/* tests/test_bf16_model.c - F32 vs BF16 vs F16 loader/dispatch parity on a tiny * synthetic model. Hermetic: builds its own model dirs. Runs via `make test`. * F16 has no fused kernel: the loader widens it to F32 once at load, so an F16 * copy must match the F32 copy as closely as the BF16 copy does. */ #include "internal.h" #include "tiny_model.h" #include #include #include #include #include int main(void) { char root[1024], f32_dir[1088], bf16_dir[1088], f16_dir[1088], prefixed_dir[1088]; snprintf(root, sizeof(root), "%s/ffwd-bf16-test-XXXXXX", getenv("TMPDIR") ? getenv("TMPDIR") : "/tmp"); if (!mkdtemp(root)) { perror("mkdtemp"); return 2; } snprintf(f32_dir, sizeof(f32_dir), "%s/f32", root); snprintf(bf16_dir, sizeof(bf16_dir), "%s/bf16", root); snprintf(f16_dir, sizeof(f16_dir), "%s/f16", root); snprintf(prefixed_dir, sizeof(prefixed_dir), "%s/prefixed", root); static const tm_dims_t dims = {4, 2, 1, 2, 8, 16}; if (mkdir(f32_dir, 0755) != 0 || mkdir(bf16_dir, 0755) != 0 || mkdir(f16_dir, 0755) != 0 || mkdir(prefixed_dir, 0755) != 0 || tm_write_model(f32_dir, "F32") != 0 || tm_write_model(bf16_dir, "BF16") != 0 || tm_write_model(f16_dir, "F16") != 0 || tm_write_prefixed_model_dims(prefixed_dir, "F32", &dims) != 0) { fprintf(stderr, "fixture creation failed\n"); return 2; } int ids[] = {1, 2, 3, 4}; ffwd_model_t *mf32 = ffwd_model_load(f32_dir); ffwd_model_t *mbf16 = ffwd_model_load(bf16_dir); ffwd_model_t *mf16 = ffwd_model_load(f16_dir); ffwd_model_t *mprefixed = ffwd_model_load(prefixed_dir); if (!mf32 || !mbf16 || !mf16 || !mprefixed) { fprintf(stderr, "model load failed\n"); return 1; } const ffwd_config_t *cfg = ffwd_model_config(mf32); int dim = cfg->hidden_size; float *a = (float *)malloc((size_t)dim * sizeof(float)); float *b = (float *)malloc((size_t)dim * sizeof(float)); float *c = (float *)malloc((size_t)dim * sizeof(float)); float *e = (float *)malloc((size_t)dim * sizeof(float)); ffwd_workspace_t *wf32 = ffwd_workspace_new(mf32); ffwd_workspace_t *wbf16 = ffwd_workspace_new(mbf16); ffwd_workspace_t *wf16 = ffwd_workspace_new(mf16); ffwd_workspace_t *wprefixed = ffwd_workspace_new(mprefixed); if (!a || !b || !c || !e || !wf32 || !wbf16 || !wf16 || !wprefixed) { fprintf(stderr, "allocation failed\n"); return 1; } if (ffwd_model_encode_into(mf32, wf32, ids, 4, a) != 0 || ffwd_model_encode_into(mbf16, wbf16, ids, 4, b) != 0 || ffwd_model_encode_into(mf16, wf16, ids, 4, e) != 0 || ffwd_model_encode_into(mprefixed, wprefixed, ids, 4, c) != 0) { fprintf(stderr, "embedding failed\n"); return 1; } float max_diff = 0.0f; float f16_diff = 0.0f; float prefix_diff = 0.0f; float norm = 0.0f; for (int i = 0; i < dim; i++) { if (!isfinite(a[i]) || !isfinite(b[i]) || !isfinite(e[i]) || !isfinite(c[i])) { fprintf(stderr, "non-finite embedding value\n"); return 1; } float d = fabsf(a[i] - b[i]); if (d > max_diff) max_diff = d; d = fabsf(a[i] - e[i]); if (d > f16_diff) f16_diff = d; d = fabsf(a[i] - c[i]); if (d > prefix_diff) prefix_diff = d; norm += b[i] * b[i]; } norm = sqrtf(norm); /* Embeddings are intentionally unnormalized; require a sane nonzero * magnitude and tight F32-vs-BF16/F16 parity, not unit norm. F16 carries * more mantissa bits than BF16, so the same 2e-5 bound holds. */ if (norm <= 1e-6f || max_diff > 2e-5f || f16_diff > 2e-5f || prefix_diff != 0.0f) { fprintf(stderr, "bad parity: norm=%.9g max_diff=%.9g f16_diff=%.9g prefix_diff=%.9g\n", norm, max_diff, f16_diff, prefix_diff); return 1; } /* Sequences longer than 16 tokens take the widen-to-F32 weight path * (linear_nobias_weight); parity must hold there too. */ enum { LONG_N = 24 }; int long_ids[LONG_N]; for (int i = 0; i < LONG_N; i++) long_ids[i] = (i % 14) + 1; if (ffwd_model_encode_into(mf32, wf32, long_ids, LONG_N, a) != 0 || ffwd_model_encode_into(mbf16, wbf16, long_ids, LONG_N, b) != 0 || ffwd_model_encode_into(mf16, wf16, long_ids, LONG_N, e) != 0) { fprintf(stderr, "long-sequence embedding failed\n"); return 1; } float long_diff = 0.0f, long_f16_diff = 0.0f; for (int i = 0; i < dim; i++) { if (!isfinite(a[i]) || !isfinite(b[i]) || !isfinite(e[i])) { fprintf(stderr, "non-finite long-sequence value\n"); return 1; } float d = fabsf(a[i] - b[i]); if (d > long_diff) long_diff = d; d = fabsf(a[i] - e[i]); if (d > long_f16_diff) long_f16_diff = d; } if (long_diff > 2e-5f || long_f16_diff > 2e-5f) { fprintf(stderr, "bad widen-path parity: bf16=%.9g f16=%.9g\n", long_diff, long_f16_diff); return 1; } printf("ok: bf16/f16 and model-prefix parity dim=%d bf16_diff=%.9g f16_diff=%.9g norm=%.9g\n", dim, max_diff, f16_diff, norm); ffwd_workspace_free(wf32); ffwd_workspace_free(wbf16); ffwd_workspace_free(wf16); ffwd_workspace_free(wprefixed); ffwd_model_free(mf32); ffwd_model_free(mbf16); ffwd_model_free(mf16); ffwd_model_free(mprefixed); free(a); free(b); free(c); free(e); return 0; }