OpenCompass/configs/models/hf_internlm/lmdeploy_internlm2_chat_7b.py
Fengzhe Zhou 7505b3cadf
[Feature] Add huggingface apply_chat_template (#1098)
* add TheoremQA with 5-shot

* add huggingface_above_v4_33 classes

* use num_worker partitioner in cli

* update theoremqa

* update TheoremQA

* add TheoremQA

* rename theoremqa -> TheoremQA

* update TheoremQA output path

* rewrite many model configs

* update huggingface

* further update

* refine configs

* update configs

* update configs

* add configs/eval_llama3_instruct.py

* add summarizer multi faceted

* update bbh datasets

* update configs/models/hf_llama/lmdeploy_llama3_8b_instruct.py

* rename class

* update readme

* update hf above v4.33
2024-05-14 14:50:16 +08:00

37 lines
942 B
Python

from opencompass.models.turbomind import TurboMindModel
_meta_template = dict(
round=[
dict(role="HUMAN", begin="<|im_start|>user\n", end="<|im_end|>\n"),
dict(role="BOT", begin="<|im_start|>assistant\n", end="<|im_end|>\n", generate=True),
],
)
models = [
dict(
type=TurboMindModel,
abbr="internlm2-chat-7b-turbomind",
path="internlm/internlm2-chat-7b",
meta_template=_meta_template,
engine_config=dict(
session_len=32768,
max_batch_size=32,
model_name="internlm2-chat-7b",
tp=1,
stop_words=[2, 92542],
),
gen_config=dict(
top_k=1,
top_p=0.8,
temperature=1.0,
max_new_tokens=2000,
),
max_out_len=2000,
max_seq_len=32768,
batch_size=32,
concurrency=8,
run_cfg=dict(num_gpus=1, num_procs=1),
)
]