theroyallab/tabbyAPI
1
0
Fork 0
mirror of https://github.com/theroyallab/tabbyAPI.git synced 2026-03-14 15:57:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

API: Fix num_experts_per_token reporting

Browse Source 
This wasn't linked to the model config. This value can be 1 if
a MoE model isn't loaded.

Signed-off-by: kingbri <bdashore3@proton.me>
This commit is contained in:
kingbri
2023-12-28 00:31:14 -05:00
parent c5bbfd97b2
commit 3622710582
2 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
config_sample.yml
View File

@@ -77,7 +77,7 @@ model:
# NOTE: Only works with chat completion message lists!
prompt_template:
# Number of experts to use per token. Loads from the model's config.json if not specified (default: None)
# Number of experts to use PER TOKEN. Fetched from the model's config.json if not specified (default: None)
# WARNING: Don't set this unless you know what you're doing!
# NOTE: For MoE models (ex. Mixtral) only!
num_experts_per_token:

1
main.py
View File

@@ -111,6 +111,7 @@ async def get_current_model():
max_seq_len=MODEL_CONTAINER.config.max_seq_len,
cache_mode="FP8" if MODEL_CONTAINER.cache_fp8 else "FP16",
prompt_template=prompt_template.name if prompt_template else None,
num_experts_per_token=MODEL_CONTAINER.config.num_experts_per_token,
),
logging=gen_logging.PREFERENCES,
)