Documentation – Model source parameters

Net: ModelFilePath

Fri, 24 Apr 2026 00:00:00 +0000

ModelFilePath is the highest-priority source for the model file. When set to a non-empty path, the engine loads directly from disk and skips all download logic.

Quick reference


Type	`string?`
Default	`null`
Priority	1 (highest; wins over `AsposeModelId` and Hugging Face fields)
Category	Model source
Field on	`ModelSourceParameters.ModelFilePath`

What it does

null or empty — skip this priority level; fall through to AsposeModelId, then Hugging Face fields.
A path — load the GGUF from that file. No download.

The file must be accessible and readable by the process.

When to change it

Scenario	Value
Default (use Hugging Face or Aspose catalog)	`null`
Air-gapped deployment with pre-downloaded model	Absolute path
Custom local GGUF	Path to the file
Testing a model before mirroring to Hugging Face	Local path

Example

var preset = new Qwen25Preset();
preset.BaseModelSourceParameters.ModelFilePath =
    @"C:\models\Qwen2.5-7B-Instruct-Q4_K_M.gguf";
// Bypass the Hugging Face values on the preset.

using var api = AsposeLLMApi.Create(preset);

Interactions

AsposeModelId — priority 2; ignored when ModelFilePath is set.
HuggingFaceRepoId + HuggingFaceFileName — priority 3; ignored when ModelFilePath is set.
EngineParameters.ModelCachePath — cache is not used when loading from an explicit ModelFilePath.

What’s next

Model source hub — resolution order overview.
Bring your own GGUF — full custom-model workflow.
Offline deployment — air-gapped setup.

Net: AsposeModelId

Fri, 24 Apr 2026 00:00:00 +0000

AsposeModelId is the second-priority source — an identifier into Aspose’s internal model catalog hosted at releases.aspose.com. The engine downloads by ID when no ModelFilePath is set.

Quick reference


Type	`string?`
Default	`null`
Priority	2 (loses to `ModelFilePath`; wins over Hugging Face fields)
Category	Model source
Field on	`ModelSourceParameters.AsposeModelId`

What it does

When set, the engine resolves the ID against the Aspose model catalog and downloads the named file into EngineParameters.ModelCachePath. Subsequent runs use the cache.

null — skip this priority; fall through to Hugging Face fields.
An ID — resolve against the Aspose catalog.

Built-in presets do not set AsposeModelId — they use Hugging Face. Use this field when Aspose ships a specific model optimized or licensed for SDK consumers.

When to change it

Scenario	Value
Default	`null`
Load from Aspose catalog by ID	Specific ID

Example

var preset = new Qwen25Preset();
preset.BaseModelSourceParameters.AsposeModelId = "aspose-llm-qwen-25-7b-q4km";
preset.BaseModelSourceParameters.HuggingFaceRepoId = null; // clear the default so priority-3 fall-through isn't used

Interactions

ModelFilePath — if set, overrides AsposeModelId.
Hugging Face fields — only used when both ModelFilePath and AsposeModelId are null.

What’s next

Model source hub — resolution order.
HuggingFaceRepoId — priority-3 source.

Net: HuggingFaceRepoId

Fri, 24 Apr 2026 00:00:00 +0000

HuggingFaceRepoId identifies a Hugging Face repository to download the model from. Used when neither ModelFilePath nor AsposeModelId is set.

Quick reference


Type	`string?`
Default	`null` (set by built-in presets)
Priority	3 (lowest)
Category	Model source
Field on	`ModelSourceParameters.HuggingFaceRepoId`

What it does

The engine downloads HuggingFaceFileName from this repository when higher-priority fields are unset. Built-in presets populate this field — for example, Qwen25Preset sets it to "bartowski/Qwen2.5-7B-Instruct-GGUF".

null — no Hugging Face download.
An Id like "owner/repo" — download from that repo.

When to change it

Scenario	Value
Default (use preset’s value)	whatever the preset sets
Switch to a different quantization publisher	Different owner/repo
Use a custom GGUF in a private Hugging Face repo	Private-repo ID (requires HF token configuration)

Example

var preset = new Qwen25Preset();
preset.BaseModelSourceParameters.HuggingFaceRepoId = "unsloth/Qwen2.5-7B-Instruct-GGUF";
preset.BaseModelSourceParameters.HuggingFaceFileName = "Qwen2.5-7B-Instruct-Q4_K_M.gguf";

Interactions

HuggingFaceFileName — the specific file within the repo.
ModelFilePath, AsposeModelId — higher-priority sources; override this.

What’s next

HuggingFaceFileName — companion file selector.
Bring your own GGUF — custom-model workflow.
Model source hub — resolution order.

Net: HuggingFaceFileName

Fri, 24 Apr 2026 00:00:00 +0000

HuggingFaceFileName names the specific file within HuggingFaceRepoId to download. Many Hugging Face repos host several GGUF variants (different quantizations) — this field picks one.

Quick reference


Type	`string?`
Default	`null` (set by built-in presets)
Category	Model source
Field on	`ModelSourceParameters.HuggingFaceFileName`

What it does

Used alongside HuggingFaceRepoId. Built-in presets pick a sensible default (usually Q4_K_M); override to switch quantization.

For Qwen25Preset, the default is "Qwen2.5-7B-Instruct-Q4_K_M.gguf". Common alternatives in the same repo: Q4_0, Q5_K_M, Q6_K, Q8_0.

When to change it

Scenario	Value
Default (preset’s value)	unchanged
Higher quality, more memory	`Q5_K_M`, `Q6_K`, `Q8_0`
Lower memory, lower quality	`Q4_0`, `IQ3_M`, `IQ2_XXS`
Testing a specific quant	Full file name from Hugging Face listing

See Understand quantization for a primer on which variant to pick.

Example

var preset = new Qwen25Preset();
// Switch to Q8 for higher quality — about 2x the file size
preset.BaseModelSourceParameters.HuggingFaceFileName = "Qwen2.5-7B-Instruct-Q8_0.gguf";

Or smaller for memory-tight hardware:

preset.BaseModelSourceParameters.HuggingFaceFileName = "Qwen2.5-7B-Instruct-IQ3_M.gguf";

Interactions

HuggingFaceRepoId — the repo containing the file.
ModelFilePath — higher priority; overrides Hugging Face download.

What’s next

Understand quantization — Q4 vs Q5 vs Q8 primer.
HuggingFaceRepoId — companion repo selector.
Model source hub — resolution order.