dkdaniz commited on
Commit
48e851c
1 Parent(s): ecf5eb4

Update constants.py

Browse files
Files changed (1) hide show
  1. constants.py +3 -3
constants.py CHANGED
@@ -28,13 +28,13 @@ CHROMA_SETTINGS = Settings(
28
  )
29
 
30
  # Context Window and Max New Tokens
31
- CONTEXT_WINDOW_SIZE = 2048
32
  MAX_NEW_TOKENS = CONTEXT_WINDOW_SIZE # int(CONTEXT_WINDOW_SIZE/4)
33
 
34
  #### If you get a "not enough space in the buffer" error, you should reduce the values below, start with half of the original values and keep halving the value until the error stops appearing
35
 
36
- N_GPU_LAYERS = 20 # Llama-2-70B has 83 layers
37
- N_BATCH = 512
38
 
39
  ### From experimenting with the Llama-2-7B-Chat-GGML model on 8GB VRAM, these values work:
40
  # N_GPU_LAYERS = 20
 
28
  )
29
 
30
  # Context Window and Max New Tokens
31
+ CONTEXT_WINDOW_SIZE = 1024
32
  MAX_NEW_TOKENS = CONTEXT_WINDOW_SIZE # int(CONTEXT_WINDOW_SIZE/4)
33
 
34
  #### If you get a "not enough space in the buffer" error, you should reduce the values below, start with half of the original values and keep halving the value until the error stops appearing
35
 
36
+ N_GPU_LAYERS = 40 # Llama-2-70B has 83 layers
37
+ N_BATCH = 1024
38
 
39
  ### From experimenting with the Llama-2-7B-Chat-GGML model on 8GB VRAM, these values work:
40
  # N_GPU_LAYERS = 20