本文へ移動
cccskills

「fp16」の検索結果

18 件 ・ 関連度順

概要と使いどころ

deepspeed

無料

Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention

日本語の概要は準備中です。原文の説明を表示しています。

davila7/claude-code-templates3.3万2026年10月11日 更新

deepspeed

無料

Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention

日本語の概要は準備中です。原文の説明を表示しています。

Orchestra-Research/AI-Research-SKILLs1.3万2026年10月11日 更新

deepspeed

無料

Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention

日本語の概要は準備中です。原文の説明を表示しています。

huang-sh/DeepScience42026年7月15日 更新

gptq

無料

Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.

日本語の概要は準備中です。原文の説明を表示しています。

davila7/claude-code-templates3.3万2026年10月11日 更新

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

日本語の概要は準備中です。原文の説明を表示しています。

davila7/claude-code-templates3.3万2026年10月11日 更新

qnn-debug

無料

MNN QNN(高通 HTP/NPU)后端的问题定位、修复与算子适配。覆盖运行时报错(1002/6000/1003/6004)、模型转换/算子校验失败、推理结果异常、精度不达标、以及性能问题;并指导为 QNN 新增/适配算子。核心方法:用 MNN2QNNModel --dump_intermediate_outputs 一次性导出 QNN 全部中间张量,与 CPU(fwd=0) 基线逐张量对比,直接定位首个出错算子(旧的 testMNNFromOnnx.py 截断二分已降为回退手段);区分“真 bug/量化/HTP fp16 精度”;新增算子先查 SDK 算子文档(MasterOpDef/HtpOpDefSupplement)再实现。

日本語の概要は準備中です。原文の説明を表示しています。

alibaba/MNN1.6万2026年10月10日 更新

bugfix

無料

MNN 各类正确性/回归 bug 的排查入口,按 bug 类别分册组织,本文件只做症状分流。分册:内存别名与生命周期(arena reuse、`MemChunk`、融合引入的别名竞争)、量化误差与导出侧权重损坏(低 bit 打包、导出分块、PyTorch MPS/CUDA 大张量静默错误)、host 侧并发/线程竞争(共享所有权的引用计数被写坏、析构链崩溃、TSAN A/B 与编译期哨兵)、fp16 表示能力不足(长序列复读、position 塌缩,以及「实时计算→预计算查表」重构的三类陷阱)、GPU shader 越界与 command buffer 故障、后端 kernel 隐式假设违反(causal mask、layout 约定)、持久化缓存误信(weight-mmap sync 自我污染、跨模型缓存复用)、逐 run 不同的非确定性(未初始化内存/堆垃圾依赖、多线程动态分发×异构 kernel)。用户报告 MNN 输出乱码/退化、单测或 golden 对不上、改动后回归、换后端结果不同、开某开关才错、结果每次跑都不一样、或崩在析构链上且只在后台线程异步释放时偶现时使用。

日本語の概要は準備中です。原文の説明を表示しています。

alibaba/MNN1.6万2026年10月10日 更新

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

日本語の概要は準備中です。原文の説明を表示しています。

Orchestra-Research/AI-Research-SKILLs1.3万2026年10月11日 更新

gptq

無料

Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.

日本語の概要は準備中です。原文の説明を表示しています。

Orchestra-Research/AI-Research-SKILLs1.3万2026年10月11日 更新

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

日本語の概要は準備中です。原文の説明を表示しています。

OpenRaiser/NanoResearch1,3402026年10月9日 更新

Use when working on imp's quantization formats, loaders, or dequant paths - GGUF Q4_0...Q8_0/Q*_K/IQ4/MXFP4, NVFP4 two-level scaling (Modelopt vs compressed-tensors layouts), FP8 E4M3, StorageTier, decode cache, KV-cache dtypes (auto/fp16/fp8/int8/int4/nvfp4/mxfp4), "which quant should I use", "which KV dtype", scale-factor layout, dequant kernel wiring, imp-quantize / AWQ, judging quant quality (PPL). Do NOT use for writing/optimizing GEMM/GEMV kernels (sm120-cuda-expert) or measuring quant perf (benchmark-cuda).

日本語の概要は準備中です。原文の説明を表示しています。

kekzl/imp442026年10月11日 更新

Use when writing, reviewing, or optimizing CUDA kernels targeting sm_120a (RTX 5090 / Consumer Blackwell, GB202) in the imp inference engine. Triggers on CUDA/PTX kernel code, shared-memory layout, bank conflicts, tensor-core MMA (mxf4nvf4, tf32, 3xTF32/3xFP16), GEMV/GEMM/attention/quantization/GDN-scan kernels, decode tok/s under expected, kernel emitting HMMA instead of mxf4nvf4, occupancy, register pressure, __launch_bounds__, spills, PDL. Pair with `benchmark-cuda` for measurement and `check-degeneration` after hot-path changes.

日本語の概要は準備中です。原文の説明を表示しています。

kekzl/imp442026年10月11日 更新

gptq

無料

Quantização pós-treinamento em 4-bits para LLMs com perda mínima de precisão. Use para implantar modelos grandes (70B, 405B) em GPUs de consumo, quando você precisa de redução de memória 4× com <2% de degradação de perplexidade, ou para inferência mais rápida (aceleração de 3-4×) vs FP16. Integra com transformers e PEFT para fine-tuning QLoRA.

日本語の概要は準備中です。原文の説明を表示しています。

artubss/SKILLS-CLAUDE-CODE112026年5月17日 更新

API de treinamento distribuído mais simples. 4 linhas para adicionar suporte distribuído a qualquer script PyTorch. API unificada para DeepSpeed/FSDP/Megatron/DDP. Posicionamento automático de device, precisão mista (FP16/BF16/FP8). Config interativo, comando de launch único. Padrão do ecossistema HuggingFace.

日本語の概要は準備中です。原文の説明を表示しています。

artubss/SKILLS-CLAUDE-CODE112026年5月17日 更新

gptq

無料

Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.

日本語の概要は準備中です。原文の説明を表示しています。

huang-sh/DeepScience42026年7月15日 更新

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

日本語の概要は準備中です。原文の説明を表示しています。

huang-sh/DeepScience42026年7月15日 更新

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

日本語の概要は準備中です。原文の説明を表示しています。

ibragimov-oasis/vibe-coder22026年6月24日 更新

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

日本語の概要は準備中です。原文の説明を表示しています。

Lord1Egypt/awesome-skill-forge22026年6月10日 更新