Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

fix: ExLlama Backend Context Size & Rope Scaling #1311

Merged
merged 2 commits into from
Nov 21, 2023
Merged

fix: ExLlama Backend Context Size & Rope Scaling #1311

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ab49d85
fix: context_size not propagated to exllama backend
ok2sh Nov 21, 2023
e3156b5
fix: exllama rope scaling
ok2sh Nov 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 13 additions & 0 deletions backend/python/exllama/exllama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,19 @@ def LoadModel(self, request, context):

config = ExLlamaConfig(model_config_path) # create config from config.json
config.model_path = model_path # supply path to model weights file
if (request.ContextSize):
config.max_seq_len = request.ContextSize # override max sequence length
config.max_attention_size = request.ContextSize**2 # Should be set to context_size^2.
# https://github.com/turboderp/exllama/issues/220#issuecomment-1720324163

# Set Rope scaling.
if (request.RopeFreqScale):
# Alpha value for Rope scaling.
# Higher value increases context but adds perplexity.
# alpha_value and compress_pos_emb are mutually exclusive.
# https://github.com/turboderp/exllama/issues/115
config.alpha_value = request.RopeFreqScale
config.calculate_rotary_embedding_base()

model = ExLlama(config) # create ExLlama instance and load the weights
tokenizer = ExLlamaTokenizer(tokenizer_path) # create tokenizer from tokenizer model file
Expand Down