1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
|
#include "ffwd.h"
#include "model_internal.h"
#include "dtype.h"
#include "kernels.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <limits.h>
/* Below these packed-sequence lengths the fused BF16 row kernels beat widening
* the weight to F32 and dispatching through BLAS; above them, widen once. */
#define FFWD_BF16_QKV_FUSE_MAX_SEQ 16
#define FFWD_BF16_PAIR_FUSE_MAX_SEQ 4
/* ========================================================================
* Linear projections (weight-dtype dispatch) and workspace scratch growth
* ======================================================================== */
static void bf16_row_to_f32(float *dst, const uint16_t *src, int n) {
for (int i = 0; i < n; i++)
dst[i] = ffwd_bf16_to_f32(src[i]);
}
static void copy_weight_row(float *dst, const ffwd_weight_ref_t *w, size_t row, int row_width) {
if (w->dtype == DTYPE_F32) {
const float *src = (const float *)w->data + row * (size_t)row_width;
memcpy(dst, src, (size_t)row_width * sizeof(float));
} else {
const uint16_t *src = (const uint16_t *)w->data + row * (size_t)row_width;
bf16_row_to_f32(dst, src, row_width);
}
}
static int ensure_bf16_widen(ffwd_workspace_t *ws, size_t count) {
if (ws->bf16_widen_count >= count)
return 0;
float *p = (float *)realloc(ws->bf16_widen, count * sizeof(float));
if (!p)
return -1;
ws->bf16_widen = p;
ws->bf16_widen_count = count;
return 0;
}
void linear_nobias_weight(ffwd_workspace_t *ws,
float *y,
const float *x,
const ffwd_weight_ref_t *w,
int seq_len,
int in_dim,
int out_dim) {
if (w->dtype == DTYPE_F32) {
linear_nobias(y, x, (const float *)w->data, seq_len, in_dim, out_dim);
return;
}
if (seq_len > 16) {
/* Widen the weight matrix once into workspace scratch and use the
* BLAS-backed F32 path; below that the fused BF16 row kernels win. */
size_t count = (size_t)out_dim * (size_t)in_dim;
if (ensure_bf16_widen(ws, count) == 0) {
bf16_to_f32_buf(ws->bf16_widen, (const uint16_t *)w->data, count);
linear_nobias(y, x, ws->bf16_widen, seq_len, in_dim, out_dim);
return;
}
}
linear_nobias_bf16(y, x, (const uint16_t *)w->data, seq_len, in_dim, out_dim);
}
static void linear_qkv_weight(ffwd_workspace_t *ws,
float *q,
float *k,
float *v,
const float *x,
const ffwd_weight_ref_t *wq,
const ffwd_weight_ref_t *wk,
const ffwd_weight_ref_t *wv,
int seq_len,
int in_dim,
int q_dim,
int kv_dim) {
if (wq->dtype == DTYPE_BF16 && wk->dtype == DTYPE_BF16 && wv->dtype == DTYPE_BF16 &&
seq_len <= FFWD_BF16_QKV_FUSE_MAX_SEQ) {
linear_nobias_bf16_qkv(q, k, v, x, (const uint16_t *)wq->data, (const uint16_t *)wk->data,
(const uint16_t *)wv->data, seq_len, in_dim, q_dim, kv_dim);
return;
}
linear_nobias_weight(ws, q, x, wq, seq_len, in_dim, q_dim);
linear_nobias_weight(ws, k, x, wk, seq_len, in_dim, kv_dim);
linear_nobias_weight(ws, v, x, wv, seq_len, in_dim, kv_dim);
}
static void add_bias_rows(float *x, const float *bias, int rows, int dim) {
if (!bias)
return;
for (int row = 0; row < rows; row++) {
float *dst = x + (size_t)row * dim;
for (int d = 0; d < dim; d++)
dst[d] += bias[d];
}
}
static void linear_pair_weight(ffwd_workspace_t *ws,
float *a,
float *b,
const float *x,
const ffwd_weight_ref_t *wa,
const ffwd_weight_ref_t *wb,
int seq_len,
int in_dim,
int a_dim,
int b_dim) {
if (wa->dtype == DTYPE_BF16 && wb->dtype == DTYPE_BF16 &&
seq_len <= FFWD_BF16_PAIR_FUSE_MAX_SEQ) {
linear_nobias_bf16_pair(a, b, x, (const uint16_t *)wa->data, (const uint16_t *)wb->data,
seq_len, in_dim, a_dim, b_dim);
return;
}
linear_nobias_weight(ws, a, x, wa, seq_len, in_dim, a_dim);
linear_nobias_weight(ws, b, x, wb, seq_len, in_dim, b_dim);
}
/* ========================================================================
* Packed forward pass
*
* The workspace-growth helpers these call (ensure_buffers, ensure_rope_cache,
* ensure_attention_scores, ensure_offsets) live in model_blas.c beside the
* workspace lifecycle and ffwd_workspace_nbytes, whose size accounting mirrors
* them.
* ======================================================================== */
static ffwd_attention_fn attention_for_config(const ffwd_config_t *cfg) {
return cfg->attention_mode == FFWD_ATTENTION_CAUSAL
? causal_gqa_attention_packed_with_scratch
: bidirectional_gqa_attention_packed_with_scratch;
}
int build_offsets(const ffwd_input_t *inputs,
int batch,
int require_ids,
int *offsets,
int *total_out,
int *max_out) {
if (!inputs || batch <= 0 || !offsets || !total_out || !max_out)
return -1;
int total = 0;
int max_seq = 0;
for (int b = 0; b < batch; b++) {
if ((require_ids && !inputs[b].ids) || inputs[b].n_tokens <= 0)
return -1;
if (total > INT32_MAX - inputs[b].n_tokens)
return -1;
offsets[b] = total;
total += inputs[b].n_tokens;
if (inputs[b].n_tokens > max_seq)
max_seq = inputs[b].n_tokens;
}
offsets[batch] = total;
*total_out = total;
*max_out = max_seq;
return 0;
}
static int forward_packed_slice_inplace(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const ffwd_input_t *inputs,
int batch,
const int *offsets,
int total_seq,
int max_seq,
const float *input_states,
int layer_start,
int layer_end,
int apply_final_norm) {
if (!model || !ws || ws->model != model || !inputs || batch <= 0 || !offsets || total_seq <= 0 ||
max_seq <= 0)
return -1;
const ffwd_config_t *cfg = &model->config;
const ffwd_weights_t *w = &model->weights;
if (layer_start < 0 || layer_start > layer_end || layer_end > cfg->n_layers ||
layer_start < model->layer_start || layer_end > model->layer_end ||
(layer_start == 0 && input_states) || (layer_start != 0 && !input_states) ||
(layer_start == 0 && !w->embed_tokens.data) ||
(apply_final_norm && (layer_end != cfg->n_layers || !w->norm)))
return -1;
int hidden = cfg->hidden_size;
int n_heads = cfg->n_heads;
int n_kv_heads = cfg->n_kv_heads;
int head_dim = cfg->head_dim;
int q_dim = cfg->q_dim;
int kv_dim = cfg->kv_dim;
int inter = cfg->intermediate_size;
float eps = cfg->rms_norm_eps;
if (ensure_buffers(ws, cfg, total_seq) != 0)
return -1;
if (ensure_rope_cache(ws, cfg, max_seq) != 0)
return -1;
ensure_attention_scores(ws, offsets, batch);
const float *rope_cos = ws->rope_cos;
const float *rope_sin = ws->rope_sin;
float *x = ws->x;
float *x_norm = ws->x_norm;
float *q_buf = ws->q;
float *k_buf = ws->k;
float *v_buf = ws->v;
float *attn_out = ws->attn_out;
float *proj_out = ws->proj_out;
float *ffn_gate = ws->ffn_gate;
float *ffn_up = ws->ffn_up;
/* A first stage performs embedding lookup. Later stages resume from a
* packed hidden-state tensor received from the preceding stage. */
if (layer_start == 0) {
for (int b = 0; b < batch; b++) {
int start = offsets[b];
for (int i = 0; i < inputs[b].n_tokens; i++) {
int id = inputs[b].ids[i];
if (id < 0 || id >= cfg->vocab_size) {
fprintf(stderr, "ffwd: invalid token id %d at input %d position %d\n", id, b, i);
return -1;
}
copy_weight_row(x + (size_t)(start + i) * hidden, &w->embed_tokens, (size_t)id,
hidden);
}
}
} else {
memcpy(x, input_states, (size_t)total_seq * hidden * sizeof(float));
}
float scale = 1.0f / sqrtf((float)head_dim);
/* 2. Transformer layers. Linear/MLP ops run over all packed tokens;
* attention is explicitly block-diagonal by sequence offsets. */
for (int layer = layer_start; layer < layer_end; layer++) {
const ffwd_layer_t *l = &w->layers[layer];
rms_norm(x_norm, x, l->input_norm, total_seq, hidden, eps);
linear_qkv_weight(ws, q_buf, k_buf, v_buf, x_norm, &l->wq, &l->wk, &l->wv, total_seq, hidden,
q_dim, kv_dim);
add_bias_rows(q_buf, l->q_bias, total_seq, q_dim);
add_bias_rows(k_buf, l->k_bias, total_seq, kv_dim);
add_bias_rows(v_buf, l->v_bias, total_seq, kv_dim);
if (cfg->qk_norm) {
rms_norm_per_head(q_buf, l->q_norm, total_seq, n_heads, head_dim, eps);
rms_norm_per_head(k_buf, l->k_norm, total_seq, n_kv_heads, head_dim, eps);
}
for (int b = 0; b < batch; b++) {
int start = offsets[b];
int len = offsets[b + 1] - start;
apply_rope_neox(q_buf + (size_t)start * q_dim, rope_cos, rope_sin, len, n_heads,
head_dim);
apply_rope_neox(k_buf + (size_t)start * kv_dim, rope_cos, rope_sin, len, n_kv_heads,
head_dim);
}
ffwd_attention_fn attention = attention_for_config(cfg);
attention(attn_out, q_buf, k_buf, v_buf, offsets, batch, n_heads, n_kv_heads, head_dim, scale,
ws->attn_scores, ws->attn_scores_bytes);
linear_nobias_weight(ws, proj_out, attn_out, &l->wo, total_seq, q_dim, hidden);
add_inplace(x, proj_out, total_seq * hidden);
rms_norm(x_norm, x, l->post_attn_norm, total_seq, hidden, eps);
linear_pair_weight(ws, ffn_gate, ffn_up, x_norm, &l->gate_proj, &l->up_proj, total_seq,
hidden, inter, inter);
silu_mul_inplace(ffn_gate, ffn_up, total_seq * inter);
linear_nobias_weight(ws, proj_out, ffn_gate, &l->down_proj, total_seq, inter, hidden);
add_inplace(x, proj_out, total_seq * hidden);
}
if (apply_final_norm)
rms_norm(x, x, w->norm, total_seq, hidden, eps);
return 0;
}
int forward_packed_inplace(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const ffwd_input_t *inputs,
int batch,
const int *offsets,
int total_seq,
int max_seq,
int apply_final_norm) {
return forward_packed_slice_inplace(model, ws, inputs, batch, offsets, total_seq, max_seq, NULL,
0, model->config.n_layers, apply_final_norm);
}
/* BERT-family forward: token + learned-position + token-type embeddings and an
* embedding LayerNorm, then post-norm encoder layers (bias on every projection,
* GeLU feed-forward, no RoPE, no per-head norm). Bidirectional attention is
* block-diagonal over the packed batch, so each sequence's positions restart at
* cfg->position_id_offset (0 for BERT, pad_id + 1 for RoBERTa/XLM-R). Leaves the
* final hidden states in ws->x; pooling reads them as-is because the last
* layer's LayerNorm is already the model's final normalization. */
static int forward_packed_bert(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const ffwd_input_t *inputs,
int batch,
const int *offsets,
int total_seq,
int max_seq) {
if (!model || !ws || ws->model != model || !inputs || batch <= 0 || !offsets || total_seq <= 0 ||
max_seq <= 0)
return -1;
const ffwd_config_t *cfg = &model->config;
const ffwd_weights_t *w = &model->weights;
int hidden = cfg->hidden_size;
int n_heads = cfg->n_heads;
int head_dim = cfg->head_dim;
int inter = cfg->intermediate_size;
float eps = cfg->layer_norm_eps;
if (!w->embed_tokens.data || !w->position_embeddings || !w->token_type_embeddings ||
!w->ffwd_ln_w || !w->ffwd_ln_b)
return -1;
if (max_seq > cfg->max_position_embeddings - cfg->position_id_offset) {
fprintf(stderr,
"ffwd: sequence length %d plus position offset %d exceeds max_position_embeddings "
"%d\n",
max_seq, cfg->position_id_offset, cfg->max_position_embeddings);
return -1;
}
if (ensure_buffers(ws, cfg, total_seq) != 0)
return -1;
ensure_attention_scores(ws, offsets, batch);
float *x = ws->x;
float *q_buf = ws->q, *k_buf = ws->k, *v_buf = ws->v;
float *attn_out = ws->attn_out, *proj_out = ws->proj_out, *ffn_up = ws->ffn_up;
/* Embeddings: token + learned absolute position + token-type[0], then the
* embedding LayerNorm over every packed token. */
for (int b = 0; b < batch; b++) {
int start = offsets[b];
for (int i = 0; i < inputs[b].n_tokens; i++) {
int id = inputs[b].ids[i];
if (id < 0 || id >= cfg->vocab_size) {
fprintf(stderr, "ffwd: invalid token id %d at input %d position %d\n", id, b, i);
return -1;
}
float *row = x + (size_t)(start + i) * hidden;
copy_weight_row(row, &w->embed_tokens, (size_t)id, hidden);
int pos = cfg->position_id_offset + i;
add_inplace(row, w->position_embeddings + (size_t)pos * hidden, hidden);
add_inplace(row, w->token_type_embeddings, hidden);
}
}
layer_norm(x, x, w->ffwd_ln_w, w->ffwd_ln_b, total_seq, hidden, eps);
float scale = 1.0f / sqrtf((float)head_dim);
for (int layer = 0; layer < cfg->n_layers; layer++) {
const ffwd_layer_t *l = &w->layers[layer];
/* Self-attention (post-norm): projections read x directly. */
linear_qkv_weight(ws, q_buf, k_buf, v_buf, x, &l->wq, &l->wk, &l->wv, total_seq, hidden,
hidden, hidden);
add_bias_rows(q_buf, l->q_bias, total_seq, hidden);
add_bias_rows(k_buf, l->k_bias, total_seq, hidden);
add_bias_rows(v_buf, l->v_bias, total_seq, hidden);
ffwd_attention_fn attention = attention_for_config(cfg);
attention(attn_out, q_buf, k_buf, v_buf, offsets, batch, n_heads, n_heads, head_dim, scale,
ws->attn_scores, ws->attn_scores_bytes);
linear_nobias_weight(ws, proj_out, attn_out, &l->wo, total_seq, hidden, hidden);
add_bias_rows(proj_out, l->o_bias, total_seq, hidden);
add_inplace(x, proj_out, total_seq * hidden);
layer_norm(x, x, l->input_norm, l->attn_ln_bias, total_seq, hidden, eps);
/* Feed-forward (post-norm): dense -> GeLU -> dense. */
linear_nobias_weight(ws, ffn_up, x, &l->up_proj, total_seq, hidden, inter);
add_bias_rows(ffn_up, l->ffn_inter_bias, total_seq, inter);
if (cfg->ffn_act == FFWD_ACT_GELU_TANH)
gelu_tanh_inplace(ffn_up, total_seq * inter);
else
gelu_inplace(ffn_up, total_seq * inter);
linear_nobias_weight(ws, proj_out, ffn_up, &l->down_proj, total_seq, inter, hidden);
add_bias_rows(proj_out, l->ffn_out_bias, total_seq, hidden);
add_inplace(x, proj_out, total_seq * hidden);
layer_norm(x, x, l->post_attn_norm, l->ffn_ln_bias, total_seq, hidden, eps);
}
return 0;
}
static int pool_embeddings(const ffwd_model_t *model,
const ffwd_workspace_t *ws,
const int *offsets,
int batch,
float *out_embeddings) {
const ffwd_config_t *cfg = &model->config;
int hidden = cfg->hidden_size;
float eps = cfg->rms_norm_eps;
/* Qwen pools pre-final-norm states and applies the final RMSNorm here; BERT
* pools already-normed states (no final norm), signalled by a NULL norm. */
const float *norm_weight = model->weights.norm;
const float *x = ws->x;
if (hidden <= 0 || !x)
return -1;
for (int b = 0; b < batch; b++) {
int start = offsets[b];
int end = offsets[b + 1];
int len = end - start;
float *emb = out_embeddings + (size_t)b * hidden;
if (cfg->pooling_mode == FFWD_POOL_LAST_TOKEN || cfg->pooling_mode == FFWD_POOL_CLS) {
/* Single-token pooling: CLS takes the first token, last-token the
* final one. */
int idx = cfg->pooling_mode == FFWD_POOL_CLS ? start : end - 1;
const float *row = x + (size_t)idx * hidden;
if (norm_weight) {
float sum_sq = 0.0f;
for (int d = 0; d < hidden; d++)
sum_sq += row[d] * row[d];
float rms_inv = 1.0f / sqrtf(sum_sq / (float)hidden + eps);
for (int d = 0; d < hidden; d++)
emb[d] = row[d] * rms_inv * norm_weight[d];
} else {
memcpy(emb, row, (size_t)hidden * sizeof(float));
}
} else {
memset(emb, 0, (size_t)hidden * sizeof(float));
for (int i = start; i < end; i++) {
const float *row = x + (size_t)i * hidden;
if (norm_weight) {
float sum_sq = 0.0f;
for (int d = 0; d < hidden; d++)
sum_sq += row[d] * row[d];
float rms_inv = 1.0f / sqrtf(sum_sq / (float)hidden + eps);
for (int d = 0; d < hidden; d++)
emb[d] += row[d] * rms_inv * norm_weight[d];
} else {
for (int d = 0; d < hidden; d++)
emb[d] += row[d];
}
}
float inv_len = 1.0f / (float)len;
for (int d = 0; d < hidden; d++)
emb[d] *= inv_len;
}
if (cfg->normalize_embeddings && ffwd_l2_normalize(emb, hidden) != 0)
return -1;
}
return 0;
}
/* ========================================================================
* Public forward/embed APIs
* ======================================================================== */
int ffwd_model_forward_into(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const int *token_ids,
int n_tokens,
float *out_states) {
if (!model || !ws || !token_ids || n_tokens <= 0 || !out_states)
return -1;
size_t count, bytes;
if (mul_size((size_t)n_tokens, (size_t)model->config.hidden_size, &count) != 0 ||
mul_size(count, sizeof(float), &bytes) != 0)
return -1;
ffwd_input_t input = {token_ids, n_tokens};
int offsets[2] = {0, n_tokens};
if (forward_packed_inplace(model, ws, &input, 1, offsets, n_tokens, n_tokens, 1) != 0)
return -1;
memcpy(out_states, ws->x, bytes);
return 0;
}
int ffwd_pool_spans(const ffwd_config_t *cfg,
const float *states,
int n_tokens,
const ffwd_span_t *spans,
int n_spans,
float *out_embeddings) {
if (!cfg || !states || n_tokens <= 0 || !spans || n_spans <= 0 || !out_embeddings)
return -1;
int hidden = cfg->hidden_size;
if (hidden <= 0)
return -1;
for (int s = 0; s < n_spans; s++) {
int start = spans[s].start;
int len = spans[s].n_tokens;
if (start < 0 || len <= 0 || start > n_tokens || len > n_tokens - start)
return -1;
float *emb = out_embeddings + (size_t)s * hidden;
memset(emb, 0, (size_t)hidden * sizeof(float));
for (int t = 0; t < len; t++) {
const float *row = states + (size_t)(start + t) * hidden;
for (int d = 0; d < hidden; d++)
emb[d] += row[d];
}
float inv_len = 1.0f / (float)len;
for (int d = 0; d < hidden; d++) {
emb[d] *= inv_len;
}
}
return 0;
}
int ffwd_model_encode_spans_batch(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const ffwd_context_input_t *inputs,
int batch,
float *out_embeddings) {
if (!model || !ws || !inputs || batch <= 0 || !out_embeddings)
return -1;
if (ensure_offsets(ws, batch) != 0)
return -1;
if ((size_t)batch > SIZE_MAX / sizeof(ffwd_input_t))
return -1;
ffwd_input_t *packed_inputs = malloc((size_t)batch * sizeof(*packed_inputs));
if (!packed_inputs)
return -1;
int *offsets = ws->offsets;
int total_seq = 0;
int max_seq = 0;
int total_spans = 0;
for (int b = 0; b < batch; b++) {
const ffwd_context_input_t *input = &inputs[b];
int n_tokens = input->input.n_tokens;
if (!input->input.ids || n_tokens <= 0 || !input->spans || input->n_spans <= 0 ||
total_seq > INT_MAX - n_tokens || total_spans > INT_MAX - input->n_spans) {
free(packed_inputs);
return -1;
}
packed_inputs[b] = input->input;
offsets[b] = total_seq;
total_seq += n_tokens;
total_spans += input->n_spans;
if (n_tokens > max_seq)
max_seq = n_tokens;
for (int s = 0; s < input->n_spans; s++) {
int start = input->spans[s].start;
int len = input->spans[s].n_tokens;
if (start < 0 || len <= 0 || start > n_tokens || len > n_tokens - start) {
free(packed_inputs);
return -1;
}
}
}
offsets[batch] = total_seq;
if (forward_packed_inplace(model, ws, packed_inputs, batch, offsets, total_seq, max_seq, 1) !=
0) {
free(packed_inputs);
return -1;
}
free(packed_inputs);
int span_offset = 0;
int hidden = model->config.hidden_size;
for (int b = 0; b < batch; b++) {
const ffwd_context_input_t *input = &inputs[b];
if (ffwd_pool_spans(&model->config, ws->x + (size_t)offsets[b] * hidden,
input->input.n_tokens, input->spans, input->n_spans,
out_embeddings + (size_t)span_offset * hidden) != 0)
return -1;
span_offset += input->n_spans;
}
return 0;
}
int ffwd_model_encode_spans(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const int *token_ids,
int n_tokens,
const ffwd_span_t *spans,
int n_spans,
float *out_embeddings) {
ffwd_context_input_t input = {
.input = {token_ids, n_tokens},
.spans = spans,
.n_spans = n_spans,
};
return ffwd_model_encode_spans_batch(model, ws, &input, 1, out_embeddings);
}
float *ffwd_model_forward(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const int *token_ids,
int n_tokens) {
if (!model || !ws || !token_ids || n_tokens <= 0)
return NULL;
size_t n;
if (mul_size((size_t)n_tokens, (size_t)model->config.hidden_size, &n) != 0)
return NULL;
float *out = malloc_floats(n);
if (!out)
return NULL;
if (ffwd_model_forward_into(model, ws, token_ids, n_tokens, out) != 0) {
free(out);
return NULL;
}
return out;
}
int ffwd_model_encode_batch(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const ffwd_input_t *inputs,
int batch,
float *out_embeddings) {
if (!model || !ws || !inputs || batch <= 0 || !out_embeddings)
return -1;
if (ensure_offsets(ws, batch) != 0)
return -1;
int *offsets = ws->offsets;
int total_seq = 0;
int max_seq = 0;
if (build_offsets(inputs, batch, 1, offsets, &total_seq, &max_seq) != 0)
return -1;
int rc = model->config.family == FFWD_FAMILY_BERT
? forward_packed_bert(model, ws, inputs, batch, offsets, total_seq, max_seq)
: forward_packed_inplace(model, ws, inputs, batch, offsets, total_seq, max_seq, 0);
if (rc == 0)
rc = pool_embeddings(model, ws, offsets, batch, out_embeddings);
return rc;
}
int ffwd_pool_batch(const ffwd_config_t *cfg,
const float *states,
const int *seq_lengths,
int batch,
float *out_embeddings) {
if (!cfg || !states || !seq_lengths || batch <= 0 || !out_embeddings || cfg->hidden_size <= 0)
return -1;
int offset = 0;
for (int b = 0; b < batch; b++) {
int len = seq_lengths[b];
if (len <= 0 || offset > INT_MAX - len)
return -1;
float *emb = out_embeddings + (size_t)b * cfg->hidden_size;
if (cfg->pooling_mode == FFWD_POOL_LAST_TOKEN || cfg->pooling_mode == FFWD_POOL_CLS) {
/* CLS takes the first token, last-token the final one; these states
* already carry the model's final RMSNorm. */
int idx = cfg->pooling_mode == FFWD_POOL_CLS ? offset : offset + len - 1;
const float *row = states + (size_t)idx * cfg->hidden_size;
memcpy(emb, row, (size_t)cfg->hidden_size * sizeof(float));
} else {
memset(emb, 0, (size_t)cfg->hidden_size * sizeof(float));
for (int t = 0; t < len; t++) {
const float *row = states + (size_t)(offset + t) * cfg->hidden_size;
for (int d = 0; d < cfg->hidden_size; d++)
emb[d] += row[d];
}
float inv_len = 1.0f / (float)len;
for (int d = 0; d < cfg->hidden_size; d++)
emb[d] *= inv_len;
}
if (cfg->normalize_embeddings && ffwd_l2_normalize(emb, cfg->hidden_size) != 0)
return -1;
offset += len;
}
return 0;
}
int ffwd_model_encode_into(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const int *token_ids,
int n_tokens,
float *out_embedding) {
ffwd_input_t input = {token_ids, n_tokens};
return ffwd_model_encode_batch(model, ws, &input, 1, out_embedding);
}
float *ffwd_model_encode(const ffwd_model_t *model,
ffwd_workspace_t *ws,
const int *token_ids,
int n_tokens) {
if (!model || !ws || !token_ids || n_tokens <= 0)
return NULL;
int hidden = model->config.hidden_size;
if (hidden <= 0)
return NULL;
float *emb = malloc_floats((size_t)hidden);
if (!emb)
return NULL;
if (ffwd_model_encode_into(model, ws, token_ids, n_tokens, emb) != 0) {
free(emb);
return NULL;
}
return emb;
}
|