yi-6b-chat-32k

<div align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/01-ai/Yi/main/assets/img/Yi_logo_icon_dark.svg" width="200px"> <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/01-ai/Yi/main/assets/img/Yi_logo_icon_light.svg" width="200px"> <img alt="specify theme context for images" src="https://raw.githubusercontent.com/01-ai/Yi/main/assets/img/Yi_logo_icon_light.svg"> </picture> </br> </br> <div style="display: inline-block;"> <a rel="noopener nofollow" href="https://github.com/01-ai/Yi/actions/workflows/build_docker_image.yml"> <img src="https://github.com/01-ai/Yi/actions/workflows/build_docker_image.yml/badge.svg" style="margin: 0 0;"> </a> </div> <div style="display: inline-block;"> <a href="https://huggingface.co/01-ai"> <img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-01--ai-blue" style="margin: 0 0;"> </a> </div> <div style="display: inline-block;"> <a rel="noopener nofollow" href="https://www.modelscope.cn/organization/01ai/"> <img src="https://img.shields.io/badge/ModelScope-01--ai-blue" style="margin: 0 0;"> </a> </div> <div style="display: inline-block;"> <a rel="noopener nofollow" href="https://wisemodel.cn/organization/01.AI"> <img src="https://img.shields.io/badge/WiseModel-01--ai-blue" style="margin: 0 0;"> </a> </div> <div style="display: inline-block;"> <a rel="noopener nofollow" href="https://replicate.com/01-ai"> <img src="https://img.shields.io/badge/Replicate-01--ai-blue?logo=data:image/svg%2bxml;base64,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" style="margin: 0 0;"> </a> </div> <div style="display: inline-block;"> <a rel="noopener nofollow" href="https://github.com/01-ai/Yi/blob/main/LICENSE"> <img src="https://img.shields.io/badge/Code_License-Apache_2.0-lightblue" style="margin: 0 0;"> </a> </div> <div style="display: inline-block;"> <a rel="noopener nofollow" href="https://github.com/01-ai/Yi/blob/main/MODEL_LICENSE_AGREEMENT.txt"> <img src="https://img.shields.io/badge/Model_License-Model_Agreement-lightblue" style="margin: 0 0;"> </a> </div> <div style="display: inline-block;"> <a rel="noopener nofollow" href="mailto:[email protected]"> <img src="https://img.shields.io/badge/✉️[email protected]" style="margin: 0 0;"> </a> </div> </div>

Introduction

The Yi series models are large language models trained from scratch by developers at 01.AI.

News

<details open> <summary>🎯 <b>2023/11/23</b>: The chat models are open to public.</summary>

This release contains two chat models based on previous released base models, two 8-bits models quantized by GPTQ, two 4-bits models quantized by AWQ.

• Yi-34B-Chat
• Yi-34B-Chat-4bits
• Yi-34B-Chat-8bits
• Yi-6B-Chat
• Yi-6B-Chat-4bits
• Yi-6B-Chat-8bits

You can try some of them interactively at:

• Hugging Face
• Replicate

</details> <details open> <summary>🔔 <b>2023/11/23</b>: The Yi Series Models Community License Agreement is updated to v2.1.</summary> </details> <details> <summary>🔥 <b>2023/11/08</b>: Invited test of Yi-34B chat model.</summary>

Application form:

• English
• Chinese

</details> <details> <summary>🎯 <b>2023/11/05</b>: The base model of <code>Yi-6B-200K</code> and <code>Yi-34B-200K</code>.</summary>

This release contains two base models with the same parameter sizes of previous release, except that the context window is extended to 200K.

</details> <details> <summary>🎯 <b>2023/11/02</b>: The base model of <code>Yi-6B</code> and <code>Yi-34B</code>.</summary>

The first public release contains two bilingual (English/Chinese) base models with the parameter sizes of 6B and 34B. Both of them are trained with 4K sequence length and can be extended to 32K during inference time.

</details>

Ecosystem

🤗 You are encouraged to create a PR and share your awesome work built on top of the Yi series models.

• Serving
  • ScaleLLM: Efficiently run Yi models locally.
• Quantization
  • TheBloke/Yi-34B-GGUF
  • TheBloke/Yi-34B-GPTQ
• Finetuning
  • NousResearch/Nous-Capybara-34B
  • SUSTech/SUS-Chat-34B: This model ranks first among all models below 70B and has outperformed the twice larger deepseek-llm-67b-chat. You can check the result in 🤗 Open LLM Leaderboard.

Model Performance

Base Model Performance

While benchmarking open-source models, we have observed a disparity between the results generated by our pipeline and those reported in public sources (e.g. OpenCompass). Upon conducting a more in-depth investigation of this difference, we have discovered that various models may employ different prompts, post-processing strategies, and sampling techniques, potentially resulting in significant variations in the outcomes. Our prompt and post-processing strategy remains consistent with the original benchmark, and greedy decoding is employed during evaluation without any post-processing for the generated content. For scores that were not reported by the original authors (including scores reported with different settings), we try to get results with our pipeline.

To evaluate the model's capability extensively, we adopted the methodology outlined in Llama2. Specifically, we included PIQA, SIQA, HellaSwag, WinoGrande, ARC, OBQA, and CSQA to assess common sense reasoning. SquAD, QuAC, and BoolQ were incorporated to evaluate reading comprehension. CSQA was exclusively tested using a 7-shot setup, while all other tests were conducted with a 0-shot configuration. Additionally, we introduced GSM8K (8-shot@1), MATH (4-shot@1), HumanEval (0-shot@1), and MBPP (3-shot@1) under the category "Math & Code". Due to technical constraints, we did not test Falcon-180 on QuAC and OBQA; the score is derived by averaging the scores on the remaining tasks. Since the scores for these two tasks are generally lower than the average, we believe that Falcon-180B's performance was not underestimated.

Chat Model Performance

We evaluated various benchmarks using both zero-shot and few-shot methods, except for TruthfulQA. Generally, the zero-shot approach is more common in chat models. Our evaluation strategy involves generating responses while following instructions explicitly or implicitly (such as using few-shot examples). We then isolate relevant answers from the generated text. Some models are not well-suited to produce output in the specific format required by instructions in few datasets, which leads to suboptimal results.

<strong>*</strong>: C-Eval results are evaluated on the validation datasets

Quantized Chat Model Performance

We also provide both 4-bit (AWQ) and 8-bit (GPTQ) quantized Yi chat models. Evaluation results on various benchmarks have shown that the quantized models have negligible losses. Additionally, they reduce the memory footprint size. After testing different configurations of prompts and generation lengths, we highly recommend following the guidelines in the memory footprint table below when selecting a device to run our models.

Note: All the numbers in the table represent the minimum recommended memory for running models of the corresponding size.

Limitations of Chat Model

The released chat model has undergone exclusive training using Supervised Fine-Tuning (SFT). Compared to other standard chat models, our model produces more diverse responses, making it suitable for various downstream tasks, such as creative scenarios. Furthermore, this diversity is expected to enhance the likelihood of generating higher quality responses, which will be advantageous for subsequent Reinforcement Learning (RL) training.

However, this higher diversity might amplify certain existing issues, including:

• Hallucination: This refers to the model generating factually incorrect or nonsensical information. With the model's responses being more varied, there's a higher chance of hallucination that are not based on accurate data or logical reasoning.
• Non-determinism in re-generation: When attempting to regenerate or sample responses, inconsistencies in the outcomes may occur. The increased diversity can lead to varying results even under similar input conditions.
• Cumulative Error: This occurs when errors in the model's responses compound over time. As the model generates more diverse responses, the likelihood of small inaccuracies building up into larger errors increases, especially in complex tasks like extended reasoning, mathematical problem-solving, etc.

To achieve more coherent and consistent responses, it is advisable to adjust generation configuration parameters such astemperature,top_p, ortop_k. These adjustments can help in the balance between creativity and coherence in the model's outputs.

Usage

Feel free to create an issue if you encounter any problem when using the Yi series models.

1. Prepare development environment

1.1 Docker

The best approach to try the Yi series models is through Docker with GPUs. We provide the following docker images to help you get started.

• registry.lingyiwanwu.com/ci/01-ai/yi:latest
• ghcr.io/01-ai/yi:latest

Note that the latest tag always points to the latest code in the main branch. To test a stable version, please replace it with a specific tag.

1.2 Local development environment

We use conda-lock to generate fully reproducible lock files for conda environments. You can refer to conda-lock.yml for the exact versions of the dependencies. Additionally, we utilize micromamba for installing these dependencies.

To install the dependencies, please follow these steps:

1. Install micromamba by following the instructions available here.
2. Execute micromamba install -y -n yi -f conda-lock.yml to create a conda environment named yi and install the necessary dependencies.

2. Download the model (optional)

By default, the model weights and tokenizer will be downloaded from Hugging Face automatically in the next step. You can also download them manually from the following places:

• ModelScope
• WiseModel

3. Examples

3.1 Use the chat model

from transformers import AutoModelForCausalLM, AutoTokenizer

model_path = '01-ai/Yi-34b-Chat'

tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)

# Since transformers 4.35.0, the GPT-Q/AWQ model can be loaded using AutoModelForCausalLM.
model = AutoModelForCausalLM.from_pretrained(
    model_path,
    device_map="auto",
    torch_dtype='auto'
).eval()

# Prompt content: "hi"
messages = [
    {"role": "user", "content": "hi"}
]

input_ids = tokenizer.apply_chat_template(conversation=messages, tokenize=True, add_generation_prompt=True, return_tensors='pt')
output_ids = model.generate(input_ids.to('cuda'))
response = tokenizer.decode(output_ids[0][input_ids.shape[1]:], skip_special_tokens=True)

# Model response: "Hello! How can I assist you today?"
print(response)

To construct the prompt template manually, you can refer the chat_template field in the tokenizer_config.json file.

<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant

3.2 Use the base model

python demo/text_generation.py

To reuse the downloaded models in the previous step, you can provide the extra --model argument:

python demo/text_generation.py  --model /path/to/model

Or if you'd like to get your hands dirty:

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("01-ai/Yi-34B", device_map="auto", torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained("01-ai/Yi-34B")
inputs = tokenizer("There's a place where time stands still. A place of breath taking wonder, but also", return_tensors="pt")
max_length = 256

outputs = model.generate(
    inputs.input_ids.cuda(),
    max_length=max_length,
    eos_token_id=tokenizer.eos_token_id,
    do_sample=True,
    repetition_penalty=1.3,
    no_repeat_ngram_size=5,
    temperature=0.7,
    top_k=40,
    top_p=0.8,
)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

<details> <summary>Output</summary>

Prompt: There's a place where time stands still. A place of breath taking wonder, but also

Generation: There's a place where time stands still. A place of breath taking wonder, but also of great danger. A place where the very air you breathe could kill you. A place where the only way to survive is to be prepared. The place is called the Arctic. The Arctic is a vast, frozen wilderness. It is a place of extremes. The temperatures can drop to -40 degrees Celsius. The winds can reach speeds of 100 kilometers per hour. The sun can shine for 24 hours a day, or not at all for weeks on end. The Arctic is also a place of great beauty. The ice and snow are a pristine white. The sky is a deep blue. The sunsets are spectacular. But the Arctic is also a place of great danger. The ice can be treacherous. The winds can be deadly. The sun can be blinding. The Arctic is a place where the only way to survive is to be prepared. The Arctic is a place of extremes. The temperatures can drop to -40 degrees Celsius. The winds can reach speeds of 100 kilometers per hour. The sun can shine for 24 hours a day, or not at all for weeks on end. The Arctic is a place of great beauty. The ice and snow are a

</details>

For more advanced usage, please refer to the doc.

3.3 Finetuning from the base model:

bash finetune/scripts/run_sft_Yi_6b.sh

Once finished, you can compare the finetuned model and the base model with the following command:

bash finetune/scripts/run_eval.sh

For more advanced usage like fine-tuning based on your custom data, please refer the doc.

3.4 Quantization

GPT-Q

python quantization/gptq/quant_autogptq.py \
  --model /base_model                      \
  --output_dir /quantized_model            \
  --trust_remote_code

Once finished, you can then evaluate the resulting model as follows:

python quantization/gptq/eval_quantized_model.py \
  --model /quantized_model                       \
  --trust_remote_code

For a more detailed explanation, please read the doc

AWQ

python quantization/awq/quant_autoawq.py \
  --model /base_model                      \
  --output_dir /quantized_model            \
  --trust_remote_code

Once finished, you can then evaluate the resulting model as follows:

python quantization/awq/eval_quantized_model.py \
  --model /quantized_model                       \
  --trust_remote_code

For more detailed explanation, please read the doc

FAQ

1. What dataset was this trained with?

   The dataset we use contains Chinese & English only. We used approximately 3T tokens. The detailed number and its construction will be described in the upcoming technical report.

Disclaimer

We use data compliance checking algorithms during the training process, to ensure the compliance of the trained model to the best of our ability. Due to complex data and the diversity of language model usage scenarios, we cannot guarantee that the model will generate correct, and reasonable output in all scenarios. Please be aware that there is still a risk of the model producing problematic outputs. We will not be responsible for any risks and issues resulting from misuse, misguidance, illegal usage, and related misinformation, as well as any associated data security concerns.

License

The source code in this repo is licensed under the Apache 2.0 license. The Yi series models are fully open for academic research and free commercial usage with permission via applications. All usage must adhere to the Model License Agreement 2.0. To apply for the official commercial license, please contact us ([email protected]).