Skip to content

Actions: lvyufeng/PocketLLM

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
178 workflow runs
178 workflow runs

Filter by Workflow

⌥ + click/return to exclude

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
Design the single-request serving path (#523) Publish documentation #178: Commit 9ed90e8 pushed by lvyufeng
26s main
main
26s
Merge pull request #522 from lvyufeng/feature/c-sampling-kernels Publish documentation #177: Commit 5928791 pushed by lvyufeng
28s main
main
28s
Merge pull request #521 from lvyufeng/feature/quantized-weights-in-th… Publish documentation #176: Commit 50df306 pushed by lvyufeng
35s main
main
35s
Merge pull request #519 from lvyufeng/feature/cpp-engine-qwen3-0p6b-cuda Publish documentation #175: Commit 4b9ce94 pushed by lvyufeng
22s main
main
22s
Merge pull request #517 from lvyufeng/refactor/python-package-into-py… Publish documentation #174: Commit d7df9a5 pushed by lvyufeng
19s main
main
19s
Move the mkdocs build inputs off the repo root into .mkdocs/ (#516) Publish documentation #173: Commit 6c418f2 pushed by lvyufeng
21s main
main
21s
Seed the rebuilt PocketLLM tree: torch-free kernel ABI and backend sk… Publish documentation #172: Commit c022859 pushed by lvyufeng
23s main
main
23s
Give PocketLLM a package name of its own (#515) Publish documentation #171: Commit 28723b1 pushed by lvyufeng
38s master
master
38s
Cut the multi-card code out of PocketLLM (#514) Publish documentation #170: Commit 4040118 pushed by lvyufeng
32s master
master
32s
Re-scope PocketLLM to single-card, edge and mobile (#513) Publish documentation #169: Commit 618f294 pushed by lvyufeng
32s master
master
32s
Fold the two run_worker bodies into one (#510) Publish documentation #168: Commit 1f22016 pushed by lvyufeng
41s master
master
41s
Measure per-method duplication across the five adapters (#508) Publish documentation #167: Commit 0864606 pushed by lvyufeng
42s master
master
42s
Serve Qwen3.8-27B BF16 on the 910B and record the ladder (#500) Publish documentation #166: Commit cf11577 pushed by lvyufeng
1m 25s master
master
1m 25s
Call the Ascend test hardware 910B, and update what runs on it (#498) Publish documentation #165: Commit 85317ca pushed by lvyufeng
2m 19s master
master
2m 19s
Refuse a budget the context cannot hold (#496) Publish documentation #164: Commit 5be243c pushed by lvyufeng
1m 1s master
master
1m 1s
Record the first serving ladder on the 910B (#492) Publish documentation #163: Commit 471436b pushed by lvyufeng
41s master
master
41s
Refuse a batch width that cannot be honoured (#490) Publish documentation #162: Commit 1ef83b0 pushed by lvyufeng
43s master
master
43s
Name the interpreter and the NCCL root in the module build recipe (#480) Publish documentation #161: Commit d2ad362 pushed by lvyufeng
44s master
master
44s
Delete the C++ binary's own HTTP front end (#479) Publish documentation #160: Commit bf3886a pushed by lvyufeng
44s master
master
44s
Serve response_format from the Python host, over the engine's vocabul… Publish documentation #159: Commit 1d91103 pushed by lvyufeng
1m 18s master
master
1m 18s
Port the last two request fields: n choices and logprobs (#469) Publish documentation #158: Commit fdf7b9e pushed by lvyufeng
52s master
master
52s
Port the native request surface to the cpp adapter (#466) Publish documentation #157: Commit d4975f5 pushed by lvyufeng
53s master
master
53s
Split --device into a platform and a card list (#475) Publish documentation #156: Commit 53c9c25 pushed by lvyufeng
48s master
master
48s
Spell each declared option as a flag, and refuse one nobody reads (#473) Publish documentation #155: Commit 8023745 pushed by lvyufeng
1m 29s master
master
1m 29s
Declare a shared option once, and let each runtime state its own answ… Publish documentation #154: Commit aa61d81 pushed by lvyufeng
1m 26s master
master
1m 26s