diff --git a/LOGO.jpg b/LOGO.jpg new file mode 100644 index 0000000..3964446 Binary files /dev/null and b/LOGO.jpg differ diff --git a/MODEL_LICENSE.md b/MODEL_LICENSE.md new file mode 100644 index 0000000..ae993e7 --- /dev/null +++ b/MODEL_LICENSE.md @@ -0,0 +1,48 @@ +# CodeFuse COMMUNITY LICENSE AGREEMENT +CodeFuse Release Date: September 8, 2023 + +By clicking to agree or by using or distributing any portion or element of the Materials, you will be deemed to have recognized and accepted the content of this Agreement, which is effective immediately. + +1. Definitions. + a. This CodeFuse COMMUNITY LICENSE AGREEMENT (this "Agreement") shall mean the terms and conditions for use, reproduction, distribution and modification of the Materials as defined by this Agreement. + b. "Ant" or "We" (or "Us") shall mean Ant Group. + c. "CodeFuse" shall mean the large language models (including CodeFuse-13B and CodeFuse-CodeLlaMa-34B), and software and algorithms, consisting of trained model weights, parameters (including optimizer states), machine-learning model code, and other elements of the foregoing distributed by Us. + d. "Documentation" shall mean the specifications, manuals and documentation accompanying CodeFuse distributed by Us. + e. "Materials" shall mean, collectively, Ant's proprietary CodeFuse and Documentation (and any portion thereof) made available under this Agreement. + f. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation, and conversions to other media types. + g. "Source" form shall mean the preferred form for making modifications, including but not limited to model source code, documentation source, and configuration files. + h. "Third Parties" (or "Third Party") shall mean individuals or legal entities that are not controlling, controlled by Us or You, or under common control with Us or You. + i. "You" (or "Your") shall mean a natural person or legal entity exercising the rights granted by this Agreement and/or using the Materials for any purpose and in any field of use. + +2. Grant of Rights. +You are granted a non-exclusive, worldwide, non-transferable and royalty-free limited license under Ant's intellectual property or other rights owned by Ant embodied in the Materials to use, reproduce, distribute, copy, create derivative works of, and make modifications to the Materials. + +3. Redistribution. +You may distribute or make the Materials or derivative works thereof available to a Third Party in any medium, with or without modifications, and in Source or Object form, provided that You meet the following conditions: + a. You shall provide a copy of this Agreement to such Third Party; + b. if You modify the CodeFuse model, You shall provide a prominent notice, stating how You have modified the CodeFuse model, to such Third Party; and + c. You shall retain in all copies of the Materials that You distribute the following attribution notices within a "Notice" text file distributed as a part of such copies: "CodeFuse is licensed under the CodeFuse COMMUNITY LICENSE AGREEMENT, Copyright (c) Ant Group. All Rights Reserved." +You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such derivative works as a whole, provided Your use, reproduction, and distribution of the work otherwise complies with the terms and conditions of this Agreement. + +4. Rules of Use. + You shall comply with applicable laws and regulations (including without limitation export controls or restrictions) in Your use of the Materials. + +5. Intellectual Property. + a. Ant retains ownership of all intellectual property rights in and to the Materials and derivatives made by or for Ant. Conditioned upon compliance with the terms and conditions of this Agreement, with respect to any derivative works and modifications of the Materials that are made by You, You are and will be the owner of such derivative works and modifications. + b. No trademark license is granted to use the trade names, trademarks, service marks, or product names of Ant, except as required to fulfill notice requirements under this Agreement or as required for reasonable and customary use in describing and redistributing the Materials. + c. If You commence a lawsuit or other proceedings (including a cross-claim or counterclaim in a lawsuit) against Ant or any entity alleging that the Materials or any output therefrom, or any part of the foregoing, infringe any intellectual property or other right owned or licensable by You, then all licences granted to You under this Agreement shall terminate as of the date such lawsuit or other proceeding is commenced or brought. + +6. Disclaimer of Warranty and Limitation of Liability. + a. Ant is not obligated to support, update, provide training for, or develop any further version of the Materials or to grant any license thereto. + b. THE MATERIALS ARE PROVIDED "AS IS" WITHOUT ANY EXPRESS OR IMPLIED WARRANTY OF ANY KIND INCLUDING WARRANTIES OF TITLE, MERCHANTABILITY, NONINFRINGEMENT, OR FITNESS FOR A PARTICULAR PURPOSE. WE MAKE NO WARRANTY AND ASSUME NO RESPONSIBILITY FOR THE SAFETY OR STABILITY OF THE MATERIALS AND ANY OUTPUT THEREFROM. + c. YOU ARE SOLELY RESPONSIBLE FOR DETERMINING THE APPROPRIATENESS OF USING OR REDISTRIBUTING THE MATERIALS AND ASSUME ANY RISKS ASSOCIATED WITH YOUR USE OF THE MATERIALS AND ANY OUTPUT AND RESULTS. IN NO EVENT SHALL WE BE LIABLE TO YOU FOR ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, TORT, NEGLIGENCE, PRODUCTS LIABILITY, OR OTHERWISE, ARISING OUT OF THIS AGREEMENT OR ARISING FROM YOUR USE OR INABILITY TO USE THE MATERIALS OR ANY OUTPUT OF IT, FOR ANY DIRECT, OR INDIRECT, SPECIAL, CONSEQUENTIAL, INCIDENTAL, EXEMPLARY OR PUNITIVE DAMAGES, NO MATTER HOW IT'S CAUSED OR EVEN IF ANT OR ITS AFFILIATES HAVE BEEN ADVISED OF THE POSSIBILITY OF ANY OF THE FOREGOING. + d. You will defend, indemnify and hold harmless Ant from and against any claim by any Third Party arising out of or related to Your use or distribution of the Materials. + +7. Survival and Termination. + a. The term of this Agreement shall commence upon Your acceptance of this Agreement or access to the Materials and will continue in full force and effect until terminated in accordance with the terms and conditions herein. + b. We may terminate this Agreement if You breach any of the terms or conditions of this Agreement. Upon termination of this Agreement, You must delete and cease use of the Materials. Sections 6 and 8 shall survive the termination of this Agreement. + +8. Governing Law and Jurisdiction. + a. This Agreement and any dispute arising out of or relating to it, whether in contract, tort, negligence, products liability, or otherwise, will be governed by the laws of China, without regard to conflict of law principles, and the UN Convention on Contracts for the International Sale of Goods does not apply to this Agreement. + b. The People's Courts in Hangzhou City shall have exclusive jurisdiction over any dispute arising out of this Agreement. + diff --git a/README.md b/README.md index 1c606c5..a30da01 100644 --- a/README.md +++ b/README.md @@ -1,3 +1,316 @@ -# CodeFuse-CodeGeeX2-6B_a13683088223236096543380 +--- +frameworks: +- Pytorch +license: other +tasks: +- text-generation +--- +# Model Card for CodeFuse-CodeGeeX2-6B +

+ +

-CodeFuse-CodeGeeX2-6B \ No newline at end of file +[[中文]](#chinese) [[English]](#english) + +#### Clone with HTTP +```bash + git clone https://www.modelscope.cn/codefuse-ai/CodeFuse-CodeGeeX2-6B.git +``` + + + +## Model Description + +CodeFuse-CodeGeeX2-6B is a 6B Code-LLM finetuned by LoRA of multiple code tasks on the base model CodeGeeX2. + +
+ +## News and Updates + + 🔥🔥 2023-11-10 CodeFuse-CodeGeeX2-6B has been released, achieving a pass@1 (greedy decoding) score of 45.12% on HumanEval, which is a 9.22% increase compared to CodeGeeX2 35.9%. + + 🔥🔥 2023-10-20 CodeFuse-QWen-14B technical documentation has been released. For those interested, please refer to the CodeFuse article on our WeChat official account via the provided link.(https://mp.weixin.qq.com/s/PCQPkvbvfxSPzsqjOILCDw) + + 🔥🔥 2023-10-16 CodeFuse-QWen-14B has been released, achieving a pass@1 (greedy decoding) score of 48.78% on HumanEval, which is a 16% increase compared to Qwen-14b's 32.3%. + + 🔥🔥 2023-09-27 CodeFuse-StarCoder-15B has been released, achieving a pass@1 (greedy decoding) score of 54.9% on HumanEval, which is a 21% increase compared to StarCoder's 33.6%. + +🔥🔥🔥 2023-09-26 We are pleased to announce the release of the [4-bit quantized version](https://modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B-4bits/summary) of [CodeFuse-CodeLlama-34B](https://modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B/summary). Despite the quantization process, the model still achieves a remarkable 73.8% accuracy (greedy decoding) on the HumanEval pass@1 metric. + +🔥🔥🔥 2023-09-11 [CodeFuse-CodeLlama34B](https://modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B/summary) has achived 74.4% of pass@1 (greedy decoding) on HumanEval, which is SOTA results for openspurced LLMs at present. + +
+ +## Code Community + +**Homepage**: 🏡 https://github.com/codefuse-ai (**Please give us your support with a Star🌟 + Fork🚀 + Watch👀**) + ++ If you wish to fine-tune the model yourself, you can visit ✨[MFTCoder](https://github.com/codefuse-ai/MFTCoder)✨✨ + ++ If you wish to deploy the model yourself, you can visit ✨[FasterTransformer4CodeFuse](https://github.com/codefuse-ai/FasterTransformer4CodeFuse)✨✨ + ++ If you wish to see a demo of the model, you can visit ✨[CodeFuse Demo](https://github.com/codefuse-ai/codefuse)✨✨ + +
+ +## Performance + + +| Model | HumanEval(pass@1) | Date | +|:----------------------------|:-----------------:|:-------:| +| **CodeFuse-CodeLlama-34B** | **74.4%** | 2023.9 | +|**CodeFuse-CodeLlama-34B-4bits** | **73.8%** | 2023.9 | +| WizardCoder-Python-34B-V1.0 | 73.2% | 2023.8 | +| GPT-4(zero-shot) | 67.0% | 2023.3 | +| PanGu-Coder2 15B | 61.6% | 2023.8 | +| CodeLlama-34b-Python | 53.7% | 2023.8 | +| CodeLlama-34b | 48.8% | 2023.8 | +| GPT-3.5(zero-shot) | 48.1% | 2022.11 | +| OctoCoder | 46.2% | 2023.8 | +| StarCoder-15B | 33.6% | 2023.5 | +| Qwen-14b | 32.3% | 2023.10 | +| **CodeFuse-StarCoder-15B** | **54.9%** | 2023.9 | +| **CodeFuse-QWen-14B** | **48.78%** | 2023.10 | +| **CodeFuse-CodeGeeX2-6B** | **45.12%** | 2023.11 | + + +
+ +## Requirements + +* python>=3.8 +* pytorch>=2.0.0 +* transformers==4.33.2 +* Sentencepiece +* CUDA 11.4 +
+ +## Inference String Format + +The inference string is a concatenated string formed by combining conversation data(system, human and bot contents) in the training data format. It is used as input during the inference process. +Here is an example format of the concatenated string: + +```python +""" +system +System instruction +human +Human 1st round input +bot +Bot 1st round output<|endoftext|> +human +Human 2nd round input +bot +Bot 2nd round output<|endoftext|> +... +... +... +human +Human nth round input +bot +{Bot output to be genreated}<|endoftext|> +""" +``` + +When applying inference, you always make your input string end with "\bot" to ask the model generating answers. + + +## Quickstart + + +```bash +pip install transformers modelscope cpm_kernels -U +pip install -r requirements.txt +``` + +```python +import torch +from modelscope import ( + AutoTokenizer, + AutoModelForCausalLM, + snapshot_download +) +model_dir = snapshot_download('codefuse-ai/CodeFuse-CodeGeeX2-6B',revision = 'v1.0.0') +tokenizer = AutoTokenizer.from_pretrained(model_dir, trust_remote_code=True) +tokenizer.padding_side = "left" +tokenizer.pad_token_id = tokenizer.convert_tokens_to_ids("") +tokenizer.eos_token_id = tokenizer.convert_tokens_to_ids("") +tokenizer.pad_token = "" +tokenizer.eos_token = "" +# try 4bit loading if cuda memory not enough +model = AutoModelForCausalLM.from_pretrained(model_dir, + trust_remote_code=True, + load_in_4bit=False, + device_map="auto", + torch_dtype=torch.bfloat16) +model.eval() + +HUMAN_ROLE_START_TAG = "human\n" +BOT_ROLE_START_TAG = "bot\n" + +text = f"{HUMAN_ROLE_START_TAG}write a python function of quick sort.\n{BOT_ROLE_START_TAG}" +inputs = tokenizer(text, return_tensors='pt', padding=True, add_special_tokens=False).to("cuda") +outputs = model.generate( + inputs=inputs["input_ids"], + attention_mask=inputs["attention_mask"], + max_new_tokens=512, + top_p=0.95, + temperature=0.1, + do_sample=True, + eos_token_id=tokenizer.eos_token_id, + pad_token_id=tokenizer.pad_token_id + ) +gen_text = tokenizer.batch_decode(outputs[:, inputs["input_ids"].shape[1]:], skip_special_tokens=True) +print(gen_text) +``` + + + + + + + + + + +## 模型简介 + +CodeFuse-CodeGeeX2-6B 是一个通过LoRA对基座模型CodeGeeeX2进行多代码任务微调的代码大模型。 +
+ +## 新闻 + + 🔥🔥 2023-11-10 开源了CodeFuse-CodeGeeX2-6B模型，在HumanEval pass@1(greedy decoding)上可以达到48.12%, 比CodeGeeX2提高了9.22%的代码能力（HumanEval） + + 🔥🔥 2023-10-20 公布了CodeFuse-QWen-14B技术文档，感兴趣详见微信公众号CodeFuse文章：https://mp.weixin.qq.com/s/PCQPkvbvfxSPzsqjOILCDw + + 🔥🔥 2023-10-16开源了CodeFuse-QWen-14B模型，在HumanEval pass@1(greedy decoding)上可以达到48.78%, 比Qwen-14b提高了16%的代码能力（HumanEval） + + 🔥🔥 2023-09-27开源了CodeFuse-StarCoder-15B模型，在HumanEval pass@1(greedy decoding)上可以达到54.9%, 比StarCoder提高了21%的代码能力（HumanEval） + +🔥🔥🔥 2023-09-26 [CodeFuse-CodeLlama-34B 4bits](https://modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B-4bits/summary)量化版本发布，量化后模型在HumanEval pass@1指标为73.8% (贪婪解码)。 + +🔥🔥🔥 2023-09-11 [CodeFuse-CodeLlama-34B](https://modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B/summary)发布，HumanEval pass@1指标达到74.4% (贪婪解码), 为当前开源SOTA。 + +
+ +## 代码社区 +**大本营**： 🏡 https://github.com/codefuse-ai （**请支持我们的项目Star🌟 + Fork🚀 + Watch👀**） + ++ 如果您想自己微调该模型，可以访问 ✨[MFTCoder](https://github.com/codefuse-ai/MFTCoder)✨✨ + ++ 如果您想自己部署该模型，可以访问 ✨[FasterTransformer4CodeFuse](https://github.com/codefuse-ai/FasterTransformer4CodeFuse)✨✨ + ++ 如果您想观看该模型示例，可以访问 ✨[CodeFuse Demo](https://github.com/codefuse-ai/codefuse)✨✨ + +
+ + +## 评测表现 + +### 代码 + + +| 模型 | HumanEval(pass@1) | 日期 | +|:----------------------------|:-----------------:|:-------:| +| **CodeFuse-CodeLlama-34B** | **74.4%** | 2023.9 | +|**CodeFuse-CodeLlama-34B-4bits** | **73.8%** | 2023.9 | +| WizardCoder-Python-34B-V1.0 | 73.2% | 2023.8 | +| GPT-4(zero-shot) | 67.0% | 2023.3 | +| PanGu-Coder2 15B | 61.6% | 2023.8 | +| CodeLlama-34b-Python | 53.7% | 2023.8 | +| CodeLlama-34b | 48.8% | 2023.8 | +| GPT-3.5(zero-shot) | 48.1% | 2022.11 | +| OctoCoder | 46.2% | 2023.8 | +| StarCoder-15B | 33.6% | 2023.5 | +| Qwen-14b | 32.3% | 2023.10 | +| **CodeFuse-StarCoder-15B** | **54.9%** | 2023.9 | +| **CodeFuse-QWen-14B** | **48.78%** | 2023.8 | +| **CodeFuse-CodeGeeX2-6B** | **45.12%** | 2023.11 | + + +## Requirements + +* python>=3.8 +* pytorch>=2.0.0 +* transformers==4.33.2 +* Sentencepiece +* CUDA 11.4 +
+ +## 推理数据格式 + +推理数据为模型在训练数据格式下拼接的字符串形式，它也是推理时输入prompt拼接的方式： + +```python +""" +system +这是System指令 +human +这是第1轮用户输入的问题 +bot +这是第1轮模型生成的内容<|endoftext|> +human +这是第2轮用户输入的问题 +bot +这是第2轮模型生成的内容<|endoftext|> +... +... +... +human +这是第n轮用户输入的问题 +bot +{模型现在要生成的内容}<|endoftext|> +""" +``` + +推理时，请确保拼接的prompt字符串以"\bot\n"结尾，引导模型生成回答。 + +## 快速使用 + + +```bash +pip install transformers modelscope cpm_kernels -U +pip install -r requirements.txt +``` + +```python +import torch +from modelscope import ( + AutoTokenizer, + AutoModelForCausalLM, + snapshot_download +) +model_dir = snapshot_download('codefuse-ai/CodeFuse-CodeGeeX2-6B',revision = 'v1.0.0') +tokenizer = AutoTokenizer.from_pretrained(model_dir, trust_remote_code=True) +tokenizer.padding_side = "left" +tokenizer.pad_token_id = tokenizer.convert_tokens_to_ids("") +tokenizer.eos_token_id = tokenizer.convert_tokens_to_ids("") +tokenizer.pad_token = "" +tokenizer.eos_token = "" +# try 4bit loading if cuda memory not enough +model = AutoModelForCausalLM.from_pretrained(model_dir, + trust_remote_code=True, + load_in_4bit=False, + device_map="auto", + torch_dtype=torch.bfloat16) +model.eval() + +HUMAN_ROLE_START_TAG = "human\n" +BOT_ROLE_START_TAG = "bot\n" + +text = f"{HUMAN_ROLE_START_TAG}write a python function of quick sort.\n{BOT_ROLE_START_TAG}" +inputs = tokenizer(text, return_tensors='pt', padding=True, add_special_tokens=False).to("cuda") +outputs = model.generate( + inputs=inputs["input_ids"], + attention_mask=inputs["attention_mask"], + max_new_tokens=512, + top_p=0.95, + temperature=0.1, + do_sample=True, + eos_token_id=tokenizer.eos_token_id, + pad_token_id=tokenizer.pad_token_id + ) +gen_text = tokenizer.batch_decode(outputs[:, inputs["input_ids"].shape[1]:], skip_special_tokens=True) +print(gen_text) +``` diff --git a/config.json b/config.json new file mode 100644 index 0000000..2cef5bf --- /dev/null +++ b/config.json @@ -0,0 +1,49 @@ +{ + "_name_or_path": "/mnt/user/qumu/download_models/codegeex2-6b", + "add_bias_linear": false, + "add_qkv_bias": true, + "apply_query_key_layer_scaling": true, + "apply_residual_connection_post_layernorm": false, + "architectures": [ + "ChatGLMForConditionalGeneration" + ], + "attention_dropout": 0.0, + "attention_softmax_in_fp32": true, + "auto_map": { + "AutoConfig": "configuration_chatglm.ChatGLMConfig", + "AutoModel": "modeling_chatglm.ChatGLMForConditionalGeneration", + "AutoModelForCausalLM": "modeling_chatglm.ChatGLMForConditionalGeneration", + "AutoModelForSeq2SeqLM": "modeling_chatglm.ChatGLMForConditionalGeneration" + }, + "bias_dropout_fusion": true, + "eos_token": "", + "eos_token_id": 2, + "ffn_hidden_size": 13696, + "fp32_residual_connection": false, + "hidden_dropout": 0.0, + "hidden_size": 4096, + "interleaved_qkv": false, + "kv_channels": 128, + "layernorm_epsilon": 1e-05, + "model_type": "chatglm", + "multi_query_attention": true, + "multi_query_group_num": 2, + "num_attention_heads": 32, + "num_layers": 28, + "original_rope": true, + "pad_token": "", + "pad_token_id": 0, + "padded_vocab_size": 65024, + "post_layer_norm": true, + "pre_seq_len": null, + "prefix_projection": false, + "quantization_bit": 0, + "rmsnorm": true, + "rotary_percent": 0.5, + "seq_length": 8192, + "tie_word_embeddings": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.33.2", + "use_cache": true, + "vocab_size": 65024 +} diff --git a/configuration.json b/configuration.json new file mode 100644 index 0000000..f9291c3 --- /dev/null +++ b/configuration.json @@ -0,0 +1 @@ +{"framework":"Pytorch","task":"text-generation"} \ No newline at end of file diff --git a/configuration_chatglm.py b/configuration_chatglm.py new file mode 100644 index 0000000..3730b4e --- /dev/null +++ b/configuration_chatglm.py @@ -0,0 +1,59 @@ +from transformers import PretrainedConfig + + +class ChatGLMConfig(PretrainedConfig): + model_type = "chatglm" + def __init__( + self, + num_layers=28, + padded_vocab_size=65024, + hidden_size=4096, + ffn_hidden_size=13696, + kv_channels=128, + num_attention_heads=32, + seq_length=2048, + hidden_dropout=0.0, + attention_dropout=0.0, + layernorm_epsilon=1e-5, + rmsnorm=True, + apply_residual_connection_post_layernorm=False, + post_layer_norm=True, + add_bias_linear=False, + add_qkv_bias=False, + bias_dropout_fusion=True, + multi_query_attention=False, + multi_query_group_num=1, + apply_query_key_layer_scaling=True, + attention_softmax_in_fp32=True, + fp32_residual_connection=False, + quantization_bit=0, + pre_seq_len=None, + prefix_projection=False, + **kwargs + ): + self.num_layers = num_layers + self.vocab_size = padded_vocab_size + self.padded_vocab_size = padded_vocab_size + self.hidden_size = hidden_size + self.ffn_hidden_size = ffn_hidden_size + self.kv_channels = kv_channels + self.num_attention_heads = num_attention_heads + self.seq_length = seq_length + self.hidden_dropout = hidden_dropout + self.attention_dropout = attention_dropout + self.layernorm_epsilon = layernorm_epsilon + self.rmsnorm = rmsnorm + self.apply_residual_connection_post_layernorm = apply_residual_connection_post_layernorm + self.post_layer_norm = post_layer_norm + self.add_bias_linear = add_bias_linear + self.add_qkv_bias = add_qkv_bias + self.bias_dropout_fusion = bias_dropout_fusion + self.multi_query_attention = multi_query_attention + self.multi_query_group_num = multi_query_group_num + self.apply_query_key_layer_scaling = apply_query_key_layer_scaling + self.attention_softmax_in_fp32 = attention_softmax_in_fp32 + self.fp32_residual_connection = fp32_residual_connection + self.quantization_bit = quantization_bit + self.pre_seq_len = pre_seq_len + self.prefix_projection = prefix_projection + super().__init__(**kwargs) \ No newline at end of file diff --git a/generation_config.json b/generation_config.json new file mode 100644 index 0000000..3264de0 --- /dev/null +++ b/generation_config.json @@ -0,0 +1,5 @@ +{ + "_from_model_config": true, + "eos_token_id": 2, + "transformers_version": "4.33.2" +} diff --git a/modeling_chatglm.py b/modeling_chatglm.py new file mode 100644 index 0000000..bfa6a55 --- /dev/null +++ b/modeling_chatglm.py @@ -0,0 +1,1197 @@ +""" PyTorch ChatGLM model. """ + +import math +import copy +import warnings +import sys +import torch +import torch.utils.checkpoint +import torch.nn.functional as F +from torch import nn +from torch.nn import CrossEntropyLoss, LayerNorm +from torch.nn.utils import skip_init +from typing import Optional, Tuple, Union, List, Callable, Dict, Any + +from transformers.modeling_outputs import ( + BaseModelOutputWithPast, + CausalLMOutputWithPast, +) + +from transformers.modeling_utils import PreTrainedModel +from transformers.utils import logging +from transformers.generation.logits_process import LogitsProcessor +from transformers.generation.utils import LogitsProcessorList, StoppingCriteriaList, GenerationConfig, ModelOutput + +from .configuration_chatglm import ChatGLMConfig + +# flags required to enable jit fusion kernels + +if sys.platform != 'darwin': + torch._C._jit_set_profiling_mode(False) + torch._C._jit_set_profiling_executor(False) + torch._C._jit_override_can_fuse_on_cpu(True) + torch._C._jit_override_can_fuse_on_gpu(True) + +logger = logging.get_logger(__name__) + +_CHECKPOINT_FOR_DOC = "THUDM/ChatGLM2-6B" +_CONFIG_FOR_DOC = "ChatGLM6BConfig" + +CHATGLM_6B_PRETRAINED_MODEL_ARCHIVE_LIST = [ + "THUDM/chatglm2-6b", + # See all ChatGLM models at https://huggingface.co/models?filter=chatglm +] + + +def default_init(cls, *args, **kwargs): + return cls(*args, **kwargs) + + +class InvalidScoreLogitsProcessor(LogitsProcessor): + def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor: + if torch.isnan(scores).any() or torch.isinf(scores).any(): + scores.zero_() + scores[..., 5] = 5e4 + return scores + + +class PrefixEncoder(torch.nn.Module): + """ + The torch.nn model to encode the prefix + Input shape: (batch-size, prefix-length) + Output shape: (batch-size, prefix-length, 2*layers*hidden) + """ + + def __init__(self, config: ChatGLMConfig): + super().__init__() + self.prefix_projection = config.prefix_projection + if self.prefix_projection: + # Use a two-layer MLP to encode the prefix + kv_size = config.num_layers * config.kv_channels * config.multi_query_group_num * 2 + self.embedding = torch.nn.Embedding(config.pre_seq_len, kv_size) + self.trans = torch.nn.Sequential( + torch.nn.Linear(kv_size, config.hidden_size), + torch.nn.Tanh(), + torch.nn.Linear(config.hidden_size, kv_size) + ) + else: + self.embedding = torch.nn.Embedding(config.pre_seq_len, + config.num_layers * config.kv_channels * config.multi_query_group_num * 2) + + def forward(self, prefix: torch.Tensor): + if self.prefix_projection: + prefix_tokens = self.embedding(prefix) + past_key_values = self.trans(prefix_tokens) + else: + past_key_values = self.embedding(prefix) + return past_key_values + + +def split_tensor_along_last_dim( + tensor: torch.Tensor, + num_partitions: int, + contiguous_split_chunks: bool = False, +) -> List[torch.Tensor]: + """Split a tensor along its last dimension. + + Arguments: + tensor: input tensor. + num_partitions: number of partitions to split the tensor + contiguous_split_chunks: If True, make each chunk contiguous + in memory. + + Returns: + A list of Tensors + """ + # Get the size and dimension. + last_dim = tensor.dim() - 1 + last_dim_size = tensor.size()[last_dim] // num_partitions + # Split. + tensor_list = torch.split(tensor, last_dim_size, dim=last_dim) + # Note: torch.split does not create contiguous tensors by default. + if contiguous_split_chunks: + return tuple(chunk.contiguous() for chunk in tensor_list) + + return tensor_list + + +class RotaryEmbedding(nn.Module): + def __init__(self, dim, original_impl=False, device=None, dtype=None): + super().__init__() + inv_freq = 1.0 / (10000 ** (torch.arange(0, dim, 2, device=device).to(dtype=dtype) / dim)) + self.register_buffer("inv_freq", inv_freq) + self.dim = dim + self.original_impl = original_impl + + def forward_impl( + self, seq_len: int, n_elem: int, dtype: torch.dtype, device: torch.device, base: int = 10000 + ): + """Enhanced Transformer with Rotary Position Embedding. + + Derived from: https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/ + transformers/rope/__init__.py. MIT License: + https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/license. + """ + # $\Theta = {\theta_i = 10000^{\frac{2(i-1)}{d}}, i \in [1, 2, ..., \frac{d}{2}]}$ + theta = 1.0 / (base ** (torch.arange(0, n_elem, 2, dtype=dtype, device=device) / n_elem)) + + # Create position indexes `[0, 1, ..., seq_len - 1]` + seq_idx = torch.arange(seq_len, dtype=dtype, device=device) + + # Calculate the product of position index and $\theta_i$ + idx_theta = torch.outer(seq_idx, theta).float() + + cache = torch.stack([torch.cos(idx_theta), torch.sin(idx_theta)], dim=-1) + + # this is to mimic the behaviour of complex32, else we will get different results + if dtype in (torch.float16, torch.bfloat16, torch.int8): + cache = cache.bfloat16() if dtype == torch.bfloat16 else cache.half() + return cache + + def forward(self, max_seq_len, offset=0): + return self.forward_impl( + max_seq_len, self.dim, dtype=self.inv_freq.dtype, device=self.inv_freq.device + ) + + +@torch.jit.script +def apply_rotary_pos_emb(x: torch.Tensor, rope_cache: torch.Tensor) -> torch.Tensor: + # x: [sq, b, np, hn] + sq, b, np, hn = x.size(0), x.size(1), x.size(2), x.size(3) + rot_dim = rope_cache.shape[-2] * 2 + x, x_pass = x[..., :rot_dim], x[..., rot_dim:] + # truncate to support variable sizes + rope_cache = rope_cache[:sq] + xshaped = x.reshape(sq, -1, np, rot_dim // 2, 2) + rope_cache = rope_cache.view(sq, -1, 1, xshaped.size(3), 2) + x_out2 = torch.stack( + [ + xshaped[..., 0] * rope_cache[..., 0] - xshaped[..., 1] * rope_cache[..., 1], + xshaped[..., 1] * rope_cache[..., 0] + xshaped[..., 0] * rope_cache[..., 1], + ], + -1, + ) + x_out2 = x_out2.flatten(3) + return torch.cat((x_out2, x_pass), dim=-1) + + +class RMSNorm(torch.nn.Module): + def __init__(self, normalized_shape, eps=1e-5, device=None, dtype=None, **kwargs): + super().__init__() + self.weight = torch.nn.Parameter(torch.empty(normalized_shape, device=device, dtype=dtype)) + self.eps = eps + + def forward(self, hidden_states: torch.Tensor): + if hidden_states.dtype == torch.bfloat16: + norm_x = torch.mean(hidden_states * hidden_states, dim=-1, keepdim=True) + x_normed = hidden_states * torch.rsqrt(norm_x + self.eps) + return self.weight * x_normed + else: + input_dtype = hidden_states.dtype + variance = hidden_states.to(torch.float32).pow(2).mean(-1, keepdim=True) + hidden_states = hidden_states * torch.rsqrt(variance + self.eps) + + return (self.weight * hidden_states).to(input_dtype) + + +class CoreAttention(torch.nn.Module): + def __init__(self, config: ChatGLMConfig, layer_number): + super(CoreAttention, self).__init__() + + self.apply_query_key_layer_scaling = config.apply_query_key_layer_scaling + self.attention_softmax_in_fp32 = config.attention_softmax_in_fp32 + if self.apply_query_key_layer_scaling: + self.attention_softmax_in_fp32 = True + self.layer_number = max(1, layer_number) + + projection_size = config.kv_channels * config.num_attention_heads + + # Per attention head and per partition values. + self.hidden_size_per_partition = projection_size + self.hidden_size_per_attention_head = projection_size // config.num_attention_heads + self.num_attention_heads_per_partition = config.num_attention_heads + + coeff = None + self.norm_factor = math.sqrt(self.hidden_size_per_attention_head) + if self.apply_query_key_layer_scaling: + coeff = self.layer_number + self.norm_factor *= coeff + self.coeff = coeff + + self.attention_dropout = torch.nn.Dropout(config.attention_dropout) + + def forward(self, query_layer, key_layer, value_layer, attention_mask): + pytorch_major_version = int(torch.__version__.split('.')[0]) + if pytorch_major_version >= 2: + query_layer, key_layer, value_layer = [k.permute(1, 2, 0, 3) for k in [query_layer, key_layer, value_layer]] + if attention_mask is None and query_layer.shape[2] == key_layer.shape[2]: + context_layer = torch.nn.functional.scaled_dot_product_attention(query_layer, key_layer, value_layer, + is_causal=True) + else: + if attention_mask is not None: + attention_mask = ~attention_mask + context_layer = torch.nn.functional.scaled_dot_product_attention(query_layer, key_layer, value_layer, + attention_mask) + context_layer = context_layer.permute(2, 0, 1, 3) + new_context_layer_shape = context_layer.size()[:-2] + (self.hidden_size_per_partition,) + context_layer = context_layer.reshape(*new_context_layer_shape) + else: + # Raw attention scores + + # [b, np, sq, sk] + output_size = (query_layer.size(1), query_layer.size(2), query_layer.size(0), key_layer.size(0)) + + # [sq, b, np, hn] -> [sq, b * np, hn] + query_layer = query_layer.view(output_size[2], output_size[0] * output_size[1], -1) + # [sk, b, np, hn] -> [sk, b * np, hn] + key_layer = key_layer.view(output_size[3], output_size[0] * output_size[1], -1) + + # preallocting input tensor: [b * np, sq, sk] + matmul_input_buffer = torch.empty( + output_size[0] * output_size[1], output_size[2], output_size[3], dtype=query_layer.dtype, + device=query_layer.device + ) + + # Raw attention scores. [b * np, sq, sk] + matmul_result = torch.baddbmm( + matmul_input_buffer, + query_layer.transpose(0, 1), # [b * np, sq, hn] + key_layer.transpose(0, 1).transpose(1, 2), # [b * np, hn, sk] + beta=0.0, + alpha=(1.0 / self.norm_factor), + ) + + # change view to [b, np, sq, sk] + attention_scores = matmul_result.view(*output_size) + + # =========================== + # Attention probs and dropout + # =========================== + + # attention scores and attention mask [b, np, sq, sk] + if self.attention_softmax_in_fp32: + attention_scores = attention_scores.float() + if self.coeff is not None: + attention_scores = attention_scores * self.coeff + if attention_mask is None and attention_scores.shape[2] == attention_scores.shape[3]: + attention_mask = torch.ones(output_size[0], 1, output_size[2], output_size[3], + device=attention_scores.device, dtype=torch.bool) + attention_mask.tril_() + attention_mask = ~attention_mask + if attention_mask is not None: + attention_scores = attention_scores.masked_fill(attention_mask, float("-inf")) + attention_probs = F.softmax(attention_scores, dim=-1) + attention_probs = attention_probs.type_as(value_layer) + + # This is actually dropping out entire tokens to attend to, which might + # seem a bit unusual, but is taken from the original Transformer paper. + attention_probs = self.attention_dropout(attention_probs) + # ========================= + # Context layer. [sq, b, hp] + # ========================= + + # value_layer -> context layer. + # [sk, b, np, hn] --> [b, np, sq, hn] + + # context layer shape: [b, np, sq, hn] + output_size = (value_layer.size(1), value_layer.size(2), query_layer.size(0), value_layer.size(3)) + # change view [sk, b * np, hn] + value_layer = value_layer.view(value_layer.size(0), output_size[0] * output_size[1], -1) + # change view [b * np, sq, sk] + attention_probs = attention_probs.view(output_size[0] * output_size[1], output_size[2], -1) + # matmul: [b * np, sq, hn] + context_layer = torch.bmm(attention_probs, value_layer.transpose(0, 1)) + # change view [b, np, sq, hn] + context_layer = context_layer.view(*output_size) + # [b, np, sq, hn] --> [sq, b, np, hn] + context_layer = context_layer.permute(2, 0, 1, 3).contiguous() + # [sq, b, np, hn] --> [sq, b, hp] + new_context_layer_shape = context_layer.size()[:-2] + (self.hidden_size_per_partition,) + context_layer = context_layer.view(*new_context_layer_shape) + + return context_layer + + +class SelfAttention(torch.nn.Module): + """Parallel self-attention layer abstract class. + + Self-attention layer takes input with size [s, b, h] + and returns output of the same size. + """ + + def __init__(self, config: ChatGLMConfig, layer_number, device=None): + super(SelfAttention, self).__init__() + self.layer_number = max(1, layer_number) + + self.projection_size = config.kv_channels * config.num_attention_heads + + # Per attention head and per partition values. + self.hidden_size_per_attention_head = self.projection_size // config.num_attention_heads + self.num_attention_heads_per_partition = config.num_attention_heads + + self.multi_query_attention = config.multi_query_attention + self.qkv_hidden_size = 3 * self.projection_size + if self.multi_query_attention: + self.num_multi_query_groups_per_partition = config.multi_query_group_num + self.qkv_hidden_size = ( + self.projection_size + 2 * self.hidden_size_per_attention_head * config.multi_query_group_num + ) + self.query_key_value = nn.Linear(config.hidden_size, self.qkv_hidden_size, + bias=config.add_bias_linear or config.add_qkv_bias, + device=device, **_config_to_kwargs(config) + ) + + self.core_attention = CoreAttention(config, self.layer_number) + + # Output. + self.dense = nn.Linear(self.projection_size, config.hidden_size, bias=config.add_bias_linear, + device=device, **_config_to_kwargs(config) + ) + + def _allocate_memory(self, inference_max_sequence_len, batch_size, device=None, dtype=None): + if self.multi_query_attention: + num_attention_heads = self.num_multi_query_groups_per_partition + else: + num_attention_heads = self.num_attention_heads_per_partition + return torch.empty( + inference_max_sequence_len, + batch_size, + num_attention_heads, + self.hidden_size_per_attention_head, + dtype=dtype, + device=device, + ) + + def forward( + self, hidden_states, attention_mask, rotary_pos_emb, kv_cache=None, use_cache=True + ): + # hidden_states: [sq, b, h] + + # ================================================= + # Pre-allocate memory for key-values for inference. + # ================================================= + # ===================== + # Query, Key, and Value + # ===================== + + # Attention heads [sq, b, h] --> [sq, b, (np * 3 * hn)] + mixed_x_layer = self.query_key_value(hidden_states) + + if self.multi_query_attention: + (query_layer, key_layer, value_layer) = mixed_x_layer.split( + [ + self.num_attention_heads_per_partition * self.hidden_size_per_attention_head, + self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head, + self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head, + ], + dim=-1, + ) + query_layer = query_layer.view( + query_layer.size()[:-1] + (self.num_attention_heads_per_partition, self.hidden_size_per_attention_head) + ) + key_layer = key_layer.view( + key_layer.size()[:-1] + (self.num_multi_query_groups_per_partition, self.hidden_size_per_attention_head) + ) + value_layer = value_layer.view( + value_layer.size()[:-1] + + (self.num_multi_query_groups_per_partition, self.hidden_size_per_attention_head) + ) + else: + new_tensor_shape = mixed_x_layer.size()[:-1] + \ + (self.num_attention_heads_per_partition, + 3 * self.hidden_size_per_attention_head) + mixed_x_layer = mixed_x_layer.view(*new_tensor_shape) + + # [sq, b, np, 3 * hn] --> 3 [sq, b, np, hn] + (query_layer, key_layer, value_layer) = split_tensor_along_last_dim(mixed_x_layer, 3) + + # apply relative positional encoding (rotary embedding) + if rotary_pos_emb is not None: + query_layer = apply_rotary_pos_emb(query_layer, rotary_pos_emb) + key_layer = apply_rotary_pos_emb(key_layer, rotary_pos_emb) + + # adjust key and value for inference + if kv_cache is not None: + cache_k, cache_v = kv_cache + key_layer = torch.cat((cache_k, key_layer), dim=0) + value_layer = torch.cat((cache_v, value_layer), dim=0) + if use_cache: + kv_cache = (key_layer, value_layer) + else: + kv_cache = None + + if self.multi_query_attention: + key_layer = key_layer.unsqueeze(-2) + key_layer = key_layer.expand( + -1, -1, -1, self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, -1 + ) + key_layer = key_layer.contiguous().view( + key_layer.size()[:2] + (self.num_attention_heads_per_partition, self.hidden_size_per_attention_head) + ) + value_layer = value_layer.unsqueeze(-2) + value_layer = value_layer.expand( + -1, -1, -1, self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, -1 + ) + value_layer = value_layer.contiguous().view( + value_layer.size()[:2] + (self.num_attention_heads_per_partition, self.hidden_size_per_attention_head) + ) + + # ================================== + # core attention computation + # ================================== + + context_layer = self.core_attention(query_layer, key_layer, value_layer, attention_mask) + + # ================= + # Output. [sq, b, h] + # ================= + + output = self.dense(context_layer) + + return output, kv_cache + + +def _config_to_kwargs(args): + common_kwargs = { + "dtype": args.torch_dtype, + } + return common_kwargs + + +class MLP(torch.nn.Module): + """MLP. + + MLP will take the input with h hidden state, project it to 4*h + hidden dimension, perform nonlinear transformation, and project the + state back into h hidden dimension. + """ + + def __init__(self, config: ChatGLMConfig, device=None): + super(MLP, self).__init__() + + self.add_bias = config.add_bias_linear + + # Project to 4h. If using swiglu double the output width, see https://arxiv.org/pdf/2002.05202.pdf + self.dense_h_to_4h = nn.Linear( + config.hidden_size, + config.ffn_hidden_size * 2, + bias=self.add_bias, + device=device, + **_config_to_kwargs(config) + ) + + def swiglu(x): + x = torch.chunk(x, 2, dim=-1) + return F.silu(x[0]) * x[1] + + self.activation_func = swiglu + + # Project back to h. + self.dense_4h_to_h = nn.Linear( + config.ffn_hidden_size, + config.hidden_size, + bias=self.add_bias, + device=device, + **_config_to_kwargs(config) + ) + + def forward(self, hidden_states): + # [s, b, 4hp] + intermediate_parallel = self.dense_h_to_4h(hidden_states) + intermediate_parallel = self.activation_func(intermediate_parallel) + # [s, b, h] + output = self.dense_4h_to_h(intermediate_parallel) + return output + + +class GLMBlock(torch.nn.Module): + """A single transformer layer. + + Transformer layer takes input with size [s, b, h] and returns an + output of the same size. + """ + + def __init__(self, config: ChatGLMConfig, layer_number, device=None): + super(GLMBlock, self).__init__() + self.layer_number = layer_number + + self.apply_residual_connection_post_layernorm = config.apply_residual_connection_post_layernorm + + self.fp32_residual_connection = config.fp32_residual_connection + + LayerNormFunc = RMSNorm if config.rmsnorm else LayerNorm + # Layernorm on the input data. + self.input_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device, + dtype=config.torch_dtype) + + # Self attention. + self.self_attention = SelfAttention(config, layer_number, device=device) + self.hidden_dropout = config.hidden_dropout + + # Layernorm on the attention output + self.post_attention_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device, + dtype=config.torch_dtype) + + # MLP + self.mlp = MLP(config, device=device) + + def forward( + self, hidden_states, attention_mask, rotary_pos_emb, kv_cache=None, use_cache=True, + ): + # hidden_states: [s, b, h] + + # Layer norm at the beginning of the transformer layer. + layernorm_output = self.input_layernorm(hidden_states) + # Self attention. + attention_output, kv_cache = self.self_attention( + layernorm_output, + attention_mask, + rotary_pos_emb, + kv_cache=kv_cache, + use_cache=use_cache + ) + + # Residual connection. + if self.apply_residual_connection_post_layernorm: + residual = layernorm_output + else: + residual = hidden_states + + layernorm_input = torch.nn.functional.dropout(attention_output, p=self.hidden_dropout, training=self.training) + layernorm_input = residual + layernorm_input + + # Layer norm post the self attention. + layernorm_output = self.post_attention_layernorm(layernorm_input) + + # MLP. + mlp_output = self.mlp(layernorm_output) + + # Second residual connection. + if self.apply_residual_connection_post_layernorm: + residual = layernorm_output + else: + residual = layernorm_input + + output = torch.nn.functional.dropout(mlp_output, p=self.hidden_dropout, training=self.training) + output = residual + output + + return output, kv_cache + + +class GLMTransformer(torch.nn.Module): + """Transformer class.""" + + def __init__(self, config: ChatGLMConfig, device=None): + super(GLMTransformer, self).__init__() + + self.fp32_residual_connection = config.fp32_residual_connection + self.post_layer_norm = config.post_layer_norm + + # Number of layers. + self.num_layers = config.num_layers + + # Transformer layers. + def build_layer(layer_number): + return GLMBlock(config, layer_number, device=device) + + self.layers = torch.nn.ModuleList([build_layer(i + 1) for i in range(self.num_layers)]) + + if self.post_layer_norm: + LayerNormFunc = RMSNorm if config.rmsnorm else LayerNorm + # Final layer norm before output. + self.final_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device, + dtype=config.torch_dtype) + + self.gradient_checkpointing = False + + def _get_layer(self, layer_number): + return self.layers[layer_number] + + def forward( + self, hidden_states, attention_mask, rotary_pos_emb, kv_caches=None, + use_cache: Optional[bool] = True, + output_hidden_states: Optional[bool] = False, + ): + if not kv_caches: + kv_caches = [None for _ in range(self.num_layers)] + presents = () if use_cache else None + if self.gradient_checkpointing and self.training: + if use_cache: + logger.warning_once( + "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..." + ) + use_cache = False + + all_self_attentions = None + all_hidden_states = () if output_hidden_states else None + for index in range(self.num_layers): + if output_hidden_states: + all_hidden_states = all_hidden_states + (hidden_states,) + + layer = self._get_layer(index) + if self.gradient_checkpointing and self.training: + layer_ret = torch.utils.checkpoint.checkpoint( + layer, + hidden_states, + attention_mask, + rotary_pos_emb, + kv_caches[index], + use_cache + ) + else: + layer_ret = layer( + hidden_states, + attention_mask, + rotary_pos_emb, + kv_cache=kv_caches[index], + use_cache=use_cache + ) + hidden_states, kv_cache = layer_ret + if use_cache: + presents = presents + (kv_cache,) + + if output_hidden_states: + all_hidden_states = all_hidden_states + (hidden_states,) + + # Final layer norm. + if self.post_layer_norm: + hidden_states = self.final_layernorm(hidden_states) + + return hidden_states, presents, all_hidden_states, all_self_attentions + + +class ChatGLMPreTrainedModel(PreTrainedModel): + """ + An abstract class to handle weights initialization and + a simple interface for downloading and loading pretrained models. + """ + + is_parallelizable = False + supports_gradient_checkpointing = True + config_class = ChatGLMConfig + base_model_prefix = "transformer" + _no_split_modules = ["GLMBlock"] + + def _init_weights(self, module: nn.Module): + """Initialize the weights.""" + return + + def get_masks(self, input_ids, past_key_values, padding_mask=None): + batch_size, seq_length = input_ids.shape + full_attention_mask = torch.ones(batch_size, seq_length, seq_length, device=input_ids.device) + full_attention_mask.tril_() + past_length = 0 + if past_key_values: + past_length = past_key_values[0][0].shape[0] + if past_length: + full_attention_mask = torch.cat((torch.ones(batch_size, seq_length, past_length, + device=input_ids.device), full_attention_mask), dim=-1) + if padding_mask is not None: + full_attention_mask = full_attention_mask * padding_mask.unsqueeze(1) + if not past_length and padding_mask is not None: + full_attention_mask -= padding_mask.unsqueeze(-1) - 1 + full_attention_mask = (full_attention_mask < 0.5).bool() + full_attention_mask.unsqueeze_(1) + return full_attention_mask + + def get_position_ids(self, input_ids, device): + batch_size, seq_length = input_ids.shape + position_ids = torch.arange(seq_length, dtype=torch.long, device=device).unsqueeze(0).repeat(batch_size, 1) + return position_ids + + def _set_gradient_checkpointing(self, module, value=False): + if isinstance(module, GLMTransformer): + module.gradient_checkpointing = value + + +class Embedding(torch.nn.Module): + """Language model embeddings.""" + + def __init__(self, config: ChatGLMConfig, device=None): + super(Embedding, self).__init__() + + self.hidden_size = config.hidden_size + # Word embeddings (parallel). + self.word_embeddings = nn.Embedding( + config.padded_vocab_size, + self.hidden_size, + dtype=config.torch_dtype, + device=device + ) + self.fp32_residual_connection = config.fp32_residual_connection + + def forward(self, input_ids): + # Embeddings. + words_embeddings = self.word_embeddings(input_ids) + embeddings = words_embeddings + # Data format change to avoid explicit tranposes : [b s h] --> [s b h]. + embeddings = embeddings.transpose(0, 1).contiguous() + # If the input flag for fp32 residual connection is set, convert for float. + if self.fp32_residual_connection: + embeddings = embeddings.float() + return embeddings + + +class ChatGLMModel(ChatGLMPreTrainedModel): + def __init__(self, config: ChatGLMConfig, device=None, empty_init=True): + super().__init__(config) + if empty_init: + init_method = skip_init + else: + init_method = default_init + init_kwargs = {} + if device is not None: + init_kwargs["device"] = device + self.embedding = init_method(Embedding, config, **init_kwargs) + self.num_layers = config.num_layers + self.multi_query_group_num = config.multi_query_group_num + self.kv_channels = config.kv_channels + + # Rotary positional embeddings + self.seq_length = config.seq_length + rotary_dim = ( + config.hidden_size // config.num_attention_heads if config.kv_channels is None else config.kv_channels + ) + + self.rotary_pos_emb = RotaryEmbedding(rotary_dim // 2, original_impl=config.original_rope, device=device, + dtype=config.torch_dtype) + self.encoder = init_method(GLMTransformer, config, **init_kwargs) + self.output_layer = init_method(nn.Linear, config.hidden_size, config.padded_vocab_size, bias=False, + dtype=config.torch_dtype, **init_kwargs) + self.pre_seq_len = config.pre_seq_len + self.prefix_projection = config.prefix_projection + if self.pre_seq_len is not None: + for param in self.parameters(): + param.requires_grad = False + self.prefix_tokens = torch.arange(self.pre_seq_len).long() + self.prefix_encoder = PrefixEncoder(config) + self.dropout = torch.nn.Dropout(0.1) + + def get_input_embeddings(self): + return self.embedding.word_embeddings + + def get_prompt(self, batch_size, device, dtype=torch.half): + prefix_tokens = self.prefix_tokens.unsqueeze(0).expand(batch_size, -1).to(device) + past_key_values = self.prefix_encoder(prefix_tokens).type(dtype) + past_key_values = past_key_values.view( + batch_size, + self.pre_seq_len, + self.num_layers * 2, + self.multi_query_group_num, + self.kv_channels + ) + # seq_len, b, nh, hidden_size + past_key_values = self.dropout(past_key_values) + past_key_values = past_key_values.permute([2, 1, 0, 3, 4]).split(2) + return past_key_values + + def forward( + self, + input_ids, + position_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.BoolTensor] = None, + full_attention_mask: Optional[torch.BoolTensor] = None, + past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None, + inputs_embeds: Optional[torch.Tensor] = None, + use_cache: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ): + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + use_cache = use_cache if use_cache is not None else self.config.use_cache + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + batch_size, seq_length = input_ids.shape + + if inputs_embeds is None: + inputs_embeds = self.embedding(input_ids) + + if self.pre_seq_len is not None: + if past_key_values is None: + past_key_values = self.get_prompt(batch_size=batch_size, device=input_ids.device, + dtype=inputs_embeds.dtype) + if attention_mask is not None: + attention_mask = torch.cat([attention_mask.new_ones((batch_size, self.pre_seq_len)), + attention_mask], dim=-1) + + if full_attention_mask is None: + if (attention_mask is not None and not attention_mask.all()) or (past_key_values and seq_length != 1): + full_attention_mask = self.get_masks(input_ids, past_key_values, padding_mask=attention_mask) + + # Rotary positional embeddings + rotary_pos_emb = self.rotary_pos_emb(self.seq_length) + if position_ids is not None: + rotary_pos_emb = rotary_pos_emb[position_ids] + else: + rotary_pos_emb = rotary_pos_emb[None, :seq_length] + rotary_pos_emb = rotary_pos_emb.transpose(0, 1).contiguous() + + # Run encoder. + hidden_states, presents, all_hidden_states, all_self_attentions = self.encoder( + inputs_embeds, full_attention_mask, rotary_pos_emb=rotary_pos_emb, + kv_caches=past_key_values, use_cache=use_cache, output_hidden_states=output_hidden_states + ) + + if not return_dict: + return tuple(v for v in [hidden_states, presents, all_hidden_states, all_self_attentions] if v is not None) + + return BaseModelOutputWithPast( + last_hidden_state=hidden_states, + past_key_values=presents, + hidden_states=all_hidden_states, + attentions=all_self_attentions, + ) + + def quantize(self, weight_bit_width: int): + from .quantization import quantize + quantize(self.encoder, weight_bit_width) + return self + + +class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel): + def __init__(self, config: ChatGLMConfig, empty_init=True, device=None): + super().__init__(config) + + self.max_sequence_length = config.max_length + self.transformer = ChatGLMModel(config, empty_init=empty_init, device=device) + self.config = config + self.quantized = False + + if self.config.quantization_bit: + self.quantize(self.config.quantization_bit, empty_init=True) + + def _update_model_kwargs_for_generation( + self, + outputs: ModelOutput, + model_kwargs: Dict[str, Any], + is_encoder_decoder: bool = False, + standardize_cache_format: bool = False, + ) -> Dict[str, Any]: + # update past_key_values + model_kwargs["past_key_values"] = self._extract_past_from_model_output( + outputs, standardize_cache_format=standardize_cache_format + ) + + # update attention mask + if "attention_mask" in model_kwargs: + attention_mask = model_kwargs["attention_mask"] + model_kwargs["attention_mask"] = torch.cat( + [attention_mask, attention_mask.new_ones((attention_mask.shape[0], 1))], dim=-1 + ) + + # update position ids + if "position_ids" in model_kwargs: + position_ids = model_kwargs["position_ids"] + new_position_id = position_ids[..., -1:].clone() + new_position_id += 1 + model_kwargs["position_ids"] = torch.cat( + [position_ids, new_position_id], dim=-1 + ) + + model_kwargs["is_first_forward"] = False + return model_kwargs + + def prepare_inputs_for_generation( + self, + input_ids: torch.LongTensor, + past_key_values: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.Tensor] = None, + is_first_forward: bool = True, + **kwargs + ) -> dict: + # only last token for input_ids if past is not None + if position_ids is None: + position_ids = self.get_position_ids(input_ids, device=input_ids.device) + if not is_first_forward: + position_ids = position_ids[..., -1:] + input_ids = input_ids[:, -1:] + return { + "input_ids": input_ids, + "past_key_values": past_key_values, + "position_ids": position_ids, + "attention_mask": attention_mask, + "return_last_logit": True + } + + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + position_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + past_key_values: Optional[Tuple[torch.FloatTensor]] = None, + inputs_embeds: Optional[torch.Tensor] = None, + labels: Optional[torch.Tensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + return_last_logit: Optional[bool] = False, + ): + use_cache = use_cache if use_cache is not None else self.config.use_cache + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + transformer_outputs = self.transformer( + input_ids=input_ids, + position_ids=position_ids, + attention_mask=attention_mask, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + hidden_states = transformer_outputs[0] + if return_last_logit: + hidden_states = hidden_states[-1:] + lm_logits = self.transformer.output_layer(hidden_states) + lm_logits = lm_logits.transpose(0, 1).contiguous() + + loss = None + if labels is not None: + lm_logits = lm_logits.to(torch.float32) + + # Shift so that tokens < n predict n + shift_logits = lm_logits[..., :-1, :].contiguous() + shift_labels = labels[..., 1:].contiguous() + # Flatten the tokens + loss_fct = CrossEntropyLoss(ignore_index=-100) + loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1)) + + lm_logits = lm_logits.to(hidden_states.dtype) + loss = loss.to(hidden_states.dtype) + + if not return_dict: + output = (lm_logits,) + transformer_outputs[1:] + return ((loss,) + output) if loss is not None else output + + return CausalLMOutputWithPast( + loss=loss, + logits=lm_logits, + past_key_values=transformer_outputs.past_key_values, + hidden_states=transformer_outputs.hidden_states, + attentions=transformer_outputs.attentions, + ) + + @staticmethod + def _reorder_cache( + past: Tuple[Tuple[torch.Tensor, torch.Tensor], ...], beam_idx: torch.LongTensor + ) -> Tuple[Tuple[torch.Tensor, torch.Tensor], ...]: + """ + This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or + [`~PreTrainedModel.beam_sample`] is called. This is required to match `past_key_values` with the correct + beam_idx at every generation step. + + Output shares the same memory storage as `past`. + """ + return tuple( + ( + layer_past[0].index_select(1, beam_idx.to(layer_past[0].device)), + layer_past[1].index_select(1, beam_idx.to(layer_past[1].device)), + ) + for layer_past in past + ) + + def process_response(self, response): + response = response.strip() + response = response.replace("[[训练时间]]", "2023年") + return response + + def build_inputs(self, tokenizer, query: str, history: List[Tuple[str, str]] = None): + prompt = tokenizer.build_prompt(query, history=history) + inputs = tokenizer([prompt], return_tensors="pt") + inputs = inputs.to(self.device) + return inputs + + def build_stream_inputs(self, tokenizer, query: str, history: List[Tuple[str, str]] = None): + if history: + prompt = "\n\n[Round {}]\n\n问：{}\n\n答：".format(len(history) + 1, query) + input_ids = tokenizer.encode(prompt, add_special_tokens=False) + input_ids = input_ids[1:] + inputs = tokenizer.batch_encode_plus([(input_ids, None)], return_tensors="pt", add_special_tokens=False) + else: + prompt = "[Round {}]\n\n问：{}\n\n答：".format(len(history) + 1, query) + inputs = tokenizer([prompt], return_tensors="pt") + inputs = inputs.to(self.device) + return inputs + + @torch.inference_mode() + def chat(self, tokenizer, query: str, history: List[Tuple[str, str]] = None, max_length: int = 8192, num_beams=1, + do_sample=True, top_p=0.8, temperature=0.8, logits_processor=None, **kwargs): + if history is None: + history = [] + if logits_processor is None: + logits_processor = LogitsProcessorList() + logits_processor.append(InvalidScoreLogitsProcessor()) + gen_kwargs = {"max_length": max_length, "num_beams": num_beams, "do_sample": do_sample, "top_p": top_p, + "temperature": temperature, "logits_processor": logits_processor, **kwargs} + inputs = self.build_inputs(tokenizer, query, history=history) + outputs = self.generate(**inputs, **gen_kwargs) + outputs = outputs.tolist()[0][len(inputs["input_ids"][0]):] + response = tokenizer.decode(outputs) + response = self.process_response(response) + history = history + [(query, response)] + return response, history + + @torch.inference_mode() + def stream_chat(self, tokenizer, query: str, history: List[Tuple[str, str]] = None, past_key_values=None, + max_length: int = 8192, do_sample=True, top_p=0.8, temperature=0.8, logits_processor=None, + return_past_key_values=False, **kwargs): + if history is None: + history = [] + if logits_processor is None: + logits_processor = LogitsProcessorList() + logits_processor.append(InvalidScoreLogitsProcessor()) + gen_kwargs = {"max_length": max_length, "do_sample": do_sample, "top_p": top_p, + "temperature": temperature, "logits_processor": logits_processor, **kwargs} + if past_key_values is None and not return_past_key_values: + inputs = self.build_inputs(tokenizer, query, history=history) + else: + inputs = self.build_stream_inputs(tokenizer, query, history=history) + if past_key_values is not None: + past_length = past_key_values[0][0].shape[0] + if self.transformer.pre_seq_len is not None: + past_length -= self.transformer.pre_seq_len + inputs.position_ids += past_length + attention_mask = inputs.attention_mask + attention_mask = torch.cat((attention_mask.new_ones(1, past_length), attention_mask), dim=1) + inputs['attention_mask'] = attention_mask + for outputs in self.stream_generate(**inputs, past_key_values=past_key_values, + return_past_key_values=return_past_key_values, **gen_kwargs): + if return_past_key_values: + outputs, past_key_values = outputs + outputs = outputs.tolist()[0][len(inputs["input_ids"][0]):] + response = tokenizer.decode(outputs) + if response and response[-1] != "�": + response = self.process_response(response) + new_history = history + [(query, response)] + if return_past_key_values: + yield response, new_history, past_key_values + else: + yield response, new_history + + @torch.inference_mode() + def stream_generate( + self, + input_ids, + generation_config: Optional[GenerationConfig] = None, + logits_processor: Optional[LogitsProcessorList] = None, + stopping_criteria: Optional[StoppingCriteriaList] = None, + prefix_allowed_tokens_fn: Optional[Callable[[int, torch.Tensor], List[int]]] = None, + return_past_key_values=False, + **kwargs, + ): + batch_size, input_ids_seq_length = input_ids.shape[0], input_ids.shape[-1] + + if generation_config is None: + generation_config = self.generation_config + generation_config = copy.deepcopy(generation_config) + model_kwargs = generation_config.update(**kwargs) + bos_token_id, eos_token_id = generation_config.bos_token_id, generation_config.eos_token_id + + if isinstance(eos_token_id, int): + eos_token_id = [eos_token_id] + + has_default_max_length = kwargs.get("max_length") is None and generation_config.max_length is not None + if has_default_max_length and generation_config.max_new_tokens is None: + warnings.warn( + f"Using `max_length`'s default ({generation_config.max_length}) to control the generation length. " + "This behaviour is deprecated and will be removed from the config in v5 of Transformers -- we" + " recommend using `max_new_tokens` to control the maximum length of the generation.", + UserWarning, + ) + elif generation_config.max_new_tokens is not None: + generation_config.max_length = generation_config.max_new_tokens + input_ids_seq_length + if not has_default_max_length: + logger.warn( + f"Both `max_new_tokens` (={generation_config.max_new_tokens}) and `max_length`(=" + f"{generation_config.max_length}) seem to have been set. `max_new_tokens` will take precedence. " + "Please refer to the documentation for more information. " + "(https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)", + UserWarning, + ) + + if input_ids_seq_length >= generation_config.max_length: + input_ids_string = "decoder_input_ids" if self.config.is_encoder_decoder else "input_ids" + logger.warning( + f"Input length of {input_ids_string} is {input_ids_seq_length}, but `max_length` is set to" + f" {generation_config.max_length}. This can lead to unexpected behavior. You should consider" + " increasing `max_new_tokens`." + ) + + # 2. Set generation parameters if not already defined + logits_processor = logits_processor if logits_processor is not None else LogitsProcessorList() + stopping_criteria = stopping_criteria if stopping_criteria is not None else StoppingCriteriaList() + + logits_processor = self._get_logits_processor( + generation_config=generation_config, + input_ids_seq_length=input_ids_seq_length, + encoder_input_ids=input_ids, + prefix_allowed_tokens_fn=prefix_allowed_tokens_fn, + logits_processor=logits_processor, + ) + + stopping_criteria = self._get_stopping_criteria( + generation_config=generation_config, stopping_criteria=stopping_criteria + ) + logits_warper = self._get_logits_warper(generation_config) + + unfinished_sequences = torch.ones(input_ids.shape[0], device=input_ids.device, dtype=input_ids.dtype) + scores = None + while True: + model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs) + # forward pass to get next token + outputs = self( + **model_inputs, + return_dict=True, + output_attentions=False, + output_hidden_states=False, + ) + + next_token_logits = outputs.logits[:, -1, :] + + # pre-process distribution + next_token_scores = logits_processor(input_ids, next_token_logits) + next_token_scores = logits_warper(input_ids, next_token_scores) + + # sample + probs = nn.functional.softmax(next_token_scores, dim=-1) + if generation_config.do_sample: + next_tokens = torch.multinomial(probs, num_samples=1).squeeze(1) + else: + next_tokens = torch.argmax(probs, dim=-1) + + # update generated ids, model inputs, and length for next step + input_ids = torch.cat([input_ids, next_tokens[:, None]], dim=-1) + model_kwargs = self._update_model_kwargs_for_generation( + outputs, model_kwargs, is_encoder_decoder=self.config.is_encoder_decoder + ) + unfinished_sequences = unfinished_sequences.mul((sum(next_tokens != i for i in eos_token_id)).long()) + if return_past_key_values: + yield input_ids, outputs.past_key_values + else: + yield input_ids + # stop when each sentence is finished, or if we exceed the maximum length + if unfinished_sequences.max() == 0 or stopping_criteria(input_ids, scores): + break + + def quantize(self, bits: int, empty_init=False, device=None, **kwargs): + if bits == 0: + return + + from .quantization import quantize + + if self.quantized: + logger.info("Already quantized.") + return self + + self.quantized = True + + self.config.quantization_bit = bits + + self.transformer.encoder = quantize(self.transformer.encoder, bits, empty_init=empty_init, device=device, + **kwargs) + return self diff --git a/pytorch_model-00001-of-00002.bin b/pytorch_model-00001-of-00002.bin new file mode 100644 index 0000000..cf060ab --- /dev/null +++ b/pytorch_model-00001-of-00002.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:445eee34a6866c25accc42b2438a4f55d9b13fbe37ac514ddaf73c6b93af2df0 +size 9986266947 diff --git a/pytorch_model-00002-of-00002.bin b/pytorch_model-00002-of-00002.bin new file mode 100644 index 0000000..ff4f939 --- /dev/null +++ b/pytorch_model-00002-of-00002.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d68be1e18f6edc3d9be31977569c407d7c5c15757f49a27e642d83be02d11762 +size 2500976015 diff --git a/pytorch_model.bin.index.json b/pytorch_model.bin.index.json new file mode 100644 index 0000000..ab25cbc --- /dev/null +++ b/pytorch_model.bin.index.json @@ -0,0 +1,207 @@ +{ + "metadata": { + "total_size": 12487168064 + }, + "weight_map": { + "transformer.embedding.word_embeddings.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.final_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.0.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.0.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.0.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.0.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.0.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.1.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.1.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.1.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.1.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.1.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.10.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.10.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.10.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.10.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.10.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.10.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.10.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.11.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.11.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.11.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.11.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.11.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.11.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.11.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.12.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.12.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.12.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.12.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.12.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.12.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.12.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.13.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.13.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.13.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.13.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.13.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.13.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.13.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.14.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.14.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.14.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.14.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.14.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.14.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.14.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.15.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.15.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.15.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.15.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.15.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.15.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.15.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.16.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.16.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.16.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.16.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.16.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.16.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.16.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.17.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.17.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.17.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.17.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.17.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.17.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.17.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.18.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.18.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.18.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.18.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.18.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.18.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.18.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.19.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.19.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.19.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.19.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.19.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.19.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.19.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.2.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.2.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.2.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.2.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.2.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.20.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.20.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.20.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.20.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.20.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.20.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.20.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.21.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.21.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.21.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.21.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.21.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.21.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.21.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.22.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.22.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.22.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.22.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.22.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.22.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.22.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.23.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.23.mlp.dense_4h_to_h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.23.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.23.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.23.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.23.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.23.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.24.input_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.24.mlp.dense_4h_to_h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.24.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.24.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.24.self_attention.dense.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.24.self_attention.query_key_value.bias": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.24.self_attention.query_key_value.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.25.input_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.25.mlp.dense_4h_to_h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.25.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.25.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.25.self_attention.dense.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.25.self_attention.query_key_value.bias": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.25.self_attention.query_key_value.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.26.input_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.26.mlp.dense_4h_to_h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.26.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.26.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.26.self_attention.dense.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.26.self_attention.query_key_value.bias": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.26.self_attention.query_key_value.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.27.input_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.27.mlp.dense_4h_to_h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.27.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.27.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.27.self_attention.dense.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.27.self_attention.query_key_value.bias": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.27.self_attention.query_key_value.weight": "pytorch_model-00002-of-00002.bin", + "transformer.encoder.layers.3.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.3.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.3.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.3.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.3.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.3.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.3.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.4.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.4.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.4.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.4.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.4.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.4.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.4.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.5.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.5.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.5.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.5.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.5.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.5.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.5.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.6.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.6.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.6.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.6.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.6.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.6.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.6.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.7.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.7.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.7.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.7.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.7.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.7.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.7.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.8.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.8.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.8.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.8.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.8.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.8.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.8.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.9.input_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.9.mlp.dense_4h_to_h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.9.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.9.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.9.self_attention.dense.weight": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.9.self_attention.query_key_value.bias": "pytorch_model-00001-of-00002.bin", + "transformer.encoder.layers.9.self_attention.query_key_value.weight": "pytorch_model-00001-of-00002.bin", + "transformer.output_layer.weight": "pytorch_model-00002-of-00002.bin", + "transformer.rotary_pos_emb.inv_freq": "pytorch_model-00001-of-00002.bin" + } +} diff --git a/quantization.py b/quantization.py new file mode 100644 index 0000000..cb95bfe --- /dev/null +++ b/quantization.py @@ -0,0 +1,188 @@ +from torch.nn import Linear +from torch.nn.parameter import Parameter + +import bz2 +import torch +import base64 +import ctypes +from transformers.utils import logging + +from typing import List +from functools import partial + +logger = logging.get_logger(__name__) + +try: + from cpm_kernels.kernels.base import LazyKernelCModule, KernelFunction, round_up + + class Kernel: + def __init__(self, code: bytes, function_names: List[str]): + self.code = code + self._function_names = function_names + self._cmodule = LazyKernelCModule(self.code) + + for name in self._function_names: + setattr(self, name, KernelFunction(self._cmodule, name)) + + quantization_code = "$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" + + kernels = Kernel( + bz2.decompress(base64.b64decode(quantization_code)), + [ + "int4WeightCompression", + "int4WeightExtractionFloat", + "int4WeightExtractionHalf", + "int8WeightExtractionFloat", + "int8WeightExtractionHalf", + ], + ) +except Exception as exception: + kernels = None + logger.warning("Failed to load cpm_kernels:" + str(exception)) + + +class W8A16Linear(torch.autograd.Function): + @staticmethod + def forward(ctx, inp: torch.Tensor, quant_w: torch.Tensor, scale_w: torch.Tensor, weight_bit_width): + ctx.inp_shape = inp.size() + ctx.weight_bit_width = weight_bit_width + out_features = quant_w.size(0) + inp = inp.contiguous().view(-1, inp.size(-1)) + weight = extract_weight_to_half(quant_w, scale_w, weight_bit_width) + ctx.weight_shape = weight.size() + output = inp.mm(weight.t()) + ctx.save_for_backward(inp, quant_w, scale_w) + return output.view(*(ctx.inp_shape[:-1] + (out_features,))) + + @staticmethod + def backward(ctx, grad_output: torch.Tensor): + inp, quant_w, scale_w = ctx.saved_tensors + weight = extract_weight_to_half(quant_w, scale_w, ctx.weight_bit_width) + grad_output = grad_output.contiguous().view(-1, weight.size(0)) + grad_input = grad_output.mm(weight) + grad_weight = grad_output.t().mm(inp) + return grad_input.view(ctx.inp_shape), grad_weight.view(ctx.weight_shape), None, None + + +def compress_int4_weight(weight: torch.Tensor): # (n, m) + with torch.cuda.device(weight.device): + n, m = weight.size(0), weight.size(1) + assert m % 2 == 0 + m = m // 2 + out = torch.empty(n, m, dtype=torch.int8, device="cuda") + stream = torch.cuda.current_stream() + + gridDim = (n, 1, 1) + blockDim = (min(round_up(m, 32), 1024), 1, 1) + + kernels.int4WeightCompression( + gridDim, + blockDim, + 0, + stream, + [ctypes.c_void_p(weight.data_ptr()), ctypes.c_void_p(out.data_ptr()), ctypes.c_int32(n), ctypes.c_int32(m)], + ) + return out + + +def extract_weight_to_half(weight: torch.Tensor, scale_list: torch.Tensor, source_bit_width: int): + assert scale_list.dtype in [torch.half, torch.bfloat16] + assert weight.dtype in [torch.int8] + if source_bit_width == 8: + return weight.to(scale_list.dtype) * scale_list[:, None] + elif source_bit_width == 4: + func = ( + kernels.int4WeightExtractionHalf if scale_list.dtype == torch.half else kernels.int4WeightExtractionBFloat16 + ) + else: + assert False, "Unsupported bit-width" + + with torch.cuda.device(weight.device): + n, m = weight.size(0), weight.size(1) + out = torch.empty(n, m * (8 // source_bit_width), dtype=scale_list.dtype, device="cuda") + stream = torch.cuda.current_stream() + + gridDim = (n, 1, 1) + blockDim = (min(round_up(m, 32), 1024), 1, 1) + + func( + gridDim, + blockDim, + 0, + stream, + [ + ctypes.c_void_p(weight.data_ptr()), + ctypes.c_void_p(scale_list.data_ptr()), + ctypes.c_void_p(out.data_ptr()), + ctypes.c_int32(n), + ctypes.c_int32(m), + ], + ) + return out + + +class QuantizedLinear(torch.nn.Module): + def __init__(self, weight_bit_width: int, weight, bias=None, device="cpu", dtype=None, empty_init=False, *args, + **kwargs): + super().__init__() + self.weight_bit_width = weight_bit_width + + shape = weight.shape + + if weight is None or empty_init: + self.weight = torch.empty(shape[0], shape[1] * weight_bit_width // 8, dtype=torch.int8, device=device) + self.weight_scale = torch.empty(shape[0], dtype=dtype, device=device) + else: + self.weight_scale = weight.abs().max(dim=-1).values / ((2 ** (weight_bit_width - 1)) - 1) + self.weight = torch.round(weight / self.weight_scale[:, None]).to(torch.int8) + if weight_bit_width == 4: + self.weight = compress_int4_weight(self.weight) + + self.weight = Parameter(self.weight.to(device), requires_grad=False) + self.weight_scale = Parameter(self.weight_scale.to(device), requires_grad=False) + self.bias = Parameter(bias.to(device), requires_grad=False) if bias is not None else None + + def forward(self, input): + output = W8A16Linear.apply(input, self.weight, self.weight_scale, self.weight_bit_width) + if self.bias is not None: + output = output + self.bias + return output + + +def quantize(model, weight_bit_width, empty_init=False, device=None): + """Replace fp16 linear with quantized linear""" + for layer in model.layers: + layer.self_attention.query_key_value = QuantizedLinear( + weight_bit_width=weight_bit_width, + weight=layer.self_attention.query_key_value.weight.to(torch.cuda.current_device()), + bias=layer.self_attention.query_key_value.bias, + dtype=layer.self_attention.query_key_value.weight.dtype, + device=layer.self_attention.query_key_value.weight.device if device is None else device, + empty_init=empty_init + ) + layer.self_attention.dense = QuantizedLinear( + weight_bit_width=weight_bit_width, + weight=layer.self_attention.dense.weight.to(torch.cuda.current_device()), + bias=layer.self_attention.dense.bias, + dtype=layer.self_attention.dense.weight.dtype, + device=layer.self_attention.dense.weight.device if device is None else device, + empty_init=empty_init + ) + layer.mlp.dense_h_to_4h = QuantizedLinear( + weight_bit_width=weight_bit_width, + weight=layer.mlp.dense_h_to_4h.weight.to(torch.cuda.current_device()), + bias=layer.mlp.dense_h_to_4h.bias, + dtype=layer.mlp.dense_h_to_4h.weight.dtype, + device=layer.mlp.dense_h_to_4h.weight.device if device is None else device, + empty_init=empty_init + ) + layer.mlp.dense_4h_to_h = QuantizedLinear( + weight_bit_width=weight_bit_width, + weight=layer.mlp.dense_4h_to_h.weight.to(torch.cuda.current_device()), + bias=layer.mlp.dense_4h_to_h.bias, + dtype=layer.mlp.dense_4h_to_h.weight.dtype, + device=layer.mlp.dense_4h_to_h.weight.device if device is None else device, + empty_init=empty_init + ) + + return model diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 0000000..8a096d1 --- /dev/null +++ b/requirements.txt @@ -0,0 +1,14 @@ +numpy +pandas +einops +sentencepiece +deepspeed==0.9.3 +transformers==4.33.2 +accelerate==0.21.0 +peft==0.4.0 +BitsAndBytes==0.40.2 +xformers==0.0.21 +ujson +jsonlines +tiktoken +transformers_stream_generator diff --git a/special_tokens_map.json b/special_tokens_map.json new file mode 100644 index 0000000..0967ef4 --- /dev/null +++ b/special_tokens_map.json @@ -0,0 +1 @@ +{} diff --git a/tokenization_chatglm.py b/tokenization_chatglm.py new file mode 100644 index 0000000..6928fe8 --- /dev/null +++ b/tokenization_chatglm.py @@ -0,0 +1,264 @@ +import os +import torch +from typing import List, Optional, Union, Dict +from sentencepiece import SentencePieceProcessor +from transformers import PreTrainedTokenizer +from transformers.utils import logging, PaddingStrategy +from transformers.tokenization_utils_base import EncodedInput, BatchEncoding + + +class SPTokenizer: + def __init__(self, model_path: str): + # reload tokenizer + assert os.path.isfile(model_path), model_path + self.sp_model = SentencePieceProcessor(model_file=model_path) + + # BOS / EOS token IDs + self.n_words: int = self.sp_model.vocab_size() + self.bos_id: int = self.sp_model.bos_id() + self.eos_id: int = self.sp_model.eos_id() + self.pad_id: int = self.sp_model.unk_id() + assert self.sp_model.vocab_size() == self.sp_model.get_piece_size() + + special_tokens = ["[MASK]", "[gMASK]", "[sMASK]", "sop", "eop"] + self.special_tokens = {} + self.index_special_tokens = {} + for token in special_tokens: + self.special_tokens[token] = self.n_words + self.index_special_tokens[self.n_words] = token + self.n_words += 1 + + def tokenize(self, s: str): + return self.sp_model.EncodeAsPieces(s) + + def encode(self, s: str, bos: bool = False, eos: bool = False) -> List[int]: + assert type(s) is str + t = self.sp_model.encode(s) + if bos: + t = [self.bos_id] + t + if eos: + t = t + [self.eos_id] + return t + + def decode(self, t: List[int]) -> str: + return self.sp_model.decode(t) + + def decode_tokens(self, tokens: List[str]) -> str: + text = self.sp_model.DecodePieces(tokens) + return text + + def convert_token_to_id(self, token): + """ Converts a token (str) in an id using the vocab. """ + if token in self.special_tokens: + return self.special_tokens[token] + return self.sp_model.PieceToId(token) + + def convert_id_to_token(self, index): + """Converts an index (integer) in a token (str) using the vocab.""" + if index in self.index_special_tokens or index in [self.eos_id, self.bos_id, self.pad_id] or index < 0: + return "" + return self.sp_model.IdToPiece(index) + + +class ChatGLMTokenizer(PreTrainedTokenizer): + vocab_files_names = {"vocab_file": "tokenizer.model"} + + model_input_names = ["input_ids", "attention_mask", "position_ids"] + + def __init__(self, vocab_file, padding_side="left", clean_up_tokenization_spaces=False, **kwargs): + super().__init__(padding_side=padding_side, clean_up_tokenization_spaces=clean_up_tokenization_spaces, **kwargs) + self.name = "GLMTokenizer" + + self.vocab_file = vocab_file + self.tokenizer = SPTokenizer(vocab_file) + self.special_tokens = { + "": self.tokenizer.bos_id, + "": self.tokenizer.eos_id, + "": self.tokenizer.pad_id + } + + def get_command(self, token): + if token in self.special_tokens: + return self.special_tokens[token] + assert token in self.tokenizer.special_tokens, f"{token} is not a special token for {self.name}" + return self.tokenizer.special_tokens[token] + + @property + def unk_token(self) -> str: + return "" + + @property + def pad_token(self) -> str: + return "" + + @property + def pad_token_id(self): + return self.get_command("") + + @property + def eos_token(self) -> str: + return "" + + @property + def eos_token_id(self): + return self.get_command("") + + @property + def vocab_size(self): + return self.tokenizer.n_words + + def get_vocab(self): + """ Returns vocab as a dict """ + vocab = {self._convert_id_to_token(i): i for i in range(self.vocab_size)} + vocab.update(self.added_tokens_encoder) + return vocab + + def _tokenize(self, text, **kwargs): + return self.tokenizer.tokenize(text) + + def _convert_token_to_id(self, token): + """ Converts a token (str) in an id using the vocab. """ + return self.tokenizer.convert_token_to_id(token) + + def _convert_id_to_token(self, index): + """Converts an index (integer) in a token (str) using the vocab.""" + return self.tokenizer.convert_id_to_token(index) + + def convert_tokens_to_string(self, tokens: List[str]) -> str: + return self.tokenizer.decode_tokens(tokens) + + def save_vocabulary(self, save_directory, filename_prefix=None): + """ + Save the vocabulary and special tokens file to a directory. + + Args: + save_directory (`str`): + The directory in which to save the vocabulary. + filename_prefix (`str`, *optional*): + An optional prefix to add to the named of the saved files. + + Returns: + `Tuple(str)`: Paths to the files saved. + """ + if os.path.isdir(save_directory): + vocab_file = os.path.join( + save_directory, self.vocab_files_names["vocab_file"] + ) + else: + vocab_file = save_directory + + with open(self.vocab_file, 'rb') as fin: + proto_str = fin.read() + + with open(vocab_file, "wb") as writer: + writer.write(proto_str) + + return (vocab_file,) + + def get_prefix_tokens(self): + prefix_tokens = [self.get_command("[gMASK]"), self.get_command("sop")] + return prefix_tokens + + def build_prompt(self, query, history=None): + if history is None: + history = [] + prompt = "" + for i, (old_query, response) in enumerate(history): + prompt += "[Round {}]\n\n问：{}\n\n答：{}\n\n".format(i + 1, old_query, response) + prompt += "[Round {}]\n\n问：{}\n\n答：".format(len(history) + 1, query) + return prompt + + def build_inputs_with_special_tokens( + self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None + ) -> List[int]: + """ + Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and + adding special tokens. A BERT sequence has the following format: + + - single sequence: `[CLS] X [SEP]` + - pair of sequences: `[CLS] A [SEP] B [SEP]` + + Args: + token_ids_0 (`List[int]`): + List of IDs to which the special tokens will be added. + token_ids_1 (`List[int]`, *optional*): + Optional second list of IDs for sequence pairs. + + Returns: + `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens. + """ + prefix_tokens = self.get_prefix_tokens() + token_ids_0 = prefix_tokens + token_ids_0 + if token_ids_1 is not None: + token_ids_0 = token_ids_0 + token_ids_1 + [self.get_command("")] + return token_ids_0 + + def _pad( + self, + encoded_inputs: Union[Dict[str, EncodedInput], BatchEncoding], + max_length: Optional[int] = None, + padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD, + pad_to_multiple_of: Optional[int] = None, + return_attention_mask: Optional[bool] = None, + ) -> dict: + """ + Pad encoded inputs (on left/right and up to predefined length or max length in the batch) + + Args: + encoded_inputs: + Dictionary of tokenized inputs (`List[int]`) or batch of tokenized inputs (`List[List[int]]`). + max_length: maximum length of the returned list and optionally padding length (see below). + Will truncate by taking into account the special tokens. + padding_strategy: PaddingStrategy to use for padding. + + - PaddingStrategy.LONGEST Pad to the longest sequence in the batch + - PaddingStrategy.MAX_LENGTH: Pad to the max length (default) + - PaddingStrategy.DO_NOT_PAD: Do not pad + The tokenizer padding sides are defined in self.padding_side: + + - 'left': pads on the left of the sequences + - 'right': pads on the right of the sequences + pad_to_multiple_of: (optional) Integer if set will pad the sequence to a multiple of the provided value. + This is especially useful to enable the use of Tensor Core on NVIDIA hardware with compute capability + `>= 7.5` (Volta). + return_attention_mask: + (optional) Set to False to avoid returning attention mask (default: set to model specifics) + """ + # Load from model defaults + # assert self.padding_side == "left" + + required_input = encoded_inputs[self.model_input_names[0]] + seq_length = len(required_input) + + if padding_strategy == PaddingStrategy.LONGEST: + max_length = len(required_input) + + if max_length is not None and pad_to_multiple_of is not None and (max_length % pad_to_multiple_of != 0): + max_length = ((max_length // pad_to_multiple_of) + 1) * pad_to_multiple_of + + needs_to_be_padded = padding_strategy != PaddingStrategy.DO_NOT_PAD and len(required_input) != max_length + + # Initialize attention mask if not present. + if "attention_mask" not in encoded_inputs: + encoded_inputs["attention_mask"] = [1] * seq_length + + if "position_ids" not in encoded_inputs: + encoded_inputs["position_ids"] = list(range(seq_length)) + + if needs_to_be_padded: + difference = max_length - len(required_input) + + if self.padding_side == "left": + if "attention_mask" in encoded_inputs: + encoded_inputs["attention_mask"] = [0] * difference + encoded_inputs["attention_mask"] + if "position_ids" in encoded_inputs: + encoded_inputs["position_ids"] = [0] * difference + encoded_inputs["position_ids"] + encoded_inputs[self.model_input_names[0]] = [self.pad_token_id] * difference + required_input + else: + if "attention_mask" in encoded_inputs: + encoded_inputs["attention_mask"] = encoded_inputs["attention_mask"] + [0] * difference + if "position_ids" in encoded_inputs: + encoded_inputs["position_ids"] = encoded_inputs["position_ids"] + [0] * difference + encoded_inputs[self.model_input_names[0]] = required_input + [self.pad_token_id] * difference + + return encoded_inputs diff --git a/tokenizer.model b/tokenizer.model new file mode 100644 index 0000000..8fa03ec --- /dev/null +++ b/tokenizer.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57d9fdbdfaa7cd8c0a3a38d7e8de2e6c31374b5dbc4dc4568d85585fe745812f +size 1018370 diff --git a/tokenizer_config.json b/tokenizer_config.json new file mode 100644 index 0000000..7fa8f74 --- /dev/null +++ b/tokenizer_config.json @@ -0,0 +1,15 @@ +{ + "auto_map": { + "AutoTokenizer": [ + "tokenization_chatglm.ChatGLMTokenizer", + null + ] + }, + "clean_up_tokenization_spaces": false, + "do_lower_case": false, + "legacy": false, + "model_max_length": 1000000000000000019884624838656, + "padding_side": "left", + "remove_space": false, + "tokenizer_class": "ChatGLMTokenizer" +}