khtsly/Mini-Coder-v2-GGUF
Image-Text-to-Text β’ 14B β’ Updated β’ 2.29k β’ 2
repo stringclasses 275
values | file_path stringclasses 1
value | text stringlengths 85 274k | tokens int64 22 52.8k | type stringclasses 2
values | source_url stringlengths 17 120 β |
|---|---|---|---|---|---|
imezx/JPSPlus | README.md | # JPSPlus
JPSPlus is a high-performance **2D grid pathfinding** library for Roblox based on **Jump Point Search Plus (JPS+)** algorithm. It preprocesses static maps to enable very fast A* queries by βjumpingβ between critical nodes while preserving **optimal** paths on uniform-cost grids.
**The Concept:**
Traditional... | 368 | readme | null |
EgoMoose/rbx-viewport-window | README.md | # rbx-viewport-window
Huge shout-out to [EmilyBendsSpace](https://x.com/EmilyBendsSpace) as their work is instrumental in ViewportWindows.
ViewportWindow provides developers with the tools to use viewport frames much like you'd use a window in real life. You can look into them and see a different space that is contai... | 246 | readme | null |
EgoMoose/rbx-wallstick | README.md | # rbx-wallstick
A system for sticking characters to surfaces within Roblox.
### Videos
https://github.com/user-attachments/assets/bd4efde2-9323-4db1-896c-6407263e458e
https://github.com/user-attachments/assets/2a0478de-6e1f-4676-b778-9709b9e3f18f
https://github.com/user-attachments/assets/c6d9a53d-f6c2-4924-9286-72... | 120 | readme | null |
EgoMoose/rbx-bufferize | README.md | [bufferize/releases]: https://github.com/EgoMoose/rbx-bufferize/releases
[bufferize/wally]: https://wally.run/package/egomoose/bufferize
[badges/github]: https://raw.githubusercontent.com/gist/cxmeel/0dbc95191f239b631c3874f4ccf114e2/raw/github.svg
[badges/wally]: https://raw.githubusercontent.com/gist/cxmeel/0dbc95191... | 1,879 | readme | null |
AnotherSubatomo/RbxShader | README.md | # RbxShader
A robust, simple, and performant fragment shader engine, for everyone.
The shader engine can run virtually any shader program that operates on a per-pixel basis, just like those on [Shadertoy](https://www.shadertoy.com/). Porting these programs are relatively easy due to their code structuring being simila... | 660 | readme | null |
Sleitnick/RbxObservers | README.md | # RbxObservers
A collection of observer utility functions.
## Installation
### Wally
Add `sleitnick/observers` to dependencies list:
```toml
[package]
name = "your_name/your_project"
version = "0.1.0"
registry = "https://github.com/UpliftGames/wally-index"
realm = "shared"
[dependencies]
Observers = "sleitnick/obs... | 168 | readme | null |
imezx/InfiniteMath | README.md | InfiniteMath is a module that allows you to surpass the double-precision floating-point number limit of `10^308`.
Here's the [installation](https://kdudedev.github.io/InfiniteMath/docs/Installation) page, an in-depth [explanation](https://kdudedev.github.io/InfiniteMath/docs/Explanation) of the module, and API [docume... | 253 | readme | null |
imezx/InfiniteMath | null | # What is this?
InfiniteMath is a module that allows you to surpass the double-precision floating-point number limit which about:
> -10^308 to 10^308
This is normally perfectly fine for games, but sometimes you might want to go past that limit, even just a little bit. InfiniteMath allows you to have practically infi... | 676 | documentation | https://kdudedev.github.io/InfiniteMath/docs/Explanation/ |
imezx/InfiniteMath | null | # Datastore Implementation
InfiniteMath uses metatables, meaning numbers created using it can't be saved in a datastore. If you do save an InfiniteMath number in a datastore, it will lose its metamethods which means no operations, comparisons, etc.
The solution is to recreate the number when loading it using `.new()`... | 115 | documentation | https://kdudedev.github.io/InfiniteMath/docs/Datastore/ |
Ultray-Studios/RBXConnectionManager | README.md | # RBXConnectionManager
This is our first open-source release. Please add a β if you like the idea of this project or if you like to use it.
Roblox Developer Forum Thread: https://devforum.roblox.com/t/3503750
Consider joining our community aswell! https://discord.gg/8ed3W53kHv/ to see progress on our very advanced p... | 799 | readme | null |
Quamatic/rbxts-profile-store | README.md | # @rbxts/profile-store
## Installation:
`npm i @rbxts/profile-store`
## Documentation
See [MadStudioRoblox/ProfileStore](https://github.com/MadStudioRoblox/ProfileStore) | 47 | readme | null |
Shiawaseu/RBLX-Whitelist | README.md | # RBLX-Whitelist
Source repo for my series on writing secure Authentication systems in Roblox
- To read the guide, visit my [website](https://shiawase.dev/blog/rblx-auth1)
## Flowchart
Image Version

Mermaid (interactive) Version
```mermaid
graph TD;
A[User execute... | 500 | readme | null |
Sleitnick/Knit | README.md | ## :warning: No Longer Maintained :warning:
Knit has been archived and will no longer receive updates.
Please [read here](/ARCHIVAL.md) for more information.
# Knit
Knit is a lightweight framework for Roblox that simplifies communication between core parts of your game and seamlessly bridges the gap between the ser... | 880 | readme | null |
imezx/Warp | README.md | # Warp
# A rapidly-fast & powerful networking library.
## Why Warp
### β‘ Performance
Warp is rapidly-fast with much less bandwidth compared to native.
### π Compact
Warp is a simple, efficient & lightweight library.
### π Dynamic
Warp is dynamic by default. It serializes and deserializes data dynamically without... | 118 | readme | null |
imezx/Warp | null | # Warp 1.1 pre-release β
The public main of the Warp library.
WARNING
This version (1.1.x) is not backward compatible with 1.0.x.
## `.Server` server side β
Get the Server operation for server-side.
lua
-- Server
local Server = Warp.Server()
## `.Client` client side β
Get the Client operation for clien... | 135 | documentation | https://imezx.github.io/Warp/api/1.1/warp |
imezx/Warp | null | # Warp 1.0 deprecated β
The public main of the Warp library.
## `.Server` server side β
Create a new event for Server-Side
lua
-- Server
local Event1 = Warp.Server("Event1")
## `.Client` client side β
Create a new event for Client-Side.
lua
-- Client
local Event1 = Warp.Client("Event1") | 85 | documentation | https://imezx.github.io/Warp/api/1.0/warp |
imezx/Warp | null | # Overview β
Warp is a powerful networking library for Roblox that comes with rapidly-fast performance and efficient with Typing to any-scale games.
## Why Warp β
### β‘ Performance β
Warp is rapidly-fast with much less bandwidth compared to native.
### π Compact β
Warp is a simple, efficient & lightweight librar... | 127 | documentation | https://imezx.github.io/Warp/guide/ |
imezx/Warp | null | # Buffer module β
For efficient data serialization and schema definition with optimized packing.
## Getting the Buffer Object β
lua
local Buffer = Warp.Buffer
## Schema System v1.1 β
Define strict data schemas for optimized serialization and type safety.
### Available Schema Types β
lua
-- Basic ty... | 956 | documentation | https://imezx.github.io/Warp/api/1.1/buffer |
imezx/Warp | null | # Client 1.1 β
For Client-sided operations.
## Getting the Client Object β
lua
local Client = Warp.Client()
## `.awaitReady` yield β
Yields the current thread until the client has successfully initialized and synchronized with the server's replication data (identifier).
INFO
Its optionally, but highly recom... | 1,000 | documentation | https://imezx.github.io/Warp/api/1.1/client |
imezx/Warp | null | # Server 1.1 β
For Server-sided operations.
## Getting the Server Object β
lua
local Server = Warp.Server()
## `.reg_namespaces` β
Register namespaces to ensure all of the namespaces is being registered earlier on the server to prevent any unexpected issues on the client.
INFO
this is optional and condition... | 931 | documentation | https://imezx.github.io/Warp/api/1.1/server |
imezx/Warp | null | # Signal utilities β
A alternative of BindableEvent.
## `.Signal` β
Create new Signal.
VariableExample
luau
(
Identifier: string
)
luau
local Signal1 = Warp.Signal("Signal1")
## `.fromSignalArray` β
Create new Signal.
VariableExample
luau
(
{ string }
)
luau
local ... | 706 | documentation | https://imezx.github.io/Warp/api/1.0/signal |
imezx/Warp | null | # Rate Limit feature β
Ratelimit is one of most useful feature.
( Configured on Server only and For Client )
## `Setup` β
When creating a event on Server, you can add second argument (optional) as table `rateLimit` to limit the number of times the event can be called and the interval for reset the counter on client... | 223 | documentation | https://imezx.github.io/Warp/api/1.0/ratelimit |
imezx/Warp | null | # Server event β
For Server-sided
## `.Server` yield β
Create new Warp event.
VariableExample
luau
(
Identifier: string,
rateLimit: {
maxEntrance: number?,
interval: number?,
}?
)
luau
local Remote = Warp.Server("Remote")
## `.fromServerArray` yield β... | 795 | documentation | https://imezx.github.io/Warp/api/1.0/server |
imezx/Warp | null | # Client event β
For Client-sided
## `.Client` yield β
Create new Warp event.
VariableExample
luau
(
Identifier: string
)
luau
local Remote = Warp.Client("Remote")
## `.fromClientArray` yield β
Create new Warp events with array.
VariableExample
luau
(
{ any }
)
luau
... | 558 | documentation | https://imezx.github.io/Warp/api/1.0/client |
imezx/Warp | null | # Getting Started 1.1 β
### `Installation` β
First, you have to require the Warp module.
lua
local Warp = require(path.to.module)
Then, you should do `.Server` or `.Client`
lua
local Server = Warp.Server() --> for Server-side only
local Client = Warp.Client() --> for Client-side only
### `Basic Usag... | 275 | documentation | https://imezx.github.io/Warp/guide/getting-started |
imezx/Warp | null | # Example 1.1 β
Let's try and play something with Warp!
SchemasServerClient
luau
local Schema = require(path.to.warp).Buffer.Schema
return {
Example = Schema.array(Schema.string),
Ping = Schema.string,
Pong = Schema.string,
PingAll = Schema.string,
luau
local Warp = re... | 544 | documentation | https://imezx.github.io/Warp/guide/example |
imezx/Warp | null | # Installation β
## wally β
wally.toml
toml
[dependencies]
warp = "imezx/warp@1.1.0"
## pesde β
pesde.tomlcli
toml
[dependencies]
warp = { name = "eternitydev/warp", version = "^1.1.0" }
bash
pesde add eternitydev/warp
## Roblox Studio β
1. Get the `.rbxm` file from the [github](https:... | 139 | documentation | https://imezx.github.io/Warp/guide/installation |
evaera/Cmdr | README.md | View Docs
**Cmdr** is a fully extensible and type safe command console for Roblox developers.
- Great for admin commands, but does much more.
- Make commands that tie in specifically with your game systems.
- Intelligent autocompletion and instant validation.
- Run commands programmatically on behalf of the local use... | 102 | readme | null |
evaera/Cmdr | null | ## Integrates with your systems
Make commands that integrate and control your existing systems. Use commands to help debug your game during development by triggering events in your game or print out targeted debug information.
## Type-Safe with Intelligent Auto-completion
Discover commands and possible values for ar... | 421 | documentation | https://eryn.io/Cmdr/ |
evaera/Cmdr | null | # Dispatcher
The Dispatcher handles parsing, validating, and evaluating commands. Exists on both client and server.
## Properties
### # `Cmdr`
`Dispatcher.Cmdr: `````[Cmdr](https://eryn.io/Cmdr/api/Cmdr.html#cmdr) | [CmdrClient](https://eryn.io/Cmdr/api/CmdrClient.html#cmdrclient)``````
A reference to Cmdr. This m... | 538 | documentation | https://eryn.io/Cmdr/api/Dispatcher.html |
evaera/Cmdr | null | # # Hooks
Hooks are callback functions that you can register which _hook_ into the command execution process. Hooks are extremely useful: they can be used for implementing a custom permission system, logging commands, or overriding command output.
Hooks can be registered on both the server and the client. Server comm... | 687 | documentation | https://eryn.io/Cmdr/guide/Hooks.html |
evaera/Cmdr | null | # # Network Event Handlers
Some commands that run on the server might need to also do something on the client, or on every client. Network event handlers are callback functions that you can set to run when a server command sends a message back to the client. Only one handler can be set to a certain event at a time, so... | 330 | documentation | https://eryn.io/Cmdr/guide/NetworkEventHandlers.html |
evaera/Cmdr | null | # # Types
By default, these types are available:
Type name | Data type | Type name | Data type
---|---|---|---
`string` | `string` | `strings` | `array<string>`
`number` | `number` | `numbers` | `array<number>`
`integer` | `number` | `integers` | `array<number>`
`boolean` | `boolean` | `booleans` | `array<boolean>`
`... | 936 | documentation | https://eryn.io/Cmdr/guide/Types.html |
evaera/Cmdr | null | # Registry
The registry handles registering commands, types, and hooks. Exists on both client and server.
## Types
### # `TypeDefinition`
`interface ``TypeDefinition {``
DisplayName: `````string`````
Optionally overrides the user-facing name of this type in the autocomplete menu. If omitted, the registered name o... | 3,728 | documentation | https://eryn.io/Cmdr/api/Registry.html |
evaera/Cmdr | null | # # Meta-Commands
The `bind`, `alias`, and `run` commands make use of _command strings_. A command string is raw text made up of a command name and possibly predefined arguments that is run in the background as a command itself. Before these command strings are run, they are preprocessed, replacing arguments (in the f... | 1,108 | documentation | https://eryn.io/Cmdr/guide/MetaCommands.html |
evaera/Cmdr | null | # Cmdr server only
## Properties
### # `Registry`
`Cmdr.Registry: `````[Registry](https://eryn.io/Cmdr/api/Registry.html#registry)``````
Refers to the current command Registry.
### # `Dispatcher`
`Cmdr.Dispatcher: `````[Dispatcher](https://eryn.io/Cmdr/api/Dispatcher.html#dispatcher)``````
Refers to the current ... | 211 | documentation | https://eryn.io/Cmdr/api/Cmdr.html |
evaera/Cmdr | null | # Util
## Types
### # `NamedObject`
`interface ``NamedObject {``
Name: `````string`````
Any object with a `Name` property.
## Static Functions
### # `MakeDictionary` static
`Util.``MakeDictionary(array: ``array<T>``) β ``dictionary<T, true>`````
Accepts an array and flips it into a dictionary, its values becom... | 2,659 | documentation | https://eryn.io/Cmdr/api/Util.html |
evaera/Cmdr | null | # CmdrClient client only
βͺ Extends [Cmdr](https://eryn.io/Cmdr/api/Cmdr.html)
## Properties
βͺ Inherited from [Cmdr](https://eryn.io/Cmdr/api/Cmdr.html)
### # `Registry`
`CmdrClient.Registry: `````[Registry](https://eryn.io/Cmdr/api/Registry.html#registry)``````
Refers to the current command Registry.
βͺ Inherited... | 1,063 | documentation | https://eryn.io/Cmdr/api/CmdrClient.html |
evaera/Cmdr | null | # # Set-up
### # Installation
Pick one of the below methods to install Cmdr:
#### # Manual
You can download the latest model file release from the [releases section](https://github.com/evaera/Cmdr/releases/latest), but this may not always be the most up to date version of Cmdr. You'll want to put this is a server d... | 628 | documentation | https://eryn.io/Cmdr/guide/Setup.html |
evaera/Cmdr | null | # # Commands
No commands are registered by default. Cmdr ships with a set of default commands, which can be loaded if you so wish by calling `Cmdr:RegisterDefaultCommands()`. See Default Commands for a list.
Custom commands are defined in ModuleScripts that return a single table.
-- Teleport.lua, inside your com... | 2,207 | documentation | https://eryn.io/Cmdr/guide/Commands.html |
evaera/Cmdr | null | # ArgumentContext
## Properties
### # `Command`
`ArgumentContext.Command: `````[CommandContext](https://eryn.io/Cmdr/api/CommandContext.html#commandcontext)``````
The command that this argument belongs to.
### # `Name`
`ArgumentContext.Name: `````string``````
The name of this argument.
### # `Type`
`ArgumentCo... | 449 | documentation | https://eryn.io/Cmdr/api/ArgumentContext.html |
evaera/Cmdr | null | # CommandContext
## Properties
### # `Cmdr`
`CommandContext.Cmdr: `````[Cmdr](https://eryn.io/Cmdr/api/Cmdr.html#cmdr) | [CmdrClient](https://eryn.io/Cmdr/api/CmdrClient.html#cmdrclient)``````
A reference to Cmdr. This may either be the server or client version of Cmdr depending on where the command is running.
##... | 1,373 | documentation | https://eryn.io/Cmdr/api/CommandContext.html |
imezx/Gradien | README.md | DevForum: https://devforum.roblox.com/t/gradien-parallelized-machine-deep-learning/4055552
Wally: https://wally.run/package/imezx/gradien | 42 | readme | null |
imezx/Gradien | null | π
## Parallel-first Compute
Heavy numeric operations (matrix multiplication, element-wise math) are dispatched to parallel threads, unlocking performance impossible in serial Luau. | 34 | documentation | https://imezx.github.io/Gradien/ |
imezx/Gradien | null | # Tensor Core β
The `Tensor` is the fundamental data structure in Gradien. It represents a multi-dimensional array and supports automatic differentiation.
## Constructors β
### `.zeros` Parallel β
Creates a new tensor filled with zeros.
Definition
lua
(shape: {number}, dtype: "f32"|"f64"|"i32"?, requiresGrad... | 687 | documentation | https://imezx.github.io/Gradien/api/core/tensor |
imezx/Gradien | null | # Getting Started β
Welcome to **Gradien**. This guide will help you set up the library and train your first neural network.
## Installation β
### `Method 1: Wally` Recommended β
If you use [Wally](https://wally.run/) for dependency management, add Gradien to your `wally.toml`.
wally.tomlTerminal
toml
[depen... | 611 | documentation | https://imezx.github.io/Gradien/guide/getting-started |
imezx/Gradien | null | # Math Operations Parallel β
Located in `Gradien.Ops.Math` and `Gradien.Ops.BLAS`. These operations are parallelized.
## Element-wise β
All element-wise operations support broadcasting semantics implicitly if implemented in the kernel.
### `.add` β
Definition
lua
(A: Tensor, B: Tensor) -> Tensor
### `.sub` ... | 229 | documentation | https://imezx.github.io/Gradien/api/core/math |
imezx/Gradien | null | # Optimizers Optimization β
Gradien includes a wide variety of optimizers for different use cases.
## Adaptive Methods β
### `Adam` β
Adaptive Moment Estimation. Good default. Supports L2 weight decay (applied to gradients).
ConstructorExample
lua
(params: {Tensor}, lr: number, b1: num?, b2: num?, eps: num?,... | 536 | documentation | https://imezx.github.io/Gradien/api/optim/optimizers |
imezx/Gradien | null | # Learning Rate Schedulers Utils β
Located in `Gradien.Optim.Schedulers`. These functions return a callback `(step) -> newLr`.
## Standard β
### `step` β
Decays LR by `gamma` every `stepSize`.
lua
local sched = Schedulers.step(lr, 0.9, 100)
### `cosine` β
Cosine annealing from `lr` down to `lrMin` over `T` ... | 293 | documentation | https://imezx.github.io/Gradien/api/optim/schedulers |
imezx/Gradien | null | # Loss Functions Parallel β
Located in `Gradien.NN.Losses`. Loss functions measure how far the model's predictions are from the targets.
## Regression β
### `mse_backward` β
Mean Squared Error.
Definition
lua
(pred: Tensor, target: Tensor) -> (number, Tensor)
### `l1_backward` β
Mean Absolute Error (L1).
... | 331 | documentation | https://imezx.github.io/Gradien/api/nn/losses |
imezx/Gradien | null | # Mamba β
Implementation of the Mamba state space model architecture.
## MambaBlock β
Creates a Mamba block with selective scan mechanism.
lua
local MambaBlock = Gradien.Experimental.NN.MambaBlock
-- Args: dModel, dState (default 16), dConv (default 4), expand (default 2)
local block = MambaBlock(64, ... | 210 | documentation | https://imezx.github.io/Gradien/api/experimental/mamba |
imezx/Gradien | null | # Data Pipeline Utils β
Tools for loading, scaling, and encoding data. Located in `Gradien.Preprocess`.
## Data Loading β
### `DataLoader` β
Creates an iterator that yields batches of data from a dataset.
Definition
lua
(
dataset: Dataset,
batchSize: number,
shuffle: boolean,
... | 632 | documentation | https://imezx.github.io/Gradien/api/utils/data |
imezx/Gradien | null | # Gradient Clipping Parallel β
Located in `Gradien.GradClip`. Used to prevent exploding gradients, especially in RNNs or deep networks.
## Functions β
### `.clipValue` β
Clips gradient values element-wise to be within `[-clip, clip]`.
Definition
lua
(params: {Tensor}, clip: number) -> ()
### `.clipNorm` β
... | 124 | documentation | https://imezx.github.io/Gradien/api/optim/gradclip |
imezx/Gradien | null | # Feudal Networks β
Implementation of Feudal Reinforcement Learning agent hierarchy.
## Feudal Agent β
Creates a hierarchical agent with Manager and Worker levels.
lua
local Feudal = Gradien.Experimental.RL.Feudal
local agent = Feudal({
inputDim = 32,
actionDim = 4,
hiddenDim = 64,... | 306 | documentation | https://imezx.github.io/Gradien/api/experimental/feudal |
imezx/Gradien | null | # Autograd (Tape) Core β
The `Tape` module records operations and manages the backward pass.
## Functions β
### `.backwardFrom` Parallel β
Computes the gradient of current tensor w.r.t. graph leaves.
DefinitionExample
lua
(y: Tensor, grad: Tensor) -> ()
lua
Tape.backwardFrom(loss, Tensor.ones(loss._sha... | 240 | documentation | https://imezx.github.io/Gradien/api/core/autograd |
imezx/Gradien | null | # State Management Important β
Since Roblox games rely on `DataStoreService` for persistence, Gradien provides utilities to serialize models and optimizers into simple tables that can be saved directly.
## Models β
### Dumping β
Converts the model's parameters into a serializable Snapshot format.
lua
local sn... | 434 | documentation | https://imezx.github.io/Gradien/guide/state |
imezx/Gradien | null | # Activations Parallel β
Activation functions introduce non-linearity to the network. Located in `Gradien.NN.Activations`.
## Standard β
Function| Definition
**`ReLU(x)`**| `max(0, x)`
**`Sigmoid(x)`**| `1 / (1 + exp(-x))`
**`Tanh(x)`**| `(exp(x) - exp(-x)) / (exp(x) + exp(-x))`
## Probability β
### `Softmax` β
C... | 467 | documentation | https://imezx.github.io/Gradien/api/nn/activations |
imezx/Gradien | null | # Initializers Parallel β
Located in `Gradien.Init`. Functions to initialize tensor weights.
Function| Description
`xavierUniform(W)`| Uniform initialization for Sigmoid/Tanh layers.
`kaimingUniform(W)`| Uniform initialization for ReLU layers (He Init).
`heUniform(W)`| Alias for `kaimingUniform`.
`heNormal(W)`| Norma... | 123 | documentation | https://imezx.github.io/Gradien/api/nn/initializers |
imezx/Gradien | null | # How to Create a Neural Network Tutorial β
This guide will walk you through creating, configuring, and training a simple model to learn a mathematical pattern.
We will teach the model the function: **`y = 2x`**.
## 1. Require the Library β
Start by requiring the Gradien module.
lua
local Gradien = require(ga... | 805 | documentation | https://imezx.github.io/Gradien/guide/create-nn |
imezx/Gradien | null | # Layers Module β
Layers are the building blocks of neural networks. They store learnable parameters (`weights` and `biases`).
## `NN.Linear` β
Applies a linear transformation to the incoming data: `y = x * W^T + b`.
DefinitionExample
lua
(
inFeatures: number,
outFeatures: number,
init... | 716 | documentation | https://imezx.github.io/Gradien/api/nn/layers |
imezx/Gradien | null | # Experimental Neural Networks Experimental β
Located in `Gradien.Experimental.NN`. These are experimental neural network architectures and layers.
## `KAN` (Kolmogorov-Arnold Network) β
A Kolmogorov-Arnold Network layer that uses learnable univariate functions (approximated via Radial Basis Functions) instead of fi... | 319 | documentation | https://imezx.github.io/Gradien/api/experimental/nn |
imezx/Gradien | null | # QIMHNN Experimental β
**Quantum-Inspired Metaheuristic Neural Network**
This module implements a neural network architecture inspired by quantum mechanics concepts (Superposition, Interference, and Entanglement) to potentially escape local minima and improve generalization.
## Constructor β
Located in `Gradien.Ex... | 398 | documentation | https://imezx.github.io/Gradien/api/experimental/qimhnn |
imezx/Gradien | null | # Model Tools Utils β
Located in `Gradien.Util`. These tools help you inspect, profile, and modify models.
## `ModelStats` β
Utilities to calculate parameter counts and memory usage.
### `.summary` β
Prints or returns a summary of the model size.
DefinitionExample
lua
(model: Module, printOut: boolean?) -> ... | 470 | documentation | https://imezx.github.io/Gradien/api/utils/model-tools |
imezx/Gradien | null | # Debugging & Metrics Tools β
## Metrics Parallel β
Located in `Gradien.Metrics`.
* **`accuracy(logits, targets)`**: Returns classification accuracy (0.0 - 1.0).
* **`topk(logits, targets, k)`**: Returns top-k accuracy.
* **`mse(pred, target)`**: Mean Squared Error.
* **`prf1(pred, target, C)`**: Returns Pre... | 370 | documentation | https://imezx.github.io/Gradien/api/utils/debug |
imezx/Gradien | null | # RL Agents Module β
Gradien provides a unified API for Reinforcement Learning agents.
## Agent Types β
### `DQL` & `DoubleDQN` Off-Policy β
Deep Q-Learning algorithms. `DoubleDQN` reduces overestimation bias.
Config
lua
actionDim: number,
batchSize: number,
gamma: number,
epsilon... | 410 | documentation | https://imezx.github.io/Gradien/api/rl/agents |
imezx/Gradien | null | # The Trainer Loop Feature β
Writing custom training loops can be repetitive. The `Trainer` class abstracts the boilerplate of zeroing gradients, forward passes, loss calculation, backpropagation, and optimization steps.
## Basic Usage β
ConfigFit
lua
local trainer = Gradien.Trainer.new({
model = myMod... | 368 | documentation | https://imezx.github.io/Gradien/guide/trainer |
imezx/Gradien | null | # Tokenizers β
The `Gradien.Tokenizers` module provides a comprehensive toolkit for text tokenization, inspired by the Hugging Face Tokenizers library. It supports Byte-Pair Encoding (BPE), normalization, pre-tokenization, and post-processing.
## Tokenizer Class β
The main entry point is the `Tokenizer` class, which... | 794 | documentation | https://imezx.github.io/Gradien/api/utils/tokenizers |
imezx/Gradien | null | # Visualization Tools β
## `Visual3D.render` β
Renders the neural network into the 3D workspace using Parts and Beams.
DefinitionOptionsExample
lua
(model: Module, origin: CFrame, options: RenderOptions?) -> RenderHandle
lua
layerSizeHints: {number}?,
maxNeuronsPerLayer: number?,
maxB... | 515 | documentation | https://imezx.github.io/Gradien/api/utils/visualization |
imezx/Gradien | null | # Core Concepts β
## Tensors Core β
The `Tensor` is the fundamental data structure in Gradien. It is a multi-dimensional array that supports automatic differentiation and parallelized operations.
## Autograd (The Tape) β
Gradien uses a **Tape-based** autograd system. Operations on tensors that have `requiresGrad = ... | 200 | documentation | https://imezx.github.io/Gradien/guide/core-concepts |
imezx/Gradien | null | # Porting PyTorch to Gradien Example β
This show you examples bridging **PyTorch** (running on a Python server) with **Gradien** (running on Roblox). This allows you to run massive models (LLMs, heavy RL policies) on a GPU and send the results to your Roblox game in real-time.
dependenciesserverscriptmodelbridge
bas... | 1,628 | documentation | https://imezx.github.io/Gradien/guide/porting-pytorch |
imezx/Gradien | null | # Optimization Wrappers Advanced β
These modules wrap an existing optimizer (the "base") to add functionality like averaging or accumulation.
## `Lookahead` β
Improves stability by keeping a set of "slow weights" that interpolate with the "fast weights" trained by the base optimizer.
ConstructorExample
lua
(
... | 346 | documentation | https://imezx.github.io/Gradien/api/optim/wrappers |
imezx/Gradien | null | # Experimental Optimizers Experimental β
Located in `Gradien.Experimental.Optim`. These optimizers use population-based metaheuristics rather than (or in addition to) standard backpropagation.
## `SwarmPSO` β
**Particle Swarm Optimization**. This is a gradient-free optimizer that maintains a swarm of model variation... | 419 | documentation | https://imezx.github.io/Gradien/api/experimental/optimizers |
Sleitnick/RbxUtil | README.md | # RbxUtil
| Module | Dependency | Description |
| -- | -- | -- |
| [BufferUtil](https://sleitnick.github.io/RbxUtil/api/BufferUtil) | `BufferUtil = "sleitnick/buffer-util@0.3.2"` | Buffer utilities |
| [Comm](https://sleitnick.github.io/RbxUtil/api/Comm) | `Comm = "sleitnick/comm@1.0.1"` | Comm library for remote comm... | 1,360 | readme | null |
Last collected: 2026-04-16 12:44
Contains 275 READMEs & 1,367 Doc pages (~1.82M tokens)