zstanjj/HTML-Pruner-Llama-1B
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: zstanjj/HTML-Pruner-Llama-1B
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-05-24 03:51:17 +08:00
commit e5004bf4c0
15 changed files with 5825 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
.gitattributes vendored Normal file
View File

@@ -0,0 +1,36 @@
*.7z filter=lfs diff=lfs merge=lfs -text
*.arrow filter=lfs diff=lfs merge=lfs -text
*.bin filter=lfs diff=lfs merge=lfs -text
*.bz2 filter=lfs diff=lfs merge=lfs -text
*.ckpt filter=lfs diff=lfs merge=lfs -text
*.ftz filter=lfs diff=lfs merge=lfs -text
*.gz filter=lfs diff=lfs merge=lfs -text
*.h5 filter=lfs diff=lfs merge=lfs -text
*.joblib filter=lfs diff=lfs merge=lfs -text
*.lfs.* filter=lfs diff=lfs merge=lfs -text
*.mlmodel filter=lfs diff=lfs merge=lfs -text
*.model filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text
*.npy filter=lfs diff=lfs merge=lfs -text
*.npz filter=lfs diff=lfs merge=lfs -text
*.onnx filter=lfs diff=lfs merge=lfs -text
*.ot filter=lfs diff=lfs merge=lfs -text
*.parquet filter=lfs diff=lfs merge=lfs -text
*.pb filter=lfs diff=lfs merge=lfs -text
*.pickle filter=lfs diff=lfs merge=lfs -text
*.pkl filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
*.rar filter=lfs diff=lfs merge=lfs -text
*.safetensors filter=lfs diff=lfs merge=lfs -text
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.tar.* filter=lfs diff=lfs merge=lfs -text
*.tar filter=lfs diff=lfs merge=lfs -text
*.tflite filter=lfs diff=lfs merge=lfs -text
*.tgz filter=lfs diff=lfs merge=lfs -text
*.wasm filter=lfs diff=lfs merge=lfs -text
*.xz filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.zst filter=lfs diff=lfs merge=lfs -text
*tfevents* filter=lfs diff=lfs merge=lfs -text
tokenizer.json filter=lfs diff=lfs merge=lfs -text

248
README.md Normal file
View File

@@ -0,0 +1,248 @@
---
language:
- en
library_name: transformers
base_model: meta-llama/Llama-3.2-1B
license: apache-2.0
---
## Model Information
We release the HTML pruner model used in **HtmlRAG: HTML is Better Than Plain Text for Modeling Retrieval Results in RAG Systems**.
<p align="left">
Useful links: 📝 <a href="https://arxiv.org/abs/2411.02959" target="_blank">Paper</a> • 🤗 <a href="https://huggingface.co/papers/2411.02959" target="_blank">Hugging Face</a> • 🧩 <a href="https://github.com/plageon/HtmlRAG" target="_blank">Github</a>
</p>
We propose HtmlRAG, which uses HTML instead of plain text as the format of external knowledge in RAG systems. To tackle the long context brought by HTML, we propose **Lossless HTML Cleaning** and **Two-Step Block-Tree-Based HTML Pruning**.
- **Lossless HTML Cleaning**: This cleaning process just removes totally irrelevant contents and compress redundant structures, retaining all semantic information in the original HTML. The compressed HTML of lossless HTML cleaning is suitable for RAG systems that have long-context LLMs and are not willing to loss any information before generation.
- **Two-Step Block-Tree-Based HTML Pruning**: The block-tree-based HTML pruning consists of two steps, both of which are conducted on the block tree structure. The first pruning step uses a embedding model to calculate scores for blocks, while the second step uses a path generative model. The first step processes the result of lossless HTML cleaning, while the second step processes the result of the first pruning step.
🌹 If you use this model, please ✨star our **[GitHub repository](https://github.com/plageon/HtmlRAG)** to support us. Your star means a lot!
## 📦 Installation
Install the package using pip:
```bash
pip install htmlrag
```
Or install the package from source:
```bash
pip install -e .
```
---
## 📖 User Guide
### 🧹 HTML Cleaning
```python
from htmlrag import clean_html
question = "When was the bellagio in las vegas built?"
html = """
<html>
<head>
<h1>Bellagio Hotel in Las</h1>
</head>
<body>
<p class="class0">The Bellagio is a luxury hotel and casino located on the Las Vegas Strip in Paradise, Nevada. It was built in 1998.</p>
</body>
<div>
<div>
<p>Some other text</p>
<p>Some other text</p>
</div>
</div>
<p class="class1"></p>
<!-- Some comment -->
<script type="text/javascript">
document.write("Hello World!");
</script>
</html>
"""
#. alternatively you can read html files and merge them
# html_files=["/path/to/html/file1.html", "/path/to/html/file2.html"]
# htmls=[open(file).read() for file in html_files]
# html = "\n".join(htmls)
simplified_html = clean_html(html)
print(simplified_html)
# <html>
# <h1>Bellagio Hotel in Las</h1>
# <p>The Bellagio is a luxury hotel and casino located on the Las Vegas Strip in Paradise, Nevada. It was built in 1998.</p>
# <div>
# <p>Some other text</p>
# <p>Some other text</p>
# </div>
# </html>
```
### 🔧 Configure Pruning Parameters
The example HTML document is rather a short one. Real-world HTML documents can be much longer and more complex. To handle such cases, we can configure the following parameters:
```python
# Maximum number of words in a node when constructing the block tree for pruning with the embedding model
MAX_NODE_WORDS_EMBED = 10
# MAX_NODE_WORDS_EMBED = 256 # a recommended setting for real-world HTML documents
# Maximum number of tokens in the output HTML document pruned with the embedding model
MAX_CONTEXT_WINDOW_EMBED = 60
# MAX_CONTEXT_WINDOW_EMBED = 6144 # a recommended setting for real-world HTML documents
# Maximum number of words in a node when constructing the block tree for pruning with the generative model
MAX_NODE_WORDS_GEN = 5
# MAX_NODE_WORDS_GEN = 128 # a recommended setting for real-world HTML documents
# Maximum number of tokens in the output HTML document pruned with the generative model
MAX_CONTEXT_WINDOW_GEN = 32
# MAX_CONTEXT_WINDOW_GEN = 4096 # a recommended setting for real-world HTML documents
```
### 🌲 Build Block Tree
```python
from htmlrag import build_block_tree
block_tree, simplified_html = build_block_tree(simplified_html, max_node_words=MAX_NODE_WORDS_EMBED)
# block_tree, simplified_html = build_block_tree(simplified_html, max_node_words=MAX_NODE_WORDS_GEN, zh_char=True) # for Chinese text
for block in block_tree:
print("Block Content: ", block[0])
print("Block Path: ", block[1])
print("Is Leaf: ", block[2])
print("")
# Block Content: <h1>Bellagio Hotel in Las</h1>
# Block Path: ['html', 'title']
# Is Leaf: True
#
# Block Content: <div>
# <p>Some other text</p>
# <p>Some other text</p>
# </div>
# Block Path: ['html', 'div']
# Is Leaf: True
#
# Block Content: <p>The Bellagio is a luxury hotel and casino located on the Las Vegas Strip in Paradise, Nevada. It was built in 1998.</p>
# Block Path: ['html', 'p']
# Is Leaf: True
```
### ✂️ Prune HTML Blocks with Embedding Model
```python
from htmlrag import EmbedHTMLPruner
embed_model = "BAAI/bge-large-en"
query_instruction_for_retrieval = "Instruct: Given a web search query, retrieve relevant passages that answer the query\nQuery: "
embed_html_pruner = EmbedHTMLPruner(embed_model=embed_model, local_inference=True, query_instruction_for_retrieval = query_instruction_for_retrieval)
# alternatively you can init a remote TEI model, refer to https://github.com/huggingface/text-embeddings-inference.
# tei_endpoint="http://YOUR_TEI_ENDPOINT"
# embed_html_pruner = EmbedHTMLPruner(embed_model=embed_model, local_inference=False, query_instruction_for_retrieval = query_instruction_for_retrieval, endpoint=tei_endpoint)
block_rankings=embed_html_pruner.calculate_block_rankings(question, simplified_html, block_tree)
print(block_rankings)
# [2, 0, 1]
#. alternatively you can use bm25 to rank the blocks
from htmlrag import BM25HTMLPruner
bm25_html_pruner = BM25HTMLPruner()
block_rankings = bm25_html_pruner.calculate_block_rankings(question, simplified_html, block_tree)
print(block_rankings)
# [2, 0, 1]
from transformers import AutoTokenizer
chat_tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-70B-Instruct")
pruned_html = embed_html_pruner.prune_HTML(simplified_html, block_tree, block_rankings, chat_tokenizer, MAX_CONTEXT_WINDOW_EMBED)
print(pruned_html)
# <html>
# <h1>Bellagio Hotel in Las</h1>
# <p>The Bellagio is a luxury hotel and casino located on the Las Vegas Strip in Paradise, Nevada. It was built in 1998.</p>
# </html>
```
### ✂️ Prune HTML Blocks with Generative Model
```python
from htmlrag import GenHTMLPruner
import torch
# construct a finer block tree
block_tree, pruned_html = build_block_tree(pruned_html, max_node_words=MAX_NODE_WORDS_GEN)
# block_tree, pruned_html = build_block_tree(pruned_html, max_node_words=MAX_NODE_WORDS_GEN, zh_char=True) # for Chinese text
for block in block_tree:
print("Block Content: ", block[0])
print("Block Path: ", block[1])
print("Is Leaf: ", block[2])
print("")
# Block Content: <h1>Bellagio Hotel in Las</h1>
# Block Path: ['html', 'title']
# Is Leaf: True
#
# Block Content: <p>The Bellagio is a luxury hotel and casino located on the Las Vegas Strip in Paradise, Nevada. It was built in 1998.</p>
# Block Path: ['html', 'p']
# Is Leaf: True
ckpt_path = "zstanjj/HTML-Pruner-Llama-1B"
if torch.cuda.is_available():
device="cuda"
else:
device="cpu"
gen_embed_pruner = GenHTMLPruner(gen_model=ckpt_path, device=device)
block_rankings = gen_embed_pruner.calculate_block_rankings(question, pruned_html, block_tree)
print(block_rankings)
# [1, 0]
pruned_html = gen_embed_pruner.prune_HTML(pruned_html, block_tree, block_rankings, chat_tokenizer, MAX_CONTEXT_WINDOW_GEN)
print(pruned_html)
# <p>The Bellagio is a luxury hotel and casino located on the Las Vegas Strip in Paradise, Nevada. It was built in 1998.</p>
```
---
## Results
- **Results for [HTML-Pruner-Phi-3.8B](https://huggingface.co/zstanjj/HTML-Pruner-Phi-3.8B) and [HTML-Pruner-Llama-1B](https://huggingface.co/zstanjj/HTML-Pruner-Llama-1B) with Llama-3.1-70B-Instruct as chat model**.
| Dataset | ASQA | HotpotQA | NQ | TriviaQA | MuSiQue | ELI5 |
|------------------|-----------|-----------|-----------|-----------|-----------|-----------|
| Metrics | EM | EM | EM | EM | EM | ROUGE-L |
| BM25 | 49.50 | 38.25 | 47.00 | 88.00 | 9.50 | 16.15 |
| BGE | 68.00 | 41.75 | 59.50 | 93.00 | 12.50 | 16.20 |
| E5-Mistral | 63.00 | 36.75 | 59.50 | 90.75 | 11.00 | 16.17 |
| LongLLMLingua | 62.50 | 45.00 | 56.75 | 92.50 | 10.25 | 15.84 |
| JinaAI Reader | 55.25 | 34.25 | 48.25 | 90.00 | 9.25 | 16.06 |
| HtmlRAG-Phi-3.8B | **68.50** | **46.25** | 60.50 | **93.50** | **13.25** | **16.33** |
| HtmlRAG-Llama-1B | 66.50 | 45.00 | **60.75** | 93.00 | 10.00 | 16.25 |
---
## 📜 Citation
```bibtex
@misc{tan2024htmlraghtmlbetterplain,
title={HtmlRAG: HTML is Better Than Plain Text for Modeling Retrieved Knowledge in RAG Systems},
author={Jiejun Tan and Zhicheng Dou and Wen Wang and Mang Wang and Weipeng Chen and Ji-Rong Wen},
year={2024},
eprint={2411.02959},
archivePrefix={arXiv},
primaryClass={cs.IR},
url={https://arxiv.org/abs/2411.02959},
}
```

47
config.json Normal file
View File

@@ -0,0 +1,47 @@
{
"_name_or_path": "HTML-Pruner-Llama-3.2-1B",
"architectures": [
"LlamaForCausalLM"
],
"auto_map": {
"AutoConfig": "configuration_llama.LlamaConfig",
"AutoModel": "modeling_llama.LlamaForCausalLM",
"AutoModelForCausalLM": "modeling_llama.LlamaForCausalLM",
"AutoModelForSeq2SeqLM": "modeling_llama.LlamaForHTMLTreeGeneration"
},
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 128000,
"eos_token_id": [
128001,
128008,
128009
],
"head_dim": 64,
"hidden_act": "silu",
"hidden_size": 2048,
"initializer_range": 0.02,
"intermediate_size": 8192,
"max_position_embeddings": 131072,
"mlp_bias": false,
"model_type": "llama",
"num_attention_heads": 32,
"num_hidden_layers": 16,
"num_key_value_heads": 8,
"pretraining_tp": 1,
"rms_norm_eps": 1e-05,
"rope_scaling": {
"factor": 32.0,
"high_freq_factor": 4.0,
"low_freq_factor": 1.0,
"original_max_position_embeddings": 8192,
"rope_type": "llama3"
},
"rope_theta": 500000.0,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.45.0.dev0",
"use_cache": true,
"vocab_size": 128256,
"attn_implementation": "flash_attention_2"
}

1
configuration.json Normal file
View File

@@ -0,0 +1 @@
{}

206
configuration_llama.py Normal file
View File

@@ -0,0 +1,206 @@
# coding=utf-8
# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
#
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
# and OPT implementations in this library. It has been modified from its
# original forms to accommodate minor architectural differences compared
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
#
# 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.
"""LLaMA model configuration"""
from transformers.configuration_utils import PretrainedConfig
from transformers.modeling_rope_utils import rope_config_validation
class LlamaConfig(PretrainedConfig):
r"""
This is the configuration class to store the configuration of a [`LlamaModel`]. It is used to instantiate an LLaMA
model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
defaults will yield a similar configuration to that of the LLaMA-7B.
Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
documentation from [`PretrainedConfig`] for more information.
Args:
vocab_size (`int`, *optional*, defaults to 32000):
Vocabulary size of the LLaMA model. Defines the number of different tokens that can be represented by the
`inputs_ids` passed when calling [`LlamaModel`]
hidden_size (`int`, *optional*, defaults to 4096):
Dimension of the hidden representations.
intermediate_size (`int`, *optional*, defaults to 11008):
Dimension of the MLP representations.
num_hidden_layers (`int`, *optional*, defaults to 32):
Number of hidden layers in the Transformer decoder.
num_attention_heads (`int`, *optional*, defaults to 32):
Number of attention heads for each attention layer in the Transformer decoder.
num_key_value_heads (`int`, *optional*):
This is the number of key_value heads that should be used to implement Grouped Query Attention. If
`num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
`num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
by meanpooling all the original heads within that group. For more details checkout [this
paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
`num_attention_heads`.
hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
The non-linear activation function (function or string) in the decoder.
max_position_embeddings (`int`, *optional*, defaults to 2048):
The maximum sequence length that this model might ever be used with. Llama 1 supports up to 2048 tokens,
Llama 2 up to 4096, CodeLlama up to 16384.
initializer_range (`float`, *optional*, defaults to 0.02):
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
rms_norm_eps (`float`, *optional*, defaults to 1e-06):
The epsilon used by the rms normalization layers.
use_cache (`bool`, *optional*, defaults to `True`):
Whether or not the model should return the last key/values attentions (not used by all models). Only
relevant if `config.is_decoder=True`.
pad_token_id (`int`, *optional*):
Padding token id.
bos_token_id (`int`, *optional*, defaults to 1):
Beginning of stream token id.
eos_token_id (`int`, *optional*, defaults to 2):
End of stream token id.
pretraining_tp (`int`, *optional*, defaults to 1):
Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this
document](https://huggingface.co/docs/transformers/main/perf_train_gpu_many#tensor-parallelism) to
understand more about it. This value is necessary to ensure exact reproducibility of the pretraining
results. Please refer to [this issue](https://github.com/pytorch/pytorch/issues/76232).
tie_word_embeddings (`bool`, *optional*, defaults to `False`):
Whether to tie weight embeddings
rope_theta (`float`, *optional*, defaults to 10000.0):
The base period of the RoPE embeddings.
rope_scaling (`Dict`, *optional*):
Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type
and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value
accordingly.
Expected contents:
`rope_type` (`str`):
The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope',
'llama3'], with 'default' being the original RoPE implementation.
`factor` (`float`, *optional*):
Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In
most scaling types, a `factor` of x will enable the model to handle sequences of length x *
original maximum pre-trained length.
`original_max_position_embeddings` (`int`, *optional*):
Used with 'dynamic', 'longrope' and 'llama3'. The original max position embeddings used during
pretraining.
`attention_factor` (`float`, *optional*):
Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention
computation. If unspecified, it defaults to value recommended by the implementation, using the
`factor` field to infer the suggested value.
`beta_fast` (`float`, *optional*):
Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear
ramp function. If unspecified, it defaults to 32.
`beta_slow` (`float`, *optional*):
Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear
ramp function. If unspecified, it defaults to 1.
`short_factor` (`List[float]`, *optional*):
Only used with 'longrope'. The scaling factor to be applied to short contexts (<
`original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
size divided by the number of attention heads divided by 2
`long_factor` (`List[float]`, *optional*):
Only used with 'longrope'. The scaling factor to be applied to long contexts (<
`original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
size divided by the number of attention heads divided by 2
`low_freq_factor` (`float`, *optional*):
Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE
`high_freq_factor` (`float`, *optional*):
Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE
attention_bias (`bool`, *optional*, defaults to `False`):
Whether to use a bias in the query, key, value and output projection layers during self-attention.
attention_dropout (`float`, *optional*, defaults to 0.0):
The dropout ratio for the attention probabilities.
mlp_bias (`bool`, *optional*, defaults to `False`):
Whether to use a bias in up_proj, down_proj and gate_proj layers in the MLP layers.
head_dim (`int`, *optional*):
The attention head dimension. If None, it will default to hidden_size // num_heads
```python
>>> from transformers import LlamaModel, LlamaConfig
>>> # Initializing a LLaMA llama-7b style configuration
>>> configuration = LlamaConfig()
>>> # Initializing a model from the llama-7b style configuration
>>> model = LlamaModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
```"""
model_type = "llama"
keys_to_ignore_at_inference = ["past_key_values"]
def __init__(
self,
vocab_size=32000,
hidden_size=4096,
intermediate_size=11008,
num_hidden_layers=32,
num_attention_heads=32,
num_key_value_heads=None,
hidden_act="silu",
max_position_embeddings=2048,
initializer_range=0.02,
rms_norm_eps=1e-6,
use_cache=True,
pad_token_id=128001,
bos_token_id=1,
eos_token_id=2,
pretraining_tp=1,
tie_word_embeddings=False,
rope_theta=10000.0,
rope_scaling=None,
attention_bias=False,
attention_dropout=0.0,
mlp_bias=False,
head_dim=None,
**kwargs,
):
self.vocab_size = vocab_size
self.max_position_embeddings = max_position_embeddings
self.hidden_size = hidden_size
self.intermediate_size = intermediate_size
self.num_hidden_layers = num_hidden_layers
self.num_attention_heads = num_attention_heads
# for backward compatibility
if num_key_value_heads is None:
num_key_value_heads = num_attention_heads
self.num_key_value_heads = num_key_value_heads
self.hidden_act = hidden_act
self.initializer_range = initializer_range
self.rms_norm_eps = rms_norm_eps
self.pretraining_tp = pretraining_tp
self.use_cache = use_cache
self.rope_theta = rope_theta
self.rope_scaling = rope_scaling
self.attention_bias = attention_bias
self.attention_dropout = attention_dropout
self.mlp_bias = mlp_bias
self.head_dim = head_dim if head_dim is not None else self.hidden_size // self.num_attention_heads
# Validate the correctness of rotary position embeddings parameters
# BC: if there is a 'type' field, copy it it to 'rope_type'.
if self.rope_scaling is not None and "type" in self.rope_scaling:
self.rope_scaling["rope_type"] = self.rope_scaling["type"]
rope_config_validation(self)
super().__init__(
pad_token_id=pad_token_id,
bos_token_id=bos_token_id,
eos_token_id=eos_token_id,
tie_word_embeddings=tie_word_embeddings,
**kwargs,
)

12
generation_config.json Normal file
View File

@@ -0,0 +1,12 @@
{
"bos_token_id": 128000,
"do_sample": true,
"eos_token_id": [
128001,
128008,
128009
],
"temperature": 0.6,
"top_p": 0.9,
"transformers_version": "4.45.0.dev0"
}

1
latest Normal file
View File

@@ -0,0 +1 @@
global_step381

3
model.safetensors Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:ef427eecfe24558d8908112b54ed9ac1d1ce6f908ca1778e481c64caa9cdada7
size 2471645608

1913
modeling_llama.py Normal file
View File

File diff suppressed because it is too large Load Diff

23
special_tokens_map.json Normal file
View File

@@ -0,0 +1,23 @@
{
"bos_token": {
"content": "<|begin_of_text|>",
"lstrip": false,
"normalized": false,
"rstrip": false,
"single_word": false
},
"eos_token": {
"content": "<|eot_id|>",
"lstrip": false,
"normalized": false,
"rstrip": false,
"single_word": false
},
"pad_token": {
"content": "<|finetune_right_pad_id|>",
"lstrip": false,
"normalized": false,
"rstrip": false,
"single_word": false
}
}

410
tokenization_llama.py Normal file
View File

@@ -0,0 +1,410 @@
# coding=utf-8
# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
#
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
# and OPT implementations in this library. It has been modified from its
# original forms to accommodate minor architectural differences compared
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
#
# 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.
"""Tokenization classes for LLaMA."""
import os
from shutil import copyfile
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple
import sentencepiece as spm
from transformers.convert_slow_tokenizer import import_protobuf
from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
from transformers.utils import logging
if TYPE_CHECKING:
from transformers.tokenization_utils_base import TextInput
logger = logging.get_logger(__name__)
VOCAB_FILES_NAMES = {"vocab_file": "tokenizer.model"}
SPIECE_UNDERLINE = "▁"
B_INST, E_INST = "[INST]", "[/INST]"
B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
DEFAULT_SYSTEM_PROMPT = """You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your \
answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure\
that your responses are socially unbiased and positive in nature.
If a question does not make any sense, or is not factually coherent, explain why instead of answering something not \
correct. If you don't know the answer to a question, please don't share false information.""" # fmt: skip
class LlamaTokenizer(PreTrainedTokenizer):
"""
Construct a Llama tokenizer. Based on byte-level Byte-Pair-Encoding. The default padding token is unset as there is
no padding token in the original model.
Args:
vocab_file (`str`):
Path to the vocabulary file.
unk_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<unk>"`):
The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
token instead.
bos_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<s>"`):
The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
eos_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"</s>"`):
The end of sequence token.
pad_token (`str` or `tokenizers.AddedToken`, *optional*):
A special token used to make arrays of tokens the same size for batching purpose. Will then be ignored by
attention mechanisms or loss computation.
sp_model_kwargs (`Dict[str, Any]`, `Optional`, *optional*):
Will be passed to the `SentencePieceProcessor.__init__()` method. The [Python wrapper for
SentencePiece](https://github.com/google/sentencepiece/tree/master/python) can be used, among other things,
to set:
- `enable_sampling`: Enable subword regularization.
- `nbest_size`: Sampling parameters for unigram. Invalid for BPE-Dropout.
- `nbest_size = {0,1}`: No sampling is performed.
- `nbest_size > 1`: samples from the nbest_size results.
- `nbest_size < 0`: assuming that nbest_size is infinite and samples from the all hypothesis (lattice)
using forward-filtering-and-backward-sampling algorithm.
- `alpha`: Smoothing parameter for unigram sampling, and dropout probability of merge operations for
BPE-dropout.
add_bos_token (`bool`, *optional*, defaults to `True`):
Whether or not to add an `bos_token` at the start of sequences.
add_eos_token (`bool`, *optional*, defaults to `False`):
Whether or not to add an `eos_token` at the end of sequences.
clean_up_tokenization_spaces (`bool`, *optional*, defaults to `False`):
Whether or not to cleanup spaces after decoding, cleanup consists in removing potential artifacts like
extra spaces.
use_default_system_prompt (`bool`, *optional*, defaults to `False`):
Whether or not the default system prompt for Llama should be used.
spaces_between_special_tokens (`bool`, *optional*, defaults to `False`):
Whether or not to add spaces between special tokens.
legacy (`bool`, *optional*):
Whether or not the `legacy` behavior of the tokenizer should be used. Legacy is before the merge of #24622
and #25224 which includes fixes to properly handle tokens that appear after special tokens.
Make sure to also set `from_slow` to `True`.
A simple example:
- `legacy=True`:
```python
>>> from transformers import LlamaTokenizerFast
>>> tokenizer = LlamaTokenizerFast.from_pretrained("huggyllama/llama-7b", legacy=True, from_slow=True)
>>> tokenizer.encode("Hello <s>.") # 869 is '▁.'
[1, 15043, 29871, 1, 869]
```
- `legacy=False`:
```python
>>> from transformers import LlamaTokenizerFast
>>> tokenizer = LlamaTokenizerFast.from_pretrained("huggyllama/llama-7b", legacy=False, from_slow=True)
>>> tokenizer.encode("Hello <s>.") # 29889 is '.'
[1, 15043, 29871, 1, 29889]
```
Checkout the [pull request](https://github.com/huggingface/transformers/pull/24565) for more details.
add_prefix_space (`bool`, *optional*, defaults to `True`):
Whether or not to add an initial space to the input. This allows to treat the leading word just as any
other word. Again, this should be set with `from_slow=True` to make sure it's taken into account.
"""
vocab_files_names = VOCAB_FILES_NAMES
model_input_names = ["input_ids", "attention_mask"]
def __init__(
self,
vocab_file,
unk_token="<unk>",
bos_token="<s>",
eos_token="</s>",
pad_token=None,
sp_model_kwargs: Optional[Dict[str, Any]] = None,
add_bos_token=True,
add_eos_token=False,
clean_up_tokenization_spaces=False,
use_default_system_prompt=False,
spaces_between_special_tokens=False,
legacy=None,
add_prefix_space=True,
**kwargs,
):
self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
bos_token = AddedToken(bos_token, normalized=False, special=True) if isinstance(bos_token, str) else bos_token
eos_token = AddedToken(eos_token, normalized=False, special=True) if isinstance(eos_token, str) else eos_token
unk_token = AddedToken(unk_token, normalized=False, special=True) if isinstance(unk_token, str) else unk_token
pad_token = AddedToken(pad_token, normalized=False, special=True) if isinstance(pad_token, str) else pad_token
if legacy is None:
logger.warning_once(
f"You are using the default legacy behaviour of the {self.__class__}. This is"
" expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you."
" If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it"
" means, and thoroughly read the reason why this was added as explained in"
" https://github.com/huggingface/transformers/pull/24565 - if you loaded a llama tokenizer from a GGUF file"
" you can ignore this message"
)
legacy = True
self.legacy = legacy
self.vocab_file = vocab_file
self.add_bos_token = add_bos_token
self.add_eos_token = add_eos_token
self.use_default_system_prompt = use_default_system_prompt
self.sp_model = self.get_spm_processor(kwargs.pop("from_slow", False))
self.add_prefix_space = add_prefix_space
super().__init__(
bos_token=bos_token,
eos_token=eos_token,
unk_token=unk_token,
pad_token=pad_token,
add_bos_token=add_bos_token,
add_eos_token=add_eos_token,
sp_model_kwargs=self.sp_model_kwargs,
clean_up_tokenization_spaces=clean_up_tokenization_spaces,
use_default_system_prompt=use_default_system_prompt,
spaces_between_special_tokens=spaces_between_special_tokens,
legacy=legacy,
add_prefix_space=add_prefix_space,
**kwargs,
)
@property
def unk_token_length(self):
return len(self.sp_model.encode(str(self.unk_token)))
# Copied from transformers.models.t5.tokenization_t5.T5Tokenizer.get_spm_processor
def get_spm_processor(self, from_slow=False):
tokenizer = spm.SentencePieceProcessor(**self.sp_model_kwargs)
if self.legacy or from_slow: # no dependency on protobuf
tokenizer.Load(self.vocab_file)
return tokenizer
with open(self.vocab_file, "rb") as f:
sp_model = f.read()
model_pb2 = import_protobuf(f"The new behaviour of {self.__class__.__name__} (with `self.legacy = False`)")
model = model_pb2.ModelProto.FromString(sp_model)
normalizer_spec = model_pb2.NormalizerSpec()
normalizer_spec.add_dummy_prefix = False
model.normalizer_spec.MergeFrom(normalizer_spec)
sp_model = model.SerializeToString()
tokenizer.LoadFromSerializedProto(sp_model)
return tokenizer
def __getstate__(self):
state = self.__dict__.copy()
state["sp_model"] = None
state["sp_model_proto"] = self.sp_model.serialized_model_proto()
return state
def __setstate__(self, d):
self.__dict__ = d
self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
self.sp_model.LoadFromSerializedProto(self.sp_model_proto)
@property
def vocab_size(self):
"""Returns vocab size"""
return self.sp_model.get_piece_size()
def get_vocab(self):
"""Returns vocab as a dict"""
vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
vocab.update(self.added_tokens_encoder)
return vocab
# Copied from transformers.models.t5.tokenization_t5.T5Tokenizer.tokenize
def tokenize(self, text: "TextInput", **kwargs) -> List[str]:
"""
Converts a string to a list of tokens. If `self.legacy` is set to `False`, a prefix token is added unless the
first token is special.
"""
if self.legacy or len(text) == 0:
return super().tokenize(text, **kwargs)
text = text.replace(SPIECE_UNDERLINE, " ")
if self.add_prefix_space:
text = SPIECE_UNDERLINE + text
tokens = super().tokenize(text, **kwargs)
if len(tokens) > 1 and tokens[0] == SPIECE_UNDERLINE and tokens[1] in self.all_special_tokens:
tokens = tokens[1:]
return tokens
# Copied from transformers.models.t5.tokenization_t5.T5Tokenizer._tokenize
def _tokenize(self, text, **kwargs):
"""
Returns a tokenized string.
We de-activated the `add_dummy_prefix` option, thus the sentencepiece internals will always strip any
SPIECE_UNDERLINE. For example: `self.sp_model.encode(f"{SPIECE_UNDERLINE}Hey", out_type = str)` will give
`['H', 'e', 'y']` instead of `['▁He', 'y']`. Thus we always encode `f"{unk_token}text"` and strip the
`unk_token`. Here is an example with `unk_token = "<unk>"` and `unk_token_length = 4`.
`self.tokenizer.sp_model.encode("<unk> Hey", out_type = str)[4:]`.
"""
if self.legacy or not text.startswith((SPIECE_UNDERLINE, " ")):
return self.sp_model.encode(text, out_type=str)
# 1. Encode string + prefix ex: "<unk> Hey"
tokens = self.sp_model.encode(self.unk_token + text, out_type=str)
# 2. Remove self.unk_token from ['<','unk','>', '▁Hey']
return tokens[self.unk_token_length :] if len(tokens) >= self.unk_token_length else tokens
def _convert_token_to_id(self, token):
"""Converts a token (str) in an id using the vocab."""
return self.sp_model.piece_to_id(token)
def _convert_id_to_token(self, index):
"""Converts an index (integer) in a token (str) using the vocab."""
token = self.sp_model.IdToPiece(index)
return token
def convert_tokens_to_string(self, tokens):
"""Converts a sequence of tokens (string) in a single string."""
# since we manually add the prefix space, we have to remove it when decoding
if tokens[0].startswith(SPIECE_UNDERLINE) and self.add_prefix_space:
tokens[0] = tokens[0][1:]
current_sub_tokens = []
out_string = ""
prev_is_special = False
for i, token in enumerate(tokens):
# make sure that special tokens are not decoded using sentencepiece model
if token in self.all_special_tokens:
if not prev_is_special and i != 0 and self.legacy:
out_string += " "
out_string += self.sp_model.decode(current_sub_tokens) + token
prev_is_special = True
current_sub_tokens = []
else:
if prev_is_special and i == 1 and self.add_prefix_space and not token.startswith(SPIECE_UNDERLINE):
out_string += " "
current_sub_tokens.append(token)
prev_is_special = False
out_string += self.sp_model.decode(current_sub_tokens)
return out_string
def save_vocabulary(self, save_directory, filename_prefix: Optional[str] = None) -> Tuple[str]:
"""
Save the vocabulary and special tokens file to a directory.
Args:
save_directory (`str`):
The directory in which to save the vocabulary.
Returns:
`Tuple(str)`: Paths to the files saved.
"""
if not os.path.isdir(save_directory):
logger.error(f"Vocabulary path ({save_directory}) should be a directory")
return
out_vocab_file = os.path.join(
save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
)
if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
copyfile(self.vocab_file, out_vocab_file)
elif not os.path.isfile(self.vocab_file):
with open(out_vocab_file, "wb") as fi:
content_spiece_model = self.sp_model.serialized_model_proto()
fi.write(content_spiece_model)
return (out_vocab_file,)
def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
bos_token_id = [self.bos_token_id] if self.add_bos_token else []
eos_token_id = [self.eos_token_id] if self.add_eos_token else []
output = bos_token_id + token_ids_0 + eos_token_id
if token_ids_1 is not None:
output = output + bos_token_id + token_ids_1 + eos_token_id
return output
def get_special_tokens_mask(
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
) -> List[int]:
"""
Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
special tokens using the tokenizer `prepare_for_model` method.
Args:
token_ids_0 (`List[int]`):
List of IDs.
token_ids_1 (`List[int]`, *optional*):
Optional second list of IDs for sequence pairs.
already_has_special_tokens (`bool`, *optional*, defaults to `False`):
Whether or not the token list is already formatted with special tokens for the model.
Returns:
`List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
"""
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
)
bos_token_id = [1] if self.add_bos_token else []
eos_token_id = [1] if self.add_eos_token else []
if token_ids_1 is None:
return bos_token_id + ([0] * len(token_ids_0)) + eos_token_id
return (
bos_token_id
+ ([0] * len(token_ids_0))
+ eos_token_id
+ bos_token_id
+ ([0] * len(token_ids_1))
+ eos_token_id
)
def create_token_type_ids_from_sequences(
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
) -> List[int]:
"""
Creates a mask from the two sequences passed to be used in a sequence-pair classification task. An ALBERT
sequence pair mask has the following format:
```
0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
| first sequence | second sequence |
```
if token_ids_1 is None, only returns the first portion of the mask (0s).
Args:
token_ids_0 (`List[int]`):
List of ids.
token_ids_1 (`List[int]`, *optional*):
Optional second list of IDs for sequence pairs.
Returns:
`List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
"""
bos_token_id = [self.bos_token_id] if self.add_bos_token else []
eos_token_id = [self.eos_token_id] if self.add_eos_token else []
output = [0] * len(bos_token_id + token_ids_0 + eos_token_id)
if token_ids_1 is not None:
output += [1] * len(bos_token_id + token_ids_1 + eos_token_id)
return output

255
tokenization_llama_fast.py Normal file
View File

@@ -0,0 +1,255 @@
# coding=utf-8
# Copyright 2020 The HuggingFace Inc. team.
#
# 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.
import os
from shutil import copyfile
from typing import Optional, Tuple
from tokenizers import processors
from transformers.tokenization_utils_fast import PreTrainedTokenizerFast
from transformers.utils import is_sentencepiece_available, logging
from transformers.utils.versions import require_version
require_version("tokenizers>=0.13.3")
if is_sentencepiece_available():
from .tokenization_llama import LlamaTokenizer
else:
LlamaTokenizer = None
logger = logging.get_logger(__name__)
VOCAB_FILES_NAMES = {"vocab_file": "tokenizer.model", "tokenizer_file": "tokenizer.json"}
B_INST, E_INST = "[INST]", "[/INST]"
B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
# fmt: off
DEFAULT_SYSTEM_PROMPT = """You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your \
answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure\
that your responses are socially unbiased and positive in nature.
If a question does not make any sense, or is not factually coherent, explain why instead of answering something not \
correct. If you don't know the answer to a question, please don't share false information."""
# fmt: on
class LlamaTokenizerFast(PreTrainedTokenizerFast):
"""
Construct a Llama tokenizer. Based on byte-level Byte-Pair-Encoding.
This uses notably ByteFallback and no normalization.
```python
>>> from transformers import LlamaTokenizerFast
>>> tokenizer = LlamaTokenizerFast.from_pretrained("hf-internal-testing/llama-tokenizer")
>>> tokenizer.encode("Hello this is a test")
[1, 15043, 445, 338, 263, 1243]
```
If you want to change the `bos_token` or the `eos_token`, make sure to specify them when initializing the model, or
call `tokenizer.update_post_processor()` to make sure that the post-processing is correctly done (otherwise the
values of the first token and final token of an encoded sequence will not be correct). For more details, checkout
[post-processors] (https://huggingface.co/docs/tokenizers/api/post-processors) documentation.
This tokenizer inherits from [`PreTrainedTokenizerFast`] which contains most of the main methods. Users should
refer to this superclass for more information regarding those methods.
Args:
vocab_file (`str`, *optional*):
[SentencePiece](https://github.com/google/sentencepiece) file (generally has a .model extension) that
contains the vocabulary necessary to instantiate a tokenizer.
tokenizer_file (`str`, *optional*):
[tokenizers](https://github.com/huggingface/tokenizers) file (generally has a .json extension) that
contains everything needed to load the tokenizer.
clean_up_tokenization_spaces (`bool`, *optional*, defaults to `False`):
Whether or not to cleanup spaces after decoding, cleanup consists in removing potential artifacts like
extra spaces.
unk_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<unk>"`):
The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
token instead.
bos_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<s>"`):
The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
eos_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"</s>"`):
The end of sequence token.
add_bos_token (`bool`, *optional*, defaults to `True`):
Whether or not to add an `bos_token` at the start of sequences.
add_eos_token (`bool`, *optional*, defaults to `False`):
Whether or not to add an `eos_token` at the end of sequences.
use_default_system_prompt (`bool`, *optional*, defaults to `False`):
Whether or not the default system prompt for Llama should be used
legacy (`bool`, *optional*):
Whether or not the `legacy` behavior of the tokenizer should be used. Legacy is before the merge of #24622
and #25224 which includes fixes to properly handle tokens that appear after special tokens.
Make sure to also set `from_slow` to `True`.
A simple example:
- `legacy=True`:
```python
>>> from transformers import LlamaTokenizerFast
>>> tokenizer = LlamaTokenizerFast.from_pretrained("huggyllama/llama-7b", legacy=True, from_slow=True)
>>> tokenizer.encode("Hello <s>.") # 869 is '▁.'
[1, 15043, 29871, 1, 869]
```
- `legacy=False`:
```python
>>> from transformers import LlamaTokenizerFast
>>> tokenizer = LlamaTokenizerFast.from_pretrained("huggyllama/llama-7b", legacy=False, from_slow=True)
>>> tokenizer.encode("Hello <s>.") # 29889 is '.'
[1, 15043, 29871, 1, 29889]
```
Checkout the [pull request](https://github.com/huggingface/transformers/pull/24565) for more details.
add_prefix_space (`bool`, *optional*):
Whether or not the tokenizer should automatically add a prefix space
"""
vocab_files_names = VOCAB_FILES_NAMES
slow_tokenizer_class = LlamaTokenizer
padding_side = "left"
model_input_names = ["input_ids", "attention_mask"]
def __init__(
self,
vocab_file=None,
tokenizer_file=None,
clean_up_tokenization_spaces=False,
unk_token="<unk>",
bos_token="<s>",
eos_token="</s>",
add_bos_token=True,
add_eos_token=False,
use_default_system_prompt=False,
legacy=None,
add_prefix_space=None,
**kwargs,
):
if legacy is None:
logger.warning_once(
f"You are using the default legacy behaviour of the {self.__class__}. This is"
" expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you."
" If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it"
" means, and thoroughly read the reason why this was added as explained in"
" https://github.com/huggingface/transformers/pull/24565 - if you loaded a llama tokenizer from a GGUF file"
" you can ignore this message."
)
legacy = True
self.legacy = legacy
if add_prefix_space is not None:
kwargs["from_slow"] = True
super().__init__(
vocab_file=vocab_file,
tokenizer_file=tokenizer_file,
clean_up_tokenization_spaces=clean_up_tokenization_spaces,
unk_token=unk_token,
bos_token=bos_token,
eos_token=eos_token,
add_bos_token=add_bos_token,
add_eos_token=add_eos_token,
use_default_system_prompt=use_default_system_prompt,
add_prefix_space=add_prefix_space,
legacy=legacy,
**kwargs,
)
self._add_bos_token = add_bos_token
self._add_eos_token = add_eos_token
self.update_post_processor()
self.use_default_system_prompt = use_default_system_prompt
self.vocab_file = vocab_file
@property
def can_save_slow_tokenizer(self) -> bool:
return os.path.isfile(self.vocab_file) if self.vocab_file else False
def update_post_processor(self):
"""
Updates the underlying post processor with the current `bos_token` and `eos_token`.
"""
bos = self.bos_token
bos_token_id = self.bos_token_id
if bos is None and self.add_bos_token:
raise ValueError("add_bos_token = True but bos_token = None")
eos = self.eos_token
eos_token_id = self.eos_token_id
if eos is None and self.add_eos_token:
raise ValueError("add_eos_token = True but eos_token = None")
single = f"{(bos+':0 ') if self.add_bos_token else ''}$A:0{(' '+eos+':0') if self.add_eos_token else ''}"
pair = f"{single}{(' '+bos+':1') if self.add_bos_token else ''} $B:1{(' '+eos+':1') if self.add_eos_token else ''}"
special_tokens = []
if self.add_bos_token:
special_tokens.append((bos, bos_token_id))
if self.add_eos_token:
special_tokens.append((eos, eos_token_id))
self._tokenizer.post_processor = processors.TemplateProcessing(
single=single, pair=pair, special_tokens=special_tokens
)
@property
def add_eos_token(self):
return self._add_eos_token
@property
def add_bos_token(self):
return self._add_bos_token
@add_eos_token.setter
def add_eos_token(self, value):
self._add_eos_token = value
self.update_post_processor()
@add_bos_token.setter
def add_bos_token(self, value):
self._add_bos_token = value
self.update_post_processor()
def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
if not self.can_save_slow_tokenizer:
raise ValueError(
"Your fast tokenizer does not have the necessary information to save the vocabulary for a slow "
"tokenizer."
)
if not os.path.isdir(save_directory):
logger.error(f"Vocabulary path ({save_directory}) should be a directory")
return
out_vocab_file = os.path.join(
save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
)
if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
copyfile(self.vocab_file, out_vocab_file)
return (out_vocab_file,)
# TODO ArthurZ let's rely on the template processor instead, refactor all fast tokenizers
# Copied from transformers.models.llama.tokenization_llama.LlamaTokenizer.build_inputs_with_special_tokens
def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
bos_token_id = [self.bos_token_id] if self.add_bos_token else []
eos_token_id = [self.eos_token_id] if self.add_eos_token else []
output = bos_token_id + token_ids_0 + eos_token_id
if token_ids_1 is not None:
output = output + bos_token_id + token_ids_1 + eos_token_id
return output

BIN
tokenizer.json (Stored with Git LFS) Normal file
View File

Binary file not shown.

2063
tokenizer_config.json Normal file
View File

File diff suppressed because it is too large Load Diff

604
zero_to_fp32.py Normal file
View File

@@ -0,0 +1,604 @@
#!/usr/bin/env python
# Copyright (c) Microsoft Corporation.
# SPDX-License-Identifier: Apache-2.0
# DeepSpeed Team
# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
# application.
#
# example: python zero_to_fp32.py . pytorch_model.bin
import argparse
import torch
import glob
import math
import os
import re
from collections import OrderedDict
from dataclasses import dataclass
# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
# DeepSpeed data structures it has to be available in the current python environment.
from deepspeed.utils import logger
from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
@dataclass
class zero_model_state:
buffers: dict()
param_shapes: dict()
shared_params: list
ds_version: int
frozen_param_shapes: dict()
frozen_param_fragments: dict()
debug = 0
# load to cpu
device = torch.device('cpu')
def atoi(text):
return int(text) if text.isdigit() else text
def natural_keys(text):
'''
alist.sort(key=natural_keys) sorts in human order
http://nedbatchelder.com/blog/200712/human_sorting.html
(See Toothy's implementation in the comments)
'''
return [atoi(c) for c in re.split(r'(\d+)', text)]
def get_model_state_file(checkpoint_dir, zero_stage):
if not os.path.isdir(checkpoint_dir):
raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
# there should be only one file
if zero_stage <= 2:
file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
elif zero_stage == 3:
file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
if not os.path.exists(file):
raise FileNotFoundError(f"can't find model states file at '{file}'")
return file
def get_checkpoint_files(checkpoint_dir, glob_pattern):
# XXX: need to test that this simple glob rule works for multi-node setup too
ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
if len(ckpt_files) == 0:
raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
return ckpt_files
def get_optim_files(checkpoint_dir):
return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
def get_model_state_files(checkpoint_dir):
return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
def parse_model_states(files):
zero_model_states = []
for file in files:
state_dict = torch.load(file, map_location=device)
if BUFFER_NAMES not in state_dict:
raise ValueError(f"{file} is not a model state checkpoint")
buffer_names = state_dict[BUFFER_NAMES]
if debug:
print("Found buffers:", buffer_names)
# recover just the buffers while restoring them to fp32 if they were saved in fp16
buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
param_shapes = state_dict[PARAM_SHAPES]
# collect parameters that are included in param_shapes
param_names = []
for s in param_shapes:
for name in s.keys():
param_names.append(name)
# update with frozen parameters
frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
if frozen_param_shapes is not None:
if debug:
print(f"Found frozen_param_shapes: {frozen_param_shapes}")
param_names += list(frozen_param_shapes.keys())
# handle shared params
shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
ds_version = state_dict.get(DS_VERSION, None)
frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
z_model_state = zero_model_state(buffers=buffers,
param_shapes=param_shapes,
shared_params=shared_params,
ds_version=ds_version,
frozen_param_shapes=frozen_param_shapes,
frozen_param_fragments=frozen_param_fragments)
zero_model_states.append(z_model_state)
return zero_model_states
def parse_optim_states(files, ds_checkpoint_dir):
total_files = len(files)
state_dicts = []
for f in files:
state_dict = torch.load(f, map_location=device)
# immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
# and also handle the case where it was already removed by another helper script
state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
state_dicts.append(state_dict)
if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
raise ValueError(f"{files[0]} is not a zero checkpoint")
zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
# For ZeRO-2 each param group can have different partition_count as data parallelism for expert
# parameters can be different from data parallelism for non-expert parameters. So we can just
# use the max of the partition_count to get the dp world_size.
if type(world_size) is list:
world_size = max(world_size)
if world_size != total_files:
raise ValueError(
f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
"Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
)
# the groups are named differently in each stage
if zero_stage <= 2:
fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
elif zero_stage == 3:
fp32_groups_key = FP32_FLAT_GROUPS
else:
raise ValueError(f"unknown zero stage {zero_stage}")
if zero_stage <= 2:
fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
elif zero_stage == 3:
# if there is more than one param group, there will be multiple flattened tensors - one
# flattened tensor per group - for simplicity merge them into a single tensor
#
# XXX: could make the script more memory efficient for when there are multiple groups - it
# will require matching the sub-lists of param_shapes for each param group flattened tensor
fp32_flat_groups = [
torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
]
return zero_stage, world_size, fp32_flat_groups
def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
"""
Returns fp32 state_dict reconstructed from ds checkpoint
Args:
- ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
"""
print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
optim_files = get_optim_files(ds_checkpoint_dir)
zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
model_files = get_model_state_files(ds_checkpoint_dir)
zero_model_states = parse_model_states(model_files)
print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
if zero_stage <= 2:
return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
exclude_frozen_parameters)
elif zero_stage == 3:
return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
exclude_frozen_parameters)
def _zero2_merge_frozen_params(state_dict, zero_model_states):
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
return
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
frozen_param_fragments = zero_model_states[0].frozen_param_fragments
if debug:
num_elem = sum(s.numel() for s in frozen_param_shapes.values())
print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
wanted_params = len(frozen_param_shapes)
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
print(f'Frozen params: Have {avail_numel} numels to process.')
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
total_params = 0
total_numel = 0
for name, shape in frozen_param_shapes.items():
total_params += 1
unpartitioned_numel = shape.numel()
total_numel += unpartitioned_numel
state_dict[name] = frozen_param_fragments[name]
if debug:
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
def _has_callable(obj, fn):
attr = getattr(obj, fn, None)
return callable(attr)
def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
param_shapes = zero_model_states[0].param_shapes
# Reconstruction protocol:
#
# XXX: document this
if debug:
for i in range(world_size):
for j in range(len(fp32_flat_groups[0])):
print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
# XXX: memory usage doubles here (zero2)
num_param_groups = len(fp32_flat_groups[0])
merged_single_partition_of_fp32_groups = []
for i in range(num_param_groups):
merged_partitions = [sd[i] for sd in fp32_flat_groups]
full_single_fp32_vector = torch.cat(merged_partitions, 0)
merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
avail_numel = sum(
[full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
if debug:
wanted_params = sum([len(shapes) for shapes in param_shapes])
wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
# not asserting if there is a mismatch due to possible padding
print(f"Have {avail_numel} numels to process.")
print(f"Need {wanted_numel} numels in {wanted_params} params.")
# params
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
# out-of-core computing solution
total_numel = 0
total_params = 0
for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
offset = 0
avail_numel = full_single_fp32_vector.numel()
for name, shape in shapes.items():
unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
total_numel += unpartitioned_numel
total_params += 1
if debug:
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
offset += unpartitioned_numel
# Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
# avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
# paddings performed in the code it's almost impossible to predict the exact numbers w/o the
# live optimizer object, so we are checking that the numbers are within the right range
align_to = 2 * world_size
def zero2_align(x):
return align_to * math.ceil(x / align_to)
if debug:
print(f"original offset={offset}, avail_numel={avail_numel}")
offset = zero2_align(offset)
avail_numel = zero2_align(avail_numel)
if debug:
print(f"aligned offset={offset}, avail_numel={avail_numel}")
# Sanity check
if offset != avail_numel:
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
exclude_frozen_parameters):
state_dict = OrderedDict()
# buffers
buffers = zero_model_states[0].buffers
state_dict.update(buffers)
if debug:
print(f"added {len(buffers)} buffers")
if not exclude_frozen_parameters:
_zero2_merge_frozen_params(state_dict, zero_model_states)
_zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
# recover shared parameters
for pair in zero_model_states[0].shared_params:
if pair[1] in state_dict:
state_dict[pair[0]] = state_dict[pair[1]]
return state_dict
def zero3_partitioned_param_info(unpartitioned_numel, world_size):
remainder = unpartitioned_numel % world_size
padding_numel = (world_size - remainder) if remainder else 0
partitioned_numel = math.ceil(unpartitioned_numel / world_size)
return partitioned_numel, padding_numel
def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
return
if debug:
for i in range(world_size):
num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
wanted_params = len(frozen_param_shapes)
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
print(f'Frozen params: Have {avail_numel} numels to process.')
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
total_params = 0
total_numel = 0
for name, shape in zero_model_states[0].frozen_param_shapes.items():
total_params += 1
unpartitioned_numel = shape.numel()
total_numel += unpartitioned_numel
param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
if debug:
print(
f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
)
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
param_shapes = zero_model_states[0].param_shapes
avail_numel = fp32_flat_groups[0].numel() * world_size
# Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
# param, re-consolidating each param, while dealing with padding if any
# merge list of dicts, preserving order
param_shapes = {k: v for d in param_shapes for k, v in d.items()}
if debug:
for i in range(world_size):
print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
wanted_params = len(param_shapes)
wanted_numel = sum(shape.numel() for shape in param_shapes.values())
# not asserting if there is a mismatch due to possible padding
avail_numel = fp32_flat_groups[0].numel() * world_size
print(f"Trainable params: Have {avail_numel} numels to process.")
print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
# params
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
# out-of-core computing solution
offset = 0
total_numel = 0
total_params = 0
for name, shape in param_shapes.items():
unpartitioned_numel = shape.numel()
total_numel += unpartitioned_numel
total_params += 1
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
if debug:
print(
f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
)
# XXX: memory usage doubles here
state_dict[name] = torch.cat(
tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
0).narrow(0, 0, unpartitioned_numel).view(shape)
offset += partitioned_numel
offset *= world_size
# Sanity check
if offset != avail_numel:
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
exclude_frozen_parameters):
state_dict = OrderedDict()
# buffers
buffers = zero_model_states[0].buffers
state_dict.update(buffers)
if debug:
print(f"added {len(buffers)} buffers")
if not exclude_frozen_parameters:
_zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
_zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
# recover shared parameters
for pair in zero_model_states[0].shared_params:
if pair[1] in state_dict:
state_dict[pair[0]] = state_dict[pair[1]]
return state_dict
def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
"""
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
via a model hub.
Args:
- ``checkpoint_dir``: path to the desired checkpoint folder
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
- ``exclude_frozen_parameters``: exclude frozen parameters
Returns:
- pytorch ``state_dict``
Note: this approach may not work if your application doesn't have sufficient free CPU memory and
you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
the checkpoint.
A typical usage might be ::
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
# do the training and checkpoint saving
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
model = model.cpu() # move to cpu
model.load_state_dict(state_dict)
# submit to model hub or save the model to share with others
In this example the ``model`` will no longer be usable in the deepspeed context of the same
application. i.e. you will need to re-initialize the deepspeed engine, since
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
"""
if tag is None:
latest_path = os.path.join(checkpoint_dir, 'latest')
if os.path.isfile(latest_path):
with open(latest_path, 'r') as fd:
tag = fd.read().strip()
else:
raise ValueError(f"Unable to find 'latest' file at {latest_path}")
ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
if not os.path.isdir(ds_checkpoint_dir):
raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
"""
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
Args:
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
- ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
- ``exclude_frozen_parameters``: exclude frozen parameters
"""
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
print(f"Saving fp32 state dict to {output_file}")
torch.save(state_dict, output_file)
def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
"""
1. Put the provided model to cpu
2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
3. Load it into the provided model
Args:
- ``model``: the model object to update
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
Returns:
- ``model`: modified model
Make sure you have plenty of CPU memory available before you call this function. If you don't
have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
conveniently placed for you in the checkpoint folder.
A typical usage might be ::
from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
# submit to model hub or save the model to share with others
Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
of the same application. i.e. you will need to re-initialize the deepspeed engine, since
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
"""
logger.info(f"Extracting fp32 weights")
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
logger.info(f"Overwriting model with fp32 weights")
model = model.cpu()
model.load_state_dict(state_dict, strict=False)
return model
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("checkpoint_dir",
type=str,
help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
parser.add_argument(
"output_file",
type=str,
help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
parser.add_argument("-t",
"--tag",
type=str,
default=None,
help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
args = parser.parse_args()
debug = args.debug
convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
args.output_file,
tag=args.tag,
exclude_frozen_parameters=args.exclude_frozen_parameters)