Issues

• tensorfold benchmark: the code workload's raw /v1/completions prompt gets empty output from DeepSeek-V4.1 at T=0

  ashhart/TensorFold#1019 · Filter by author BobClawblaw opened on Oct 17, 2026
  1

• NVIDIA's GLM-5.3-Flash-FP4-Local on the native CUDA engine (tracker)

  ashhart/TensorFold#1015 · Filter by author Bizuayeu opened on Oct 17, 2026
  2

• Proposal: TensorFold Notes, memory inside the server (for 1.1.0)

  ashhart/TensorFold#1012 · Filter by author ThinkOffApp opened on Oct 17, 2026
  1

• aot: a launch with a runtime argument of 1 fails against a capture that lacks the value-1 constexpr variant (the unspecialized cubin computes the same bits)

  ashhart/TensorFold#1003 · Filter by author GustavBlack opened on Oct 17, 2026
  2

• Benchmarks: accept ROCm receipts on the board (Strix Halo)

  ashhart/TensorFold#1001 · Filter by author ThinkOffApp opened on Oct 17, 2026
  1

• GLM-5.3-Flash TP=2 (0.6.0 lineage): with uptime a 13-token prefill reaches 18.8 s and decode falls 40 to 14 tok/s, an idle rank 1 reads 96% GPU, and a restart clears both

  ashhart/TensorFold#993 · Filter by author ss-cong opened on Oct 17, 2026
  9

• Tensorfold App fails to load Qwen3.8 27B on M5 Pro with 48GB ram

  ashhart/TensorFold#989 · Filter by author Deesha08 opened on Oct 17, 2026

• tensorfold benchmark: let a managed run take --kv-cache, so an fp8-served recipe can submit ranked

  ashhart/TensorFold#983 · Filter by author BobClawblaw opened on Oct 17, 2026
  1

• Nemotron 3.5 Super VL on two GB10s: TP=2 serial decode is no faster than TP=1 here, and what TP=3/4 would need

  ashhart/TensorFold#981 · Filter by author tovereast opened on Oct 17, 2026
  3

• Proposal: GLM-5.3-Flash image/video input on the native CUDA engine (TP over hosts) — request + test hardware

  ashhart/TensorFold#971 · Filter by author gilesco opened on Oct 16, 2026
  3

• tensorfold benchmark: one sampled reply that stops before 256 tokens makes the cell incomplete (Nemotron oQ8, code T=1)

  ashhart/TensorFold#969 · Filter by author BlivionIaG opened on Oct 16, 2026
  4

• GLM-5.3-Flash: gather 5-, 6- and 8-bit routed experts in prompt chunks

  ashhart/TensorFold#968 · Filter by author harrisonfriia opened on Oct 16, 2026
  1

• GLM-5.3-Flash on Mac (native 1.0.12): a continuing turn's cache restore fails and re-prefills 100k+ tokens, with nothing logged

  ashhart/TensorFold#966 · Filter by author tpischke opened on Oct 16, 2026
  3

• GLM-5.3-Flash oQ3.5e-mtp on Mac: prefill ~270 tok/s, about 6x slower than the uniform 4-bit checkpoint (1.0.12)

  ashhart/TensorFold#965 · Filter by author tpischke opened on Oct 16, 2026
  1

• GLM-5.3-Flash on Mac (native 1.0.12): a decoding stream drops to ~2 tok/s while another stream prefills

  ashhart/TensorFold#964 · Filter by author tpischke opened on Oct 16, 2026
  1

• Nemotron Puzzle / LatentMoE on Metal (e.g. Bolt-74B-A9B)

  ashhart/TensorFold#962 · Filter by author trickyfalcon opened on Oct 16, 2026
  4

• Native 1.0.12 grammar vocabulary lookup misses llm_config for Nemotron Super VL

  ashhart/TensorFold#951 · Filter by author swbiggart opened on Oct 16, 2026
  1

• RFC: Nemotron 3.5 Bolt (nemotron_h_puzzle) on CUDA, and Nemotron Omni video

  ashhart/TensorFold#937 · Filter by author MiaAI-Lab opened on Oct 16, 2026
  1

• tensorfold benchmark refuses Nemotron 3.5 Super VL: config.json over the 4 MiB read cap reports "Checkpoint is incomplete"

  ashhart/TensorFold#934 · Filter by author jkuepker opened on Oct 16, 2026
  1

• Qwen3.6-35B-A3B (qwen3_5_moe) on the native Metal engine: proposal and gate

  ashhart/TensorFold#931 · Filter by author cshintov opened on Oct 16, 2026
  3

• Native 27B on M3 Ultra: cold prefill still 1.7x the python engine (open since 1.0.3); uniform-4bit checkpoints bypass 1.0.10's faster tiles

  ashhart/TensorFold#918 · Filter by author astro-15-alive opened on Oct 16, 2026
  4

• Drama 3 (fishaudio/drama-3) text-to-speech on CUDA: family proposal and gate

  ashhart/TensorFold#914 · Filter by author plotarmordev opened on Oct 16, 2026
  2

• App 1.0.0-beta.6: notes from a first run against a remote OpenAI-compatible GLM server

  ashhart/TensorFold#884 · Filter by author Gogo6969 opened on Oct 15, 2026
  4

• GLM-5.3-Flash on one Mac: a short request waits until another stream's prompt is read, even with free streams

  ashhart/TensorFold#880 · Filter by author Gogo6969 opened on Oct 15, 2026
  2

• Living Weights: a later lesson's quiet refit can silence an earlier fact on its own questions, and nothing asks it again

  ashhart/TensorFold#870 · Filter by author ZackSample opened on Oct 15, 2026
  3