Skip to content

GitLab

Commit 7d06d0f9 authored by yangzhong's avatar yangzhong
Browse files

Update files

parent 2f320edb
Pipeline #2827 failed with stages
in 0 seconds
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 2700 additions and 0 deletions
CITATION.cff 0 → 100644
View file @ 7d06d0f9
cff-version: 1.2.0
message: "If you find our resources useful, please cite our paper as below."
authors:
- family-names: "Cui"
given-names: "Yiming"
orcid: "https://orcid.org/0000-0002-2452-375X"
- family-names: "Yang"
given-names: "Ziqing"
- family-names: "Yao"
given-names: "Xin"
title: "Chinese LLaMA and Alpaca 2"
version: 1.0
date-released: 2023-07-28
url: "https://github.com/ymcui/Chinese-LLaMA-Alpaca-2"
preferred-citation:
type: article
authors:
- family-names: "Cui"
given-names: "Yiming"
orcid: "https://orcid.org/0000-0002-2452-375X"
- family-names: "Yang"
given-names: "Ziqing"
- family-names: "Yao"
given-names: "Xin"
title: "Efficient and Effective Text Encoding for Chinese LLaMA and Alpaca"
journal: "arXiv pre-print"
year: 2023
url: "https://arxiv.org/abs/2304.08177"
\ No newline at end of file
LICENSE 0 → 100644
View file @ 7d06d0f9
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright 2023 Yiming Cui, Ziqing Yang, Xin Yao
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
\ No newline at end of file
README_EN.md 0 → 100644
View file @ 7d06d0f9
This diff is collapsed.
download_model.py 0 → 100644
View file @ 7d06d0f9
import os
os.environ['CURL_CA_BUNDLE'] = ''
os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'
from huggingface_hub import hf_hub_download, snapshot_download
snapshot_download(repo_id="hfl/chinese-llama-2-7b", local_dir='./pre_model')
examples/README.md 0 → 100644
View file @ 7d06d0f9
## 输出示例
本目录针对Chinese-Alpaca-2模型给出参考输出样例，其目的是帮助用户快速了解模型输出情况，同时也有助于排查下载的模型是否和预期输出一致。输出样本来自于模型在线对战题库（共10个类别），每个类别选择3道题进行展示。
- [Chinese-Alpaca-2-7B输出样例](./alpaca-2-7b.md)
- [Chinese-Alpaca-2-13B输出样例](./alpaca-2-13b.md)
**📊 模型在线对战**：[http://llm-arena.ymcui.com](http://llm-arena.ymcui.com/)
examples/alpaca-2-13b.md 0 → 100644
View file @ 7d06d0f9
This diff is collapsed.
examples/alpaca-2-7b.md 0 → 100644
View file @ 7d06d0f9
This diff is collapsed.
notebooks/gradio_web_demo.ipynb 0 → 100644
View file @ 7d06d0f9
This diff is collapsed.
prompts/README.md 0 → 100644
View file @ 7d06d0f9
## 系统指令 System Prompts
### alpaca-2.txt (default)
这个文件是训练时采用的默认系统指令，内容极简，因此回复长度上略短于一代Pro系列模型。
This file is the default system prompt used in the SFT phase, which is simple. Thus, the length of the response may be shorter than 1st-gen Pro series models.
### alpaca-2-long.txt
这个文件是增加模型回复内容长度的系统指令示例，用户可根据实际情况自行参照修改。但建议保留最原始的`alpaca-2.txt`中的内容，在此基础上进行自定义系统指令的编写。
This file is an improved system prompt sample to extend the response length. The users can modify this prompt if necessary. However, we suggest keep the original content in `alpaca-2.txt` and add your customized prompt based on this.
prompts/alpaca-2-long.txt 0 → 100644
View file @ 7d06d0f9
You are a helpful assistant. 你是一个乐于助人的助手。请你提供专业、有逻辑、内容真实、有价值的详细回复。
\ No newline at end of file
prompts/alpaca-2.txt 0 → 100644
View file @ 7d06d0f9
You are a helpful assistant. 你是一个乐于助人的助手。
\ No newline at end of file
requirements.txt 0 → 100644
View file @ 7d06d0f9
peft==0.3.0
#torch==2.0.1
transformers==4.35.0
sentencepiece==0.1.99
bitsandbytes==0.41.1
scripts/README.md 0 → 100644
View file @ 7d06d0f9
# 代码与脚本 Code and Scripts
### training/
预训练与指令精调代码，Wiki：
- 预训练：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/pt_scripts_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/pt_scripts_zh)
- 指令精调：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/sft_scripts_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/sft_scripts_zh)
Pre-training and instruction finetuning code, Wiki:
- Pre-training: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/pt_scripts_en
- Instruction finetuning: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/sft_scripts_en
### inference/
使用🤗transformers进行推理，Wiki：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/inference_with_transformers_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/inference_with_transformers_zh)
Inference using 🤗transformers, Wiki: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/inference_with_transformers_en
### openai_server_demo/
使用fastapi实现的仿OPENAI API风格的服务器，Wiki：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/api_calls_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/api_calls_zh)
A server that implements OPENAI API using fastapi, Wiki: [https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/api_calls_en](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/api_calls_en)
### ceval/
C-Eval评测脚本，Wiki：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/ceval_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/ceval_zh)
Inference script for C-Eval, Wiki: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/ceval_en
### cmmlu/
CMMLU评测脚本，Wiki：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/cmmlu_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/cmmlu_zh)
Inference script for CMMLU, Wiki: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/cmmlu_en
### longbench/
LongBench评测脚本，Wiki：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/longbench_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/longbench_zh)
Inference script for LongBench, Wiki: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/longbench_en
### llama-cpp/
llama.cpp启动脚本、server脚本，Wiki：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/llamacpp_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/llamacpp_zh)
launch script and server script for llama.cpp, Wiki: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/llamacpp_en
### attn_ang_long_ctx_patches.py
Memory efficient attention补丁和NTK上下文拓展方法补丁。
Patches for memory efficient attention and NTK context size scaling.
### merge_llama2_with_chinese_lora_low_mem.py
低资源版合并LLaMA-2/Alpaca-2 LoRA脚本，Wiki：[https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/manual_conversion_zh](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/manual_conversion_zh)
Script for merging LLaMA-2/Alpaca-2 LoRA (low-resource version). Wiki: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/wiki/manual_conversion_en
### tokenizer/
Chinese-LLaMA-2 & Chinese-Alpaca-2 tokenizer
\ No newline at end of file
scripts/attn_and_long_ctx_patches.py 0 → 100644
View file @ 7d06d0f9
import torch
from torch import nn
from typing import Optional, Tuple, Union
import transformers
from transformers.models.llama.modeling_llama import apply_rotary_pos_emb, rotate_half
import math
try:
from xformers import ops as xops
except ImportError:
xops = None
print(
"Xformers is not installed correctly. If you want to use memory_efficient_attention use the following command to install Xformers\npip install xformers."
)
STORE_KV_BEFORE_ROPE = False
USE_MEM_EFF_ATTENTION = False
ALPHA = 1.0
AUTO_COEFF = 1.0
SCALING_FACTOR = None
def apply_rotary_pos_emb_single(q, cos, sin, position_ids):
# The first two dimensions of cos and sin are always 1, so we can `squeeze` them.
cos = cos.squeeze(1).squeeze(0) # [seq_len, dim]
sin = sin.squeeze(1).squeeze(0) # [seq_len, dim]
cos = cos[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim]
sin = sin[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim]
q_embed = (q * cos) + (rotate_half(q) * sin)
return q_embed
def xformers_forward(
self,
hidden_states: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_value: Optional[Tuple[torch.Tensor]] = None,
output_attentions: bool = False,
use_cache: bool = False,
padding_mask=None,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
bsz, q_len, _ = hidden_states.size()
query_states = self.q_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
key_states = self.k_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
value_states = self.v_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
kv_seq_len = key_states.shape[-2]
past_kv_len = 0
if past_key_value is not None:
past_kv_len = past_key_value[0].shape[-2]
kv_seq_len += past_kv_len
if STORE_KV_BEFORE_ROPE is False:
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
# [bsz, nh, t, hd]
if past_key_value is not None:
# reuse k, v, self_attention
key_states = torch.cat([past_key_value[0], key_states], dim=2)
value_states = torch.cat([past_key_value[1], value_states], dim=2)
past_key_value = (key_states, value_states) if use_cache else None
else:
if past_key_value is not None:
# reuse k, v, self_attention
key_states = torch.cat([past_key_value[0], key_states], dim=2)
value_states = torch.cat([past_key_value[1], value_states], dim=2)
past_key_value = (key_states, value_states) if use_cache else None
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
query_states = apply_rotary_pos_emb_single(query_states, cos, sin, position_ids)
position_ids = torch.arange(kv_seq_len, dtype=torch.long, device=cos.device)
position_ids = position_ids.unsqueeze(0).view(-1, kv_seq_len)
key_states = apply_rotary_pos_emb_single(key_states, cos, sin, position_ids)
pad_query = False
if xops is not None and USE_MEM_EFF_ATTENTION:
attn_weights = None
query_states = query_states.transpose(1, 2)
key_states = key_states.transpose(1, 2)
value_states = value_states.transpose(1, 2)
if query_states.size(1)==1 and key_states.size(1)>1:
attn_bias = None
elif query_states.size(1)<key_states.size(1) and key_states.size(1)>1 and past_kv_len > 0:
attn_bias = xops.LowerTriangularMask()
query_states = torch.cat(
(
torch.full(
(bsz, past_kv_len, self.num_heads, self.head_dim),
0.0,
dtype=query_states.dtype,
device=query_states.device,
),
query_states,
),
dim=1,
)
pad_query = True
else:
attn_bias = xops.LowerTriangularMask()
attn_output = xops.memory_efficient_attention(
query_states, key_states, value_states, attn_bias=attn_bias, p=0)
else:
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
raise ValueError(
f"Attention weights should be of size {(bsz * self.num_heads, q_len, kv_seq_len)}, but is"
f" {attn_weights.size()}"
)
if attention_mask is not None:
if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
raise ValueError(
f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
)
attn_weights = attn_weights + attention_mask
attn_weights = torch.max(
attn_weights, torch.tensor(torch.finfo(attn_weights.dtype).min, device=attn_weights.device)
)
# upcast attention to fp32
attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
attn_output = torch.matmul(attn_weights, value_states)
if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
raise ValueError(
f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
f" {attn_output.size()}"
)
attn_output = attn_output.transpose(1, 2)
if pad_query:
attn_output = attn_output[:,past_kv_len:]
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
attn_output = self.o_proj(attn_output)
if not output_attentions:
attn_weights = None
return attn_output, attn_weights, past_key_value
old_init = transformers.models.llama.modeling_llama.LlamaRotaryEmbedding.__init__
def _set_cos_sin_cache(self, seq_len, device, dtype):
self.max_seq_len_cached = seq_len
t = torch.arange(self.max_seq_len_cached, device=device, dtype=torch.float32)
t = t / self.scaling_factor
freqs = torch.einsum("i,j->ij", t, self.ntk_inv_freq.to(device))
# Different from paper, but it uses a different permutation in order to obtain the same calculation
emb = torch.cat((freqs, freqs), dim=-1)
self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
def adaptive_ntk_init(self, dim, max_position_embeddings=2048, base=10000, device=None, scaling_factor=None):
self.alpha = ALPHA
if SCALING_FACTOR is None:
self.scaling_factor = scaling_factor or 1.0
else:
self.scaling_factor = SCALING_FACTOR
if isinstance(ALPHA,(float,int)):
base = base * ALPHA ** (dim / (dim-2))
self.base = base
elif ALPHA=='auto':
self.base = base
else:
raise ValueError(ALPHA)
old_init(self, dim, max_position_embeddings, base, device)
self.ntk_inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim))
self._set_cos_sin_cache = _set_cos_sin_cache
self._set_cos_sin_cache(
self, seq_len=max_position_embeddings, device=self.ntk_inv_freq.device, dtype=torch.get_default_dtype()
)
def adaptive_ntk_forward(self, x, seq_len=None):
if seq_len > self.max_seq_len_cached:
if isinstance(self.alpha,(float,int)):
self._set_cos_sin_cache(self, seq_len=seq_len, device=x.device, dtype=x.dtype)
elif self.alpha=='auto':
t = torch.arange(seq_len, device=x.device, dtype=torch.float32)
t = t / self.scaling_factor
dim = self.dim
alpha = (seq_len / (self.max_position_embeddings/2) - 1) * AUTO_COEFF
base = self.base * alpha ** (dim / (dim-2))
ntk_inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(x.device) / dim ))
freqs = torch.einsum("i,j->ij", t, ntk_inv_freq)
emb = torch.cat((freqs, freqs), dim=-1).to(x.device)
cos_cached = emb.cos()
sin_cached = emb.sin()
return (
cos_cached[:seq_len].to(dtype=x.dtype),
sin_cached[:seq_len].to(dtype=x.dtype)
)
return (
self.cos_cached[:seq_len].to(dtype=x.dtype),
self.sin_cached[:seq_len].to(dtype=x.dtype)
)
def apply_attention_patch(
use_memory_efficient_attention=False,
store_kv_before_rope=False
):
global USE_MEM_EFF_ATTENTION, STORE_KV_BEFORE_ROPE
if use_memory_efficient_attention is True and xops is not None:
USE_MEM_EFF_ATTENTION = use_memory_efficient_attention
print("USE_XFORMERS_ATTENTION: ", USE_MEM_EFF_ATTENTION)
STORE_KV_BEFORE_ROPE = store_kv_before_rope
print("STORE_KV_BEFORE_ROPE:", STORE_KV_BEFORE_ROPE)
transformers.models.llama.modeling_llama.LlamaAttention.forward = xformers_forward
def apply_ntk_scaling_patch(alpha: Union[float,str], scaling_factor: Optional[float] = None):
global ALPHA
global SCALING_FACTOR
ALPHA = alpha
SCALING_FACTOR = scaling_factor
try:
ALPHA = float(ALPHA)
except ValueError:
if ALPHA!="auto":
raise ValueError(f"Alpha can only be a float or 'auto', but given {ALPHA}")
print(f"Apply NTK scaling with ALPHA={ALPHA}")
if scaling_factor is None:
print(f"The value of scaling factor will be read from model config file, or set to 1.")
else:
print(f"Warning: scaling factor is set to {SCALING_FACTOR}. \
If you set the value by hand, do not forget to update \
max_position_embeddings in the model config file.")
transformers.models.llama.modeling_llama.LlamaRotaryEmbedding.__init__ = adaptive_ntk_init
if hasattr(transformers.models.llama.modeling_llama,'LlamaLinearScalingRotaryEmbedding'):
transformers.models.llama.modeling_llama.LlamaLinearScalingRotaryEmbedding.__init__ = adaptive_ntk_init
transformers.models.llama.modeling_llama.LlamaRotaryEmbedding.forward = adaptive_ntk_forward
\ No newline at end of file
scripts/ceval/eval.py 0 → 100644
View file @ 7d06d0f9
# This code is modified from C-Eval Project: https://github.com/SJTU-LIT/ceval
import os
import argparse
import pandas as pd
import torch
import json
from llama_evaluator import Llama_Evaluator
import time
choices = ["A", "B", "C", "D"]
def main(args, evaluator,take):
assert os.path.exists("subject_mapping.json"), "subject_mapping.json not found!"
with open("subject_mapping.json") as f:
subject_mapping = json.load(f)
filenames = os.listdir("data/val")
subject_list = [val_file.replace("_val.csv","") for val_file in filenames]
accuracy, summary = {}, {}
run_date=time.strftime('%Y-%m-%d_%H-%M-%S',time.localtime(time.time()))
output_dir = args.output_dir
save_result_dir=os.path.join(output_dir,f"take{take}")
if not os.path.exists(save_result_dir):
os.makedirs(save_result_dir,exist_ok=True)
all_answers = {}
for index,subject_name in enumerate(subject_list):
print(f"{index/len(subject_list)} Inference starts at {run_date} on {args.model_path} with subject of {subject_name}!")
val_file_path=os.path.join('data/val',f'{subject_name}_val.csv')
dev_file_path=os.path.join('data/dev',f'{subject_name}_dev.csv')
test_file_path=os.path.join('data/test',f'{subject_name}_test.csv')
val_df=pd.read_csv(val_file_path) if args.do_test is False else pd.read_csv(test_file_path)
dev_df=pd.read_csv(dev_file_path) if args.few_shot else None
correct_ratio, answers = evaluator.eval_subject(subject_name, val_df, dev_df,
save_result_dir=save_result_dir if args.do_save_csv else None,
few_shot=args.few_shot,
cot=args.cot,
with_prompt=args.with_prompt,
constrained_decoding=args.constrained_decoding,
do_test=args.do_test)
print(f"Subject: {subject_name}")
print(f"Acc: {correct_ratio}")
accuracy[subject_name] = correct_ratio
summary[subject_name] = {"score":correct_ratio,
"num":len(val_df),
"correct":correct_ratio*len(val_df)/100}
all_answers[subject_name] = answers
json.dump(all_answers,open(save_result_dir+'/submission.json','w'),ensure_ascii=False,indent=4)
print("Accuracy:")
for k, v in accuracy.items():
print(k, ": ", v)
total_num = 0
total_correct = 0
summary['grouped'] = {
"STEM": {"correct": 0.0, "num": 0},
"Social Science": {"correct": 0.0, "num": 0},
"Humanities": {"correct": 0.0, "num": 0},
"Other": {"correct": 0.0, "num": 0}
}
for subj, info in subject_mapping.items():
group = info[2]
summary['grouped'][group]["num"] += summary[subj]['num']
summary['grouped'][group]["correct"] += summary[subj]['correct']
for group, info in summary['grouped'].items():
info['score'] = info["correct"] / info["num"]
total_num += info["num"]
total_correct += info["correct"]
summary['All'] = {"score": total_correct / total_num, "num": total_num, "correct": total_correct}
json.dump(summary,open(save_result_dir+'/summary.json','w'),ensure_ascii=False,indent=2)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--model_path", type=str)
parser.add_argument("--cot",choices=["False","True"], default="False")
parser.add_argument("--few_shot", choices=["False","True"], default="True")
parser.add_argument("--ntrain", "-k", type=int, default=5)
parser.add_argument("--with_prompt", choices=["False","True"], default="False")
parser.add_argument("--constrained_decoding", choices=["False","True"], default="True")
parser.add_argument("--temperature",type=float,default=0.2)
parser.add_argument("--n_times", default=1,type=int)
parser.add_argument("--do_save_csv", choices=["False","True"], default="False")
parser.add_argument("--output_dir", type=str)
parser.add_argument("--do_test", choices=["False","True"], default="False")
parser.add_argument("--verbose", action="store_true", help="Print detailed information of each example.")
args = parser.parse_args()
args.cot = args.cot == "True"
args.few_shot = args.few_shot == "True"
args.with_prompt = args.with_prompt == "True"
args.constrained_decoding = args.constrained_decoding == "True"
args.do_test = args.do_test == "True"
args.do_save_csv = args.do_save_csv == "True"
if args.constrained_decoding is True:
args.n_times=max(args.n_times,1)
print(args)
device = torch.device(0)
print(device)
evaluator=Llama_Evaluator(
choices=choices,
k=args.ntrain,
model_path=args.model_path,
device=device,
temperature = args.temperature,
verbose = args.verbose
)
for i in range(args.n_times):
main(args,evaluator=evaluator,take=i)
scripts/ceval/evaluator.py 0 → 100644
View file @ 7d06d0f9
# This code is modified from C-Eval Project: https://github.com/SJTU-LIT/ceval
import string
class Evaluator:
def __init__(self, choices, model_name, k=-1):
self.choices = choices
self.model_name = model_name
self.k = k
self.puncs = list(string.punctuation)
def format_example(self, line, include_answer=True):
example = line['question']
for choice in self.choices:
example += f'\n{choice}. {line[f"{choice}"]}'
example += '\n答案：'
if include_answer:
example += f'{line["answer"]}\n\n'
return example
def generate_few_shot_prompt(self, subject, dev_df):
prompt = f"以下是中国关于{subject}考试的单项选择题，请选出其中的正确答案。\n\n"
k = self.k
if self.k == -1:
k = dev_df.shape[0]
for i in range(k):
prompt += self.format_example(dev_df.iloc[i, :])
return prompt
def eval_subject(self, subject_name, test_df, dev_df=None, few_shot=False, save_result_dir=None):
pass
def normalize_answer(self,s):
def white_space_fix(text):
return ' '.join(text.split())
def remove_punc(text):
exclude=set(self.puncs)
return ''.join(ch for ch in text if ch not in exclude)
def lower(text):
return text.lower()
return white_space_fix(remove_punc(lower(s)))
def exact_match(self,pred, target):
return self.normalize_answer(pred)==self.normalize_answer(target)
scripts/ceval/llama_evaluator.py 0 → 100644
View file @ 7d06d0f9
# This code is modified from C-Eval Project: https://github.com/SJTU-LIT/ceval
import os
import re
from tqdm import tqdm
import random
import numpy as np
import torch
from transformers import AutoModelForCausalLM, LlamaTokenizer
from transformers import GenerationConfig
from evaluator import Evaluator
DEFAULT_SYSTEM_PROMPT = """You are a helpful assistant. 你是一个乐于助人的助手。"""
class Llama_Evaluator(Evaluator):
def __init__(self, choices, k, model_path, device, temperature=0.2, verbose=False):
super(Llama_Evaluator, self).__init__(choices, model_path, k)
load_type = torch.float16
self.model_path = model_path
self.device = device
self.verbose = verbose
self.tokenizer = LlamaTokenizer.from_pretrained(model_path, legacy=True)
self.model = AutoModelForCausalLM.from_pretrained(
model_path,
load_in_8bit=False,
torch_dtype=load_type,
low_cpu_mem_usage=True,
device_map='auto',
trust_remote_code=True)
self.generation_config = GenerationConfig(
temperature=temperature,
top_k=40,
top_p=0.9,
do_sample=True,
num_beams=1,
repetition_penalty=1.1,
max_new_tokens=20
)
self.sA_id = self.tokenizer.encode("A", add_special_tokens=False)[0]
self.sB_id = self.tokenizer.encode("B", add_special_tokens=False)[0]
self.sC_id = self.tokenizer.encode("C", add_special_tokens=False)[0]
self.sD_id = self.tokenizer.encode("D", add_special_tokens=False)[0]
self.A_id = self.tokenizer.encode("：A")[-1]
self.B_id = self.tokenizer.encode("：B")[-1]
self.C_id = self.tokenizer.encode("：C")[-1]
self.D_id = self.tokenizer.encode("：D")[-1]
def eval_subject(self, subject_name,
test_df,
dev_df=None,
few_shot=False,
cot=False,
save_result_dir=None,
with_prompt=False,
constrained_decoding=False,
do_test=False):
all_answers = {}
if constrained_decoding is True:
self.generation_config.output_scores = True
self.generation_config.return_dict_in_generate = True
self.generation_config.max_new_tokens = 1
self.generation_config.top_p = 1.0
self.generation_config.top_k = 0
correct_num = 0
if save_result_dir:
result = []
score = []
if few_shot:
if with_prompt:
history = self.generate_alpaca2_few_shot_prompt(subject_name, dev_df, cot=cot)
else:
history = self.generate_llama2_few_shot_prompt(subject_name, dev_df, cot=cot)
else:
history = ''
answers = ['NA'] * len(test_df) if do_test is True else list(test_df['answer'])
for row_index, row in tqdm(test_df.iterrows(), total=len(test_df)):
question = self.format_example(row, include_answer=False, cot=cot,with_prompt=with_prompt)
instruction = question
if with_prompt:
prompt_template = (
"[INST] <<SYS>>\n"
"{system_prompt}\n"
"<</SYS>>\n\n"
"{instruction} [/INST]"
)
instruction = prompt_template.format_map({'instruction': instruction,'system_prompt':DEFAULT_SYSTEM_PROMPT})
instruction = history + instruction
inputs = self.tokenizer(instruction, return_tensors="pt")
generation_output = self.model.generate(
input_ids = inputs["input_ids"].to(self.device),
attention_mask = inputs['attention_mask'].to(self.device),
eos_token_id=self.tokenizer.eos_token_id,
pad_token_id=self.tokenizer.pad_token_id,
generation_config = self.generation_config
)
batch_size, length = inputs.input_ids.shape
if constrained_decoding is True:
logits = generation_output.scores[0][0]
logits = logits.float().cpu().detach()
choices1_logits = logits[[self.sA_id,self.sB_id,self.sC_id,self.sD_id]]
choices2_logits = logits[[self.A_id,self.B_id,self.C_id,self.D_id]]
choicesAll_logits = (choices1_logits + choices2_logits).numpy()
assert not (np.any(np.isinf(choicesAll_logits)) or np.any(np.isnan(choicesAll_logits)))
ans = {0: "A", 1: "B", 2: "C", 3: "D"}[np.argmax(choicesAll_logits)]
response = self.tokenizer.decode([logits.argmax(-1).item()])
else:
response = self.tokenizer.decode(generation_output[0, length:], skip_special_tokens=True)
ans, direct_extract = self.extract_answer(row, response)
if ans == answers[row_index]:
correct_num += 1
correct = 1
else:
correct = 0
if self.verbose is True:
print(f"\n======={str(row_index)}=======")
print(f"question: {question}\n")
print(f"response: {response}\n")
print(f"extracted answer: {ans}")
print(f"ground truth: {answers[row_index]} \n")
if save_result_dir:
result.append(response)
score.append(correct)
all_answers[str(row_index)] = ans
correct_ratio = 100*correct_num/len(answers)
if save_result_dir:
test_df['model_output'] = result
test_df['correctness'] = score
test_df.to_csv(os.path.join(save_result_dir, f'{subject_name}_test.csv'))
return correct_ratio, all_answers
def format_example(self, line, include_answer=True, cot=False, with_prompt=False):
example = line['question']
for choice in self.choices:
example += f'\n{choice}. {line[f"{choice}"]}'
if include_answer:
if cot:
example += "\n答案：让我们一步一步思考，\n" + \
line["explanation"] + f"\n所以答案是{line['answer']}。\n\n"
else:
example += '\n答案：' + line["answer"] + '\n\n'
else:
if with_prompt is False:
if cot:
example += "\n答案：让我们一步一步思考，\n1."
else:
example += '\n答案：'
else:
if cot:
example += "\n答案是什么？让我们一步一步思考，\n1."
else:
example += '\n答案：'
return example
def generate_llama2_few_shot_prompt(self, subject, dev_df, cot=False):
prompt = f"以下是中国关于{subject}考试的单项选择题，请选出其中的正确答案。\n\n"
k = self.k
if self.k == -1:
k = dev_df.shape[0]
for i in range(k):
prompt += self.format_example(
dev_df.iloc[i, :],
include_answer=True,
cot=cot
)
return prompt
def generate_alpaca2_few_shot_prompt(self, subject, dev_df, cot=False):
prompt = f"以下是中国关于{subject}考试的单项选择题，请选出其中的正确答案。\n\n"
prompt_template = (
"[INST] <<SYS>>\n"
"{system_prompt}\n"
"<</SYS>>\n\n"
"{instruction} [/INST]好的，我会结合{subject}相关知识回答"
)
prompt = prompt_template.format_map({'instruction':prompt,'system_prompt':DEFAULT_SYSTEM_PROMPT,'subject':subject})
k = self.k
if self.k == -1:
k = dev_df.shape[0]
for i in range(k):
line = dev_df.iloc[i, :]
q=line['question']
for choice in self.choices:
q += f'\n{choice}. {line[f"{choice}"]}'
a = line['answer']
prompt += "[INST] "+q+"\n答案：[/INST]"+a+"\n"
return prompt
def extract_answer(self, line, gen_ans):
m = re.findall(r'所以答案是(.+?)。', gen_ans, re.M)
if len(m) > 0 and m[-1] in self.choices:
return m[-1], True
answer_patterns = [
r'([ABCD])是正确的',
r'选项([ABCD])正确',
r'答案为([ABCD])',
r'答案是([ABCD])',
r'答案([ABCD])',
r'选择([ABCD])',
r'答案：([ABCD])',
r'选择答案([ABCD])'
]
# RE extraction
for answer_pattern in answer_patterns:
m = re.search(answer_pattern, gen_ans, re.M)
if m:
answer = m.group(1)
return answer, False
# only containing one choice-character
m = re.findall(r'[ABCD]', gen_ans, re.M)
if len(m) >= 1:
answer = m[0]
return answer, False
# only containing one choice-context
choices_dict = {}
pattern = ""
for c in self.choices:
choices_dict[str(line[f'{c}'])] = c
pattern += re.escape(str(line[f'{c}']))+"|"
pattern = pattern[:-1]
m = re.findall(pattern, gen_ans, re.M)
print("w/ escape:",repr(pattern),gen_ans,(len(m)>=1))
if len(m) >= 1:
answer = choices_dict[m[0]]
return answer, False
return random.choice('ABCD'), False
scripts/ceval/subject_mapping.json 0 → 100644
View file @ 7d06d0f9
{
"computer_network": [
"Computer Network",
"\u8ba1\u7b97\u673a\u7f51\u7edc",
"STEM"
],
"operating_system": [
"Operating System",
"\u64cd\u4f5c\u7cfb\u7edf",
"STEM"
],
"computer_architecture": [
"Computer Architecture",
"\u8ba1\u7b97\u673a\u7ec4\u6210",
"STEM"
],
"college_programming": [
"College Programming",
"\u5927\u5b66\u7f16\u7a0b",
"STEM"
],
"college_physics": [
"College Physics",
"\u5927\u5b66\u7269\u7406",
"STEM"
],
"college_chemistry": [
"College Chemistry",
"\u5927\u5b66\u5316\u5b66",
"STEM"
],
"advanced_mathematics": [
"Advanced Mathematics",
"\u9ad8\u7b49\u6570\u5b66",
"STEM"
],
"probability_and_statistics": [
"Probability and Statistics",
"\u6982\u7387\u7edf\u8ba1",
"STEM"
],
"discrete_mathematics": [
"Discrete Mathematics",
"\u79bb\u6563\u6570\u5b66",
"STEM"
],
"electrical_engineer": [
"Electrical Engineer",
"\u6ce8\u518c\u7535\u6c14\u5de5\u7a0b\u5e08",
"STEM"
],
"metrology_engineer": [
"Metrology Engineer",
"\u6ce8\u518c\u8ba1\u91cf\u5e08",
"STEM"
],
"high_school_mathematics": [
"High School Mathematics",
"\u9ad8\u4e2d\u6570\u5b66",
"STEM"
],
"high_school_physics": [
"High School Physics",
"\u9ad8\u4e2d\u7269\u7406",
"STEM"
],
"high_school_chemistry": [
"High School Chemistry",
"\u9ad8\u4e2d\u5316\u5b66",
"STEM"
],
"high_school_biology": [
"High School Biology",
"\u9ad8\u4e2d\u751f\u7269",
"STEM"
],
"middle_school_mathematics": [
"Middle School Mathematics",
"\u521d\u4e2d\u6570\u5b66",
"STEM"
],
"middle_school_biology": [
"Middle School Biology",
"\u521d\u4e2d\u751f\u7269",
"STEM"
],
"middle_school_physics": [
"Middle School Physics",
"\u521d\u4e2d\u7269\u7406",
"STEM"
],
"middle_school_chemistry": [
"Middle School Chemistry",
"\u521d\u4e2d\u5316\u5b66",
"STEM"
],
"veterinary_medicine": [
"Veterinary Medicine",
"\u517d\u533b\u5b66",
"STEM"
],
"college_economics": [
"College Economics",
"\u5927\u5b66\u7ecf\u6d4e\u5b66",
"Social Science"
],
"business_administration": [
"Business Administration",
"\u5de5\u5546\u7ba1\u7406",
"Social Science"
],
"marxism": [
"Marxism",
"\u9a6c\u514b\u601d\u4e3b\u4e49\u57fa\u672c\u539f\u7406",
"Social Science"
],
"mao_zedong_thought": [
"Mao Zedong Thought",
"\u6bdb\u6cfd\u4e1c\u601d\u60f3\u548c\u4e2d\u56fd\u7279\u8272\u793e\u4f1a\u4e3b\u4e49\u7406\u8bba\u4f53\u7cfb\u6982\u8bba",
"Social Science"
],
"education_science": [
"Education Science",
"\u6559\u80b2\u5b66",
"Social Science"
],
"teacher_qualification": [
"Teacher Qualification",
"\u6559\u5e08\u8d44\u683c",
"Social Science"
],
"high_school_politics": [
"High School Politics",
"\u9ad8\u4e2d\u653f\u6cbb",
"Social Science"
],
"high_school_geography": [
"High School Geography",
"\u9ad8\u4e2d\u5730\u7406",
"Social Science"
],
"middle_school_politics": [
"Middle School Politics",
"\u521d\u4e2d\u653f\u6cbb",
"Social Science"
],
"middle_school_geography": [
"Middle School Geography",
"\u521d\u4e2d\u5730\u7406",
"Social Science"
],
"modern_chinese_history": [
"Modern Chinese History",
"\u8fd1\u4ee3\u53f2\u7eb2\u8981",
"Humanities"
],
"ideological_and_moral_cultivation": [
"Ideological and Moral Cultivation",
"\u601d\u60f3\u9053\u5fb7\u4fee\u517b\u4e0e\u6cd5\u5f8b\u57fa\u7840",
"Humanities"
],
"logic": [
"Logic",
"\u903b\u8f91\u5b66",
"Humanities"
],
"law": [
"Law",
"\u6cd5\u5b66",
"Humanities"
],
"chinese_language_and_literature": [
"Chinese Language and Literature",
"\u4e2d\u56fd\u8bed\u8a00\u6587\u5b66",
"Humanities"
],
"art_studies": [
"Art Studies",
"\u827a\u672f\u5b66",
"Humanities"
],
"professional_tour_guide": [
"Professional Tour Guide",
"\u5bfc\u6e38\u8d44\u683c",
"Humanities"
],
"legal_professional": [
"Legal Professional",
"\u6cd5\u5f8b\u804c\u4e1a\u8d44\u683c",
"Humanities"
],
"high_school_chinese": [
"High School Chinese",
"\u9ad8\u4e2d\u8bed\u6587",
"Humanities"
],
"high_school_history": [
"High School History",
"\u9ad8\u4e2d\u5386\u53f2",
"Humanities"
],
"middle_school_history": [
"Middle School History",
"\u521d\u4e2d\u5386\u53f2",
"Humanities"
],
"civil_servant": [
"Civil Servant",
"\u516c\u52a1\u5458",
"Other"
],
"sports_science": [
"Sports Science",
"\u4f53\u80b2\u5b66",
"Other"
],
"plant_protection": [
"Plant Protection",
"\u690d\u7269\u4fdd\u62a4",
"Other"
],
"basic_medicine": [
"Basic Medicine",
"\u57fa\u7840\u533b\u5b66",
"Other"
],
"clinical_medicine": [
"Clinical Medicine",
"\u4e34\u5e8a\u533b\u5b66",
"Other"
],
"urban_and_rural_planner": [
"Urban and Rural Planner",
"\u6ce8\u518c\u57ce\u4e61\u89c4\u5212\u5e08",
"Other"
],
"accountant": [
"Accountant",
"\u6ce8\u518c\u4f1a\u8ba1\u5e08",
"Other"
],
"fire_engineer": [
"Fire Engineer",
"\u6ce8\u518c\u6d88\u9632\u5de5\u7a0b\u5e08",
"Other"
],
"environmental_impact_assessment_engineer": [
"Environmental Impact Assessment Engineer",
"\u73af\u5883\u5f71\u54cd\u8bc4\u4ef7\u5de5\u7a0b\u5e08",
"Other"
],
"tax_accountant": [
"Tax Accountant",
"\u7a0e\u52a1\u5e08",
"Other"
],
"physician": [
"Physician",
"\u533b\u5e08\u8d44\u683c",
"Other"
]
}
\ No newline at end of file
scripts/cmmlu/categories.py 0 → 100644
View file @ 7d06d0f9
# This code is modified from CMMLU Project: https://github.com/haonan-li/CMMLU
name_en2zh = {
"agronomy": "农学",
"anatomy": "解剖学",
"ancient_chinese": "古汉语",
"arts": "艺术学",
"astronomy": "天文学",
"business_ethics": "商业伦理",
"chinese_civil_service_exam": "中国公务员考试",
"chinese_driving_rule": "中国驾驶规则",
"chinese_food_culture": "中国饮食文化",
"chinese_foreign_policy": "中国外交政策",
"chinese_history":"中国历史",
"chinese_literature": "中国文学",
"chinese_teacher_qualification": "中国教师资格",
"clinical_knowledge": "临床知识",
"college_actuarial_science":"大学精算学",
"college_education":"大学教育学",
"college_engineering_hydrology": "大学工程水文学",
"college_law": "大学法律",
"college_mathematics": "大学数学",
"college_medical_statistics":"大学医学统计",
"college_medicine": "大学医学",
"computer_science": "计算机科学",
"computer_security": "计算机安全",
"conceptual_physics": "概念物理学",
"construction_project_management": "建设工程管理",
"economics": "经济学",
"education": "教育学",
"electrical_engineering": "电气工程",
"elementary_chinese":"小学语文",
"elementary_commonsense":"小学常识",
"elementary_information_and_technology": "小学信息技术",
"elementary_mathematics": "初等数学",
"ethnology": "民族学",
"food_science": "食品科学",
"genetics": "遗传学",
"global_facts": "全球事实",
"high_school_biology": "高中生物",
"high_school_chemistry": "高中化学",
"high_school_geography": "高中地理",
"high_school_mathematics": "高中数学",
"high_school_physics": "高中物理学",
"high_school_politics": "高中政治",
"human_sexuality": "人类性行为",
"international_law": "国际法学",
"journalism": "新闻学",
"jurisprudence": "法理学",
"legal_and_moral_basis": "法律与道德基础",
"logical": "逻辑学",
"machine_learning": "机器学习",
"management": "管理学",
"marketing": "市场营销",
"marxist_theory": "马克思主义理论",
"modern_chinese": "现代汉语",
"nutrition": "营养学",
"philosophy": "哲学",
"professional_accounting": "专业会计",
"professional_law": "专业法学",
"professional_medicine": "专业医学",
"professional_psychology": "专业心理学",
"public_relations": "公共关系",
"security_study":"安全研究",
"sociology": "社会学",
"sports_science": "体育学",
"traditional_chinese_medicine": "中医中药",
"virology": "病毒学",
"world_history":"世界历史",
"world_religions": "世界宗教",
}
subcategories = {
"agronomy": ['other'],
"anatomy": ['biology'],
"ancient_chinese": ['linguistics','china specific'],
"arts": ['arts'],
"astronomy": ['physics'],
"business_ethics": ['business'],
"chinese_civil_service_exam": ['politics','china specific'],
"chinese_driving_rule": ['other','china specific'],
"chinese_food_culture": ['culture','china specific'],
"chinese_foreign_policy": ['politics','china specific'],
"chinese_history":['history','china specific'],
"chinese_literature": ['literature','china specific'],
"chinese_teacher_qualification": ['education','china specific'],
"college_actuarial_science":['math'],
"college_education":['education'],
"college_engineering_hydrology": ['engineering'],
"college_law": ['law'],
"college_mathematics": ['math'],
"college_medical_statistics":['statistics'],
"clinical_knowledge": ['other'],
"college_medicine": ['other'],
"computer_science": ['computer science'],
"computer_security": ['other'],
"conceptual_physics": ['physics'],
"construction_project_management": ['other','china specific'],
"economics": ['economics'],
"education": ['education'],
"elementary_chinese":['linguistics','china specific'],
"elementary_commonsense":['other','china specific'],
"elementary_information_and_technology": ['other'],
"electrical_engineering": ['engineering'],
"elementary_mathematics": ['math'],
"ethnology": ['culture','china specific'],
"food_science": ['other'],
"genetics": ['biology'],
"global_facts": ['global'],
"high_school_biology": ['biology'],
"high_school_chemistry": ['chemistry'],
"high_school_geography": ['geography'],
"high_school_mathematics": ['math'],
"high_school_physics": ['physics'],
"high_school_politics": ['politics','china specific'],
"human_sexuality": ['other'],
"international_law": ['law'],
"journalism": ['sociology'],
"jurisprudence": ['law'],
"legal_and_moral_basis": ['other'],
"logical": ['philosophy'],
"machine_learning": ['computer science'],
"management": ['business'],
"marketing": ['business'],
"marxist_theory": ['philosophy'],
"modern_chinese": ['linguistics','china specific'],
"nutrition": ['other'],
"philosophy": ['philosophy'],
"professional_accounting": ['business'],
"professional_law": ['law'],
"professional_medicine": ['other'],
"professional_psychology": ['psychology'],
"public_relations": ['politics'],
"security_study": ['politics'],
"sociology": ['culture'],
"sports_science": ['other'],
"traditional_chinese_medicine": ['other','china specific'],
"virology": ['biology'],
"world_history":['history'],
"world_religions": ['global'],
}
categories = {
"STEM": ["physics", "chemistry", "biology", "computer science", "math", "engineering", "statistics"],
"Humanities": ["history", "philosophy", "law", "arts", "literature", "global"],
"Social Science": ['linguistics',"business", "politics", "culture", "economics", "geography", "psychology", "education", "sociology"],
"Other":["other"],
"China specific": ["china specific"],
}
scripts/cmmlu/eval.py 0 → 100644
View file @ 7d06d0f9
# This code is modified from C-Eval Project: https://github.com/SJTU-LIT/ceval
import os
import argparse
import pandas as pd
import torch
import json
from llama2_evaluator import Llama_Evaluator
from glob import glob
import time
from collections import defaultdict
from categories import name_en2zh, subcategories, categories
choices = ["A", "B", "C", "D"]
category2subject = defaultdict(list)
for k,v in categories.items():
for subject, subcat in subcategories.items():
for c in subcat:
if c in v:
category2subject[k].append(subject)
category2subject_list = defaultdict(list)
for key,value in category2subject.items():
for val in value:
category2subject_list[val]=[val,name_en2zh[val],key]
category2subject=category2subject_list
choices = ["A", "B", "C", "D"]
def main(args, evaluator,take):
subject_mapping = category2subject #json.load(f)
filenames = [s.split('/')[-1] for s in glob(args.input_dir+"/test/*csv")]
subject_list = [val_file.replace(".csv","") for val_file in filenames]
accuracy, summary = {}, {}
run_date=time.strftime('%Y-%m-%d_%H-%M-%S',time.localtime(time.time()))
output_dir = args.output_dir
save_result_dir=os.path.join(output_dir,f"take{take}")
if not os.path.exists(save_result_dir):
os.makedirs(save_result_dir,exist_ok=True)
all_answers = {}
for index,subject_name in enumerate(subject_list):
print(f"{index/len(subject_list)} Inference starts at {run_date} on {args.model_path} with subject of {subject_name}!")
val_file_path=os.path.join(args.input_dir+'/test',f'{subject_name}.csv')
dev_file_path=os.path.join(args.input_dir+'/dev',f'{subject_name}.csv')
val_df=pd.read_csv(val_file_path)
dev_df=pd.read_csv(dev_file_path) if args.few_shot else None
correct_ratio, answers = evaluator.eval_subject(subject_name, val_df, dev_df,
save_result_dir=save_result_dir if args.do_save_csv else None,
few_shot=args.few_shot,
cot=args.cot,
with_prompt=args.with_prompt,
constrained_decoding=args.constrained_decoding,
do_test=False)
print(f"Subject: {subject_name}")
print(f"Acc: {correct_ratio}")
accuracy[subject_name] = correct_ratio
summary[subject_name] = {"score":correct_ratio,
"num":len(val_df),
"correct":correct_ratio*len(val_df)/100}
all_answers[subject_name] = answers
json.dump(all_answers,open(save_result_dir+'/submission.json','w'),ensure_ascii=False,indent=4)
print("\n\nModel:",args.model_path)
print("Accuracy:")
for k, v in accuracy.items():
print(k, ": ", v)
total_num = 0
total_correct = 0
summary['grouped'] = {
"China specific": {"correct": 0.0, "num": 0},
"STEM": {"correct": 0.0, "num": 0},
"Social Science": {"correct": 0.0, "num": 0},
"Humanities": {"correct": 0.0, "num": 0},
"Other": {"correct": 0.0, "num": 0}
}
for subj, info in subject_mapping.items():
group = info[2]
summary['grouped'][group]["num"] += summary[subj]['num']
summary['grouped'][group]["correct"] += summary[subj]['correct']
for group, info in summary['grouped'].items():
info['score'] = info["correct"] / info["num"]
total_num += info["num"]
total_correct += info["correct"]
summary['All'] = {"score": total_correct / total_num, "num": total_num, "correct": total_correct}
json.dump(summary,open(save_result_dir+'/summary.json','w'),ensure_ascii=False,indent=2)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--ntrain", "-k", type=int, default=5)
parser.add_argument("--model_path", type=str)
parser.add_argument("--cot",choices=["False","True"], default="False")
parser.add_argument("--few_shot", choices=["False","True"], default="True")
parser.add_argument("--with_prompt", choices=["False","True"], default="False")
parser.add_argument("--constrained_decoding", choices=["False","True"], default="False")
parser.add_argument("--temperature",type=float,default=0.2)
parser.add_argument("--n_times", default=1,type=int)
parser.add_argument("--do_save_csv", choices=["False","True"], default="False")
parser.add_argument("--output_dir", type=str)
parser.add_argument("--input_dir", type=str)
parser.add_argument("--verbose", action="store_true", help="Print detailed information of each example.")
args = parser.parse_args()
args.cot = args.cot == "True"
args.few_shot = args.few_shot == "True"
args.with_prompt = args.with_prompt == "True"
args.do_save_csv = args.do_save_csv == "True"
args.constrained_decoding = args.constrained_decoding == "True"
if args.constrained_decoding is True:
args.n_times=max(args.n_times,1)
print(args)
device = torch.device(0)
print(device)
evaluator=Llama_Evaluator(
choices=choices,
k=args.ntrain,
model_path=args.model_path,
device=device,
temperature = args.temperature,
verbose = args.verbose
)
for i in range(args.n_times):
main(args,evaluator=evaluator,take=i)
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