Issues

• HIP: on gfx1151 the matrix-core prompt tiles are up to 2.9x slower on short passes when a weight row is 1,024, 2,048 or 4,096 bytes (K 2048, 4096, 8192 at 4-bit); 64 bytes of padding a row removes it

  ashhart/TensorFold#3001 · Filter by author AbrahamCete opened on Nov 12, 2026

• CUDA on unified memory: what the reserve, the KV pool and the prompt cache each decide, and at which scope

  ashhart/TensorFold#3000 · Filter by author Bizuayeu opened on Nov 12, 2026

• Native SAM3/SAM3.1 segmentation and tracking on Metal

  ashhart/TensorFold#2991 · Filter by author CerebralCoding opened on Nov 12, 2026
  1

• H3 Metal: read checkpoint shards ahead of their first mapped access

  ashhart/TensorFold#2988 · Filter by author plotarmordev opened on Nov 12, 2026
  3

• GLM CUDA: a named TF_KV_SHARE is planned after every window the room holds, so at TP=3 on a lighter checkpoint it keeps 8 of 32 states

  ashhart/TensorFold#2987 · Filter by author Bizuayeu opened on Nov 12, 2026

• DeepSeek-V4.1 on Metal: decode and prompt speed on M3 Ultra and M5 Ultra, token-exact (gate and order for the drafts)

  ashhart/TensorFold#2984 · Filter by author gilby opened on Nov 12, 2026
  2

• Flash Next on CUDA: flashnext-rank timeout stops both ranks inside NCCL initialization

  ashhart/TensorFold#2975 · Filter by author swbiggart opened on Nov 12, 2026
  2

• Flash Next on CUDA: TF_FLASHNEXT_DECODE_GRAPH=1 at --tp 2 fails the first plain round

  ashhart/TensorFold#2974 · Filter by author swbiggart opened on Nov 12, 2026
  2

• Flash Next on CUDA: a --tp 2 start fails when one rank loads over 30 s later

  ashhart/TensorFold#2970 · Filter by author swbiggart opened on Nov 12, 2026
  1

• Nemotron on Metal without an MTP head: concurrent streams share one round

  ashhart/TensorFold#2968 · Filter by author ashhart opened on Nov 12, 2026

• Nemotron CUDA: a failed image encode leaves its rows to splice into the next request

  ashhart/TensorFold#2943 · Filter by author ashhart opened on Nov 11, 2026
  4

• Qwen 3.5 CUDA: a prompt chunk nothing reads stops its last layer at the keys

  ashhart/TensorFold#2941 · Filter by author ashhart opened on Nov 11, 2026

• qwen27 Metal: K and V read in place from the mixed projection, no unstack copies

  ashhart/TensorFold#2940 · Filter by author ashhart opened on Nov 11, 2026

• qwen27 Metal (M5): out_proj group sums written by the GDN post and attention merge kernels

  ashhart/TensorFold#2931 · Filter by author ashhart opened on Nov 11, 2026

• qwen27 Metal: greedy picks ride the verify's last command buffer instead of their own

  ashhart/TensorFold#2930 · Filter by author ashhart opened on Nov 11, 2026

• HIP: on gfx1151 a plain prompt pass of 1 or 2 rows runs on kp0; on kp3 a resumed first token is 18-24% sooner on Qwen3.6-35B-A3B and 48-64% on Qwen3.8-27B, same bytes

  ashhart/TensorFold#2923 · Filter by author AbrahamCete opened on Nov 11, 2026
  6

• flashnext CUDA sampled draws: one launch and one wait a step, and #2753's check inside the top-k pass

  ashhart/TensorFold#2920 · Filter by author ashhart opened on Nov 11, 2026

• qwen27 Metal prompt attention: threadgroups past the causal edge still walk masked keys

  ashhart/TensorFold#2912 · Filter by author ashhart opened on Nov 11, 2026

• qwen3_5 CUDA prompts: the 48-column a and b projections launch alone from 2,048 rows

  ashhart/TensorFold#2911 · Filter by author ashhart opened on Nov 11, 2026

• nemotron CUDA: draw every stream's draft and target tokens in a fixed number of launches

  ashhart/TensorFold#2910 · Filter by author ashhart opened on Nov 11, 2026

• qwen3_5 CUDA: windows of 2-8 rows run padded to 16

  ashhart/TensorFold#2909 · Filter by author ashhart opened on Nov 11, 2026

• qwen3_5 CUDA: served one-row steps copy the whole recurrent state for a rollback nothing reads

  ashhart/TensorFold#2908 · Filter by author ashhart opened on Nov 11, 2026

• qwen27 Metal (M1-M4): solo prompt passes capped at 128 rows though the frame is 512

  ashhart/TensorFold#2907 · Filter by author ashhart opened on Nov 11, 2026

• qwen27 Metal prompts: drop the GDN tails buffer nothing reads

  ashhart/TensorFold#2905 · Filter by author ashhart opened on Nov 11, 2026

• CUDA release server: refuse a GPU it has no kernels for at startup, by name (today: CUDA error 209 at the first kernel load)

  ashhart/TensorFold#2904 · Filter by author ashhart opened on Nov 11, 2026