Create a Model FileSet | NVIDIA NeMo Platform

Create a FileSet containing your base model checkpoint before creating a Model Entity.

Prerequisites

Obtained the base URL of your NeMo Platform.
For gated or private HuggingFace models: Created a secret with your HF token. Refer to Manage Secrets.
Set the NMP_BASE_URL environment variable.

$ export NMP_BASE_URL="https://your-nemo-platform-url"

From HuggingFace Hub

The most common method is downloading directly from HuggingFace. The example below uses Qwen/Qwen3-1.7B, a public model that requires no token:

1 import os
2 from nemo_platform import ConflictError, NeMoPlatform
3 from nemo_platform.types.files import HuggingfaceStorageConfigParam
4 
5 client = NeMoPlatform(
6     base_url=os.environ.get("NMP_BASE_URL", "http://localhost:8080"),
7     workspace="default",
8 )
9 
10 HF_REPO_ID = "Qwen/Qwen3-1.7B"
11 MODEL_NAME = "qwen3-1.7b"
12 
13 # Create FileSet from HuggingFace
14 try:
15     fileset = client.files.filesets.create(
16         workspace="default",
17         name=MODEL_NAME,
18         description="Qwen3 1.7B from HuggingFace",
19         purpose="model",
20         storage=HuggingfaceStorageConfigParam(
21             type="huggingface",
22             repo_id=HF_REPO_ID,
23             repo_type="model",
24         ),
25     )
26     print(f"Created FileSet: {fileset.name}")
27 except ConflictError:
28     print(f"FileSet '{MODEL_NAME}' already exists, retrieving...")
29     fileset = client.files.filesets.retrieve(workspace="default", name=MODEL_NAME)
30 
31 print(f"FileSet ready: {fileset.name}")

Gated models (such as Llama) require a HuggingFace token. To use one:

Accept the model license on the HuggingFace model page.
Create a HuggingFace token with read access.
Store the token as a secret in the platform (see Manage Secrets), then pass it as token_secret in the storage config:

1 storage=HuggingfaceStorageConfigParam(
2     type="huggingface",
3     repo_id="meta-llama/Llama-3.2-1B-Instruct",
4     repo_type="model",
5     token_secret="my-hf-token",
6 )

From NGC

For models from NVIDIA NGC:

1 import os
2 from nemo_platform import ConflictError, NeMoPlatform
3 from nemo_platform.types.files import NGCStorageConfigParam
4 
5 client = NeMoPlatform(
6     base_url=os.environ.get("NMP_BASE_URL", "http://localhost:8080"),
7     workspace="default",
8 )
9 
10 MODEL_NAME = "nemotron-mini-4b"
11 NGC_RESOURCE = "nemotron-mini-4b-instruct"
12 NGC_ORG = "nvidia"
13 NGC_TEAM = "nemo"
14 NGC_VERSION = "1.0"
15 NGC_API_KEY_SECRET = "my-ngc-key"
16 NGC_API_KEY = os.environ.get("NGC_API_KEY")
17 if not NGC_API_KEY:
18     raise RuntimeError("Set NGC_API_KEY before creating the NGC secret.")
19 
20 # First, create a secret for your NGC API key (if not already created)
21 try:
22     ngc_secret = client.secrets.create(
23         name=NGC_API_KEY_SECRET,
24         workspace="default",
25         value=NGC_API_KEY,
26     )
27     print(f"Created secret: {NGC_API_KEY_SECRET}")
28 except ConflictError:
29     print(f"Secret '{NGC_API_KEY_SECRET}' already exists, continuing...")
30     ngc_secret = client.secrets.retrieve(name=NGC_API_KEY_SECRET, workspace="default")
31 
32 # Create FileSet from NGC
33 try:
34     fileset = client.files.filesets.create(
35         workspace="default",
36         name=MODEL_NAME,
37         description="Nemotron Mini 4B from NGC",
38         purpose="model",
39         storage=NGCStorageConfigParam(
40             type="ngc",
41             org=NGC_ORG,
42             team=NGC_TEAM,
43             resource=NGC_RESOURCE,  # NGC resource name
44             version=NGC_VERSION,
45             api_key_secret=ngc_secret.name,
46         ),
47     )
48     print(f"Created FileSet: {fileset.name}")
49 except ConflictError:
50     print("FileSet already exists, retrieving...")
51     fileset = client.files.filesets.retrieve(workspace="default", name=MODEL_NAME)

Check FileSet Status

Files are downloaded in the background after you create a FileSet. Check the status:

1 # List files in the FileSet
2 files = client.files.list(
3     workspace="default",
4     fileset="qwen3-1.7b",
5 )
6 
7 print(f"Files in FileSet ({len(files)} total):")
8 for f in files[:10]:  # Show first 10
9     print(f" - {f.path} ({f.size:,} bytes)")

Example Response

1 {
2   "data": [
3     {
4       "file_ref": "/v2/workspaces/default/filesets/qwen3-1.7b/-/config.json",
5       "path": "config.json",
6       "size": 1234,
7       "cache_status": null
8     },
9     {
10       "file_ref": "/v2/workspaces/default/filesets/qwen3-1.7b/-/model.safetensors",
11       "path": "model.safetensors",
12       "size": 2400000000,
13       "cache_status": null
14     },
15     {
16       "file_ref": "/v2/workspaces/default/filesets/qwen3-1.7b/-/tokenizer.json",
17       "path": "tokenizer.json",
18       "size": 9085657,
19       "cache_status": null
20     },
21     {
22       "file_ref": "/v2/workspaces/default/filesets/qwen3-1.7b/-/tokenizer_config.json",
23       "path": "tokenizer_config.json",
24       "size": 901,
25       "cache_status": null
26     }
27   ]
28 }

Next Steps

Proceed to create a Model Entity.