Spaces:
Running
on
Zero
Running
on
Zero
improve model management
Browse files
app.py
CHANGED
@@ -12,10 +12,10 @@ MODELS = {
|
|
12 |
"filename": "qwen2.5-7b-instruct-q2_k.gguf",
|
13 |
"description": "Qwen2.5-7B Instruct (Q2_K)"
|
14 |
},
|
15 |
-
"Gemma-3-4B-IT (
|
16 |
"repo_id": "unsloth/gemma-3-4b-it-GGUF",
|
17 |
-
"filename": "gemma-3-4b-it-
|
18 |
-
"description": "Gemma 3 4B IT (
|
19 |
},
|
20 |
"Phi-4-mini-Instruct (Q4_K_M)": {
|
21 |
"repo_id": "unsloth/Phi-4-mini-instruct-GGUF",
|
@@ -38,19 +38,37 @@ with st.sidebar:
|
|
38 |
selected_model = MODELS[selected_model_name]
|
39 |
model_path = os.path.join("models", selected_model["filename"])
|
40 |
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
# Make sure models dir exists
|
42 |
os.makedirs("models", exist_ok=True)
|
43 |
|
44 |
-
#
|
45 |
-
|
46 |
-
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
try:
|
49 |
-
os.remove(os.path.join("models",
|
50 |
except Exception as e:
|
51 |
-
st.warning(f"
|
52 |
|
53 |
-
|
54 |
with st.spinner(f"Downloading {selected_model['filename']}..."):
|
55 |
hf_hub_download(
|
56 |
repo_id=selected_model["repo_id"],
|
@@ -59,29 +77,28 @@ if not os.path.exists(model_path):
|
|
59 |
local_dir_use_symlinks=False,
|
60 |
)
|
61 |
|
62 |
-
|
63 |
-
if "model_name" not in st.session_state:
|
64 |
-
st.session_state.model_name = None
|
65 |
-
if "llm" not in st.session_state:
|
66 |
-
st.session_state.llm = None
|
67 |
|
68 |
# Load model if changed
|
69 |
if st.session_state.model_name != selected_model_name:
|
70 |
if st.session_state.llm is not None:
|
71 |
del st.session_state.llm
|
72 |
gc.collect()
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
|
|
|
|
|
|
85 |
st.session_state.model_name = selected_model_name
|
86 |
|
87 |
llm = st.session_state.llm
|
|
|
12 |
"filename": "qwen2.5-7b-instruct-q2_k.gguf",
|
13 |
"description": "Qwen2.5-7B Instruct (Q2_K)"
|
14 |
},
|
15 |
+
"Gemma-3-4B-IT (Q4_K_M)": {
|
16 |
"repo_id": "unsloth/gemma-3-4b-it-GGUF",
|
17 |
+
"filename": "gemma-3-4b-it-Q4_K_M.gguf",
|
18 |
+
"description": "Gemma 3 4B IT (Q4_K_M)"
|
19 |
},
|
20 |
"Phi-4-mini-Instruct (Q4_K_M)": {
|
21 |
"repo_id": "unsloth/Phi-4-mini-instruct-GGUF",
|
|
|
38 |
selected_model = MODELS[selected_model_name]
|
39 |
model_path = os.path.join("models", selected_model["filename"])
|
40 |
|
41 |
+
# Init state
|
42 |
+
if "model_name" not in st.session_state:
|
43 |
+
st.session_state.model_name = None
|
44 |
+
if "llm" not in st.session_state:
|
45 |
+
st.session_state.llm = None
|
46 |
+
|
47 |
# Make sure models dir exists
|
48 |
os.makedirs("models", exist_ok=True)
|
49 |
|
50 |
+
# If the selected model file does not exist or is invalid, clean up and re-download
|
51 |
+
def validate_or_download_model():
|
52 |
+
if not os.path.exists(model_path):
|
53 |
+
cleanup_old_models()
|
54 |
+
download_model()
|
55 |
+
return
|
56 |
+
try:
|
57 |
+
_ = Llama(model_path=model_path, n_ctx=16, n_threads=1) # dummy check
|
58 |
+
except Exception as e:
|
59 |
+
st.warning(f"Model file was invalid or corrupt: {e}\nRedownloading...")
|
60 |
+
cleanup_old_models()
|
61 |
+
download_model()
|
62 |
+
|
63 |
+
def cleanup_old_models():
|
64 |
+
for f in os.listdir("models"):
|
65 |
+
if f.endswith(".gguf") and f != selected_model["filename"]:
|
66 |
try:
|
67 |
+
os.remove(os.path.join("models", f))
|
68 |
except Exception as e:
|
69 |
+
st.warning(f"Couldn't delete old model {f}: {e}")
|
70 |
|
71 |
+
def download_model():
|
72 |
with st.spinner(f"Downloading {selected_model['filename']}..."):
|
73 |
hf_hub_download(
|
74 |
repo_id=selected_model["repo_id"],
|
|
|
77 |
local_dir_use_symlinks=False,
|
78 |
)
|
79 |
|
80 |
+
validate_or_download_model()
|
|
|
|
|
|
|
|
|
81 |
|
82 |
# Load model if changed
|
83 |
if st.session_state.model_name != selected_model_name:
|
84 |
if st.session_state.llm is not None:
|
85 |
del st.session_state.llm
|
86 |
gc.collect()
|
87 |
+
try:
|
88 |
+
st.session_state.llm = Llama(
|
89 |
+
model_path=model_path,
|
90 |
+
n_ctx=1024,
|
91 |
+
n_threads=2,
|
92 |
+
n_threads_batch=2,
|
93 |
+
n_batch=4,
|
94 |
+
n_gpu_layers=0,
|
95 |
+
use_mlock=False,
|
96 |
+
use_mmap=True,
|
97 |
+
verbose=False,
|
98 |
+
)
|
99 |
+
except Exception as e:
|
100 |
+
st.error(f"Failed to load model: {e}")
|
101 |
+
st.stop()
|
102 |
st.session_state.model_name = selected_model_name
|
103 |
|
104 |
llm = st.session_state.llm
|