Skip to content

C
coderai
Commit 015c6908 authored by Your Name's avatar Your Name
Browse files
Options

Add GGUF magic bytes validation 

- Check if downloaded file is valid GGUF (magic bytes = 'GGUF')
- If not valid, show clear error that URL is wrong (returns HTML instead)
- Explain that URL must be direct download link ending in .gguf
parent 611bfd8f
Hide whitespace changes
Inline Side-by-side
Showing with 47 additions and 34 deletions
coderai
View file @ 015c6908
......@@ -4467,28 +4467,33 @@ def main():
n_ctx = 2048
print(f"Loading GGUF model from: {model_path}")
print(f"GGUF model file size: {os.path.getsize(model_path) / (1024*1024):.1f} MB")
print(f"File exists: {os.path.exists(model_path)}")
file_size = os.path.getsize(model_path)
print(f"GGUF model file size: {file_size / (1024*1024):.1f} MB")
# Try to verify it's a valid GGUF file
# Verify it's a valid GGUF file (check magic bytes)
with open(model_path, 'rb') as f:
magic = f.read(8)
print(f"GGUF magic bytes: {magic.hex()}")
try:
llama_model = Llama(
model_path=model_path,
n_gpu_layers=n_gpu_layers,
n_ctx=n_ctx,
verbose=True, # Enable verbose to see errors
)
multi_model_manager.add_model(model_key, llama_model)
print(f"GGUF image model loaded successfully: {original_model_name}")
except Exception as llama_error:
print(f"llama.cpp load error: {llama_error}")
import traceback
traceback.print_exc()
print(f"Will try loading image model on first request instead")
print(f"File magic bytes: {magic}")
if magic != b'GGUF':
print(f"ERROR: File is NOT a valid GGUF! Expected 'GGUF', got: {magic}")
print(f"This means the download returned an HTML error page instead of the model.")
print(f"The URL must be a DIRECT download link (ends with .gguf, not a model page)")
print(f"Example: https://huggingface.co/owner/repo/resolve/main/model.gguf")
print(f"Image model will load on first request")
else:
# Valid GGUF, try to load
try:
llama_model = Llama(
model_path=model_path,
n_gpu_layers=n_gpu_layers,
n_ctx=n_ctx,
verbose=True,
)
multi_model_manager.add_model(model_key, llama_model)
print(f"GGUF image model loaded successfully: {original_model_name}")
except Exception as llama_error:
print(f"llama.cpp load error: {llama_error}")
print(f"Will try loading image model on first request instead")
else:
print(f"Could not load GGUF image model: no valid model path")
......@@ -4863,22 +4868,30 @@ def main():
n_ctx = 2048
print(f"Loading GGUF model from: {model_path}")
print(f"GGUF model file size: {os.path.getsize(model_path) / (1024*1024):.1f} MB")
file_size = os.path.getsize(model_path)
print(f"GGUF model file size: {file_size / (1024*1024):.1f} MB")
try:
llama_model = Llama(
model_path=model_path,
n_gpu_layers=n_gpu_layers,
n_ctx=n_ctx,
verbose=True,
)
multi_model_manager.add_model(model_key, llama_model)
print(f"GGUF image model loaded successfully: {original_model_name}")
except Exception as llama_error:
print(f"llama.cpp load error: {llama_error}")
import traceback
traceback.print_exc()
print(f"Will try loading image model on first request instead")
# Verify it's a valid GGUF file (check magic bytes)
with open(model_path, 'rb') as f:
magic = f.read(8)
print(f"File magic bytes: {magic}")
if magic != b'GGUF':
print(f"ERROR: File is NOT a valid GGUF! Expected 'GGUF', got: {magic}")
print(f"The URL must be a DIRECT download link (ends with .gguf)")
print(f"Image model will load on first request")
else:
try:
llama_model = Llama(
model_path=model_path,
n_gpu_layers=n_gpu_layers,
n_ctx=n_ctx,
verbose=True,
)
multi_model_manager.add_model(model_key, llama_model)
print(f"GGUF image model loaded successfully: {original_model_name}")
except Exception as llama_error:
print(f"llama.cpp load error: {llama_error}")
print(f"Will try loading image model on first request instead")
else:
print(f"Could not load GGUF image model: no valid model path")
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment