gi111f11cb7f354a74a8c123f5b155498e
/
Doge-160M-Instruct
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

first commit

This commit is contained in:
xxl 2025-03-14 14:56:33 +08:00
parent 2d4485c273
commit e718bd4f4c
9 changed files with 167917 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

143
README.md
View File

@ -1,3 +1,142 @@
# Doge-160M-Instruct
---
library_name: transformers
license: apache-2.0
datasets:
- HuggingFaceTB/smoltalk
- HuggingFaceH4/ultrafeedback_binarized
base_model:
- SmallDoge/Doge-160M
language:
- en
pipeline_tag: question-answering
tags:
- trl
- sft
- dpo
- doge
---
Doge-160M-Instruct
# **Doge 160M Instruct**
<div align="center">
<img src="https://huggingface.co/spaces/SmallDoge/README/resolve/main/org_icon.png" width="100%" alt="SmallDoge" />
</div>
<hr>
<div align="center">
<a href="https://discord.gg/P2yYH95N" target="_blank" style="margin: 2px;">
<img alt="Discord" src="https://img.shields.io/badge/Discord-Small%20Doges-7289da?logo=discord&logoColor=white&color=7289da" style="display: inline-block; vertical-align: middle;"/>
</a>
<a href="https://arxiv.org/abs/2412.11834" target="_blank" style="margin: 2px;">
<img alt="arXiv" src="https://img.shields.io/static/v1?label=arXiv&message=2412.11834&color=B31B1B&logo=arXiv" style="display: inline-block; vertical-align: middle;"/>
</a>
<a href="https://github.com/SmallDoges/small-doge" target="_blank" style="margin: 2px;">
<img alt="GitHub" src="https://img.shields.io/badge/GitHub-SmallDoge-181717?logo=github" style="display: inline-block; vertical-align: middle;"/>
</a>
<a href="https://github.com/SmallDoges/small-doge/blob/main/LICENSE" style="margin: 2px;">
<img alt="License" src="https://img.shields.io/badge/License-Apache--2.0-blue.svg" style="display: inline-block; vertical-align: middle;"/>
</a>
</div>
Doge uses Dynamic Mask Attention as sequence transformation and can use Multi-Layer Perceptron or Cross Domain Mixture of Experts as state transformation. Dynamic Mask Attention allows the Transformer to use self-attention during training and state space during inference, and Cross Domain Mixture of Experts can directly inherit the weights of Multi-Layer Perceptron for further training. This model is trained by [SmallDoge](https://huggingface.co/SmallDoge) community, for detailed algorithm and model architecture, please refer to [Wonderful Matrices](https://arxiv.org/abs/2412.11834), all training details and code are publicly available on the [small-doge](https://github.com/SmallDoges/small-doge) repository.
## Uses
```python
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig, TextStreamer
tokenizer = AutoTokenizer.from_pretrained("SmallDoge/Doge-160M-Instruct")
model = AutoModelForCausalLM.from_pretrained("SmallDoge/Doge-160M-Instruct", trust_remote_code=True)
generation_config = GenerationConfig(
max_new_tokens=100,
use_cache=True,
do_sample=True,
temperature=0.8,
top_p=0.9,
repetition_penalty=1.0
)
steamer = TextStreamer(
tokenizer=tokenizer,
skip_prompt=True
)
prompt = "Hi, how are you doing today?"
conversation = [
{"role": "user", "content": prompt}
]
inputs = tokenizer.apply_chat_template(
conversation=conversation,
tokenize=True,
return_tensors="pt",
)
outputs = model.generate(
inputs,
tokenizer=tokenizer,
generation_config=generation_config,
streamer=steamer
)
```
## Model Details
We build the Doge-Instruct by first SFT on [SmolTalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk) and then DPO on [UltraFeedback Binarized](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized).
> TODO: The larger model is under training and will be uploaded soon.
**SFT**:
| Model | Training Data | Epochs | Content Length | LR | Batch Size | Precision |
|---|---|---|---|---|---|---|
| [Doge-20M-Instruct-SFT](https://huggingface.co/SmallDoge/Doge-20M-Instruct-SFT) | [HuggingFaceTB/smoltalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk) | 2 | 2048 | 8e-4 | 0.25M | bfloat16 |
| [Doge-60M-Instruct-SFT](https://huggingface.co/SmallDoge/Doge-60M-Instruct-SFT) | [HuggingFaceTB/smoltalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk) | 2 | 2048 | 6e-4 | 0.25M | bfloat16 |
| [Doge-160M-Instruct-SFT](https://huggingface.co/SmallDoge/Doge-160M-Instruct-SFT) | [HuggingFaceTB/smoltalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk) | 2 | 2048 | 4e-4 | 0.25M | bfloat16 |
**DPO**:
| Model | Training Data | Epochs | Content Length | LR | Batch Size | Precision |
|---|---|---|---|---|---|---|
| [Doge-20M-Instruct](https://huggingface.co/SmallDoge/Doge-20M-Instruct) | [HuggingFaceH4/ultrafeedback_binarized](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized) | 2 | 1024 | 8e-5 | 0.125M | bfloat16 |
| [Doge-60M-Instruct](https://huggingface.co/SmallDoge/Doge-60M-Instruct) | [HuggingFaceH4/ultrafeedback_binarized](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized) | 2 | 1024 | 6e-5 | 0.125M | bfloat16 |
| [Doge-160M-Instruct](https://huggingface.co/SmallDoge/Doge-160M-Instruct) | [HuggingFaceH4/ultrafeedback_binarized](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized) | 2 | 1024 | 4e-5 | 0.125M | bfloat16 |
**Evaluation**:
| Model | IFEval (Prompt Strict Acc) | MMLU | BBH | ARC | PIQA | HellaSwag | tokens / s on i7-11 CPU |
|---|---|---|---|---|---|---|---|
| [Doge-20M-Instruct](https://huggingface.co/SmallDoge/Doge-20M-Instruct) | 7.3 | 26.3 | 18.3 | 29.2 | 57.8 | 27.8 | 142 |
| [Doge-60M-Instruct](https://huggingface.co/SmallDoge/Doge-60M-Instruct) | 7.4 | 27.5 | 27.7 | 37.5 | 61.4 | 32.1 | 62 |
| [SmolLM-135M-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM-135M-Instruct) | 12.2 | 28.3 | 25.2 | 33.9 | 64.0 | 38.9 | 30 |
| [Doge-160M-Instruct](https://huggingface.co/SmallDoge/Doge-160M-Instruct) | 16.8 | 29.7 | 29.1 | 42.8 | 64.1 | 37.1 | 28 |
**Procedure**:
**SFT**:
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/loser_cheems/huggingface/runs/0jht5dro)
**DPO**:
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/loser_cheems/huggingface/runs/m5onn07v)
**Environment**:
- Image: nvcr.io/nvidia/pytorch:24.12-py3
- Hardware: 1x NVIDIA RTX 4090
- Software: Transformers, TRL
## Citation
```bibtex
@misc{shi2024wonderfulmatrices,
title={Wonderful Matrices: Combining for a More Efficient and Effective Foundation Model Architecture},
author={Jingze Shi and Bingheng Wu},
year={2024},
eprint={2412.11834},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2412.11834},
}
```

43
config.json Normal file
View File

@ -0,0 +1,43 @@
{
"_name_or_path": "SmallDoge/Doge-160M-Instruct",
"architectures": [
"DogeForCausalLM"
],
"attention_dropout": 0.0,
"auto_map": {
"AutoConfig": "configuration_doge.DogeConfig",
"AutoModelForCausalLM": "modeling_doge.DogeForCausalLM"
},
"bos_token_id": 0,
"dynamic_mask_ratio": 0.0,
"eos_token_id": 1,
"expert_retrieval_size": 64,
"hidden_act": "silu",
"hidden_bias": false,
"hidden_dropout": 0.0,
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 1536,
"is_moe": false,
"max_position_embeddings": 2048,
"model_type": "doge",
"num_attention_heads": 6,
"num_cdmoe_experts": 16348,
"num_cdmoe_experts_per_head": 8,
"num_cdmoe_heads": 4,
"num_hidden_layers": 24,
"num_key_value_heads": 3,
"pad_token_id": 2,
"rms_norm_eps": 1e-06,
"rope_scaling": {
"factor": 4.0,
"original_max_position_embeddings": 2048,
"rope_type": "dynamic"
},
"rope_theta": 10000.0,
"tie_word_embeddings": true,
"torch_dtype": "float32",
"transformers_version": "4.48.3",
"use_cache": true,
"vocab_size": 32768
}

228
configuration_doge.py Normal file
View File

@ -0,0 +1,228 @@
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# This file was automatically generated from src/transformers/models/doge/modular_doge.py.
# Do NOT edit this file manually as any edits will be overwritten by the generation of
# the file from the modular. If any change should be done, please apply the change to the
# modular_doge.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# coding=utf-8
# Copyright 2024 Jingze Shi and the HuggingFace Inc. team. All rights reserved.
#
# This code is based on the Wonderful Matrices paper implementation.
# The Doge family of small language models is trained by Jingze Shi.
#
# 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.
from transformers.configuration_utils import PretrainedConfig
from transformers.modeling_rope_utils import rope_config_validation
class DogeConfig(PretrainedConfig):
r"""
This is the configuration class to store the configuration of a [`DogeModel`]. It is used to instantiate an Doge
model according to the specified arguments, defining the model architecture like [SmallDoge/Doge-20M](https://huggingface.co/SmallDoge/Doge-20M).
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 32768):
Vocabulary size of the Doge model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`DogeModel`]
hidden_size (`int`, *optional*, defaults to 1024):
Dimension of the hidden representations.
intermediate_size (`int`, *optional*, defaults to 2048):
Dimension of the MLP representations.
num_hidden_layers (`int`, *optional*, defaults to 32):
Number of hidden layers in the Transformer decoder.
hidden_bias (`bool`, *optional*, defaults to `False`):
Whether to use bias in the hidden layers.
hidden_dropout (`float`, *optional*, defaults to 0.0):
Dropout probability for each sequence transformation and state transformation module.
hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
The non-linear activation function (function or string) in the decoder.
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`.
bos_token_id (`int`, *optional*, defaults to 0):
Beginning of stream token id.
eos_token_id (`int`, *optional*, defaults to 1):
End of stream token id.
pad_token_id (`int`, *optional*, defaults to 2):
Padding token id.
tie_word_embeddings (`bool`, *optional*, defaults to `False`):
Whether to tie weight embeddings
max_position_embeddings (`int`, *optional*, defaults to 2048):
The maximum sequence length that this model might ever be used with.
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.
Doge family of small models use `{ 'rope_type': 'dynamic', 'factor': 4.0, 'original_max_position_embeddings': 2048 }` as the default value.
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
num_attention_heads (`int`, *optional*, defaults to 8):
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`.
attention_dropout (`float`, *optional*, defaults to 0.0):
The dropout ratio for the attention probabilities.
dynamic_mask_ratio (`float`, *optional*, defaults to 0.0):
The ratio to control the proportion of the dynamic mask filled with the minimum value. For more details checkout [this paper](https://arxiv.org/pdf/2412.11834).
is_moe (`bool`, *optional*, defaults to `False`):
Whether to use the Cross Domain Mixture of Experts, if `True`, the MoE will inherit the MLP to initialize. For more details checkout [this paper](https://arxiv.org/pdf/2412.11834).
num_cdmoe_experts (`int`, *optional*, defaults to 16348):
Number of Experts for the Cross Domain Mixture of Experts.
num_cdmoe_heads (`int`, *optional*, defaults to 4):
Number of retrieval heads, used to mix multi-head experts.
num_cdmoe_experts_per_head (`int`, *optional*, defaults to 8):
Number of Experts per retrieval head, used to mix multi-head experts.
expert_retrieval_size (`int`, *optional*, defaults to 64):
Dimension of the Expert retrieval states for calculating the dot product of query and key to determine the expert index.
```python
>>> from transformers import DogeConfig, DogeModel
>>> # Initializing a Doge-320M style configuration
>>> configuration = DogeConfig()
>>> # Initializing a model from the Doge-320M style configuration
>>> model = DogeModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
```"""
model_type = "doge"
keys_to_ignore_at_inference = ["past_key_values"]
# Default tensor parallel plan for base model `DogeModel`
base_model_tp_plan = {
"layers.*.self_attn.q_proj": "colwise",
"layers.*.self_attn.k_proj": "colwise",
"layers.*.self_attn.v_proj": "colwise",
"layers.*.self_attn.dt_proj": "rowwise",
"layers.*.self_attn.o_proj": "rowwise",
"layers.*.mlp.gate_proj": "colwise",
"layers.*.mlp.up_proj": "colwise",
"layers.*.mlp.down_proj": "rowwise",
}
def __init__(
self,
vocab_size=32768,
hidden_size=1024,
intermediate_size=2048,
num_hidden_layers=32,
hidden_bias=False,
hidden_dropout=0.0,
hidden_act="silu",
initializer_range=0.02,
rms_norm_eps=1e-06,
use_cache=True,
bos_token_id=0,
eos_token_id=1,
pad_token_id=2,
tie_word_embeddings=False,
max_position_embeddings=2048,
rope_theta=10000.0,
rope_scaling=None,
num_attention_heads=8,
num_key_value_heads=None,
attention_dropout=0.0,
dynamic_mask_ratio=0.0,
is_moe=False,
num_cdmoe_experts=16348,
num_cdmoe_heads=4,
num_cdmoe_experts_per_head=8,
expert_retrieval_size=64,
**kwargs,
):
self.vocab_size = vocab_size
self.hidden_size = hidden_size
self.intermediate_size = intermediate_size
self.num_hidden_layers = num_hidden_layers
self.hidden_bias = hidden_bias
self.hidden_dropout = hidden_dropout
self.hidden_act = hidden_act
self.initializer_range = initializer_range
self.rms_norm_eps = rms_norm_eps
self.use_cache = use_cache
self.max_position_embeddings = max_position_embeddings
self.rope_theta = rope_theta
self.rope_scaling = rope_scaling
self.num_attention_heads = num_attention_heads
self.num_key_value_heads = num_key_value_heads
self.attention_dropout = attention_dropout
self.dynamic_mask_ratio = dynamic_mask_ratio
self.is_moe = is_moe
self.num_cdmoe_experts = num_cdmoe_experts
self.num_cdmoe_heads = num_cdmoe_heads
self.num_cdmoe_experts_per_head = num_cdmoe_experts_per_head
self.expert_retrieval_size = expert_retrieval_size
# 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)
# for backward compatibility
if num_key_value_heads is None:
self.num_key_value_heads = num_attention_heads
super().__init__(
bos_token_id=bos_token_id,
eos_token_id=eos_token_id,
pad_token_id=pad_token_id,
tie_word_embeddings=tie_word_embeddings,
**kwargs,
)
__all__ = ["DogeConfig"]

7
generation_config.json Normal file
View File

@ -0,0 +1,7 @@
{
"_from_model_config": true,
"bos_token_id": 0,
"eos_token_id": 1,
"pad_token_id": 2,
"transformers_version": "4.48.3"
}

BIN
model.safetensors (Stored with Git LFS) Normal file
View File

Binary file not shown.

1198
modeling_doge.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": "<|end_of_text|>",
"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
}
}

164208
tokenizer.json Normal file
View File

File diff suppressed because it is too large Load Diff

2066
tokenizer_config.json Normal file
View File

File diff suppressed because it is too large Load Diff