BackendError: POST failed with: {"errors":["Internal error"],"error":{"code":13,"details":[]},"wasSuccessful":false}

I am using newly launched kaggle packages but it is giving me internal error when doing kagglehub.model_download
___
My code ->
#| export
class Model:
def __init__(self):

quantization_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_use_double_quant=True,
bnb_4bit_compute_dtype=torch.float16,
)
#self.model_path = "/kaggle/input/deepseek-r1/transformers/deepseek-r1-distill-qwen-7b/2"
self.model_path = kagglehub.model_download('deepseek-r1/transformers/deepseek-r1-distill-qwen-7b/2')
self.tokenizer = AutoTokenizer.from_pretrained(self.model_path)
self.model = AutoModelForCausalLM.from_pretrained(self.model_path,
device_map="auto",
quantization_config=quantization_config,
)

def predict(self, description):
# Prompt the model to generate SVG code
prompt = f"""
I changed prompt here
"""
inputs = self.tokenizer.encode(prompt, return_tensors='pt').to('cuda')
outputs = self.model.generate(inputs, max_length=5000)
response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)

# Extract valid SVG code if possible
start = response.find('', start) + 6 if start != -1 else -1
if start != -1 and end != -1:
svg_code = response[start:end]
else:
svg_code = "Failed to generate valid SVG"
return svg_code, response
___________________________________________________________________________
Complete error -->
---------------------------------------------------------------------------
BackendError Traceback (most recent call last)
in ()
1 # Test the SVGGenerator
----> 2 generator = Model()

in __init__(self)
10 )
11 #self.model_path = "/kaggle/input/deepseek-r1/transformers/deepseek-r1-distill-qwen-7b/2"
---> 12 self.model_path = kagglehub.model_download('deepseek-r1/transformers/deepseek-r1-distill-qwen-7b/2')
13 self.tokenizer = AutoTokenizer.from_pretrained(self.model_path)
14 self.model = AutoModelForCausalLM.from_pretrained(self.model_path,

/usr/local/lib/python3.10/dist-packages/kagglehub/models.py in model_download(handle, path, force_download)
33 h = parse_model_handle(handle)
34 logger.info(f"Downloading Model: {h.to_url()} ...", extra={**EXTRA_CONSOLE_BLOCK})
---> 35 path, _ = registry.model_resolver(h, path, force_download=force_download)
36 return path
37

/usr/local/lib/python3.10/dist-packages/kagglehub/registry.py in __call__(self, *args, **kwargs)
26 for impl in reversed(self._impls):
27 if impl.is_supported(*args, **kwargs):
---> 28 return impl(*args, **kwargs)
29 else:
30 fails.append(type(impl).__name__)

/usr/local/lib/python3.10/dist-packages/kagglehub/resolver.py in __call__(self, handle, path, force_download)
27 Some cases where version number might be missing: Competition datasource, API-based models.
28 """
---> 29 path, version = self._resolve(handle, path, force_download=force_download)
30
31 # Note handles are immutable, so _resolve() could not have altered our reference

/usr/local/lib/python3.10/dist-packages/kagglehub/kaggle_cache_resolver.py in _resolve(self, h, path, force_download)
200 model_ref["VersionNumber"] = str(version_from_package_scope)
201
--> 202 result = client.post(
203 ATTACH_DATASOURCE_REQUEST_NAME,
204 {

/usr/local/lib/python3.10/dist-packages/kagglehub/clients.py in post(self, request_name, data, timeout)
389 if not json_response["wasSuccessful"]:
390 msg = f"POST failed with: {response.text!s}"
--> 391 raise BackendError(msg)
392 if "result" not in json_response:
393 msg = "'result' field missing from response"

BackendError: POST failed with: {"errors":["Internal error"],"error":{"code":13,"details":[]},"wasSuccessful":false}
_________________________________________________________________________________________

Do I need to download kaggle api key and use kaggle.json somewhere?

Start by reproducing the failure at kagglehub.model_download using the model handle shown in the issue. Read kagglehub/models.py, registry.py, resolver.py, kaggle_cache_resolver.py, and clients.py around the reported POST request, then determine whether the error depends on authentication or the model package. Done means identifying the cause and confirming a reproducible fix or documenting the required authentication setup.

Written by the indexing model from the issue text.