Skip to content

fix: gate turbo V unpad on V type, not K type #91

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
TheTom merged 2 commits into from
Apr 20, 2026
Merged

fix: gate turbo V unpad on V type, not K type #91

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6112eb4
fix: gate turbo V unpad on V type, not K type
TheTom Apr 20, 2026
a1bcb34
fix(metal): disable TurboFlash by default — corrupt output on Apple10
TheTom Apr 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions ggml/src/ggml-metal/ggml-metal-ops.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2731,8 +2731,10 @@ static bool ggml_metal_op_flash_attn_ext_use_turbo_flash(const ggml_tensor * op)
// Check environment variable to force enable (bypasses other checks)
if (turbo_flash_env && turbo_flash_env[0] == '1') return true;

// Default: enabled for all qualifying configurations
return true;
// Default: disabled — TurboFlash two-pass kernel produces corrupt output
// on Apple10 (M5 Max) and possibly other Metal4 GPUs. Use TURBO_FLASH=1
// to opt-in for testing. See PR #91.
return false;
}

size_t ggml_metal_op_flash_attn_ext_extra_pad(const ggml_tensor * op) {
Expand Down
6 changes: 4 additions & 2 deletions src/llama-graph.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2189,7 +2189,8 @@ ggml_tensor * llm_graph_context::build_attn(

// TurboQuant: if V was padded, the output has padded dimensions.
// Extract original V head_dim after inverse WHT (applied inside build_attn_mha).
if (k->type == GGML_TYPE_TURBO3_0 || k->type == GGML_TYPE_TURBO4_0 || k->type == GGML_TYPE_TURBO2_0) {
// NOTE: gate on v->type (not k->type) for asymmetric configs where K=q8_0 but V=turbo
if (v->type == GGML_TYPE_TURBO3_0 || v->type == GGML_TYPE_TURBO4_0 || v->type == GGML_TYPE_TURBO2_0) {
const int64_t orig_v_head = hparams.n_embd_head_v(il);
// cur is 2D: (n_embd_head * n_head, n_tokens) after build_attn_mha
const int64_t padded_v_head = v->ne[0];
Expand Down Expand Up @@ -2415,7 +2416,8 @@ ggml_tensor * llm_graph_context::build_attn(
cb(cur, "kqv_out", il);

// TurboQuant: if V was padded, extract original V head_dim after inverse WHT
if (k->type == GGML_TYPE_TURBO3_0 || k->type == GGML_TYPE_TURBO4_0 || k->type == GGML_TYPE_TURBO2_0) {
// NOTE: gate on v->type (not k->type) for asymmetric configs where K=q8_0 but V=turbo
if (v->type == GGML_TYPE_TURBO3_0 || v->type == GGML_TYPE_TURBO4_0 || v->type == GGML_TYPE_TURBO2_0) {
const int64_t orig_v_head = hparams.n_embd_head_v(il);
const int64_t padded_v_head = v->ne[0];
if (padded_v_head != orig_v_head) {
Expand Down
Loading