core/cli
8
0
Fork 0
Code Issues 1 Pull requests Projects Releases Packages Wiki Activity Actions 1

debug: add shape logging and stderr error handler for inference debugging

Browse source
This commit is contained in:
Claude 2026-02-16 02:16:30 +00:00 • committed by Snider
parent d92d097a7f
commit f416f7e529
2 changed files with 12 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
pkg/mlx/mlx.go
View file

@ -23,12 +23,14 @@ package mlx
#cgo darwin LDFLAGS: -framework Foundation -framework Metal -framework Accelerate #cgo darwin LDFLAGS: -framework Foundation -framework Metal -framework Accelerate
#cgo darwin LDFLAGS: -Wl,-rpath,${SRCDIR}/dist/lib #cgo darwin LDFLAGS: -Wl,-rpath,${SRCDIR}/dist/lib
#include <stdio.h>
#include <stdlib.h> #include <stdlib.h>
#include "mlx/c/mlx.h" #include "mlx/c/mlx.h"
static const char *last_mlx_error = NULL; static const char *last_mlx_error = NULL;
static void mlx_go_error_handler(const char *msg, void *data) { static void mlx_go_error_handler(const char *msg, void *data) {
fprintf(stderr, "MLX ERROR: %s\n", msg);
last_mlx_error = msg; last_mlx_error = msg;
} }

10
pkg/mlx/model/gemma3.go
View file

@ -95,6 +95,9 @@ type MLP struct {
DownProj *mlx.Linear DownProj *mlx.Linear
} }
// debugOnce is a temporary flag for shape debugging (remove after fix).
var debugOnce = true
// compiledGELU is a singleton for the compiled GELU function. // compiledGELU is a singleton for the compiled GELU function.
var compiledGELU *mlx.CompiledFunc var compiledGELU *mlx.CompiledFunc
@ -356,6 +359,13 @@ func (a *Attention) forward(x *mlx.Array, c cache.Cache, B, L int32, isSliding b
k := a.KProj.Forward(x) k := a.KProj.Forward(x)
v := a.VProj.Forward(x) v := a.VProj.Forward(x)
if debugOnce {
slog.Info("mlx: debug", "x_shape", x.Shape(), "q_shape", q.Shape(), "k_shape", k.Shape(), "v_shape", v.Shape(),
"B", B, "L", L, "num_heads", cfg.NumAttentionHeads, "num_kv_heads", cfg.NumKeyValueHeads, "head_dim", cfg.HeadDim,
"q_scales", a.QProj.Scales != nil)
debugOnce = false
}
// Reshape to [B, num_heads, L, head_dim] // Reshape to [B, num_heads, L, head_dim]
q = mlx.AsStrided(q, []int32{B, cfg.NumAttentionHeads, L, cfg.HeadDim}, q = mlx.AsStrided(q, []int32{B, cfg.NumAttentionHeads, L, cfg.HeadDim},
[]int64{int64(L * cfg.NumAttentionHeads * cfg.HeadDim), int64(cfg.HeadDim), int64(cfg.NumAttentionHeads * cfg.HeadDim), 1}, 0) []int64{int64(L * cfg.NumAttentionHeads * cfg.HeadDim), int64(cfg.HeadDim), int64(cfg.NumAttentionHeads * cfg.HeadDim), 1}, 0)