mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-01-29 23:23:48 +08:00
32 lines
699 B
YAML
32 lines
699 B
YAML
model:
|
|
model_type: llm
|
|
llm_type: openai-apicompatible
|
|
llm_name: <llm-name>
|
|
container: null
|
|
use_cache: false
|
|
use_chat_endpoint: false
|
|
is_hf_model: false
|
|
tokenizer_path: <tokenizer-path>
|
|
base_url: http://localhost:12478/v1
|
|
|
|
evaluations:
|
|
- eval_type: automatic
|
|
eval_subtype: lm_eval_harness
|
|
native_args: null
|
|
hf_token: null
|
|
tasks:
|
|
- task_name: wikilingua_english
|
|
task_config: null
|
|
num_fewshot: 3
|
|
batch_size: 4
|
|
bootstrap_iters: 100000
|
|
limit: 0.05
|
|
inference_params:
|
|
use_greedy: true
|
|
top_p: 0.00001
|
|
top_k: 1
|
|
temperature: 0.00001
|
|
stop:
|
|
- <|endoftext|>
|
|
tokens_to_generate: 256
|