All workflows

Showing runs from all workflows

26 workflow runs

prepare: progressive F16 input deletion under RUSTYLLM_DELETE_F16_INP…
failed:
test#26:Commit6c8b8d7pushedbyeinyx
main
May 24, 20265m 1s
spec_decode_70b: model IDs env-configurable
failed:
test#25:Commit4daf92cpushedbyeinyx
main
May 24, 20265m 0s
README: document M3 Pro 36 GB recipe and Q2K-target trick
failed:
test#24:Commitd1be0d7pushedbyeinyx
main
May 24, 20265m 3s
Add mixed quantization (attn vs MLP) + slurp-based GgufLayer
failed:
test#23:Commit5325df9pushedbyeinyx
main
May 24, 20265m 8s
RustyLLM: 70B inference in pure Rust on a single small GPU
failed:
test#22:Commit9b3ebd7pushedbyeinyx
main
May 23, 20265m 9s
RustyLLM: 70B inference in pure Rust on a single small GPU
failed:
test#21:Commit1bc09d4pushedbyeinyx
main
May 23, 20264m 49s
RustyLLM: 70B inference in pure Rust on a single small GPU
failed:
test#20:Commit7db6576pushedbyeinyx
main
May 23, 20264m 51s
RustyLLM: 70B inference in pure Rust on a single small GPU
failed:
test#19:Commit1bb1c66pushedbyeinyx
main
May 23, 20264m 58s
RustyLLM: 70B LLM inference in pure Rust on a single small GPU
failed:
test#18:Commit19fc0edpushedbyeinyx
main
May 22, 20265m 15s
RustyLLM: 70B LLM inference in pure Rust on a single small GPU
failed:
test#17:Commit792e32apushedbyeinyx
main
May 22, 20265m 6s
RustyLLM: 70B inference in Rust on a single small GPU
failed:
test#16:Commitba9d0c6pushedbyeinyx
main
May 22, 20264m 56s
Async H2D F16 streaming: 1.43x over sync (1033s -> 723s)
failed:
test#15:Commit0d8fa82pushedbyeinyx
main
May 22, 20263m 5s
Bench Q4_0 hybrid: 2x slower than Q4_K, despite same bits/param
failed:
test#14:Commit9c0d12bpushedbyeinyx
main
May 22, 20263m 8s
README: 20 GPU + 60 CPU hybrid = 22.3s (81x over baseline)
failed:
test#13:Commit49247edpushedbyeinyx
main
May 22, 20264m 53s
README: update benchmark table with hybrid GPU+CPU residency win
failed:
test#12:Commit89f8fcfpushedbyeinyx
main
May 22, 20265m 9s
Add Phase 3a spike: real 1.6 GiB layer async-uploaded, header offset …
failed:
test#11:Commitd750584pushedbyeinyx
main
May 22, 20265m 0s
Hybrid GPU + CPU residency: 78x over CPU baseline, beats Ollama 1.5x
failed:
test#10:Commit697f0d8pushedbyeinyx
main
May 22, 20266m 30s
Add Phase 2 spike: cudarc CudaSlice → candle Tensor bridge validated
failed:
test#9:Commitc949e4apushedbyeinyx
main
May 22, 20265m 0s
Add async H2D spike: validates cudarc multi-stream overlap (37% savings)
failed:
test#8:Commitd9351d4pushedbyeinyx
main
May 22, 20265m 0s
Document q4_kernel microbench result: candle QMatMul wins on wide shapes
failed:
test#7:Commit5b5c7depushedbyeinyx
main
May 22, 20264m 49s
README: add CUDA streaming benchmark — rustyllm CUDA beats Python Air…
failed:
test#6:Commitcc683e6pushedbyeinyx
main
May 22, 20265m 34s
Initial commit
failed:
test#5:Commita1a7057pushedbyeinyx
main
May 22, 20265m 28s
StreamingMixtral::from_pretrained + StreamingFamily Mixtral arm
failed:
test#4:Commiteaf0a51pushedbyeinyx
main
May 22, 20265m 0s
Add CONTRIBUTING.md
failed:
test#3:Commit64aa083pushedbyeinyx
main
May 22, 20265m 4s
Add direct AVX2 Q4_0×F32 matmul kernel (q4_kernel.rs)
failed:
test#2:Commit8179ae4pushedbyeinyx
main
May 22, 20265m 14s