Hacktoberfest 2026：维护者为十月标记出来的 issue，仍然开放、适合新手。 浏览 Hacktoberfest issue

[feat] FP8 training

未关闭
  • 哪里不对？
#63 4 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
8/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
停滞
技术栈
python, pytorch
领域
ai-infra-agents

调研方向

This is an open design question spanning matmul kernels and attention integration, with no linked pull request and no assignee. The payload names no concrete file or test to start from, so a newcomer would first need to read the training stack and the FlashAttention 3 FP8 discussion it links. Done would mean a recorded decision on FP8 attention plus a BF16-versus-FP8 loss comparison, which is maintainer-level scope.

由索引模型根据 Issue 内容生成。

描述

enhancement need update

🧐 Problem Description

FP8 training can significantly improve training throughput by reducing memory requirements and improving computational efficiency. However, challenges remain in integrating FP8 across all components of the training stack, such as FlashAttention, while maintaining model performance and stability. For example, the FP8 implementation in FlashAttention 3 raises questions about quantization accuracy and latency between operations.

💡 Proposed Solution

Introduce FP8 training in Fast-LLM with the following steps:

  1. Matmul Operations: Implement FP8-compatible matrix multiplication using Transformer Engine or Triton kernels.
  2. FlashAttention Compatibility:
    • Evaluate the feasibility of integrating FlashAttention 3, which supports FP8.
    • Address concerns about quantization accuracy and potential pipeline latency.
    • Explore alternatives like FlexAttention if FlashAttention 3 proves unsuitable.
  3. Validation:
    • Compare loss curves and convergence behavior between FP8 and BF16 to ensure model stability and performance.

🔄 Alternatives Considered

  1. Continue with BF16/FP16 Precision:
    • Pros: Stable and widely supported.
    • Cons: Misses throughput and memory benefits of FP8.
  2. Partial FP8 Integration:
    • Pros: Focuses on high-impact components like matrix multiplications.
    • Cons: Could complicate debugging due to mixed precision across components.
  3. FlexAttention:
    • Pros: Potentially better support for FP8.
    • Cons: May require more development and testing for integration. May be slower than FlashAttention 3.

📈 Potential Benefits

  1. Increased Throughput: FP8 training has demonstrated up to 1.5x speedup compared to BF16 on NVIDIA H100 GPUs.
  2. Enhanced Efficiency: Reducing memory overhead enables larger batch sizes and improves GPU utilization.
  3. Scalability: Lower precision facilitates training on larger clusters without hitting memory limits.
  4. Comparable Convergence: Loss curves for FP8 have been shown to closely track those for BF16, ensuring minimal compromise in training stability.

📝 Additional Context

  1. Key Findings from Databricks:
    • FP8 training achieves ~1.5x throughput improvement.
    • Integrated with PyTorch FSDP and activation checkpointing for memory efficiency.
    • Loss curves for FP8 closely match BF16 during training.
    • Databricks Blog: Turbocharged Training with FP8.
  2. FlashAttention 3 Insights:
  3. Known Challenges:
    • Managing FP8 weight storage and quantization scaling.
    • Addressing pipeline latency issues in FP8 operations.
    • Ensuring alignment between different FP8-compatible components.
主要语言
Python
星标
337
派生
52
30 天内没有已合并 PR

环境准备

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

ServiceNow/Fast-LLM 的其他 Issue

查看 ServiceNow/Fast-LLM 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。