Commit
•
967fe65
1
Parent(s):
b6bbbe0
Fix things. (#12)
Browse files- Fix things. (de5a40ce88944af8efe23d4e7e3b36f931634a74)
Co-authored-by: Lewdiculous <Lewdiculous@users.noreply.huggingface.co>
- gguf-imat.py +21 -15
gguf-imat.py
CHANGED
@@ -89,8 +89,11 @@ def download_model_repo():
|
|
89 |
# If the model already exists, prompt the user if they want to delete the model directory
|
90 |
delete_model_dir = input("Remove HF model folder after converting original model to GGUF? (yes/no) (default: no): ").strip().lower()
|
91 |
|
|
|
|
|
|
|
92 |
# Convert the existing model to GGUF F16 format and generate imatrix.dat
|
93 |
-
convert_model_to_gguf_f16(base_dir, model_dir, model_name, delete_model_dir)
|
94 |
|
95 |
else:
|
96 |
revision = input("Enter the revision (branch, tag, or commit) to download (default: main): ") or "main"
|
@@ -102,11 +105,14 @@ def download_model_repo():
|
|
102 |
snapshot_download(repo_id=model_id, local_dir=model_dir, revision=revision)
|
103 |
print("Model repository downloaded successfully.")
|
104 |
|
|
|
|
|
|
|
105 |
# Convert the downloaded model to GGUF F16 format and generate imatrix.dat
|
106 |
-
convert_model_to_gguf_f16(base_dir, model_dir, model_name, delete_model_dir)
|
107 |
|
108 |
# Convert the downloaded model to GGUF F16 format
|
109 |
-
def convert_model_to_gguf_f16(base_dir, model_dir, model_name, delete_model_dir):
|
110 |
convert_script = os.path.join(base_dir, "llama.cpp", "convert.py")
|
111 |
gguf_dir = os.path.join(base_dir, "models", f"{model_name}-GGUF")
|
112 |
gguf_model_path = os.path.join(gguf_dir, f"{model_name}-F16.gguf")
|
@@ -126,21 +132,21 @@ def convert_model_to_gguf_f16(base_dir, model_dir, model_name, delete_model_dir)
|
|
126 |
else:
|
127 |
print(f"Original model directory '{model_dir}' was not deleted. You can remove it manually.")
|
128 |
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
|
|
137 |
shutil.move(os.path.join(gguf_dir, "imatrix.dat"), gguf_dir)
|
138 |
print("imatrix.dat generated successfully.")
|
139 |
else:
|
140 |
-
print("
|
141 |
-
|
142 |
else:
|
143 |
-
print("Skipping
|
144 |
|
145 |
# Quantize the models
|
146 |
quantize_models(base_dir, model_name)
|
@@ -177,4 +183,4 @@ def main():
|
|
177 |
print("Finished preparing resources.")
|
178 |
|
179 |
if __name__ == "__main__":
|
180 |
-
main()
|
|
|
89 |
# If the model already exists, prompt the user if they want to delete the model directory
|
90 |
delete_model_dir = input("Remove HF model folder after converting original model to GGUF? (yes/no) (default: no): ").strip().lower()
|
91 |
|
92 |
+
# Ask for the name of the imatrix.txt file
|
93 |
+
imatrix_file_name = input("Enter the name of the imatrix.txt file (default: imatrix.txt): ").strip() or "imatrix.txt"
|
94 |
+
|
95 |
# Convert the existing model to GGUF F16 format and generate imatrix.dat
|
96 |
+
convert_model_to_gguf_f16(base_dir, model_dir, model_name, delete_model_dir, imatrix_file_name)
|
97 |
|
98 |
else:
|
99 |
revision = input("Enter the revision (branch, tag, or commit) to download (default: main): ") or "main"
|
|
|
105 |
snapshot_download(repo_id=model_id, local_dir=model_dir, revision=revision)
|
106 |
print("Model repository downloaded successfully.")
|
107 |
|
108 |
+
# Ask for the name of the imatrix.txt file
|
109 |
+
imatrix_file_name = input("Enter the name of the imatrix.txt file (default: imatrix.txt): ").strip() or "imatrix.txt"
|
110 |
+
|
111 |
# Convert the downloaded model to GGUF F16 format and generate imatrix.dat
|
112 |
+
convert_model_to_gguf_f16(base_dir, model_dir, model_name, delete_model_dir, imatrix_file_name)
|
113 |
|
114 |
# Convert the downloaded model to GGUF F16 format
|
115 |
+
def convert_model_to_gguf_f16(base_dir, model_dir, model_name, delete_model_dir, imatrix_file_name):
|
116 |
convert_script = os.path.join(base_dir, "llama.cpp", "convert.py")
|
117 |
gguf_dir = os.path.join(base_dir, "models", f"{model_name}-GGUF")
|
118 |
gguf_model_path = os.path.join(gguf_dir, f"{model_name}-F16.gguf")
|
|
|
132 |
else:
|
133 |
print(f"Original model directory '{model_dir}' was not deleted. You can remove it manually.")
|
134 |
|
135 |
+
# Generate imatrix.dat if it doesn't exist
|
136 |
+
imatrix_exe = os.path.join(base_dir, "bin", "imatrix.exe")
|
137 |
+
imatrix_output = os.path.join(gguf_dir, "imatrix.dat")
|
138 |
+
imatrix_txt = os.path.join(base_dir, "imatrix", imatrix_file_name)
|
139 |
+
if not os.path.exists(imatrix_output):
|
140 |
+
# Execute the imatrix command
|
141 |
+
subprocess.run([imatrix_exe, "-m", gguf_model_path, "-f", imatrix_txt, "-ngl", "13"], cwd=gguf_dir)
|
142 |
+
# Move the imatrix.dat file to the GGUF folder
|
143 |
+
if os.path.exists(os.path.join(gguf_dir, "imatrix.dat")):
|
144 |
shutil.move(os.path.join(gguf_dir, "imatrix.dat"), gguf_dir)
|
145 |
print("imatrix.dat generated successfully.")
|
146 |
else:
|
147 |
+
print("Failed to generate imatrix.dat file.")
|
|
|
148 |
else:
|
149 |
+
print("Skipping imatrix generation as imatrix.dat already exists.")
|
150 |
|
151 |
# Quantize the models
|
152 |
quantize_models(base_dir, model_name)
|
|
|
183 |
print("Finished preparing resources.")
|
184 |
|
185 |
if __name__ == "__main__":
|
186 |
+
main()
|