Skip to content

GitLab

Unverified Commit 7505b3ca authored by Fengzhe Zhou's avatar Fengzhe Zhou Committed by GitHub
Browse files

[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
parent 6c711cb2
Hide whitespace changes
Inline Side-by-side
Showing with 120 additions and 334 deletions
configs/models/hf_internlm/hf_internlm2_chat_20b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
_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),
],
)
from opencompass.models import HuggingFacewithChatTemplate
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFacewithChatTemplate,
abbr='internlm2-chat-20b-hf',
path="internlm/internlm2-chat-20b",
tokenizer_path='internlm/internlm2-chat-20b',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
path='internlm/internlm2-chat-20b',
max_out_len=1024,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=2, num_procs=1),
end_str='<|im_end|>',
generation_kwargs = {"eos_token_id": [2, 92542]},
batch_padding=True,
run_cfg=dict(num_gpus=2),
stop_words=['</s>', '<|im_end|>'],
)
]
configs/models/hf_internlm/hf_internlm2_chat_20b_sft.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
_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),
],
)
from opencompass.models import HuggingFacewithChatTemplate
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFacewithChatTemplate,
abbr='internlm2-chat-20b-sft-hf',
path="internlm/internlm2-chat-20b-sft",
tokenizer_path='internlm/internlm2-chat-20b-sft',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
path='internlm/internlm2-chat-20b-sft',
max_out_len=1024,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=2, num_procs=1),
end_str='<|im_end|>',
generation_kwargs = {"eos_token_id": [2, 92542]},
batch_padding=True,
run_cfg=dict(num_gpus=2),
stop_words=['</s>', '<|im_end|>'],
)
]
configs/models/hf_internlm/hf_internlm2_chat_7b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
_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),
],
)
from opencompass.models import HuggingFacewithChatTemplate
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFacewithChatTemplate,
abbr='internlm2-chat-7b-hf',
path="internlm/internlm2-chat-7b",
tokenizer_path='internlm/internlm2-chat-7b',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
path='internlm/internlm2-chat-7b',
max_out_len=1024,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=1, num_procs=1),
end_str='<|im_end|>',
generation_kwargs = {"eos_token_id": [2, 92542]},
batch_padding=True,
run_cfg=dict(num_gpus=1),
stop_words=['</s>', '<|im_end|>'],
)
]
configs/models/hf_internlm/hf_internlm2_chat_7b_sft.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
_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),
],
)
from opencompass.models import HuggingFacewithChatTemplate
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFacewithChatTemplate,
abbr='internlm2-chat-7b-sft-hf',
path="internlm/internlm2-chat-7b-sft",
tokenizer_path='internlm/internlm2-chat-7b-sft',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
path='internlm/internlm2-chat-7b-sft',
max_out_len=1024,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=1, num_procs=1),
end_str='<|im_end|>',
generation_kwargs = {"eos_token_id": [2, 92542]},
batch_padding=True,
run_cfg=dict(num_gpus=1),
stop_words=['</s>', '<|im_end|>'],
)
]
configs/models/hf_internlm/hf_internlm2_chat_math_20b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
_meta_template = dict(
round=[
dict(role='HUMAN', begin='[UNUSED_TOKEN_146]user\n', end='[UNUSED_TOKEN_145]\n'),
dict(role='BOT', begin='[UNUSED_TOKEN_146]assistant\n', end='[UNUSED_TOKEN_145]\n', generate=True),
],
eos_token_id=92542
)
from opencompass.models import HuggingFacewithChatTemplate
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFacewithChatTemplate,
abbr='internlm2-chat-math-20b-hf',
path="internlm/internlm2-math-20b",
tokenizer_path='internlm/internlm2-math-20b',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
path='internlm/internlm2-math-20b',
max_out_len=1024,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=2, num_procs=1),
end_str='[UNUSED_TOKEN_145]',
run_cfg=dict(num_gpus=2),
stop_words=['</s>', '<|im_end|>'],
)
]
configs/models/hf_internlm/hf_internlm2_chat_math_20b_with_system.py
View file @ 7505b3ca
......@@ -7,7 +7,6 @@ _meta_template = dict(
dict(role='SYSTEM', begin='[UNUSED_TOKEN_146]system\n', end='[UNUSED_TOKEN_145]\n'),
dict(role='BOT', begin='[UNUSED_TOKEN_146]assistant\n', end='[UNUSED_TOKEN_145]\n', generate=True),
],
eos_token_id=92542
)
models = [
......
configs/models/hf_internlm/hf_internlm2_chat_math_7b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
_meta_template = dict(
round=[
dict(role='HUMAN', begin='[UNUSED_TOKEN_146]user\n', end='[UNUSED_TOKEN_145]\n'),
dict(role='BOT', begin='[UNUSED_TOKEN_146]assistant\n', end='[UNUSED_TOKEN_145]\n', generate=True),
],
eos_token_id=92542
)
from opencompass.models import HuggingFacewithChatTemplate
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFacewithChatTemplate,
abbr='internlm2-chat-math-7b-hf',
path="internlm/internlm2-math-7b",
tokenizer_path='internlm/internlm2-math-7b',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
path='internlm/internlm2-math-7b',
max_out_len=1024,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=1, num_procs=1),
end_str='[UNUSED_TOKEN_145]',
run_cfg=dict(num_gpus=1),
stop_words=['</s>', '<|im_end|>'],
)
]
configs/models/hf_internlm/hf_internlm2_chat_math_7b_with_system.py
View file @ 7505b3ca
......@@ -7,7 +7,6 @@ _meta_template = dict(
dict(role='SYSTEM', begin='[UNUSED_TOKEN_146]system\n', end='[UNUSED_TOKEN_145]\n'),
dict(role='BOT', begin='[UNUSED_TOKEN_146]assistant\n', end='[UNUSED_TOKEN_145]\n', generate=True),
],
eos_token_id=92542
)
models = [
......
configs/models/hf_internlm/hf_internlm2_math_20b.py 0 → 100644
View file @ 7505b3ca
from opencompass.models import HuggingFaceBaseModel
models = [
dict(
type=HuggingFaceBaseModel,
abbr='internlm2-math-20b-hf',
path="internlm/internlm2-math-20b",
max_out_len=1024,
batch_size=8,
run_cfg=dict(num_gpus=2),
)
]
configs/models/hf_internlm/hf_internlm2_math_7b.py 0 → 100644
View file @ 7505b3ca
from opencompass.models import HuggingFaceBaseModel
models = [
dict(
type=HuggingFaceBaseModel,
abbr='internlm2-math-7b-hf',
path="internlm/internlm2-math-7b",
max_out_len=1024,
batch_size=8,
run_cfg=dict(num_gpus=1),
)
]
configs/models/hf_internlm/hf_internlm_20b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
from opencompass.models import HuggingFaceBaseModel
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFaceBaseModel,
abbr='internlm-20b-hf',
path="internlm/internlm-20b",
tokenizer_path='internlm/internlm-20b',
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
max_out_len=1024,
batch_size=8,
model_kwargs=dict(trust_remote_code=True, device_map='auto'),
run_cfg=dict(num_gpus=2, num_procs=1),
run_cfg=dict(num_gpus=2),
)
]
configs/models/hf_internlm/hf_internlm_7b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
from opencompass.models import HuggingFaceBaseModel
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFaceBaseModel,
abbr='internlm-7b-hf',
path="internlm/internlm-7b",
tokenizer_path='internlm/internlm-7b',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
max_out_len=1024,
batch_size=8,
run_cfg=dict(num_gpus=1, num_procs=1),
run_cfg=dict(num_gpus=1),
)
]
configs/models/hf_internlm/hf_internlm_chat_7b_8k.py deleted 100644 → 0
View file @ 6c711cb2
from opencompass.models import HuggingFaceCausalLM
_meta_template = dict(
round=[
dict(role='HUMAN', begin='<|User|>:', end='\n'),
dict(role='BOT', begin='<|Bot|>:', end='<eoa>\n', generate=True),
],
)
models = [
dict(
type=HuggingFaceCausalLM,
abbr='internlm-chat-7b-8k-hf',
path="internlm/internlm-chat-7b-8k",
tokenizer_path='internlm/internlm-chat-7b-8k',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=1, num_procs=1),
end_str='<eoa>',
)
]
configs/models/hf_internlm/hf_internlm_chat_7b_v1_1.py deleted 100644 → 0
View file @ 6c711cb2
from opencompass.models import HuggingFaceCausalLM
_meta_template = dict(
round=[
dict(role='HUMAN', begin='<|User|>:', end='\n'),
dict(role='BOT', begin='<|Bot|>:', end='<eoa>\n', generate=True),
],
)
models = [
dict(
type=HuggingFaceCausalLM,
abbr='internlm-chat-7b-v1.1-hf',
path="internlm/internlm-chat-7b-v1_1",
tokenizer_path='internlm/internlm-chat-7b-v1_1',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
max_seq_len=2048,
batch_size=8,
meta_template=_meta_template,
run_cfg=dict(num_gpus=1, num_procs=1),
end_str='<eoa>',
)
]
configs/models/hf_internlm/lmdeploy_internlm2_20b.py 0 → 100644
View file @ 7505b3ca
from opencompass.models.turbomind import TurboMindModel
models = [
dict(
type=TurboMindModel,
abbr="internlm2-20b-turbomind",
path="internlm/internlm2-20b",
engine_config=dict(
session_len=32768,
max_batch_size=32,
model_name="internlm2-20b",
tp=2,
),
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=2, num_procs=1),
)
]
configs/models/hf_internlm/lmdeploy_internlm2_chat_20b.py
View file @ 7505b3ca
......@@ -15,9 +15,8 @@ models = [
path="internlm/internlm2-chat-20b",
meta_template=_meta_template,
engine_config=dict(
session_len=210000,
max_batch_size=8,
rope_scaling_factor=3.0,
session_len=32768,
max_batch_size=32,
model_name="internlm2-chat-20b",
tp=2,
stop_words=[2, 92542],
......@@ -29,8 +28,8 @@ models = [
max_new_tokens=2000,
),
max_out_len=2000,
max_seq_len=210000,
batch_size=1,
max_seq_len=32768,
batch_size=32,
concurrency=8,
run_cfg=dict(num_gpus=2, num_procs=1),
)
......
configs/models/hf_internlm/lmdeploy_internlm2_chat_7b.py
View file @ 7505b3ca
......@@ -15,9 +15,8 @@ models = [
path="internlm/internlm2-chat-7b",
meta_template=_meta_template,
engine_config=dict(
session_len=210000,
max_batch_size=8,
rope_scaling_factor=2.0,
session_len=32768,
max_batch_size=32,
model_name="internlm2-chat-7b",
tp=1,
stop_words=[2, 92542],
......@@ -29,8 +28,8 @@ models = [
max_new_tokens=2000,
),
max_out_len=2000,
max_seq_len=210000,
batch_size=1,
max_seq_len=32768,
batch_size=32,
concurrency=8,
run_cfg=dict(num_gpus=1, num_procs=1),
)
......
configs/models/hf_llama/hf_llama2_13b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
from opencompass.models import HuggingFaceBaseModel
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFaceBaseModel,
abbr='llama-2-13b-hf',
path="meta-llama/Llama-2-13b-hf",
tokenizer_path='meta-llama/Llama-2-13b-hf',
tokenizer_kwargs=dict(padding_side='left',
truncation_side='left',
use_fast=False,
),
max_out_len=100,
max_seq_len=2048,
path='meta-llama/Llama-2-13b-hf',
max_out_len=1024,
batch_size=8,
model_kwargs=dict(device_map='auto'),
batch_padding=False, # if false, inference with for-loop without batch padding
run_cfg=dict(num_gpus=2, num_procs=1),
run_cfg=dict(num_gpus=1),
)
]
configs/models/hf_llama/hf_llama2_13b_chat.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
_meta_template = dict(
round=[
dict(role="HUMAN", begin='[INST] ', end=' [/INST]'),
dict(role="BOT", begin=' ', end=' ', generate=True),
],
)
from opencompass.models import HuggingFacewithChatTemplate
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFacewithChatTemplate,
abbr='llama-2-13b-chat-hf',
path="meta-llama/Llama-2-13b-chat-hf",
tokenizer_path='meta-llama/Llama-2-13b-chat-hf',
model_kwargs=dict(
device_map='auto'
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
),
meta_template=_meta_template,
max_out_len=100,
max_seq_len=2048,
path='meta-llama/Llama-2-13b-chat-hf',
max_out_len=1024,
batch_size=8,
run_cfg=dict(num_gpus=2, num_procs=1),
end_str='[INST]',
batch_padding=True,
run_cfg=dict(num_gpus=1),
)
]
configs/models/hf_llama/hf_llama2_70b.py
View file @ 7505b3ca
from opencompass.models import HuggingFaceCausalLM
from opencompass.models import HuggingFaceBaseModel
models = [
dict(
type=HuggingFaceCausalLM,
type=HuggingFaceBaseModel,
abbr='llama-2-70b-hf',
path="meta-llama/Llama-2-70b-hf",
tokenizer_path='meta-llama/Llama-2-70b-hf',
tokenizer_kwargs=dict(padding_side='left',
truncation_side='left',
use_fast=False,
),
max_out_len=100,
max_seq_len=2048,
path='meta-llama/Llama-2-70b-hf',
max_out_len=1024,
batch_size=8,
model_kwargs=dict(device_map='auto'),
batch_padding=False, # if false, inference with for-loop without batch padding
run_cfg=dict(num_gpus=4, num_procs=1),
run_cfg=dict(num_gpus=4),
)
]
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment