Hugging Face's logo

qualcomm
/
Baichuan2-7B

Text Generation
PyTorch
llm
generative_ai
quantized
android
Model card Files Community
Baichuan2-7B / README.md
qaihm-bot's picture
qaihm-bot
Upload README.md with huggingface_hub
aca534a verified
preview code
|
raw
history blame
5.32 kB
metadata
library_name: pytorch
license: apache-2.0
pipeline_tag: text-generation
tags:
  - llm
  - generative_ai
  - quantized
  - android

Baichuan2-7B: Optimized for Mobile Deployment

State-of-the-art large language model useful on a variety of language understanding and generation tasks

Baichuan2-7B is a family of LLMs. It achieves the state-of-the-art performance of its size on standard Chinese and English authoritative benchmarks (C-EVAL/MMLU). 4-bit weights and 16-bit activations making it suitable for on-device deployment. For Prompt and output length specified below, the time to first token is Baichuan2-PromptProcessor-Quantized's latency and average time per addition token is Baichuan2-TokenGenerator-Quantized's latency.

This is based on the implementation of Baichuan2-7B found here. More details on model performance accross various devices, can be found here.

Model Details

  • Model Type: Text generation
  • Model Stats:
    • Input sequence length for Prompt Processor: 128
    • Context length: 4096
    • Number of parameters: 7.07B
    • Precision: w4a16 + w8a16 (few layers)
    • Num of key-value heads: 8
    • Information about the model parts: Prompt Processor and Token Generator are split into 5 parts each. Each corresponding Prompt Processor and Token Generator part share weights.
    • Prompt processor model size: 5.06 GB
    • Prompt processor input (part1): 128 tokens
    • Prompt processor output (part1): Embeddings output
    • Prompt processor input (other parts): 128 tokens + KVCache initialized with pad token
    • Prompt processor output (other parts): 128 output tokens + KVCache for token generator
    • Token generator model size: 5.06 GB
    • Token generator input (part1): 128 tokens
    • Token generator output (part1): Embeddings output
    • Token generator input (other parts): 1 input token + past KVCache
    • Token generator output (other parts): 1 output token + KVCache for next iteration
    • Use: Initiate conversation with prompt-processor and then token generator for subsequent iterations.
    • Supported languages: Chinese and English.
    • Minimum QNN SDK version required: 2.27.7
    • TTFT: Time To First Token is the time it takes to generate the first response token. This is expressed as a range because it varies based on the length of the prompt. The lower bound is for a short prompt (up to 128 tokens, i.e., one iteration of the prompt processor) and the upper bound is for a prompt using the full context length (4096 tokens).
    • Response Rate: Rate of response generation after the first response token.
    • Tiny MMLU: Tiny MMLU (Massive Multitask Language Understanding) is an English language benchmark designed to measure knowledge acquired during pretraining by evaluating models exclusively in zero-shot and few-shot settings. This makes the benchmark more challenging and more similar to how we evaluate humans.
Model Device Chipset Target Runtime Response Rate (tokens per second) Time To First Token (range, seconds) Tiny MMLU
Baichuan2-7B Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 7.72 0.20804799999999998 - 6.6575359999999995 49.34%

Deploying Baichuan2-7B on-device

Please follow the LLM on-device deployment tutorial.

License

  • The license for the original implementation of Baichuan2-7B can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Usage and Limitations

Model may not be used for or in connection with any of the following applications:

  • Accessing essential private and public services and benefits;
  • Administration of justice and democratic processes;
  • Assessing or recognizing the emotional state of a person;
  • Biometric and biometrics-based systems, including categorization of persons based on sensitive characteristics;
  • Education and vocational training;
  • Employment and workers management;
  • Exploitation of the vulnerabilities of persons resulting in harmful behavior;
  • General purpose social scoring;
  • Law enforcement;
  • Management and operation of critical infrastructure;
  • Migration, asylum and border control management;
  • Predictive policing;
  • Real-time remote biometric identification in public spaces;
  • Recommender systems of social media platforms;
  • Scraping of facial images (from the internet or otherwise); and/or
  • Subliminal manipulation