Spaces:
Sleeping
Sleeping
increase n_ctx to 8192
Browse files
app.py
CHANGED
@@ -92,7 +92,7 @@ def update_llm(size, model_file, clip_file):
|
|
92 |
if (model_cache['size'], model_cache['model_file'], model_cache['clip_file']) != (size, model_file, clip_file):
|
93 |
mf, cf = ensure_weights(size, model_file, clip_file)
|
94 |
handler = SmolVLM2ChatHandler(clip_model_path=cf, verbose=False)
|
95 |
-
llm = Llama(model_path=mf, chat_handler=handler, n_ctx=
|
96 |
verbose=False, n_threads=max(2, os.cpu_count()))
|
97 |
model_cache.update({'size': size, 'model_file': mf, 'clip_file': cf, 'llm': llm})
|
98 |
return None # no UI output
|
|
|
92 |
if (model_cache['size'], model_cache['model_file'], model_cache['clip_file']) != (size, model_file, clip_file):
|
93 |
mf, cf = ensure_weights(size, model_file, clip_file)
|
94 |
handler = SmolVLM2ChatHandler(clip_model_path=cf, verbose=False)
|
95 |
+
llm = Llama(model_path=mf, chat_handler=handler, n_ctx=8192,
|
96 |
verbose=False, n_threads=max(2, os.cpu_count()))
|
97 |
model_cache.update({'size': size, 'model_file': mf, 'clip_file': cf, 'llm': llm})
|
98 |
return None # no UI output
|