transcribe-cpp-sys 0.0.2

Native FFI bindings for transcribe.cpp, a C/C++ speech-to-text library built on ggml
Documentation 
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

// arch/gigaam/weights.cpp - read_gigaam_hparams + build_gigaam_weights.
//
// Pattern follows arch/parakeet/weights.cpp: read every required KV
// from the GGUF, then validate the tensor catalog by name + shape using
// transcribe::weights::find_tensor.

#include "weights.h"

#include "transcribe-meta.h"
#include "transcribe-weights-util.h"
#include "transcribe-log.h"

#include "ggml.h"
#include "gguf.h"

#include <cstdio>
#include <cstring>

namespace transcribe::gigaam {

namespace {

constexpr const char * kFamilyTag = "gigaam";

} // namespace

transcribe_status read_gigaam_hparams(const gguf_context * gguf,
                                      GigaamHParams &      hp)
{
    if (gguf == nullptr) return TRANSCRIBE_ERR_INVALID_ARG;

    // Head discriminator (drives RNN-T vs CTC reads below).
    {
        std::string head_kind_str;
        if (auto st = read_required_string_kv(gguf, "stt.gigaam.head_kind",
                                              kFamilyTag, head_kind_str);
            st != TRANSCRIBE_OK) return st;
        if (head_kind_str == "rnnt") {
            hp.head_kind = HeadKind::RNNT;
        } else if (head_kind_str == "ctc") {
            hp.head_kind = HeadKind::CTC;
        } else {
            log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                         "gigaam: unsupported stt.gigaam.head_kind=%s",
                         head_kind_str.c_str());
            return TRANSCRIBE_ERR_GGUF;
        }
    }

    // Encoder.
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.n_layers",           kFamilyTag, hp.enc_n_layers);           st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.d_model",            kFamilyTag, hp.enc_d_model);            st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.n_heads",            kFamilyTag, hp.enc_n_heads);            st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.d_ff",               kFamilyTag, hp.enc_d_ff);               st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.conv_kernel",        kFamilyTag, hp.enc_conv_kernel);        st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.subsampling_factor", kFamilyTag, hp.enc_subsampling_factor); st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.subs_kernel_size",   kFamilyTag, hp.enc_subs_kernel_size);   st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.pos_emb_max_len",    kFamilyTag, hp.enc_pos_emb_max_len);    st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv(gguf, "stt.gigaam.encoder.feat_in",            kFamilyTag, hp.enc_feat_in);            st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_string_kv(gguf, "stt.gigaam.encoder.self_attention_model", kFamilyTag, hp.enc_self_attention_model); st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_string_kv(gguf, "stt.gigaam.encoder.conv_norm_type",  kFamilyTag, hp.enc_conv_norm_type);  st != TRANSCRIBE_OK) return st;

    if (hp.enc_n_heads <= 0 || hp.enc_d_model % hp.enc_n_heads != 0) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                     "gigaam: invariant d_model %% n_heads != 0 "
                     "(d_model=%d, n_heads=%d)",
                     hp.enc_d_model, hp.enc_n_heads);
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.enc_self_attention_model != "rotary") {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                     "gigaam: unsupported self_attention_model=%s (expected rotary)",
                     hp.enc_self_attention_model.c_str());
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.enc_conv_norm_type != "layer_norm") {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                     "gigaam: unsupported conv_norm_type=%s (expected layer_norm)",
                     hp.enc_conv_norm_type.c_str());
        return TRANSCRIBE_ERR_GGUF;
    }

    if (hp.head_kind == HeadKind::RNNT) {
        if (auto st = read_required_u32_kv(gguf, "stt.gigaam.predictor.hidden",   kFamilyTag, hp.pred_hidden);   st != TRANSCRIBE_OK) return st;
        if (auto st = read_required_u32_kv(gguf, "stt.gigaam.predictor.n_layers", kFamilyTag, hp.pred_n_layers); st != TRANSCRIBE_OK) return st;
        if (auto st = read_required_u32_kv(gguf, "stt.gigaam.predictor.vocab",    kFamilyTag, hp.pred_vocab);    st != TRANSCRIBE_OK) return st;
        if (auto st = read_required_u32_kv(gguf, "stt.gigaam.joint.hidden",       kFamilyTag, hp.joint_hidden);  st != TRANSCRIBE_OK) return st;
        if (auto st = read_required_u32_kv(gguf, "stt.gigaam.joint.num_classes",  kFamilyTag, hp.joint_n_classes); st != TRANSCRIBE_OK) return st;
        if (auto st = read_required_string_kv(gguf, "stt.gigaam.joint.activation", kFamilyTag, hp.joint_activation); st != TRANSCRIBE_OK) return st;
        if (hp.pred_vocab != hp.joint_n_classes) {
            log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                         "gigaam: pred_vocab (%d) != joint_n_classes (%d)",
                         hp.pred_vocab, hp.joint_n_classes);
            return TRANSCRIBE_ERR_GGUF;
        }
    } else { // CTC
        if (auto st = read_required_u32_kv(gguf, "stt.gigaam.head.feat_in",     kFamilyTag, hp.head_feat_in);   st != TRANSCRIBE_OK) return st;
        if (auto st = read_required_u32_kv(gguf, "stt.gigaam.head.num_classes", kFamilyTag, hp.head_n_classes); st != TRANSCRIBE_OK) return st;
        if (hp.head_feat_in != hp.enc_d_model) {
            log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                         "gigaam: ctc head.feat_in (%d) != encoder.d_model (%d)",
                         hp.head_feat_in, hp.enc_d_model);
            return TRANSCRIBE_ERR_GGUF;
        }
    }

    // Frontend.
    if (auto st = read_required_string_kv(gguf, "stt.frontend.type",        kFamilyTag, hp.fe_type);        st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv   (gguf, "stt.frontend.num_mels",    kFamilyTag, hp.fe_num_mels);    st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv   (gguf, "stt.frontend.sample_rate", kFamilyTag, hp.fe_sample_rate); st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv   (gguf, "stt.frontend.n_fft",       kFamilyTag, hp.fe_n_fft);       st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv   (gguf, "stt.frontend.win_length",  kFamilyTag, hp.fe_win_length);  st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_u32_kv   (gguf, "stt.frontend.hop_length",  kFamilyTag, hp.fe_hop_length);  st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_string_kv(gguf, "stt.frontend.window",      kFamilyTag, hp.fe_window);      st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_string_kv(gguf, "stt.frontend.normalize",   kFamilyTag, hp.fe_normalize);   st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_f32_kv   (gguf, "stt.frontend.dither",      kFamilyTag, hp.fe_dither);      st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_f32_kv   (gguf, "stt.frontend.pre_emphasis",kFamilyTag, hp.fe_pre_emphasis);st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_f32_kv   (gguf, "stt.frontend.f_min",       kFamilyTag, hp.fe_f_min);       st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_f32_kv   (gguf, "stt.frontend.f_max",       kFamilyTag, hp.fe_f_max);       st != TRANSCRIBE_OK) return st;
    if (auto st = read_optional_bool_kv  (gguf, "stt.frontend.center",      kFamilyTag, false, hp.fe_center); st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_string_kv(gguf, "stt.frontend.mel_norm",    kFamilyTag, hp.fe_mel_norm);    st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_f32_kv   (gguf, "stt.frontend.log_clamp_min", kFamilyTag, hp.fe_log_clamp_min); st != TRANSCRIBE_OK) return st;
    if (auto st = read_required_f32_kv   (gguf, "stt.frontend.log_clamp_max", kFamilyTag, hp.fe_log_clamp_max); st != TRANSCRIBE_OK) return st;

    return TRANSCRIBE_OK;
}

// ---------------------------------------------------------------------------
// Weights catalog
// ---------------------------------------------------------------------------

#define GET_LN(slot, name) do { \
    ggml_tensor * _t = transcribe::weights::find_tensor( \
        ctx_meta, (name), {GGML_TYPE_F32}, \
        {hp.enc_d_model}, kFamilyTag); \
    if (_t == nullptr) return TRANSCRIBE_ERR_GGUF; \
    (slot) = _t; \
} while (0)

#define GET_LIN_W(slot, name, in_dim, out_dim) do { \
    ggml_tensor * _t = transcribe::weights::find_tensor( \
        ctx_meta, (name), {TRANSCRIBE_QUANT_LINEAR_TYPES}, \
        {(in_dim), (out_dim)}, kFamilyTag); \
    if (_t == nullptr) return TRANSCRIBE_ERR_GGUF; \
    (slot) = _t; \
} while (0)

#define GET_LIN_B(slot, name, out_dim) do { \
    ggml_tensor * _t = transcribe::weights::find_tensor( \
        ctx_meta, (name), {GGML_TYPE_F32}, \
        {(out_dim)}, kFamilyTag); \
    if (_t == nullptr) return TRANSCRIBE_ERR_GGUF; \
    (slot) = _t; \
} while (0)

#define GET_CONV_W(slot, name, k, in_dim, out_dim) do { \
    ggml_tensor * _t = transcribe::weights::find_tensor( \
        ctx_meta, (name), {TRANSCRIBE_QUANT_CONV_TYPES}, \
        {(k), (in_dim), (out_dim)}, kFamilyTag); \
    if (_t == nullptr) return TRANSCRIBE_ERR_GGUF; \
    (slot) = _t; \
} while (0)

#define GET_F32_1D(slot, name, n) do { \
    ggml_tensor * _t = transcribe::weights::find_tensor( \
        ctx_meta, (name), {GGML_TYPE_F32}, \
        {(n)}, kFamilyTag); \
    if (_t == nullptr) return TRANSCRIBE_ERR_GGUF; \
    (slot) = _t; \
} while (0)

#define GET_F32_2D(slot, name, n0, n1) do { \
    ggml_tensor * _t = transcribe::weights::find_tensor( \
        ctx_meta, (name), {GGML_TYPE_F32}, \
        {(n0), (n1)}, kFamilyTag); \
    if (_t == nullptr) return TRANSCRIBE_ERR_GGUF; \
    (slot) = _t; \
} while (0)

using transcribe::weights::lname;

namespace {

transcribe_status build_block(ggml_context *        ctx_meta,
                              const GigaamHParams & hp,
                              int                   i,
                              GigaamBlock &         b)
{
    const int d = hp.enc_d_model;
    const int dff = hp.enc_d_ff;
    const int dw_k = hp.enc_conv_kernel;

    // Macaron FF1.
    GET_LN  (b.norm_ff1_w, lname("enc.blocks.%d.norm_ff1.weight", i));
    GET_LN  (b.norm_ff1_b, lname("enc.blocks.%d.norm_ff1.bias",   i));
    GET_LIN_W(b.ff1_lin1_w, lname("enc.blocks.%d.ff1.linear1.weight", i), d,   dff);
    GET_LIN_B(b.ff1_lin1_b, lname("enc.blocks.%d.ff1.linear1.bias",   i), dff);
    GET_LIN_W(b.ff1_lin2_w, lname("enc.blocks.%d.ff1.linear2.weight", i), dff, d);
    GET_LIN_B(b.ff1_lin2_b, lname("enc.blocks.%d.ff1.linear2.bias",   i), d);

    // Conv module.
    GET_LN  (b.norm_conv_w, lname("enc.blocks.%d.norm_conv.weight", i));
    GET_LN  (b.norm_conv_b, lname("enc.blocks.%d.norm_conv.bias",   i));
    GET_CONV_W(b.conv_pw1_w, lname("enc.blocks.%d.conv.pointwise1.weight", i), 1, d, 2 * d);
    GET_LIN_B (b.conv_pw1_b, lname("enc.blocks.%d.conv.pointwise1.bias",   i), 2 * d);
    GET_CONV_W(b.conv_dw_w,  lname("enc.blocks.%d.conv.depthwise.weight", i), dw_k, 1, d);
    GET_LIN_B (b.conv_dw_b,  lname("enc.blocks.%d.conv.depthwise.bias",   i), d);
    GET_LN  (b.conv_ln_w,    lname("enc.blocks.%d.conv.ln.weight", i));
    GET_LN  (b.conv_ln_b,    lname("enc.blocks.%d.conv.ln.bias",   i));
    GET_CONV_W(b.conv_pw2_w, lname("enc.blocks.%d.conv.pointwise2.weight", i), 1, d, d);
    GET_LIN_B (b.conv_pw2_b, lname("enc.blocks.%d.conv.pointwise2.bias",   i), d);

    // Self-attention (rotary, no linear_pos / pos_bias_*).
    GET_LN  (b.norm_attn_w, lname("enc.blocks.%d.norm_attn.weight", i));
    GET_LN  (b.norm_attn_b, lname("enc.blocks.%d.norm_attn.bias",   i));
    GET_LIN_W(b.attn_q_w,   lname("enc.blocks.%d.attn.linear_q.weight",   i), d, d);
    GET_LIN_B(b.attn_q_b,   lname("enc.blocks.%d.attn.linear_q.bias",     i), d);
    GET_LIN_W(b.attn_k_w,   lname("enc.blocks.%d.attn.linear_k.weight",   i), d, d);
    GET_LIN_B(b.attn_k_b,   lname("enc.blocks.%d.attn.linear_k.bias",     i), d);
    GET_LIN_W(b.attn_v_w,   lname("enc.blocks.%d.attn.linear_v.weight",   i), d, d);
    GET_LIN_B(b.attn_v_b,   lname("enc.blocks.%d.attn.linear_v.bias",     i), d);
    GET_LIN_W(b.attn_out_w, lname("enc.blocks.%d.attn.linear_out.weight", i), d, d);
    GET_LIN_B(b.attn_out_b, lname("enc.blocks.%d.attn.linear_out.bias",   i), d);

    // Macaron FF2.
    GET_LN  (b.norm_ff2_w, lname("enc.blocks.%d.norm_ff2.weight", i));
    GET_LN  (b.norm_ff2_b, lname("enc.blocks.%d.norm_ff2.bias",   i));
    GET_LIN_W(b.ff2_lin1_w, lname("enc.blocks.%d.ff2.linear1.weight", i), d,   dff);
    GET_LIN_B(b.ff2_lin1_b, lname("enc.blocks.%d.ff2.linear1.bias",   i), dff);
    GET_LIN_W(b.ff2_lin2_w, lname("enc.blocks.%d.ff2.linear2.weight", i), dff, d);
    GET_LIN_B(b.ff2_lin2_b, lname("enc.blocks.%d.ff2.linear2.bias",   i), d);

    GET_LN  (b.norm_out_w, lname("enc.blocks.%d.norm_out.weight", i));
    GET_LN  (b.norm_out_b, lname("enc.blocks.%d.norm_out.bias",   i));

    return TRANSCRIBE_OK;
}

} // namespace

transcribe_status build_gigaam_weights(ggml_context *        ctx_meta,
                                       const GigaamHParams & hp,
                                       GigaamWeights &       weights)
{
    if (ctx_meta == nullptr) return TRANSCRIBE_ERR_INVALID_ARG;

    const int d   = hp.enc_d_model;
    const int feat_in = hp.enc_feat_in;
    const int sub_k = hp.enc_subs_kernel_size;
    const int n_freq = hp.fe_n_fft / 2 + 1;

    // Frontend buffers (baked into the GGUF by the converter).
    GET_F32_2D(weights.frontend_mel_filterbank,
               "frontend.mel_filterbank", n_freq, hp.fe_num_mels);
    GET_F32_1D(weights.frontend_window, "frontend.window", hp.fe_win_length);

    // Pre-encode (2 stride-2 conv1d).
    GET_CONV_W(weights.pre_encode.conv0_w, "enc.pre_encode.conv.0.weight",
               sub_k, feat_in, d);
    GET_LIN_B (weights.pre_encode.conv0_b, "enc.pre_encode.conv.0.bias", d);
    GET_CONV_W(weights.pre_encode.conv2_w, "enc.pre_encode.conv.2.weight",
               sub_k, d, d);
    GET_LIN_B (weights.pre_encode.conv2_b, "enc.pre_encode.conv.2.bias", d);

    // Conformer blocks.
    weights.blocks.resize(hp.enc_n_layers);
    for (int i = 0; i < hp.enc_n_layers; ++i) {
        if (auto st = build_block(ctx_meta, hp, i, weights.blocks[i]);
            st != TRANSCRIBE_OK) return st;
    }

    // Head.
    if (hp.head_kind == HeadKind::RNNT) {
        // pred.embed.weight is a token-id lookup table consumed by the
        // host RNN-T decoder. The quantizer treats it as a linear-weight
        // tile (TRANSCRIBE_QUANT_LINEAR_TYPES: F32/F16/Q*), so the loader
        // must too. decoder.cpp's readback_f32 dequantizes once at load
        // time via ggml type traits.
        GET_LIN_W(weights.predictor.embed_w, "pred.embed.weight",
                  hp.pred_hidden, hp.pred_vocab);

        weights.predictor.lstm.resize(hp.pred_n_layers);
        for (int i = 0; i < hp.pred_n_layers; ++i) {
            auto & L = weights.predictor.lstm[i];
            // PyTorch nn.LSTM packs 4 gates along the row dim.
            GET_LIN_W(L.Wx, lname("pred.lstm.%d.Wx", i),
                      hp.pred_hidden, 4 * hp.pred_hidden);
            GET_LIN_W(L.Wh, lname("pred.lstm.%d.Wh", i),
                      hp.pred_hidden, 4 * hp.pred_hidden);
            GET_LIN_B(L.b,  lname("pred.lstm.%d.bias", i),
                      4 * hp.pred_hidden);
        }
        GET_LIN_W(weights.joint.enc_w,  "joint.enc.weight",  d,   hp.joint_hidden);
        GET_LIN_B(weights.joint.enc_b,  "joint.enc.bias",         hp.joint_hidden);
        GET_LIN_W(weights.joint.pred_w, "joint.pred.weight", hp.pred_hidden, hp.joint_hidden);
        GET_LIN_B(weights.joint.pred_b, "joint.pred.bias",         hp.joint_hidden);
        GET_LIN_W(weights.joint.out_w,  "joint.out.weight",  hp.joint_hidden, hp.joint_n_classes);
        GET_LIN_B(weights.joint.out_b,  "joint.out.bias",         hp.joint_n_classes);
    } else { // CTC
        GET_CONV_W(weights.ctc_head.weight, "head.ctc.weight",
                   1, d, hp.head_n_classes);
        GET_LIN_B (weights.ctc_head.bias,   "head.ctc.bias", hp.head_n_classes);
    }

    return TRANSCRIBE_OK;
}

} // namespace transcribe::gigaam