Repository navigation
Expand file tree
/
Copy pathtext_encoder.cpp
More file actions
265 lines (228 loc) · 8.86 KB
/
Copy pathtext_encoder.cpp
File metadata and controls
265 lines (228 loc) · 8.86 KB
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
#include "sam3/text_encoder.h"
#include "ggml-alloc.h"
#include <cmath>
#include <stdexcept>
#include <vector>
namespace sam3 {
namespace {
constexpr int32_t kTextWidth = 1024;
constexpr int32_t kTextHeads = 16;
constexpr int32_t kTextLayers = 24;
constexpr float kLayerNormEps = 1e-5f;
ggml_tensor * require_tensor(const GgufModel & model, const std::string & name) {
ggml_tensor * tensor = model.find_weight(name);
if (tensor == nullptr) {
throw std::runtime_error("missing tensor: " + name);
}
return tensor;
}
ggml_tensor * ensure_f32(ggml_context * ctx, ggml_tensor * tensor) {
if (tensor->type == GGML_TYPE_F32) {
return tensor;
}
return ggml_cast(ctx, tensor, GGML_TYPE_F32);
}
ggml_tensor * layer_norm(
ggml_context * ctx,
ggml_tensor * x,
ggml_tensor * weight,
ggml_tensor * bias
) {
ggml_tensor * y = ggml_norm(ctx, ensure_f32(ctx, x), kLayerNormEps);
y = ggml_mul(ctx, y, ensure_f32(ctx, weight));
y = ggml_add(ctx, y, ensure_f32(ctx, bias));
return y;
}
ggml_tensor * linear(
ggml_context * ctx,
ggml_tensor * weight,
ggml_tensor * bias,
ggml_tensor * x
) {
ggml_tensor * y = ggml_mul_mat(ctx, weight, x);
y = ensure_f32(ctx, y);
if (bias != nullptr) {
y = ggml_add(ctx, y, ensure_f32(ctx, bias));
}
return y;
}
ggml_tensor * self_attention(
ggml_context * ctx,
ggml_tensor * q,
ggml_tensor * k,
ggml_tensor * v,
int32_t seq_len
) {
const int32_t head_dim = kTextWidth / kTextHeads;
ggml_tensor * qh = ggml_permute(ctx, ggml_cont_3d(ctx, q, head_dim, kTextHeads, seq_len), 0, 2, 1, 3);
ggml_tensor * kh = ggml_permute(ctx, ggml_cont_3d(ctx, k, head_dim, kTextHeads, seq_len), 0, 2, 1, 3);
ggml_tensor * vh = ggml_cont_3d(
ctx,
ggml_permute(ctx, ggml_cont_3d(ctx, v, head_dim, kTextHeads, seq_len), 1, 2, 0, 3),
seq_len,
head_dim,
kTextHeads
);
ggml_tensor * kq = ensure_f32(ctx, ggml_mul_mat(ctx, kh, qh));
ggml_tensor * scaled = ggml_scale(ctx, kq, 1.0f / std::sqrt(static_cast<float>(head_dim)));
ggml_tensor * masked = ggml_diag_mask_inf(ctx, scaled, 0);
ggml_tensor * probs = ggml_soft_max(ctx, masked);
ggml_tensor * kqv = ensure_f32(ctx, ggml_mul_mat(ctx, vh, probs));
ggml_tensor * merged = ggml_permute(ctx, kqv, 0, 2, 1, 3);
return ggml_cont_2d(ctx, merged, kTextWidth, seq_len);
}
std::string block_prefix(int layer) {
return "backbone.language_backbone.encoder.transformer.resblocks." + std::to_string(layer);
}
} // namespace
TextEncoder::TextEncoder(const GgufModel & model) : model_(model) {}
TextEncodeOutput TextEncoder::encode(const std::vector<int32_t> & tokens) const {
if (tokens.empty()) {
throw std::runtime_error("no tokens provided");
}
const int32_t seq_len = static_cast<int32_t>(tokens.size());
std::vector<int32_t> positions(tokens.size());
for (int32_t i = 0; i < seq_len; ++i) {
positions[i] = i;
}
const size_t graph_size = 32768;
const size_t ctx_size =
ggml_tensor_overhead() * graph_size +
ggml_graph_overhead_custom(graph_size, false);
std::vector<uint8_t> ctx_buf(ctx_size);
ggml_init_params params {
/*.mem_size =*/ ctx_size,
/*.mem_buffer =*/ ctx_buf.data(),
/*.no_alloc =*/ true,
};
ggml_context * ctx = ggml_init(params);
if (ctx == nullptr) {
throw std::runtime_error("ggml_init failed");
}
ggml_backend_t backend = model_.backend();
const bool is_cpu = ggml_backend_dev_type(ggml_backend_get_device(backend)) == GGML_BACKEND_DEVICE_TYPE_CPU;
ggml_backend_t cpu_backend = is_cpu ? backend : model_.cpu_backend();
ggml_backend_t backends[2] = { backend, cpu_backend };
const int n_backends = (is_cpu || cpu_backend == nullptr) ? 1 : 2;
ggml_backend_sched_t sched = ggml_backend_sched_new(backends, nullptr, n_backends, graph_size, false, true);
if (sched == nullptr) {
ggml_free(ctx);
throw std::runtime_error("failed to create backend scheduler");
}
ggml_cgraph * gf = ggml_new_graph_custom(ctx, graph_size, false);
ggml_tensor * token_ids = ggml_new_tensor_1d(ctx, GGML_TYPE_I32, seq_len);
ggml_set_name(token_ids, "token_ids");
ggml_backend_sched_set_tensor_backend(sched, token_ids, cpu_backend);
ggml_tensor * pos_ids = ggml_new_tensor_1d(ctx, GGML_TYPE_I32, seq_len);
ggml_set_name(pos_ids, "position_ids");
ggml_backend_sched_set_tensor_backend(sched, pos_ids, cpu_backend);
ggml_tensor * tok = ggml_get_rows(ctx, require_tensor(model_, "backbone.language_backbone.encoder.token_embedding.weight"), token_ids);
ggml_tensor * pos = ggml_get_rows(ctx, require_tensor(model_, "backbone.language_backbone.encoder.positional_embedding"), pos_ids);
ggml_tensor * cur = ggml_add(ctx, tok, pos);
const bool mixed_backend = n_backends > 1;
for (int layer = 0; layer < kTextLayers; ++layer) {
const std::string prefix = block_prefix(layer);
ggml_tensor * ln1 = layer_norm(
ctx,
cur,
require_tensor(model_, prefix + ".ln_1.weight"),
require_tensor(model_, prefix + ".ln_1.bias")
);
if (mixed_backend) {
ggml_backend_sched_set_tensor_backend(sched, ln1, cpu_backend);
}
ggml_tensor * q = linear(
ctx,
require_tensor(model_, prefix + ".attn.query_proj.weight"),
require_tensor(model_, prefix + ".attn.query_proj.bias"),
ln1
);
ggml_tensor * k = linear(
ctx,
require_tensor(model_, prefix + ".attn.key_proj.weight"),
require_tensor(model_, prefix + ".attn.key_proj.bias"),
ln1
);
ggml_tensor * v = linear(
ctx,
require_tensor(model_, prefix + ".attn.value_proj.weight"),
require_tensor(model_, prefix + ".attn.value_proj.bias"),
ln1
);
ggml_tensor * attn = self_attention(ctx, q, k, v, seq_len);
attn = linear(
ctx,
require_tensor(model_, prefix + ".attn.out_proj.weight"),
require_tensor(model_, prefix + ".attn.out_proj.bias"),
attn
);
cur = ggml_add(ctx, cur, attn);
ggml_tensor * ln2 = layer_norm(
ctx,
cur,
require_tensor(model_, prefix + ".ln_2.weight"),
require_tensor(model_, prefix + ".ln_2.bias")
);
if (mixed_backend) {
ggml_backend_sched_set_tensor_backend(sched, ln2, cpu_backend);
}
ggml_tensor * mlp = linear(
ctx,
require_tensor(model_, prefix + ".mlp.c_fc.weight"),
require_tensor(model_, prefix + ".mlp.c_fc.bias"),
ln2
);
mlp = ggml_gelu_erf(ctx, mlp);
if (mixed_backend) {
ggml_backend_sched_set_tensor_backend(sched, mlp, cpu_backend);
}
mlp = linear(
ctx,
require_tensor(model_, prefix + ".mlp.c_proj.weight"),
require_tensor(model_, prefix + ".mlp.c_proj.bias"),
mlp
);
cur = ggml_add(ctx, cur, mlp);
}
ggml_tensor * ln_final = layer_norm(
ctx,
cur,
require_tensor(model_, "backbone.language_backbone.encoder.ln_final.weight"),
require_tensor(model_, "backbone.language_backbone.encoder.ln_final.bias")
);
if (mixed_backend) {
ggml_backend_sched_set_tensor_backend(sched, ln_final, cpu_backend);
}
ggml_tensor * resized = linear(
ctx,
require_tensor(model_, "backbone.language_backbone.resizer.weight"),
require_tensor(model_, "backbone.language_backbone.resizer.bias"),
ln_final
);
ggml_build_forward_expand(gf, resized);
ggml_backend_sched_alloc_graph(sched, gf);
ggml_backend_tensor_set(token_ids, tokens.data(), 0, tokens.size() * sizeof(int32_t));
ggml_backend_tensor_set(pos_ids, positions.data(), 0, positions.size() * sizeof(int32_t));
const ggml_status status = ggml_backend_sched_graph_compute(sched, gf);
if (status != GGML_STATUS_SUCCESS) {
ggml_backend_sched_free(sched);
ggml_free(ctx);
throw std::runtime_error("graph compute failed");
}
std::vector<float> raw(static_cast<size_t>(ggml_nelements(resized)));
ggml_backend_tensor_get(resized, raw.data(), 0, ggml_nbytes(resized));
TextEncodeOutput output;
output.seq_len = seq_len;
output.hidden_dim = 256;
output.memory.resize(static_cast<size_t>(seq_len) * output.hidden_dim);
for (int32_t s = 0; s < seq_len; ++s) {
for (int32_t d = 0; d < output.hidden_dim; ++d) {
output.memory[static_cast<size_t>(s) * output.hidden_dim + d] =
raw[static_cast<size_t>(s) * output.hidden_dim + d];
}
}
ggml_backend_sched_free(sched);
ggml_free(ctx);
return output;
}
} // namespace sam3