mirror of
https://github.com/ikawrakow/ik_llama.cpp.git
synced 2026-01-26 17:20:01 +00:00
135 lines
13 KiB
Markdown
135 lines
13 KiB
Markdown
# ik_llama.cpp: llama.cpp fork with better CPU performance
|
||
|
||
[](https://opensource.org/licenses/MIT)
|
||
|
||
## TL;DR
|
||
|
||
This repository is a fork of [llama.cpp](https://github.com/ggerganov/llama.cpp) with better CPU and hybrid GPU/CPU performance, new SOTA quantization types, first-class Bitnet support, better DeepSeek performance via MLA, FlashMLA, fused MoE operations and tensor overrides for hybrid GPU/CPU inference, row-interleaved quant packing, etc.
|
||
|
||
## Latest News
|
||
|
||
### Model Support
|
||
|
||
LlaMA-3-Nemotron [PR 377](https://github.com/ikawrakow/ik_llama.cpp/pull/377), Qwen3 [PR 355](https://github.com/ikawrakow/ik_llama.cpp/pull/355), GLM-4 [PR 344](https://github.com/ikawrakow/ik_llama.cpp/pull/344), Command-A [PR 341](https://github.com/ikawrakow/ik_llama.cpp/pull/341), bitnet-b1.58-2B-4T [PR 337](https://github.com/ikawrakow/ik_llama.cpp/pull/337), LLaMA-4 [PR 321](https://github.com/ikawrakow/ik_llama.cpp/pull/321), Gemma3 [PR 276](https://github.com/ikawrakow/ik_llama.cpp/pull/276), DeepSeek-V3 [PR 176](https://github.com/ikawrakow/ik_llama.cpp/pull/176), Kimi-2 [PR 609](https://github.com/ikawrakow/ik_llama.cpp/pull/609), dots.llm1 [PR 573](https://github.com/ikawrakow/ik_llama.cpp/pull/573), Hunyuan [PR 565](https://github.com/ikawrakow/ik_llama.cpp/pull/565)
|
||
|
||
### Quantization
|
||
|
||
#### Quantization additions
|
||
|
||
##### Trellis quants (`IQ1_KT`, `IQ2_KT`, `IQ3_KT`, `IQ4_KT`)
|
||
|
||
Information and the original CUDA implementation in [PR 113](https://github.com/ikawrakow/ik_llama.cpp/pull/113). Additional implementations: Metal [PR 475](https://github.com/ikawrakow/ik_llama.cpp/pull/475), Neon [PR 471](https://github.com/ikawrakow/ik_llama.cpp/pull/471), CPU [PR 441](https://github.com/ikawrakow/ik_llama.cpp/pull/441). `IQ1_KT` was added more recently in [PR 616](https://github.com/ikawrakow/ik_llama.cpp/pull/616). Note: these are base on a novel, integer-base trellis, which allows to achieve reasonable CPU performance, see [PR 529](https://github.com/ikawrakow/ik_llama.cpp/pull/529) and PRs quoted there for details.
|
||
|
||
##### IQK quants
|
||
|
||
Information can be found in [Discussion 8](https://github.com/ikawrakow/ik_llama.cpp/discussions/8).
|
||
|
||
Initial implementations (Zen4, AVX2, NEON): `IQ5_KS_R4` [PR 426](https://github.com/ikawrakow/ik_llama.cpp/pull/426), `IQ5_KS` [PR 422](https://github.com/ikawrakow/ik_llama.cpp/pull/422), `IQ4_KS_R4` [PR 150](https://github.com/ikawrakow/ik_llama.cpp/pull/150), `IQ5_K_R4` [PR 149](https://github.com/ikawrakow/ik_llama.cpp/pull/149), `IQ2_K_R4` [PR 146](https://github.com/ikawrakow/ik_llama.cpp/pull/146), `IQ3_K_R4` [PR 145](https://github.com/ikawrakow/ik_llama.cpp/pull/145), `IQ4_K_R4` [PR 138](https://github.com/ikawrakow/ik_llama.cpp/pull/138), `IQ4_KSS` [PR 89](https://github.com/ikawrakow/ik_llama.cpp/pull/89), `IQ2_KS` [PR 85](https://github.com/ikawrakow/ik_llama.cpp/pull/85), `IQ4_KS` [PR 83](https://github.com/ikawrakow/ik_llama.cpp/pull/83), `IQ6_K` [PR 14](https://github.com/ikawrakow/ik_llama.cpp/pull/14), `IQ2_K, IQ3_K and IQ5_K` [PR 7](https://github.com/ikawrakow/ik_llama.cpp/pull/7), `IQ4_K` [PR 6](https://github.com/ikawrakow/ik_llama.cpp/pull/6)
|
||
|
||
Cuda implementations: `IQ4_KS_R4` and `IQ5_KS_R4` [PR 493](https://github.com/ikawrakow/ik_llama.cpp/pull/493), `IQ1_S_R4` [PR 492](https://github.com/ikawrakow/ik_llama.cpp/pull/492), `IQ1_M_R4` [PR 494](https://github.com/ikawrakow/ik_llama.cpp/pull/494). `IQ4_KS_R4` and `IQ5_KS_R4` [PR 462](https://github.com/ikawrakow/ik_llama.cpp/pull/462), `IQ2_K_R4`, `IQ3_K_R4`, `IQ4_K_R4`, `IQ5_K_R4` [PR 461](https://github.com/ikawrakow/ik_llama.cpp/pull/461), `IQ4_K, IQ5_K, IQ6_K` [PR 417](https://github.com/ikawrakow/ik_llama.cpp/pull/417), `IQ2_KS, IQ2_K, IQ3_K` [PR 418](https://github.com/ikawrakow/ik_llama.cpp/pull/417)
|
||
|
||
`IQ2_KL` is a more recent addition in [PR 602](https://github.com/ikawrakow/ik_llama.cpp/pull/602)
|
||
|
||
#### Quantization improvements
|
||
|
||
`IQ1_M` [PR 327](https://github.com/ikawrakow/ik_llama.cpp/pull/327), `IQ2_XS` [PR 312](https://github.com/ikawrakow/ik_llama.cpp/pull/312), `Q2_K, Q4_K, Q5_K, Q4_1, Q5_1` [PR 302](https://github.com/ikawrakow/ik_llama.cpp/pull/302), `Q4_0, Q5_0, Q6_0, Q3_K, Q6_K, IQ4_XS, IQ4_NL` [PR 295](https://github.com/ikawrakow/ik_llama.cpp/pull/295)
|
||
|
||
#### Quantization performance improvements
|
||
|
||
* Much faster CPU prompt processing for all non-interleaved quants. Initial idea in [PR 515](https://github.com/ikawrakow/ik_llama.cpp/pull/515) and [PR 531](https://github.com/ikawrakow/ik_llama.cpp/pull/531), with many follow up PRs to apply to all quantization types for the 3 supported CPU platforms.
|
||
* All quantization types now have quantized matrix multiplication CUDA kernels, see [PR 557](https://github.com/ikawrakow/ik_llama.cpp/pull/515) and several others
|
||
* Faster CPU prompt processing for Trellis quants and MoE models. [PR 488](https://github.com/ikawrakow/ik_llama.cpp/pull/488)
|
||
* Trellis quants: faster CPU prompt processing [PR 482](https://github.com/ikawrakow/ik_llama.cpp/pull/482).
|
||
* Minor (~2%) `iq2_ks` TG performance improvement on CUDA [PR 468](https://github.com/ikawrakow/ik_llama.cpp/pull/468)
|
||
* Faster `IQ3_KT` and `IQ4_KT` [PR 453](https://github.com/ikawrakow/ik_llama.cpp/pull/453)
|
||
* Zen4: Faster PP for `IQ2_KS, IQ4_KS, IQ5_KS` [PR 428](https://github.com/ikawrakow/ik_llama.cpp/pull/428)
|
||
* Fast GEMM/GEMV for `IQ1_S` [PR 212](https://github.com/ikawrakow/ik_llama.cpp/pull/212)
|
||
|
||
### Features
|
||
|
||
* Function call support [PR 628](https://github.com/ikawrakow/ik_llama.cpp/pull/628)
|
||
* Webui: New Features for Conversations, Settings, and Chat Messages [PR 618](https://github.com/ikawrakow/ik_llama.cpp/pull/618)
|
||
* Legacy quants conversion schemes in `convert_hf_to_gguf.py` [PR 449](https://github.com/ikawrakow/ik_llama.cpp/pull/449), `Q6_0` in [PR 483](https://github.com/ikawrakow/ik_llama.cpp/pull/483)
|
||
* June 8 2025: Webui updated (legacy still available when `--path ./examples/server/public_legacy` is passed) [PR 481](https://github.com/ikawrakow/ik_llama.cpp/pull/481)
|
||
* June 8 2025: RPC improvements [PR 480](https://github.com/ikawrakow/ik_llama.cpp/pull/480)
|
||
* June 7 2025: Add an endpoint that lists all the saved prompt caches to server [PR 502](https://github.com/ikawrakow/ik_llama.cpp/pull/502)
|
||
* June 6 2025: Make prompt cache saving and restoring MLA aware [PR 497](https://github.com/ikawrakow/ik_llama.cpp/pull/497)
|
||
* June 3 2025: Added samplers, XTC [PR 486](https://github.com/ikawrakow/ik_llama.cpp/pull/486), top-n σ [PR 489](https://github.com/ikawrakow/ik_llama.cpp/pull/489).
|
||
* May 22 2025: Refactor `iqk_mul_mat.cpp` which speeds up compilation time significantly. [PR 435](https://github.com/ikawrakow/ik_llama.cpp/pull/435)
|
||
* May 17 2025: Option to enable or disable the CPU FA kernels [PR 429](https://github.com/ikawrakow/ik_llama.cpp/pull/429).
|
||
* May 12 2025: User can now control if/which operations with tensors held in RAM are offloaded to the GPU. See [PR 405](https://github.com/ikawrakow/ik_llama.cpp/pull/405)
|
||
* May 12 2025: Compatibility issues with mainline `llama.cpp` GGUFs for DeepSeek models with MLA enabled were resolved in [PR 394](https://github.com/ikawrakow/ik_llama.cpp/pull/394). The lower prompt processing performance resulting from using `llama.cpp`-style MLA GGUFs was recovered in [PR 409](https://github.com/ikawrakow/ik_llama.cpp/pull/409).
|
||
* April 21 2025: ik_llama.cpp builds and runs successfully on Android (using termux), see [PR 336](https://github.com/ikawrakow/ik_llama.cpp/pull/336)
|
||
* March 1 2025: Smart Expert Reduction for faster DeepSeek inference [PR 239](https://github.com/ikawrakow/ik_llama.cpp/pull/239)
|
||
* Feb 25 2025: Tensor overrides for better control where model weights are stored (GPU or CPU) [PR 232](https://github.com/ikawrakow/ik_llama.cpp/pull/232)
|
||
* Feb 23 2025: `sweep-bench` - better performance benchmarking [PR 225](https://github.com/ikawrakow/ik_llama.cpp/pull/225)
|
||
* Feb 19 2025: `Q8_KV` - new type for 8-bit KV-cache quantization [PR 208](https://github.com/ikawrakow/ik_llama.cpp/pull/208)
|
||
* March 7 2025: Custom quantization mixes using regular expressions [PR 244](https://github.com/ikawrakow/ik_llama.cpp/pull/244)
|
||
|
||
### Performance improvements
|
||
|
||
* Better GPU offload strategy for MoE models when using hybrid HPU/CPU inference, see [PR 520](https://github.com/ikawrakow/ik_llama.cpp/pull/520)
|
||
* May 13 2025: Better CPU FA performance for DeepSeek-Lite. [PR 410](https://github.com/ikawrakow/ik_llama.cpp/pull/410)
|
||
* May 11 2025: Slightly faster flash attention for DeepSeek models on CUDA, along with extending compatibility to Touring or newer GPUs. [PR 408](https://github.com/ikawrakow/ik_llama.cpp/pull/408)
|
||
* May 4 2025: Significant token generation performance improvement on CUDA with Flash Attention for GQA models. For details and benchmarks. [PR 370](https://github.com/ikawrakow/ik_llama.cpp/pull/370)
|
||
* April 17 2025: Better CPU Flash Attention token generation performance. [PR 332](https://github.com/ikawrakow/ik_llama.cpp/pull/332)
|
||
* April 3 2025: Much faster MoE implementation on Metal. [PR 307](https://github.com/ikawrakow/ik_llama.cpp/pull/307)
|
||
* March 25 2025: Better MoE performance on CUDA [PR 283](https://github.com/ikawrakow/ik_llama.cpp/pull/283)
|
||
* March 23 2025: Better batched processing speed for DeepSeek models [PR 282](https://github.com/ikawrakow/ik_llama.cpp/pull/282)
|
||
* March 18 2025: Reduce compute buffer size [PR 237](https://github.com/ikawrakow/ik_llama.cpp/pull/237)
|
||
* March 10 2025: Better TG performance for MoE models on CUDA [PR 248](https://github.com/ikawrakow/ik_llama.cpp/pull/248)
|
||
* Feb 23 2025: Fused FFN ops for faster MoE inference [PR 229](https://github.com/ikawrakow/ik_llama.cpp/pull/229)
|
||
|
||
### Flash-MLA
|
||
|
||
* May 7 2025: 🚀 FlashMLA-3 for DeepSeek models on CUDA. [PR 386](https://github.com/ikawrakow/ik_llama.cpp/pull/386). Caveat: Ampere or newer Nvidia GPU required
|
||
* March 21 2025: 🚀 FlashMLA-3: fastest CPU-only inference for DeepSeek models [PR 273](https://github.com/ikawrakow/ik_llama.cpp/pull/273)
|
||
* March 17 2025: 🚀 FlashMLA-2 performance improvements [PR 253](https://github.com/ikawrakow/ik_llama.cpp/pull/253)
|
||
* March 12 2025: Allow `Q8_0` KV cache with FlashMLA-2 on CUDA [PR 265](https://github.com/ikawrakow/ik_llama.cpp/pull/265)
|
||
* March 9 2025: 🚀 FlashMLA on CUDA [PR 247](https://github.com/ikawrakow/ik_llama.cpp/pull/247)
|
||
* March 8 2025: 🚀 Faster FlashMLA CPU implementation [PR 243](https://github.com/ikawrakow/ik_llama.cpp/pull/243)
|
||
* March 3 2025: 🚀 Introducing FlashMLA - MLA with Flash Attention [PR 240](https://github.com/ikawrakow/ik_llama.cpp/pull/240)
|
||
* Feb 27 2025: MLA without transposed cache [PR 235](https://github.com/ikawrakow/ik_llama.cpp/pull/235)
|
||
* Feb 13 2025: Allow `Q8_0` quantized cache with MLA [PR 206](https://github.com/ikawrakow/ik_llama.cpp/pull/206)
|
||
* Feb 11 2025: 🚀 Flash Attention support for DeepSeek models [PR 200](https://github.com/ikawrakow/ik_llama.cpp/pull/200)
|
||
* Feb 9 2025: 🚀 MLA for DeepSeek models [PR 188](https://github.com/ikawrakow/ik_llama.cpp/pull/188)
|
||
|
||
### Fixes
|
||
|
||
* Fix bug in MMVQ kernel [PR 446](https://github.com/ikawrakow/ik_llama.cpp/pull/446)
|
||
* Fix AVX2 implementation of `IQ4_K, IQ4_KS, IQ5_K, IQ6_K` [PR 427](https://github.com/ikawrakow/ik_llama.cpp/pull/427)
|
||
* Fix standard attention on the CPU [PR 421](https://github.com/ikawrakow/ik_llama.cpp/pull/421)
|
||
* Fix imatrix calculation for MLA models [PR 411](https://github.com/ikawrakow/ik_llama.cpp/pull/411)
|
||
* Fix new CUDA FA on Touring [PR 413](https://github.com/ikawrakow/ik_llama.cpp/pull/413)
|
||
* Fix SER. CPU: [PR 415](https://github.com/ikawrakow/ik_llama.cpp/pull/415) CUDA: [PR 416](https://github.com/ikawrakow/ik_llama.cpp/pull/416)
|
||
|
||
## Resources
|
||
|
||
There is no single point of reference describing all new `ik_llama.cpp` features. Pull requests often contain detailed information, so browsing the PRs is often the best way to learn about new features and how to use them. In addition
|
||
* [The Wiki page](https://github.com/ikawrakow/ik_llama.cpp/wiki) has performance comparisons to mainline `llama.cpp`
|
||
* [This guide](https://github.com/ikawrakow/ik_llama.cpp/discussions/258) is a good place to start if you came here because of DeepSeek models
|
||
* [This discussion](https://github.com/ikawrakow/ik_llama.cpp/discussions/266) is about running DeepSeek-V3/R1 on a 16 x 3090 setup
|
||
* [This discussion](https://github.com/ikawrakow/ik_llama.cpp/discussions/8) describes the new quantization types available in `ik_llama.cpp`
|
||
|
||
## Testing
|
||
|
||
### Function Calls Tests
|
||
|
||
To run the function calls test suite:
|
||
|
||
```bash
|
||
cd build
|
||
cmake --build . --target test-function-calls
|
||
./bin/test-function-calls
|
||
```
|
||
|
||
The test suite covers parser functionality, streaming, error handling, content cleaning, and server integration. All tests should pass to ensure production readiness.
|
||
|
||
## Contributing
|
||
|
||
Contributions in form of pull requests, issue submissions (bug reports, feature requests), or general discussions, are welcome.
|
||
|
||
## License
|
||
|
||
MIT
|