Skip to content

Actions: qualcomm/llama.cpp

Actions

Check Pre-Tokenizer Hashes

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
pre-tokenizer-hashes.yml
18 workflow runs
18 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

opencl: do sqrt in f32 then convert to f16 for better precision Check Pre-Tokenizer Hashes #18: Commit 338d9ac pushed by lhez
31s lh/sqr-sqrt-mean-ssm-conv
lh/sqr-sqrt-mean-ssm-conv
31s
opencl: refine condition for kqv mm Check Pre-Tokenizer Hashes #17: Commit a8b6189 pushed by lhez
31s lh/fix-gemm-kqv-cond
lh/fix-gemm-kqv-cond
31s
vulkan: add LOG operation support for F32 and F16 (#17183) Check Pre-Tokenizer Hashes #16: Commit dbed612 pushed by max-krasnyansky
31s master
master
31s
CANN: Add cross_entropy_loss op support (#16886) Check Pre-Tokenizer Hashes #15: Commit 97d5117 pushed by max-krasnyansky
33s master
master
33s
cpu: skip NOPs to avoid barriers Check Pre-Tokenizer Hashes #14: Commit 011b2f9 pushed by max-krasnyansky
32s cpu-skip-nops
cpu-skip-nops
32s
fix whitespace Check Pre-Tokenizer Hashes #13: Commit dada517 pushed by shaofeiqi
28s sq/mul-mat-kq-kqv-opt-new
sq/mul-mat-kq-kqv-opt-new
28s
opencl: update doc Check Pre-Tokenizer Hashes #12: Commit 40d2095 pushed by lhez
35s lh/update-doc
lh/update-doc
35s
Fix garbled output with REPACK at high thread counts (#16956) Check Pre-Tokenizer Hashes #11: Commit 1f5accb pushed by max-krasnyansky
31s hexagon
hexagon
31s
Fix garbled output with REPACK at high thread counts (#16956) Check Pre-Tokenizer Hashes #10: Commit 1f5accb pushed by max-krasnyansky
30s master
master
30s
opencl: support imrope Check Pre-Tokenizer Hashes #9: Commit bec743f pushed by lhez
34s lh/imrope
lh/imrope
34s
server : bump request URI max length to 32768 (#16862) Check Pre-Tokenizer Hashes #8: Commit 16724b5 pushed by max-krasnyansky
56s master
master
56s
server : remove n_past (#16818) Check Pre-Tokenizer Hashes #7: Commit b52edd2 pushed by max-krasnyansky
51s hexagon
hexagon
51s
cpu: introduce chunking for flash attention Check Pre-Tokenizer Hashes #6: Commit e2364c9 pushed by max-krasnyansky
54s flashattn-chunking
flashattn-chunking
54s
hexagon: bump the thread count in the adb wrapper scripts Check Pre-Tokenizer Hashes #5: Commit 5cd4880 pushed by max-krasnyansky
52s hexagon-dspqueue-opts
hexagon-dspqueue-opts
52s
cpu: introduce chunking for flash attention Check Pre-Tokenizer Hashes #4: Commit e078508 pushed by max-krasnyansky
55s flashattn-chunking
flashattn-chunking
55s
convert : handle mmproj filename/path properly (#16760) Check Pre-Tokenizer Hashes #3: Commit 5d195f1 pushed by max-krasnyansky
54s hexagon
hexagon
54s
hexagon: remove custom can_repeat function and use ggml_can_repeat Check Pre-Tokenizer Hashes #2: Commit f7d7411 pushed by max-krasnyansky
56s hexagon
hexagon
56s
Add experimental ggml-hexagon backend for the Hexagon NPU (#16547) Check Pre-Tokenizer Hashes #1: Commit 63d2fc4 pushed by max-krasnyansky
55s master
master
55s