Any-to-Any
Transformers
Safetensors
PyTorch
NemotronH_Nano_Omni_Reasoning_V3
feature-extraction
nvidia
multimodal
custom_code
Eval Results
Instructions to use nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Fix "download the weights" rendering
Browse files
README.md
CHANGED
|
@@ -145,24 +145,25 @@ hf auth login
|
|
| 145 |
# Sanity check: should print your username and orgs
|
| 146 |
hf auth whoami
|
| 147 |
```
|
| 148 |
-
|
| 149 |
#### Download the weights
|
| 150 |
|
| 151 |
Pick a target directory on a volume with ≥70 GB free (the model is ~62 GB).
|
| 152 |
|
| 153 |
```bash
|
| 154 |
WEIGHTS=/path/to/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16
|
| 155 |
-
|
| 156 |
hf download nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 \
|
| 157 |
--local-dir "$WEIGHTS" \
|
| 158 |
--max-workers 8
|
| 159 |
```
|
| 160 |
|
| 161 |
-
|
| 162 |
> - `hf download` is **resumable** — re-run the same command if the connection drops.
|
| 163 |
> - `--max-workers 8` parallelizes downloads; tune up on fast networks.
|
| 164 |
> - The `hf_xet` extra enables native Xet-protocol transfers for Xet-backed repos; no need for `git-xet` or `git-lfs` when using `hf download`.
|
| 165 |
-
|
|
|
|
| 166 |
#### Verify the download
|
| 167 |
|
| 168 |
```bash
|
|
|
|
| 145 |
# Sanity check: should print your username and orgs
|
| 146 |
hf auth whoami
|
| 147 |
```
|
| 148 |
+
|
| 149 |
#### Download the weights
|
| 150 |
|
| 151 |
Pick a target directory on a volume with ≥70 GB free (the model is ~62 GB).
|
| 152 |
|
| 153 |
```bash
|
| 154 |
WEIGHTS=/path/to/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16
|
| 155 |
+
|
| 156 |
hf download nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 \
|
| 157 |
--local-dir "$WEIGHTS" \
|
| 158 |
--max-workers 8
|
| 159 |
```
|
| 160 |
|
| 161 |
+
**Notes:**
|
| 162 |
> - `hf download` is **resumable** — re-run the same command if the connection drops.
|
| 163 |
> - `--max-workers 8` parallelizes downloads; tune up on fast networks.
|
| 164 |
> - The `hf_xet` extra enables native Xet-protocol transfers for Xet-backed repos; no need for `git-xet` or `git-lfs` when using `hf download`.
|
| 165 |
+
|
| 166 |
+
|
| 167 |
#### Verify the download
|
| 168 |
|
| 169 |
```bash
|