Instructions to use MiniMaxAI/MiniMax-H3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use MiniMaxAI/MiniMax-H3 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("MiniMaxAI/MiniMax-H3", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
README: fix hf download command dropping model_index.json
Browse filesWith huggingface_hub >= 1.0, `--include` takes one pattern per flag. In `hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" "Ref2VA/*"`, only `model_index.json` goes to `--include`; the other two become positional file names, and the CLI warns "Ignoring `--include` since filenames have being explicitly set". The download then lacks the root `model_index.json`.
Passing all patterns as positional arguments works on both 0.x and 1.x, since several positional names are used as allow patterns. Checked on huggingface_hub 1.17.0 with `--dry-run` (both families: 163 files including `model_index.json`; FL2VA only: 82 files) and on 0.35.3 with a small real download. Repeating `--include` per pattern would fix 1.x but keep only the last pattern on 0.x.
|
@@ -206,10 +206,10 @@ Download the model. The repository hosts the original checkpoint (`FL2VA/`, `Ref
|
|
| 206 |
|
| 207 |
```bash
|
| 208 |
# Original checkpoint, both task families (SGLang, vLLM):
|
| 209 |
-
hf download MiniMaxAI/MiniMax-H3
|
| 210 |
|
| 211 |
# Or a single task family:
|
| 212 |
-
hf download MiniMaxAI/MiniMax-H3
|
| 213 |
```
|
| 214 |
|
| 215 |
diffusers users do not need a manual download: `ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` fetches exactly the components it needs. See the [diffusers documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/minimax_h3) for loading recipes.
|
|
|
|
| 206 |
|
| 207 |
```bash
|
| 208 |
# Original checkpoint, both task families (SGLang, vLLM):
|
| 209 |
+
hf download MiniMaxAI/MiniMax-H3 model_index.json "FL2VA/*" "Ref2VA/*" --local-dir MiniMax-H3
|
| 210 |
|
| 211 |
# Or a single task family:
|
| 212 |
+
hf download MiniMaxAI/MiniMax-H3 model_index.json "FL2VA/*" --local-dir MiniMax-H3
|
| 213 |
```
|
| 214 |
|
| 215 |
diffusers users do not need a manual download: `ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` fetches exactly the components it needs. See the [diffusers documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/minimax_h3) for loading recipes.
|