Spaces:
Runtime error
Runtime error
lalanikarim
commited on
Commit
•
e67a4d6
1
Parent(s):
1e62eb5
added documentation for GPU and Mac support
Browse files
main.py
CHANGED
@@ -38,19 +38,25 @@ def create_chain(system_prompt):
|
|
38 |
# callback_manager = CallbackManager([stream_handler])
|
39 |
|
40 |
(repo_id, model_file_name) = ("TheBloke/Mistral-7B-Instruct-v0.1-GGUF",
|
41 |
-
"mistral-7b-instruct-v0.1.
|
42 |
|
43 |
model_path = hf_hub_download(repo_id=repo_id,
|
44 |
filename=model_file_name,
|
45 |
repo_type="model")
|
46 |
|
47 |
# initialize LlamaCpp llm model
|
|
|
|
|
|
|
48 |
llm = LlamaCpp(
|
49 |
model_path=model_path,
|
50 |
temperature=0,
|
51 |
max_tokens=512,
|
52 |
top_p=1,
|
53 |
# callback_manager=callback_manager,
|
|
|
|
|
|
|
54 |
verbose=False,
|
55 |
streaming=True,
|
56 |
stop=["Human:"]
|
|
|
38 |
# callback_manager = CallbackManager([stream_handler])
|
39 |
|
40 |
(repo_id, model_file_name) = ("TheBloke/Mistral-7B-Instruct-v0.1-GGUF",
|
41 |
+
"mistral-7b-instruct-v0.1.Q4_0.gguf")
|
42 |
|
43 |
model_path = hf_hub_download(repo_id=repo_id,
|
44 |
filename=model_file_name,
|
45 |
repo_type="model")
|
46 |
|
47 |
# initialize LlamaCpp llm model
|
48 |
+
# n_gpu_layers, n_batch, and n_ctx are for GPU support.
|
49 |
+
# When not set, CPU will be used.
|
50 |
+
# set 1 for mac m2, and higher numbers based on your GPU support
|
51 |
llm = LlamaCpp(
|
52 |
model_path=model_path,
|
53 |
temperature=0,
|
54 |
max_tokens=512,
|
55 |
top_p=1,
|
56 |
# callback_manager=callback_manager,
|
57 |
+
# n_gpu_layers=1,
|
58 |
+
# n_batch=512,
|
59 |
+
# n_ctx=4096,
|
60 |
verbose=False,
|
61 |
streaming=True,
|
62 |
stop=["Human:"]
|