[Request] Enable the retrieval of more detailed repository data using `huggingface_hub.hf_api`

[suzukimain]

Is your feature request related to a problem? Please describe.
There is no option in huggingface_hub.hf_api to get the type of files in the repository or to get the base model of the checkpoint files.

Describe the solution you'd like
Please make it possible to retrieve using hf_api

Describe alternatives you've considered
baseModels should be in hf_api.model_info, but it is currently not available for use.

huggingface_hub/src/huggingface_hub/hf_api.py

Line 2500 in 2702ec2

Possible values are `"author"`, `"baseModels"`, `"cardData"`, `"childrenModelCount"`, `"config"`, `"createdAt"`, `"disabled"`, `"downloads"`, `"downloadsAllTime"`, `"gated"`, `"gguf"`, `"inference"`, `"lastModified"`, `"library_name"`, `"likes"`, `"mask_token"`, `"model-index"`, `"pipeline_tag"`, `"private"`, `"safetensors"`, `"sha"`, `"siblings"`, `"spaces"`, `"tags"`, `"transformersInfo"`, `"trendingScore"`, `"widgetData"`, `"usedStorage"` and `"resourceGroup"`.

Example

from huggingface_hub import hf_api

info = hf_api.model_info(
    "stable-diffusion-v1-5/stable-diffusion-v1-5",
    expand=["baseModels"]
)

print(info.baseModels)
# AttributeError: 'ModelInfo' object has no attribute 'baseModels'

Additional context
I was creating a function to search for models using diffusers, but due to the lack of a method to retrieve data from hf_api, I couldn't implement it. Therefore, I would like to use it for the creation of the functions that couldn't be implemented
Also related to #2039

[Wauplin]

Hi @suzukimain, if I understand correctly this issue could be renamed "Can I get baseModels information using HfApi?", correct? Asking just to be sure you are not referring to something else in "more detailed repository".

Also, in your example above using "stable-diffusion-v1-5/stable-diffusion-v1-5", what would be the expected value for baseModels ?

[suzukimain]

Hi @suzukimain, if I understand correctly this issue could be renamed "Can I get baseModels information using HfApi?", correct? Asking just to be sure you are not referring to something else in "more detailed repository".

Also, in your example above using "stable-diffusion-v1-5/stable-diffusion-v1-5", what would be the expected value for baseModels ?

Hello. Thank you for your prompt response. There are two options I would like to request:

1. An option to retrieve the base models.
2. An option to retrieve the file types.

The expected value for baseModels in the example above is as follows:

# print(info.baseModels)

{
    "v1-5-pruned.safetensors": "SD 1.5",
    "v1-5-pruned-emaonly.safetensors": "SD 1.5",
}

---

Regarding the option to retrieve file types (number 2):

from huggingface_hub import hf_api

info = hf_api.model_info("gsdf/CounterfeitXL")

print(info.fileTypes)

result

{
    "CounterfeitXL_α.safetensors": "Checkpoint",
    "CounterfeitXL_β.safetensors": "Checkpoint",
    "embeddings/negativeXL_A.safetensors": "TextualInversion",
    "embeddings/negativeXL_B.safetensors": "TextualInversion",
}

An example of the expected values for this option is:

• Checkpoint,TextualInversion,Controlnet,LORA

[Wauplin]

Oh ok I see. So quite related to #2039 indeed. As we explained there, the Hub repo is not opinionated on the file format i.e. it is not the Hub's job to tell which file is meant for what (checkpoint, textualinversion, etc.). Usually this information can be inferred from either a config file or the file name itself. If you have more questions about the format and convention, I'd recommend you to check in diffusers docs and discussion page directly.

The Hub is meant to be host model repositories in a library-agnostic way. Any library can define their own convention, the same way GitHub don't enforce anything on file convention and content (except very generic things like READMEs). For now, I'll close this issue as a duplicate.