Spaces:
Runtime error
Runtime error
Sean-Case
commited on
Commit
•
8249fd3
1
Parent(s):
994ad90
Made GPU layer bar invisible. Changed default GPU layers when CUDA available
Browse files- app.py +6 -4
- chatfuncs/chatfuncs.py +1 -1
app.py
CHANGED
@@ -80,9 +80,11 @@ def load_model(model_type, gpu_layers, gpu_config=None, cpu_config=None, torch_d
|
|
80 |
torch_device = chatf.torch_device
|
81 |
|
82 |
if model_type == "Mistral Open Orca (larger, slow)":
|
83 |
-
|
84 |
-
|
85 |
-
|
|
|
|
|
86 |
|
87 |
print("Loading with", cpu_config.gpu_layers, "model layers sent to GPU.")
|
88 |
|
@@ -231,7 +233,7 @@ with block:
|
|
231 |
with gr.Tab("Advanced features"):
|
232 |
model_choice = gr.Radio(label="Choose a chat model", value="Flan Alpaca (small, fast)", choices = ["Flan Alpaca (small, fast)", "Mistral Open Orca (larger, slow)"])
|
233 |
with gr.Row():
|
234 |
-
gpu_layer_choice = gr.Slider(label="Choose number of model layers to send to GPU (WARNING: please don't modify unless you have a GPU).", value=0, minimum=0, maximum=5, step = 1, visible=
|
235 |
change_model_button = gr.Button(value="Load model", scale=0)
|
236 |
load_text = gr.Text(label="Load status")
|
237 |
|
|
|
80 |
torch_device = chatf.torch_device
|
81 |
|
82 |
if model_type == "Mistral Open Orca (larger, slow)":
|
83 |
+
if torch_device == "cuda":
|
84 |
+
gpu_config.update_gpu(12)
|
85 |
+
else:
|
86 |
+
gpu_config.update_gpu(gpu_layers)
|
87 |
+
cpu_config.update_gpu(gpu_layers)
|
88 |
|
89 |
print("Loading with", cpu_config.gpu_layers, "model layers sent to GPU.")
|
90 |
|
|
|
233 |
with gr.Tab("Advanced features"):
|
234 |
model_choice = gr.Radio(label="Choose a chat model", value="Flan Alpaca (small, fast)", choices = ["Flan Alpaca (small, fast)", "Mistral Open Orca (larger, slow)"])
|
235 |
with gr.Row():
|
236 |
+
gpu_layer_choice = gr.Slider(label="Choose number of model layers to send to GPU (WARNING: please don't modify unless you have a GPU).", value=0, minimum=0, maximum=5, step = 1, visible=False)
|
237 |
change_model_button = gr.Button(value="Load model", scale=0)
|
238 |
load_text = gr.Text(label="Load status")
|
239 |
|
chatfuncs/chatfuncs.py
CHANGED
@@ -69,7 +69,7 @@ kw_model = pipeline("feature-extraction", model="sentence-transformers/all-MiniL
|
|
69 |
|
70 |
if torch.cuda.is_available():
|
71 |
torch_device = "cuda"
|
72 |
-
gpu_layers =
|
73 |
else:
|
74 |
torch_device = "cpu"
|
75 |
gpu_layers = 0
|
|
|
69 |
|
70 |
if torch.cuda.is_available():
|
71 |
torch_device = "cuda"
|
72 |
+
gpu_layers = 12
|
73 |
else:
|
74 |
torch_device = "cpu"
|
75 |
gpu_layers = 0
|