime	Tip	Opis
`module`	`any`	Neural network module
`input`	`any`	Input tensor

Vrne: tensor

Primer

lolo

(nn/forward {"layers":[{"type":"dense","units":2}]} [0.1,0.2])

`nn/getWeights`

Get network weights as a flat tensor

lolo

(nn/getWeights module)

Parametri

Ime	Tip	Opis
`module`	`any`	Neural network module

Vrne: tensor

Primer

lolo

(nn/getWeights {"layers":[{"type":"dense","units":2}]})

`nn/layernorm`

Layer normalization

lolo

(nn/layernorm input)

Parametri

Ime	Tip	Opis
`input`	`any`	Input layer or config

Vrne: nn/layer

Primer

lolo

(nn/layernorm 64)

`nn/linear`

Fully connected linear layer (output = input * weights + bias)

lolo

(nn/linear inSize outSize)

Parametri

Ime	Tip	Opis
`inSize`	`number`	Input dimension
`outSize`	`number`	Output dimension

Vrne: nn/layer

Primer

lolo

(nn/linear 128 64)

`nn/paramCount`

Get total number of trainable parameters

lolo

(nn/paramCount module)

Parametri

Ime	Tip	Opis
`module`	`any`	Neural network module

Vrne: number

Primer

lolo

(nn/paramCount {"layers":[{"type":"dense","units":2}]})

`nn/relu`

ReLU activation function: max(0, x)

lolo

(nn/relu)

Vrne: nn/layer

Primer

lolo

(nn/relu)

`nn/sequential`

Create a sequential neural network from layers

lolo

(nn/sequential layer)

Parametri

Ime	Tip	Opis
`layer`	`any`	First layer (variadic)

Vrne: nn/module

Primer

lolo

(nn/sequential ["nn/linear",16,64] ["nn/relu"] ["nn/linear",64,4])

`nn/setWeights`

Set network weights from a flat tensor

lolo

(nn/setWeights module weights)

Parametri

Ime	Tip	Opis
`module`	`any`	Neural network module
`weights`	`any`	Weight tensor (variadic)

Vrne: nn/module · Teče na: any

Primer

lolo

(nn/setWeights {"layers":[{"type":"dense","units":2}]} [0.1,0.2,0.3])

`nn/sigmoid`

Sigmoid activation function: 1 / (1 + e^-x)

lolo

(nn/sigmoid)

Vrne: nn/layer

Primer

lolo

(nn/sigmoid)

`nn/softmax`

Softmax activation function (normalizes to probability distribution)

lolo

(nn/softmax dim)

Parametri

Ime	Tip	Opis
`dim`	`number`	Dimension to apply softmax

Vrne: nn/layer

Primer

lolo

(nn/softmax)

`nn/tanh`

Tanh activation function: (e^x - e^-x) / (e^x + e^-x)

lolo

(nn/tanh)

Vrne: nn/layer

Primer

lolo

(nn/tanh)

tensor/

`tensor/add`

Element-wise addition

lolo

(tensor/add a b)

Parametri

Ime	Tip	Opis
`a`	`any`	First tensor
`b`	`any`	Second tensor

Vrne: tensor

Primer

lolo

(tensor/add [1,2] [3,4])

`tensor/allInRange`

Check if all elements are within range [min, max]

lolo

(tensor/allInRange tensor range)

Parametri

Ime	Tip	Opis
`tensor`	`any`	Tensor to check
`range`	`array`	Range [min, max]

Vrne: boolean

Primer

lolo

(tensor/allInRange [0.5,-0.2] [-1.0,1.0]) // => true

`tensor/argmax`

Index of maximum value

lolo

(tensor/argmax tensor)

Parametri

Ime	Tip	Opis
`tensor`	`any`	Tensor to search (variadic)

Vrne: number | tensor

Primer

lolo

(tensor/argmax [0.1,0.9,0.3]) // => 1

`tensor/cat`

Concatenate two tensors along the last dimension

lolo

(tensor/cat a b)

Parametri

Ime	Tip	Opis
`a`	`tensor`	First tensor
`b`	`tensor`	Second tensor

Vrne: tensor

Primer

lolo

(tensor/cat [1,2] [3,4])

`tensor/clamp`

Clamp all elements to range [min, max]

lolo

(tensor/clamp t min max)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to clamp
`min`	`number`	Minimum value
`max`	`number`	Maximum value

Vrne: tensor

Primer

lolo

(tensor/clamp [5,-15,3] -10.0 10.0)

`tensor/clampPerDim`

Clamp each dimension to its specified range

lolo

(tensor/clampPerDim t ranges)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to clamp
`ranges`	`object`	Per-dimension range specifications

Vrne: tensor

Primer

lolo

(tensor/clampPerDim [5,-15] {"0":{"min":0,"max":10},"1":{"min":-10,"max":10}})

`tensor/div`

Element-wise division

lolo

(tensor/div a b)

Parametri

Ime	Tip	Opis
`a`	`tensor`	Dividend tensor
`b`	`tensor`	Divisor tensor

Vrne: tensor

Primer

lolo

(tensor/div [10,20] 2)

`tensor/dot`

Dot product of two 1D tensors

lolo

(tensor/dot a b)

Parametri

Ime	Tip	Opis
`a`	`tensor`	First 1D tensor
`b`	`tensor`	Second 1D tensor

Vrne: number

Primer

lolo

(tensor/dot [1,2] [3,4]) // => 11

`tensor/flatten`

Flatten tensor to 1D

lolo

(tensor/flatten t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to flatten

Vrne: tensor

Primer

lolo

(tensor/flatten [[1,2],[3,4]])

`tensor/from`

Create tensor from array

lolo

(tensor/from arr)

Parametri

Ime	Tip	Opis
`arr`	`array`	Array to convert

Vrne: tensor

Primer

lolo

(tensor/from [1 2 3 4])

`tensor/get`

Get element at index

lolo

(tensor/get t idx)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to index
`idx`	`number`	Index position

Vrne: number

Primer

lolo

(tensor/get [10,20,30,40] 3) // => 3.14

`tensor/matmul`

Matrix multiplication

lolo

(tensor/matmul a b)

Parametri

Ime	Tip	Opis
`a`	`tensor`	First matrix
`b`	`tensor`	Second matrix

Vrne: tensor

Primer

lolo

(tensor/matmul [[1,2],[3,4]] [[5,6],[7,8]])

`tensor/max`

Maximum value in tensor

lolo

(tensor/max t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor

Vrne: number | tensor

Primer

lolo

(tensor/max [0.1,0.9,0.3]) // => 99

`tensor/mean`

Mean of tensor elements

lolo

(tensor/mean t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor

Vrne: number | tensor

Primer

lolo

(tensor/mean [0.2,0.4,0.6]) // => 50

`tensor/min`

Minimum value in tensor

lolo

(tensor/min t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor

Vrne: number | tensor

Primer

lolo

(tensor/min [3,1,2]) // => 1

`tensor/mul`

Element-wise multiplication

lolo

(tensor/mul a b)

Parametri

Ime	Tip	Opis
`a`	`tensor`	First tensor
`b`	`tensor`	Second tensor

Vrne: tensor

Primer

lolo

(tensor/mul [0.5,0.7] 0.99)

`tensor/norm`

L2 norm of tensor

lolo

(tensor/norm t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor

Vrne: number

Primer

lolo

(tensor/norm [3,4]) // => 5.2

`tensor/ones`

Create tensor filled with ones

lolo

(tensor/ones shape)

Parametri

Ime	Tip	Opis
`shape`	`array`	Tensor shape

Vrne: tensor

Primer

lolo

(tensor/ones [2 3])

`tensor/outOfRangeDims`

Get dimensions that exceed their specified ranges

lolo

(tensor/outOfRangeDims t ranges)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to check
`ranges`	`object`	Dimension range specifications

Vrne: array

Primer

lolo

(tensor/outOfRangeDims [5,-15] {"0":{"min":0,"max":10},"1":{"min":-10,"max":10}}) // => [0]

`tensor/outOfRangeIndices`

Get indices of elements outside range

lolo

(tensor/outOfRangeIndices t ranges)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to check
`ranges`	`object`	Value range specifications

Vrne: number[]

Primer

lolo

(tensor/outOfRangeIndices [0.5,1.5] [-1.0,1.0]) // => [0 2]

`tensor/rand`

Create tensor with random values in [0, 1)

lolo

(tensor/rand shape)

Parametri

Ime	Tip	Opis
`shape`	`array`	Tensor shape

Vrne: tensor

Primer

lolo

(tensor/rand [2 3])

`tensor/randn`

Create tensor with random values from standard normal distribution

lolo

(tensor/randn shape)

Parametri

Ime	Tip	Opis
`shape`	`array`	Tensor shape

Vrne: tensor

Primer

lolo

(tensor/randn [2 3])

`tensor/reshape`

Reshape tensor to new shape (total elements must match)

lolo

(tensor/reshape t shape)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to reshape
`shape`	`array`	New shape

Vrne: tensor

Primer

lolo

(tensor/reshape [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16] [4,4])

`tensor/shape`

Get tensor shape as array

lolo

(tensor/shape t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor

Vrne: number[]

Primer

lolo

(tensor/shape [[1,2],[3,4]]) // => [3 4 5]

`tensor/slice`

Get slice of tensor

lolo

(tensor/slice t start end)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to slice
`start`	`number`	Start index
`end`	`number`	End index

Vrne: tensor

Primer

lolo

(tensor/slice [10,20,30,40] 0 3)

`tensor/squeeze`

Remove all dimensions of size one

lolo

(tensor/squeeze t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to squeeze

Vrne: tensor

Primer

lolo

(tensor/squeeze [[1,2,3]])

`tensor/stack`

Stack two tensors along a new leading dimension

lolo

(tensor/stack a b)

Parametri

Ime	Tip	Opis
`a`	`tensor`	First tensor
`b`	`tensor`	Second tensor

Vrne: tensor

Primer

lolo

(tensor/stack [1,2] [3,4])

`tensor/sub`

Element-wise subtraction

lolo

(tensor/sub a b)

Parametri

Ime	Tip	Opis
`a`	`tensor`	Minuend tensor
`b`	`tensor`	Subtrahend tensor

Vrne: tensor

Primer

lolo

(tensor/sub [1,2] [3,4])

`tensor/sum`

Sum of tensor elements

lolo

(tensor/sum t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor

Vrne: number | tensor

Primer

lolo

(tensor/sum [1,0,1]) // => 500

`tensor/toArray`

Convert tensor to nested array

lolo

(tensor/toArray t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to convert

Vrne: array

Primer

lolo

(tensor/toArray [[1,2],[3,4]]) // => [[1 2] [3 4]]

`tensor/toList`

Convert 1D tensor to flat array

lolo

(tensor/toList t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	1D tensor

Vrne: number[]

Primer

lolo

(tensor/toList [1,2,3]) // => [1 2 3]

`tensor/transpose`

Transpose a tensor (optionally swapping two named dimensions)

lolo

(tensor/transpose t)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to transpose

Vrne: tensor

Primer

lolo

(tensor/transpose [[1,2],[3,4]] 0 1)

`tensor/unsqueeze`

Insert a dimension of size one at the given position

lolo

(tensor/unsqueeze t dim)

Parametri

Ime	Tip	Opis
`t`	`tensor`	Tensor to unsqueeze
`dim`	`number`	Dimension position

Vrne: tensor

Primer

lolo

(tensor/unsqueeze [1,2,3] 0)

`tensor/zeros`

Create tensor filled with zeros

lolo

(tensor/zeros shape)

Parametri

Ime	Tip	Opis
`shape`	`array`	Tensor shape

Vrne: tensor

Primer

lolo

(tensor/zeros [2 3])

train/

`train/adam`

Adam optimizer step

lolo

(train/adam params grads)

Parametri

Ime	Tip	Opis
`params`	`object`	Model parameters
`grads`	`object`	Gradient values

Vrne: void · Teče na: any

Primer

lolo

(train/adam {"layers":[{"type":"dense","units":2}]} 0.001)

`train/checkConstraints`

Check if network weights satisfy all constraints

lolo

(train/checkConstraints weights constraints)

Parametri

Ime	Tip	Opis
`weights`	`object`	Network weights
`constraints`	`object`	Constraint specifications

Vrne: train/constraintResult

Primer

lolo

(train/checkConstraints [0.1,0.2] {"maxMagnitude":10})

`train/checkForbiddenOutputs`

Check if model produces outputs in forbidden regions

lolo

(train/checkForbiddenOutputs model outputs ranges)

Parametri

Ime	Tip	Opis
`model`	`object`	Model
`outputs`	`array`	Model outputs
`ranges`	`object`	Forbidden output ranges

Vrne: train/forbiddenResult

Primer

lolo

(train/checkForbiddenOutputs {"layers":[{"type":"dense","units":2}]} [[0],[1]] {"0":{"min":100,"max":200}})

`train/checkRegression`

Check if new model regresses on required invariants

lolo

(train/checkRegression oldModel newModel testData)

Parametri

Ime	Tip	Opis
`oldModel`	`object`	Old model
`newModel`	`object`	New model
`testData`	`array`	Test dataset

Vrne: train/regressionResult

Primer

lolo

(train/checkRegression [0.1,0.2] {"layers":[{"type":"dense","units":2}]} [{"input":[0],"expected":[1]}])

`train/checkWeightMagnitude`

Check if all weights are within magnitude limit

lolo

(train/checkWeightMagnitude weights maxMag)

Parametri

Ime	Tip	Opis
`weights`	`object`	Network weights
`maxMag`	`number`	Maximum magnitude

Vrne: boolean

Primer

lolo

(train/checkWeightMagnitude {"layers":[{"type":"dense","units":2}]} 10.0) // => true

`train/clipGradients`

Clip gradients to max norm (modifies in place)

lolo

(train/clipGradients grads maxNorm)

Parametri

Ime	Tip	Opis
`grads`	`object`	Gradients to clip
`maxNorm`	`number`	Maximum norm

Vrne: number · Teče na: any

Primer

lolo

(train/clipGradients {"layers":[{"type":"dense","units":2}]} 1.0)

`train/clipWeights`

Clip weights to max magnitude (modifies in place)

lolo

(train/clipWeights weights maxMag)

Parametri

Ime	Tip	Opis
`weights`	`object`	Weights to clip
`maxMag`	`number`	Maximum magnitude

Vrne: void · Teče na: any

Primer

lolo

(train/clipWeights {"layers":[{"type":"dense","units":2}]} 10.0)

`train/computeAdvantages`

Compute GAE advantages for policy gradient

lolo

(train/computeAdvantages rewards baseline gamma)

Parametri

Ime	Tip	Opis
`rewards`	`array`	Reward trajectory
`baseline`	`array`	Baseline values
`gamma`	`number`	Discount factor

Vrne: tensor

Primer

lolo

(train/computeAdvantages [1,0,1] [0.5,0.4,0.3] {"gamma":0.99,"lambda":0.95})

`train/computeReturns`

Compute discounted returns from rewards

lolo

(train/computeReturns rewards gamma)

Parametri

Ime	Tip	Opis
`rewards`	`array`	Reward trajectory
`gamma`	`number`	Discount factor

Vrne: tensor

Primer

lolo

(train/computeReturns [1,0,1] 0.99)

`train/crossEntropy`

Cross-entropy loss for classification

lolo

(train/crossEntropy predictions targets)

Parametri

Ime	Tip	Opis
`predictions`	`array`	Predicted probabilities
`targets`	`array`	Target labels

Vrne: number

Primer

lolo

(train/crossEntropy [0.1 0.9] [0 1]) // => 0.105

`train/getGradientNorm`

Get current gradient norm

lolo

(train/getGradientNorm gradients)

Parametri

Ime	Tip	Opis
`gradients`	`object`	Gradient structure

Vrne: number

Primer

lolo

(train/getGradientNorm {"layers":[{"type":"dense","units":2}]})

`train/getMaxWeightMagnitude`

Get maximum weight magnitude in network

lolo

(train/getMaxWeightMagnitude model)

Parametri

Ime	Tip	Opis
`model`	`object`	Neural network model

Vrne: number

Primer

lolo

(train/getMaxWeightMagnitude {"layers":[{"type":"dense","units":2}]})

`train/huber`

Huber loss (smooth L1, robust to outliers)

lolo

(train/huber predicted target)

Parametri

Ime	Tip	Opis
`predicted`	`number`	Predicted value
`target`	`number`	Target value (variadic)

Vrne: number

Primer

lolo

(train/huber 0.5 0.7 1.0)

`train/loop`

Execute training loop with constraints

lolo

(train/loop model config callback)

Parametri

Ime	Tip	Opis
`model`	`object`	Neural network model
`config`	`object`	Training configuration
`callback`	`lambda`	Epoch callback

Vrne: train/result · Teče na: any

Primer

lolo

(train/loop {"layers":[{"type":"dense","units":2}]} [{"state":[0],"action":1,"reward":1}] {"epochs":10})

`train/mse`

Mean squared error loss

lolo

(train/mse predicted target)

Parametri

Ime	Tip	Opis
`predicted`	`array`	Predicted values
`target`	`array`	Target values

Vrne: number

Primer

lolo

(train/mse [1.0 2.0] [1.1 2.1]) // => 0.01

`train/sampleBatch`

Sample random batch from experience buffer

lolo

(train/sampleBatch buffer batch_size)

Parametri

Ime	Tip	Opis
`buffer`	`array`	Experience buffer
`batch_size`	`number`	Batch size

Vrne: array

Primer

lolo

(train/sampleBatch [{"state":[0],"reward":1},{"state":[1],"reward":0}] 2)

`train/sgd`

Stochastic gradient descent optimizer step

lolo

(train/sgd gradients learning_rate)

Parametri

Ime	Tip	Opis
`gradients`	`object`	Gradient structure
`learning_rate`	`number`	Learning rate (variadic)

Vrne: void · Teče na: any

Primer

lolo

(train/sgd {"layers":[{"type":"dense","units":2}]} 0.01 0.9)

`train/step`

Execute single training step (forward, loss, backward, update)

lolo

(train/step model input target optimizer)

Parametri

Ime	Tip	Opis
`model`	`object`	Neural network model
`input`	`any`	Input data
`target`	`any`	Target data
`optimizer`	`object`	Optimizer state

Vrne: train/stepResult · Teče na: any

Primer

lolo

(train/step {"layers":[{"type":"dense","units":2}]} [0.1,0.2] [1] {"lr":0.01})

`train/validate`

Validate model on test cases, returns pass/fail metrics

lolo

(train/validate model test_data)

Parametri

Ime	Tip	Opis
`model`	`object`	Neural network model
`test_data`	`array`	Test dataset

Vrne: train/validationResult

Primer

lolo

(train/validate {"layers":[{"type":"dense","units":2}]} [{"input":[0],"expected":[1]}])

Prostor

Gostitelj

Orb

Jezik, v katerem je pravilo program.

Učenje

Začni Dokumentacija Blog

Orodja

Peskovnik Standardna knjižnica Raziščite Igrajte Prenosi Studio

Skupnost

GitHub Discord Almadar