oobabooga/textgen · error · FileNotFoundError
No .gguf models found in directory: {path}
Error message
No .gguf models found in directory: {path} What it means
Error "No .gguf models found in directory: {path}" thrown in oobabooga/textgen.
Source
Thrown at modules/models_settings.py:59
model_settings,
hf_quant_method=quant_method
)
# Default bos/eos tokens (may be overridden by GGUF metadata or tokenizer_config.json)
shared.bos_token = '<s>'
shared.eos_token = '</s>'
# GGUF metadata
if model_settings['loader'] == 'llama.cpp':
path = model_path
if path.is_file():
model_file = path
else:
gguf_files = list(path.glob('*.gguf'))
if not gguf_files:
error_msg = f"No .gguf models found in directory: {path}"
logger.error(error_msg)
raise FileNotFoundError(error_msg)
model_file = gguf_files[0]
metadata = load_gguf_metadata_with_cache(model_file)
for k in metadata:
if k.endswith('.context_length'):
model_settings['ctx_size'] = 0
model_settings['truncation_length_info'] = metadata[k]
elif k.endswith('.block_count'):
model_settings['gpu_layers'] = -1
model_settings['max_gpu_layers'] = metadata[k] + 1
elif k.endswith('.nextn_predict_layers') and metadata[k] > 0:
model_settings['spec_type'] = 'draft-mtp'
if 'tokenizer.chat_template' in metadata:
template = metadata['tokenizer.chat_template']
if 'tokenizer.ggml.eos_token_id' in metadata:View on GitHub (pinned to ed888c71f2)
Solutions
- Place at least one .gguf model file in the reported directory.
- Point the model directory setting at the folder that actually contains your .gguf files.
When it happens
Trigger: Raised when scanning a model directory for llama.cpp models and no .gguf files are found. Triggers when the model folder path is wrong, the model was not downloaded, or the folder only contains sharded/non-GGUF weights.
Common situations: See trigger scenarios.
AI-assisted analysis of oobabooga/textgen@ed888c71f2 (2026-08-15).
Data as JSON: /api/errors/d77a1c6e643b2e80.
Report an issue: GitHub.