guanwenyu1995 commited on
Commit
2774e76
·
verified ·
1 Parent(s): 4e73e9c

Upload 11 files

Browse files

Add model files of
MiniCPM4-0.5B-QAT-Int4-unquantized

README.md CHANGED
@@ -1,3 +1,152 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ language:
4
+ - zh
5
+ - en
6
+ pipeline_tag: text-generation
7
+ library_name: transformers
8
+ ---
9
+ <div align="center">
10
+ <img src="https://github.com/OpenBMB/MiniCPM/blob/main/assets/minicpm_logo.png?raw=true" width="500em" ></img>
11
+ </div>
12
+
13
+ <p align="center">
14
+ <a href="https://github.com/OpenBMB/MiniCPM/" target="_blank">GitHub Repo</a> |
15
+ <a href="https://github.com/OpenBMB/MiniCPM/tree/main/report/MiniCPM_4_Technical_Report.pdf" target="_blank">Technical Report</a>
16
+ </p>
17
+ <p align="center">
18
+ 👋 Join us on <a href="https://discord.gg/3cGQn9b3YM" target="_blank">Discord</a> and <a href="https://github.com/OpenBMB/MiniCPM/blob/main/assets/wechat.jpg" target="_blank">WeChat</a>
19
+ </p>
20
+
21
+ ## What's New
22
+ - [2025.06.06] **MiniCPM4** series are released! This model achieves ultimate efficiency improvements while maintaining optimal performance at the same scale! It can achieve over 5x generation acceleration on typical end-side chips! You can find technical report [here](https://github.com/OpenBMB/MiniCPM/tree/main/report/MiniCPM_4_Technical_Report.pdf).🔥🔥🔥
23
+
24
+ ## MiniCPM4 Series
25
+ MiniCPM4 series are highly efficient large language models (LLMs) designed explicitly for end-side devices, which achieves this efficiency through systematic innovation in four key dimensions: model architecture, training data, training algorithms, and inference systems.
26
+ - [MiniCPM4-8B](https://huggingface.co/openbmb/MiniCPM4-8B): The flagship of MiniCPM4, with 8B parameters, trained on 8T tokens.
27
+ - [MiniCPM4-0.5B](https://huggingface.co/openbmb/MiniCPM4-0.5B): The small version of MiniCPM4, with 0.5B parameters, trained on 1T tokens.
28
+ - [MiniCPM4-8B-Eagle-FRSpec](https://huggingface.co/openbmb/MiniCPM4-8B-Eagle-FRSpec): Eagle head for FRSpec, accelerating speculative inference for MiniCPM4-8B.
29
+ - [MiniCPM4-8B-Eagle-FRSpec-QAT-cpmcu](https://huggingface.co/openbmb/MiniCPM4-8B-Eagle-FRSpec-QAT-cpmcu): Eagle head trained with QAT for FRSpec, efficiently integrate speculation and quantization to achieve ultra acceleration for MiniCPM4-8B.
30
+ - [MiniCPM4-8B-Eagle-vLLM](https://huggingface.co/openbmb/MiniCPM4-8B-Eagle-vLLM): Eagle head in vLLM format, accelerating speculative inference for MiniCPM4-8B.
31
+ - [MiniCPM4-8B-marlin-Eagle-vLLM](https://huggingface.co/openbmb/MiniCPM4-8B-marlin-Eagle-vLLM): Quantized Eagle head for vLLM format, accelerating speculative inference for MiniCPM4-8B.
32
+ - [BitCPM4-0.5B](https://huggingface.co/openbmb/BitCPM4-0.5B): Extreme ternary quantization applied to MiniCPM4-0.5B compresses model parameters into ternary values, achieving a 90% reduction in bit width.
33
+ - [BitCPM4-1B](https://huggingface.co/openbmb/BitCPM4-1B): Extreme ternary quantization applied to MiniCPM3-1B compresses model parameters into ternary values, achieving a 90% reduction in bit width.
34
+ - [MiniCPM4-Survey](https://huggingface.co/openbmb/MiniCPM4-Survey): Based on MiniCPM4-8B, accepts users' quiries as input and autonomously generate trustworthy, long-form survey papers.
35
+ - [MiniCPM4-MCP](https://huggingface.co/openbmb/MiniCPM4-MCP): Based on MiniCPM4-8B, accepts users' queries and available MCP tools as input and autonomously calls relevant MCP tools to satisfy users' requirements.
36
+ - [MiniCPM4-0.5B-QAT-Int4-unquantized](https://huggingface.co/openbmb/MiniCPM4-0.5B-QAT-Int4-unquantized): Int4 version of MiniCPM4-0.5B, trained by QAT and stored in fake quantization style. (**<-- you are here**)
37
+ - [MiniCPM4-0.5B-QAT-Int4-GPTQ-format](https://huggingface.co/openbmb/MiniCPM4-0.5B-QAT-Int4-GPTQ-format): Int4 version of MiniCPM4-0.5B, trained by QAT and stored in GPTQ format.
38
+ - [MiniCPM4-0.5B-QAT-Int4-GGUF](https://huggingface.co/openbmb/MiniCPM4-0.5B-QAT-Int4-GGUF): Int4 version of MiniCPM4-0.5B in GGUF.
39
+
40
+ ## Introduction
41
+ MiniCPM 4 is an extremely efficient edge-side large model that has undergone efficient optimization across four dimensions: model architecture, learning algorithms, training data, and inference systems, achieving ultimate efficiency improvements.
42
+
43
+ - 🏗️ **Efficient Model Architecture:**
44
+ - InfLLM v2 -- Trainable Sparse Attention Mechanism: Adopts a trainable sparse attention mechanism architecture where each token only needs to compute relevance with less than 5% of tokens in 128K long text processing, significantly reducing computational overhead for long texts
45
+
46
+ - 🧠 **Efficient Learning Algorithms:**
47
+ - Model Wind Tunnel 2.0 -- Efficient Predictable Scaling: Introduces scaling prediction methods for performance of downstream tasks, enabling more precise model training configuration search
48
+ - BitCPM -- Ultimate Ternary Quantization: Compresses model parameter bit-width to 3 values, achieving 90% extreme model bit-width reduction
49
+ - Efficient Training Engineering Optimization: Adopts FP8 low-precision computing technology combined with Multi-token Prediction training strategy
50
+
51
+ - 📚 **High-Quality Training Data:**
52
+ - UltraClean -- High-quality Pre-training Data Filtering and Generation: Builds iterative data cleaning strategies based on efficient data verification, open-sourcing high-quality Chinese and English pre-training dataset [UltraFinweb](https://huggingface.co/datasets/openbmb/Ultra-FineWeb)
53
+ - UltraChat v2 -- High-quality Supervised Fine-tuning Data Generation: Constructs large-scale high-quality supervised fine-tuning datasets covering multiple dimensions including knowledge-intensive data, reasoning-intensive data, instruction-following data, long text understanding data, and tool calling data
54
+
55
+ - ⚡ **Efficient Inference System:**
56
+ - CPM.cu -- Lightweight and Efficient CUDA Inference Framework: Integrates sparse attention, model quantization, and speculative sampling to achieve efficient prefilling and decoding
57
+ - ArkInfer -- Cross-platform Deployment System: Supports efficient deployment across multiple backend environments, providing flexible cross-platform adaptation capabilities
58
+
59
+ ## Usage
60
+ ### Inference with Transformers
61
+ ```python
62
+ from transformers import AutoModelForCausalLM, AutoTokenizer
63
+ import torch
64
+
65
+ path = "openbmb/MiniCPM4-0.5B-QAT-Int4-unquantized"
66
+ device = "cuda"
67
+
68
+ tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True)
69
+ model = AutoModelForCausalLM.from_pretrained(path, torch_dtype=torch.bfloat16, device_map=device, trust_remote_code=True)
70
+
71
+ messages = [
72
+ {"role": "user", "content": "推荐5个北京的景点。"},
73
+ ]
74
+ model_inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True).to(device)
75
+
76
+ model_outputs = model.generate(
77
+ model_inputs,
78
+ max_new_tokens=1024,
79
+ top_p=0.7,
80
+ temperature=0.7
81
+ )
82
+
83
+ output_token_ids = [
84
+ model_outputs[i][len(model_inputs[i]):] for i in range(len(model_inputs))
85
+ ]
86
+
87
+ responses = tokenizer.batch_decode(output_token_ids, skip_special_tokens=True)[0]
88
+ print(responses)
89
+
90
+ ```
91
+
92
+ ### Inference with [vLLM](https://github.com/vllm-project/vllm)
93
+
94
+ You can inference MiniCPM4-0.5B-QAT-Int4-unquantized with vLLM:
95
+ ```python
96
+ from transformers import AutoTokenizer
97
+ from vllm import LLM, SamplingParams
98
+
99
+ model_name = "openbmb/MiniCPM4-0.5B-QAT-Int4-unquantized"
100
+ prompt = [{"role": "user", "content": "推荐5个北京的景点。"}]
101
+
102
+ tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
103
+ input_text = tokenizer.apply_chat_template(prompt, tokenize=False, add_generation_prompt=True)
104
+
105
+ llm = LLM(
106
+ model=model_name,
107
+ trust_remote_code=True,
108
+ max_num_batched_tokens=32768,
109
+ dtype="bfloat16",
110
+ gpu_memory_utilization=0.8,
111
+ )
112
+ sampling_params = SamplingParams(top_p=0.7, temperature=0.7, max_tokens=1024, repetition_penalty=1.02)
113
+
114
+ outputs = llm.generate(prompts=input_text, sampling_params=sampling_params)
115
+
116
+ print(outputs[0].outputs[0].text)
117
+ ```
118
+
119
+ ## Evaluation Results
120
+ | Model | Qwen3 | Llama3.2 | Gemma3 | MiniCPM4 | MiniCPM4 | MiniCPM4 |
121
+ |----------------|-------|----------|--------|----------|----------|----------|
122
+ | #Paramete | 0.6B | 1B | 1B | 0.5B | 0.5B | 0.5B |
123
+ | #Precision | BF16 | BF16 | BF16 | BF16 |Int4(Fake)|Int4(GPTQ)|
124
+ | MMLU | 42.95 | 46.89 | 41.64 | 55.55 | 55.46 | 53.93 |
125
+ | CMMLU | 42.05 | 23.73 | 25.09 | 65.22 | 63.91 | 63.73 |
126
+ | Ceval | 45.53 | 36.74 | 31.83 | 66.11 | 64.85 | 65.22 |
127
+ | BBH | 28.32 | 25.42 | 33.21 | 49.87 | 48.81 | 49.09 |
128
+ | GSM8K | 61.71 | 39.76 | 61.26 | 52.08 | 45.41 | 45.49 |
129
+ | MBPP | 47.86 | 47.47 | 59.92 | 59.14 | 55.64 | 55.25 |
130
+ | AVERAGE | 44.73 | 36.66 | 42.15 | 58.00 | 55.68 | 55.45 |
131
+
132
+
133
+
134
+ ## Statement
135
+ - As a language model, MiniCPM generates content by learning from a vast amount of text.
136
+ - However, it does not possess the ability to comprehend or express personal opinions or value judgments.
137
+ - Any content generated by MiniCPM does not represent the viewpoints or positions of the model developers.
138
+ - Therefore, when using content generated by MiniCPM, users should take full responsibility for evaluating and verifying it on their own.
139
+
140
+ ## LICENSE
141
+ - This repository and MiniCPM models are released under the [Apache-2.0](https://github.com/OpenBMB/MiniCPM/blob/main/LICENSE) License.
142
+
143
+ ## Citation
144
+ - Please cite our [paper](https://github.com/OpenBMB/MiniCPM/tree/main/report/MiniCPM_4_Technical_Report.pdf) if you find our work valuable.
145
+
146
+ ```bibtex
147
+ @article{minicpm4,
148
+ title={{MiniCPM4}: Ultra-Efficient LLMs on End Devices},
149
+ author={MiniCPM Team},
150
+ year={2025}
151
+ }
152
+ ```
added_tokens.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "<|execute_end|>": 73444,
3
+ "<|execute_start|>": 73443,
4
+ "<|fim_middle|>": 73446,
5
+ "<|fim_prefix|>": 73445,
6
+ "<|fim_suffix|>": 73447,
7
+ "<|im_end|>": 73440,
8
+ "<|im_start|>": 73441,
9
+ "<|tool_call|>": 73442
10
+ }
config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "openbmb/MiniCPM4-0.5B-QAT-Int4-unquantized",
3
+ "architectures": [
4
+ "MiniCPMForCausalLM"
5
+ ],
6
+ "auto_map": {
7
+ "AutoConfig": "configuration_minicpm.MiniCPMConfig",
8
+ "AutoModel": "modeling_minicpm.MiniCPMModel",
9
+ "AutoModelForCausalLM": "modeling_minicpm.MiniCPMForCausalLM",
10
+ "AutoModelForSeq2SeqLM": "modeling_minicpm.MiniCPMForCausalLM",
11
+ "AutoModelForSequenceClassification": "modeling_minicpm.MiniCPMForSequenceClassification"
12
+ },
13
+ "bos_token_id": 1,
14
+ "eos_token_id": [2, 73440],
15
+ "hidden_act": "silu",
16
+ "hidden_size": 1024,
17
+ "initializer_range": 0.1,
18
+ "intermediate_size": 4096,
19
+ "max_position_embeddings": 32768,
20
+ "num_attention_heads": 16,
21
+ "num_hidden_layers": 24,
22
+ "num_key_value_heads": 2,
23
+ "rms_norm_eps": 1e-05,
24
+ "rope_scaling": {
25
+ "rope_type": "longrope",
26
+ "long_factor": [1.0004360675811768, 1.0668443441390991, 1.1631425619125366, 1.3025742769241333, 1.5040205717086792, 1.7941505908966064, 2.2101221084594727, 2.802666664123535, 3.6389970779418945, 4.804192543029785, 6.39855432510376, 8.527148246765137, 11.277542114257812, 14.684998512268066, 18.69317054748535, 23.13019371032715, 27.72362518310547, 32.1606559753418, 36.168827056884766, 39.57627868652344, 42.32667541503906, 44.45526885986328, 46.04962921142578, 47.21482849121094, 48.05115509033203, 48.64370346069336, 49.05967712402344, 49.34980392456055, 49.551246643066406, 49.69068145751953, 49.78697967529297, 49.85338592529297],
27
+ "short_factor": [1.0004360675811768, 1.0668443441390991, 1.1631425619125366, 1.3025742769241333, 1.5040205717086792, 1.7941505908966064, 2.2101221084594727, 2.802666664123535, 3.6389970779418945, 4.804192543029785, 6.39855432510376, 8.527148246765137, 11.277542114257812, 14.684998512268066, 18.69317054748535, 23.13019371032715, 27.72362518310547, 32.1606559753418, 36.168827056884766, 39.57627868652344, 42.32667541503906, 44.45526885986328, 46.04962921142578, 47.21482849121094, 48.05115509033203, 48.64370346069336, 49.05967712402344, 49.34980392456055, 49.551246643066406, 49.69068145751953, 49.78697967529297, 49.85338592529297],
28
+ "original_max_position_embeddings": 32768
29
+ },
30
+ "torch_dtype": "bfloat16",
31
+ "transformers_version": "4.46.3",
32
+ "use_cache": true,
33
+ "vocab_size": 73448,
34
+ "scale_emb": 12,
35
+ "dim_model_base": 256,
36
+ "scale_depth": 1.4
37
+ }
configuration_minicpm.py ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
2
+ #
3
+ # This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
4
+ # and OPT implementations in this library. It has been modified from its
5
+ # original forms to accommodate minor architectural differences compared
6
+ # to GPT-NeoX and OPT used by the Meta AI team that trained the model.
7
+ #
8
+ # Licensed under the Apache License, Version 2.0 (the "License");
9
+ # you may not use this file except in compliance with the License.
10
+ # You may obtain a copy of the License at
11
+ #
12
+ # http://www.apache.org/licenses/LICENSE-2.0
13
+ #
14
+ # Unless required by applicable law or agreed to in writing, software
15
+ # distributed under the License is distributed on an "AS IS" BASIS,
16
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
17
+ # See the License for the specific language governing permissions and
18
+ # limitations under the License.
19
+ """ MiniCPM model configuration"""
20
+
21
+ from transformers.configuration_utils import PretrainedConfig
22
+ from transformers.utils import logging
23
+
24
+ logger = logging.get_logger(__name__)
25
+
26
+ MINICPM_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
27
+
28
+
29
+ class MiniCPMConfig(PretrainedConfig):
30
+ r"""
31
+ This is the configuration class to store the configuration of a [`MiniCPMModel`]. It is used to instantiate an MiniCPM
32
+ model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
33
+ defaults will yield a similar configuration to that of the MiniCPM-7B.
34
+
35
+ Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
36
+ documentation from [`PretrainedConfig`] for more information.
37
+
38
+
39
+ Args:
40
+ vocab_size (`int`, *optional*, defaults to 32000):
41
+ Vocabulary size of the MiniCPM model. Defines the number of different tokens that can be represented by the
42
+ `inputs_ids` passed when calling [`MiniCPMModel`]
43
+ hidden_size (`int`, *optional*, defaults to 4096):
44
+ Dimension of the hidden representations.
45
+ intermediate_size (`int`, *optional*, defaults to 11008):
46
+ Dimension of the MLP representations.
47
+ num_hidden_layers (`int`, *optional*, defaults to 32):
48
+ Number of hidden layers in the Transformer decoder.
49
+ num_attention_heads (`int`, *optional*, defaults to 32):
50
+ Number of attention heads for each attention layer in the Transformer decoder.
51
+ num_key_value_heads (`int`, *optional*):
52
+ This is the number of key_value heads that should be used to implement Grouped Query Attention. If
53
+ `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
54
+ `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
55
+ converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
56
+ by meanpooling all the original heads within that group. For more details checkout [this
57
+ paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
58
+ `num_attention_heads`.
59
+ hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
60
+ The non-linear activation function (function or string) in the decoder.
61
+ max_position_embeddings (`int`, *optional*, defaults to 2048):
62
+ The maximum sequence length that this model might ever be used with. MiniCPM 1 supports up to 2048 tokens,
63
+ MiniCPM 2 up to 4096, CodeMiniCPM up to 16384.
64
+ initializer_range (`float`, *optional*, defaults to 0.02):
65
+ The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
66
+ rms_norm_eps (`float`, *optional*, defaults to 1e-06):
67
+ The epsilon used by the rms normalization layers.
68
+ use_cache (`bool`, *optional*, defaults to `True`):
69
+ Whether or not the model should return the last key/values attentions (not used by all models). Only
70
+ relevant if `config.is_decoder=True`.
71
+ pad_token_id (`int`, *optional*):
72
+ Padding token id.
73
+ bos_token_id (`int`, *optional*, defaults to 1):
74
+ Beginning of stream token id.
75
+ eos_token_id (`int`, *optional*, defaults to 2):
76
+ End of stream token id.
77
+ pretraining_tp (`int`, *optional*, defaults to 1):
78
+ Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this
79
+ document](https://huggingface.co/docs/transformers/parallelism) to understand more about it. This value is
80
+ necessary to ensure exact reproducibility of the pretraining results. Please refer to [this
81
+ issue](https://github.com/pytorch/pytorch/issues/76232).
82
+ tie_word_embeddings (`bool`, *optional*, defaults to `False`):
83
+ Whether to tie weight embeddings
84
+ rope_theta (`float`, *optional*, defaults to 10000.0):
85
+ The base period of the RoPE embeddings.
86
+ rope_scaling (`Dict`, *optional*):
87
+ Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling
88
+ strategies: linear and dynamic. Their scaling factor must be a float greater than 1. The expected format is
89
+ `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update
90
+ `max_position_embeddings` to the expected new maximum. See the following thread for more information on how
91
+ these scaling strategies behave:
92
+ https://www.reddit.com/r/LocalMiniCPM/comments/14mrgpr/dynamically_scaled_rope_further_increases/. This is an
93
+ experimental feature, subject to breaking API changes in future versions.
94
+ attention_bias (`bool`, defaults to `False`, *optional*, defaults to `False`):
95
+ Whether to use a bias in the query, key, value and output projection layers during self-attention.
96
+ attention_dropout (`float`, *optional*, defaults to 0.0):
97
+ The dropout ratio for the attention probabilities.
98
+
99
+ ```python
100
+ >>> from transformers import MiniCPMModel, MiniCPMConfig
101
+
102
+ >>> # Initializing a MiniCPM minicpm-7b style configuration
103
+ >>> configuration = MiniCPMConfig()
104
+
105
+ >>> # Initializing a model from the minicpm-7b style configuration
106
+ >>> model = MiniCPMModel(configuration)
107
+
108
+ >>> # Accessing the model configuration
109
+ >>> configuration = model.config
110
+ ```"""
111
+
112
+ model_type = 'minicpm'
113
+ keys_to_ignore_at_inference = ['past_key_values']
114
+
115
+ def __init__(
116
+ self,
117
+ vocab_size=32000,
118
+ hidden_size=4096,
119
+ intermediate_size=11008,
120
+ num_hidden_layers=32,
121
+ num_attention_heads=32,
122
+ num_key_value_heads=None,
123
+ hidden_act='silu',
124
+ max_position_embeddings=2048,
125
+ initializer_range=0.02,
126
+ rms_norm_eps=1e-6,
127
+ use_cache=True,
128
+ pad_token_id=None,
129
+ bos_token_id=1,
130
+ eos_token_id=2,
131
+ pretraining_tp=1,
132
+ tie_word_embeddings=True,
133
+ rope_theta=10000.0,
134
+ rope_scaling=None,
135
+ attention_bias=False,
136
+ attention_dropout=0.0,
137
+ scale_emb=1,
138
+ dim_model_base=1,
139
+ scale_depth=1,
140
+ mup_denominator=None,
141
+ sparse_config=None,
142
+ **kwargs):
143
+
144
+ self.vocab_size = vocab_size
145
+ self.max_position_embeddings = max_position_embeddings
146
+ self.hidden_size = hidden_size
147
+ self.intermediate_size = intermediate_size
148
+ self.num_hidden_layers = num_hidden_layers
149
+ self.num_attention_heads = num_attention_heads
150
+
151
+ # for backward compatibility
152
+ if num_key_value_heads is None:
153
+ num_key_value_heads = num_attention_heads
154
+
155
+ self.num_key_value_heads = num_key_value_heads
156
+ self.hidden_act = hidden_act
157
+ self.initializer_range = initializer_range
158
+ self.rms_norm_eps = rms_norm_eps
159
+ self.pretraining_tp = pretraining_tp
160
+ self.use_cache = use_cache
161
+ self.rope_theta = rope_theta
162
+ self.rope_scaling = rope_scaling
163
+ # self._rope_scaling_validation()
164
+ self.attention_bias = attention_bias
165
+ self.attention_dropout = attention_dropout
166
+ self.scale_emb = scale_emb
167
+ self.dim_model_base = dim_model_base
168
+ self.scale_depth = scale_depth
169
+ # only used for Eagle Head
170
+ self.mup_denominator = mup_denominator
171
+
172
+ # sparse config
173
+ self.sparse_config = sparse_config
174
+
175
+ super().__init__(
176
+ pad_token_id=pad_token_id,
177
+ bos_token_id=bos_token_id,
178
+ eos_token_id=eos_token_id,
179
+ tie_word_embeddings=tie_word_embeddings,
180
+ **kwargs,
181
+ )
182
+ try:
183
+ import flash_attn
184
+ self._attn_implementation = 'flash_attention_2'
185
+ except:
186
+ pass
187
+
188
+ def _rope_scaling_validation(self):
189
+ """
190
+ Validate the `rope_scaling` configuration.
191
+ """
192
+ if self.rope_scaling is None:
193
+ return
194
+
195
+ if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 2:
196
+ raise ValueError(
197
+ '`rope_scaling` must be a dictionary with with two fields, `type` and `factor`, '
198
+ f'got {self.rope_scaling}'
199
+ )
200
+ rope_scaling_type = self.rope_scaling.get('type', None)
201
+ rope_scaling_factor = self.rope_scaling.get('factor', None)
202
+ if rope_scaling_type is None or rope_scaling_type not in ['linear', 'dynamic']:
203
+ raise ValueError(
204
+ f"`rope_scaling`'s type field must be one of ['linear', 'dynamic'], got {rope_scaling_type}"
205
+ )
206
+ if rope_scaling_factor is None or not isinstance(rope_scaling_factor, float) or rope_scaling_factor <= 1.0:
207
+ raise ValueError(f"`rope_scaling`'s factor field must be a float > 1, got {rope_scaling_factor}")
generation_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 1,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 2,
6
+ 73440
7
+ ],
8
+ "pad_token_id": 2,
9
+ "temperature": 0.8,
10
+ "top_p": 0.8,
11
+ "transformers_version": "4.46.1"
12
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0fa5ac145cb76786bc274f65325695eabdd3ef6e718bac1080db1093f6af44b
3
+ size 1735520504
modeling_minicpm.py ADDED
The diff for this file is too large to render. See raw diff
 
special_tokens_map.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_end|>",
4
+ "<|im_start|>",
5
+ "<|tool_call|>",
6
+ "<|execute_start|>",
7
+ "<|execute_end|>",
8
+ "<|fim_prefix|>",
9
+ "<|fim_middle|>",
10
+ "<|fim_suffix|>"
11
+ ],
12
+ "bos_token": {
13
+ "content": "<s>",
14
+ "lstrip": false,
15
+ "normalized": false,
16
+ "rstrip": false,
17
+ "single_word": false
18
+ },
19
+ "eos_token": {
20
+ "content": "<|im_end|>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false
25
+ },
26
+ "unk_token": {
27
+ "content": "<unk>",
28
+ "lstrip": false,
29
+ "normalized": false,
30
+ "rstrip": false,
31
+ "single_word": false
32
+ }
33
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb74d51116831c3bf65db812c553f94ab0c88dcf97a5bbb37e3504f6d359c530
3
+ size 1181204
tokenizer_config.json ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": false,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ },
30
+ "73440": {
31
+ "content": "<|im_end|>",
32
+ "lstrip": false,
33
+ "normalized": false,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": true
37
+ },
38
+ "73441": {
39
+ "content": "<|im_start|>",
40
+ "lstrip": false,
41
+ "normalized": false,
42
+ "rstrip": false,
43
+ "single_word": false,
44
+ "special": true
45
+ },
46
+ "73442": {
47
+ "content": "<|tool_call|>",
48
+ "lstrip": false,
49
+ "normalized": false,
50
+ "rstrip": false,
51
+ "single_word": false,
52
+ "special": true
53
+ },
54
+ "73443": {
55
+ "content": "<|execute_start|>",
56
+ "lstrip": false,
57
+ "normalized": false,
58
+ "rstrip": false,
59
+ "single_word": false,
60
+ "special": true
61
+ },
62
+ "73444": {
63
+ "content": "<|execute_end|>",
64
+ "lstrip": false,
65
+ "normalized": false,
66
+ "rstrip": false,
67
+ "single_word": false,
68
+ "special": true
69
+ },
70
+ "73445": {
71
+ "content": "<|fim_prefix|>",
72
+ "lstrip": false,
73
+ "normalized": false,
74
+ "rstrip": false,
75
+ "single_word": false,
76
+ "special": true
77
+ },
78
+ "73446": {
79
+ "content": "<|fim_middle|>",
80
+ "lstrip": false,
81
+ "normalized": false,
82
+ "rstrip": false,
83
+ "single_word": false,
84
+ "special": true
85
+ },
86
+ "73447": {
87
+ "content": "<|fim_suffix|>",
88
+ "lstrip": false,
89
+ "normalized": false,
90
+ "rstrip": false,
91
+ "single_word": false,
92
+ "special": true
93
+ }
94
+ },
95
+ "additional_special_tokens": [
96
+ "<|im_end|>",
97
+ "<|im_start|>",
98
+ "<|tool_call|>",
99
+ "<|execute_start|>",
100
+ "<|execute_end|>",
101
+ "<|fim_prefix|>",
102
+ "<|fim_middle|>",
103
+ "<|fim_suffix|>"
104
+ ],
105
+ "bos_token": "<s>",
106
+ "chat_template": "{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
107
+ "clean_up_tokenization_spaces": false,
108
+ "eos_token": "<|im_end|>",
109
+ "legacy": true,
110
+ "model_max_length": 1000000000000000019884624838656,
111
+ "pad_token": null,
112
+ "sp_model_kwargs": {},
113
+ "spaces_between_special_tokens": false,
114
+ "tokenizer_class": "LlamaTokenizer",
115
+ "unk_token": "<unk>",
116
+ "use_default_system_prompt": false
117
+ }