CPU Tests

Workflow runs for CPU Tests

• server: release the connection slot exactly once (fixes permanent connection-limit lockout after client hang-ups)

  completed successfully:
  CPU Tests#67:Pull request#911openedbytawei-amd
  fix/connection-slot-leak

  Nov 1, 20261m 7s
  More actions for workflow run 67

• Merge pull request #907 from ROCm/feat/ignore-eos-min-tokens

  completed successfully:
  CPU Tests#66:Commite0e1645pushedbytawei-amd
  main

  Nov 1, 20261m 8s
  More actions for workflow run 66

• Merge pull request #906 from ROCm/fix/887-min-p-float

  completed successfully:
  CPU Tests#65:Commit65a5f64pushedbytawei-amd
  main

  Nov 1, 20261m 13s
  More actions for workflow run 65

• Merge pull request #909 from ROCm/feat/qwen38-proj-mmul-xclbins

  completed successfully:
  CPU Tests#64:Commit1c39813pushedbytawei-amd
  main

  Nov 1, 20261m 9s
  More actions for workflow run 64

• qwen3.8-mtp: vendor mmul proj-engine xclbins (decode 1.77 -> 2.84 tok/s)

  completed successfully:
  CPU Tests#63:Pull request#909synchronizebytawei-amd
  feat/qwen38-proj-mmul-xclbins

  Nov 1, 20261m 12s
  More actions for workflow run 63

• qwen3.8-mtp: vendor mmul proj-engine xclbins (decode 1.77 -> 2.84 tok/s)

  completed successfully:
  CPU Tests#62:Pull request#909openedbytawei-amd
  feat/qwen38-proj-mmul-xclbins

  Nov 1, 20261m 5s
  More actions for workflow run 62

• Merge pull request #908 from ROCm/fix/gemma4-channel-opener-886

  completed successfully:
  CPU Tests#61:Commita2b0b6fpushedbytawei-amd
  main

  Nov 1, 20261m 11s
  More actions for workflow run 61

• fix(gemma4 tool calls): block the bare <|channel> token before a tool call (#886)

  completed successfully:
  CPU Tests#60:Pull request#908openedbytawei-amd
  fix/gemma4-channel-opener-886

  Nov 1, 20261m 13s
  More actions for workflow run 60

• Merge pull request #905 from ROCm/fix/gemma4-quote-closer

  completed successfully:
  CPU Tests#59:Commit0ce8a71pushedbytawei-amd
  main

  Nov 1, 20261m 3s
  More actions for workflow run 59

• feat(server): honour ignore_eos and min_tokens (#881)

  completed successfully:
  CPU Tests#58:Pull request#907openedbytawei-amd
  feat/ignore-eos-min-tokens

  Nov 1, 20261m 10s
  More actions for workflow run 58

• server: read min_p as float (fixes #887)

  completed successfully:
  CPU Tests#57:Pull request#906openedbytawei-amd
  fix/887-min-p-float

  Oct 31, 20261m 19s
  More actions for workflow run 57

• fix(gemma4 tool calls): a quote inside a string value no longer ends the value (#900)

  completed successfully:
  CPU Tests#56:Pull request#905openedbytawei-amd
  fix/gemma4-quote-closer

  Oct 31, 20261m 12s
  More actions for workflow run 56

• Merge pull request #902 from ROCm/docs/qwen35-footprint

  completed successfully:
  CPU Tests#55:Commitec2395bpushedbytawei-amd
  main

  Oct 31, 20261m 5s
  More actions for workflow run 55

• models: lower qwen3.5 0.8b/2b/4b footprint, note the mmap scope in the docs

  completed successfully:
  CPU Tests#54:Pull request#902openedbytawei-amd
  docs/qwen35-footprint

  Oct 31, 20261m 10s
  More actions for workflow run 54

• Merge pull request #901 from ROCm/feat/qwen35-mmap-embedding

  completed successfully:
  CPU Tests#53:Commit002ff1apushedbytawei-amd
  main

  Oct 31, 20261m 7s
  More actions for workflow run 53

• chore: vendor qwen3_5vl libs from IRON#134 — map the tied lm_head embedding instead of a bf16 table (0.8B/2B/4B)

  completed successfully:
  CPU Tests#52:Pull request#901openedbytawei-amd
  feat/qwen35-mmap-embedding

  Oct 31, 20261m 14s
  More actions for workflow run 52

• Merge pull request #899 from ROCm/feat/bolt-dflash-flash

  completed successfully:
  CPU Tests#51:Commitc42e223pushedbytawei-amd
  main

  Oct 31, 20261m 11s
  More actions for workflow run 51

• feat(gemma4e-dflash-flash): nemotron3.5-bolt-dflash-flash:e4b (single-turn flash + DFlash speculation)

  completed successfully:
  CPU Tests#50:Pull request#899synchronizebytawei-amd
  feat/bolt-dflash-flash

  Oct 31, 20261m 24s
  More actions for workflow run 50

• feat(server): log a 'Request done:' timing summary for every request …

  completed successfully:
  CPU Tests#49:Commit64d330apushedbytawei-amd
  main

  Oct 30, 20261m 16s
  More actions for workflow run 49

• feat(gemma4e-dflash-flash): nemotron3.5-bolt-dflash-flash:e4b (single-turn flash + DFlash speculation)

  completed successfully:
  CPU Tests#48:Pull request#899openedbytawei-amd
  feat/bolt-dflash-flash

  Oct 30, 20261m 17s
  More actions for workflow run 48

• feat(server): log a 'Request done:' timing summary for every request (#841)

  completed successfully:
  CPU Tests#47:Pull request#898openedbytawei-amd
  feat/request-timing-log

  Oct 30, 20261m 23s
  More actions for workflow run 47

• Minicpm ocr

  completed successfully:
  CPU Tests#46:Pull request#876synchronizebymichyu-amd
  minicpm_ocr

  Oct 30, 20261m 13s
  More actions for workflow run 46

• ci: map the minicpm-v-4.7-moe engine to its catalog family (#897)

  completed successfully:
  CPU Tests#45:Commit11a1452pushedbytawei-amd
  main

  Oct 30, 20261m 19s
  More actions for workflow run 45

• ci: map the minicpm-v-4.7-moe engine to its catalog family

  completed successfully:
  CPU Tests#44:Pull request#897openedbytawei-amd
  fix/ci-minicpm-moe-table

  Oct 30, 20261m 24s
  More actions for workflow run 44

• Add IRON backend and EmbeddingGemma-2 Model

  completed successfully:
  CPU Tests#43:Pull request#896synchronizebyandrej
  andrej:iron-backend

  Oct 30, 20261m 14s
  More actions for workflow run 43