transcribe-cpp-sys 0.0.11

// arch/cohere/weights.cpp - implementation of read_cohere_hparams
// and build_cohere_weights.
//
// Pattern follows parakeet/weights.cpp exactly:
//   - read every required hparam from KV explicitly
//   - build the tensor catalog as a sequence of get_tensor() calls
//     with explicit expected shapes
//   - missing tensor or shape mismatch -> TRANSCRIBE_ERR_GGUF

#include "weights.h"

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

#include <cstdio>
#include <cstring>
#include <initializer_list>

namespace transcribe::cohere {

namespace {

constexpr const char * kFamilyTag = "cohere";

}  // namespace

transcribe_status read_cohere_hparams(const gguf_context * gguf, CohereHParams & hp) {
    if (gguf == nullptr) {
        return TRANSCRIBE_ERR_INVALID_ARG;
    }

    // Encoder.
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.encoder.n_layers", kFamilyTag, hp.enc_n_layers);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.encoder.d_model", kFamilyTag, hp.enc_d_model);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.encoder.n_heads", kFamilyTag, hp.enc_n_heads);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.encoder.d_ff", kFamilyTag, hp.enc_d_ff); st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.encoder.conv_kernel", kFamilyTag, hp.enc_conv_kernel);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st =
            read_required_u32_kv(gguf, "stt.cohere.encoder.subsampling_factor", kFamilyTag, hp.enc_subsampling_factor);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.encoder.subsampling_channels", kFamilyTag,
                                       hp.enc_subsampling_channels);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.encoder.pos_emb_max_len", kFamilyTag, hp.enc_pos_emb_max_len);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_optional_bool_kv(gguf, "stt.cohere.encoder.use_bias", kFamilyTag, true, hp.enc_use_bias);
        st != TRANSCRIBE_OK) {
        return st;
    }

    // Decoder.
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.decoder.n_layers", kFamilyTag, hp.dec_n_layers);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.decoder.hidden_size", kFamilyTag, hp.dec_hidden);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.decoder.n_heads", kFamilyTag, hp.dec_n_heads);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.decoder.inner_size", kFamilyTag, hp.dec_inner);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_u32_kv(gguf, "stt.cohere.decoder.max_seq_len", kFamilyTag, hp.dec_max_seq);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_required_string_kv(gguf, "stt.cohere.decoder.activation", kFamilyTag, hp.dec_activation);
        st != TRANSCRIBE_OK) {
        return st;
    }

    // Token IDs. vocab_size comes from the tokenizer; decoder_start_token_id is explicit.
    if (auto st =
            read_required_u32_kv(gguf, "stt.cohere.decoder_start_token_id", kFamilyTag, hp.decoder_start_token_id);
        st != TRANSCRIBE_OK) {
        return st;
    }

    // Head.
    if (auto st = read_optional_bool_kv(gguf, "stt.cohere.head.log_softmax", kFamilyTag, true, hp.head_log_softmax);
        st != TRANSCRIBE_OK) {
        return st;
    }
    if (auto st = read_optional_bool_kv(gguf, "stt.cohere.head.tied_weights", kFamilyTag, true, hp.head_tied_weights);
        st != TRANSCRIBE_OK) {
        return st;
    }

    // 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_string_kv(gguf, "stt.frontend.pad_mode", kFamilyTag, "reflect", hp.fe_pad_mode);
        st != TRANSCRIBE_OK) {
        return st;
    }

    // Cross-field invariants.
    if (hp.enc_n_layers <= 0 || hp.enc_d_model <= 0 || hp.enc_n_heads <= 0 || hp.enc_d_ff <= 0 ||
        hp.enc_conv_kernel <= 0 || hp.enc_subsampling_factor <= 0 || hp.enc_subsampling_channels <= 0) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: encoder hparams must be positive");
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.enc_d_model % hp.enc_n_heads != 0) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: encoder d_model (%d) not divisible by n_heads (%d)",
                hp.enc_d_model, hp.enc_n_heads);
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.dec_n_layers <= 0 || hp.dec_hidden <= 0 || hp.dec_n_heads <= 0 || hp.dec_inner <= 0 || hp.dec_max_seq <= 0) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: decoder hparams must be positive");
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.dec_hidden % hp.dec_n_heads != 0) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: decoder hidden (%d) not divisible by n_heads (%d)", hp.dec_hidden,
                hp.dec_n_heads);
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.dec_activation != "relu" && hp.dec_activation != "silu" && hp.dec_activation != "swish") {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                "cohere: unsupported decoder activation \"%s\" "
                "(only relu, silu, swish are implemented)",
                hp.dec_activation.c_str());
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.fe_num_mels <= 0 || hp.fe_sample_rate <= 0 || hp.fe_n_fft <= 0 || hp.fe_win_length <= 0 ||
        hp.fe_hop_length <= 0) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: frontend dimensions must be positive");
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.fe_win_length > hp.fe_n_fft) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: frontend win_length (%d) > n_fft (%d)", hp.fe_win_length,
                hp.fe_n_fft);
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.fe_f_min < 0.0f || hp.fe_f_max <= hp.fe_f_min) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: frontend mel band invalid: f_min=%f f_max=%f", hp.fe_f_min,
                hp.fe_f_max);
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.fe_type != "mel") {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: unsupported frontend type \"%s\" (only \"mel\")",
                hp.fe_type.c_str());
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.fe_window != "hann") {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                "cohere: unsupported frontend window \"%s\" "
                "(only \"hann\" is implemented)",
                hp.fe_window.c_str());
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.fe_normalize != "per_feature") {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                "cohere: unsupported frontend normalize \"%s\" "
                "(only \"per_feature\" is implemented)",
                hp.fe_normalize.c_str());
        return TRANSCRIBE_ERR_GGUF;
    }
    if ((hp.fe_n_fft & (hp.fe_n_fft - 1)) != 0) {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: frontend n_fft (%d) must be a power of 2", hp.fe_n_fft);
        return TRANSCRIBE_ERR_GGUF;
    }
    if (hp.fe_pad_mode != "reflect" && hp.fe_pad_mode != "constant") {
        log_msg(TRANSCRIBE_LOG_LEVEL_ERROR,
                "cohere: unsupported frontend pad_mode \"%s\" "
                "(only \"reflect\" and \"constant\" are implemented)",
                hp.fe_pad_mode.c_str());
        return TRANSCRIBE_ERR_GGUF;
    }

    return TRANSCRIBE_OK;
}

namespace {

using transcribe::weights::lname;

// find_tensor() + lname() live in src/transcribe-weights-util.{h,cpp};
// the GET_* macros stay here so the family log tag lands in diagnostics.
constexpr const char * kTag = kFamilyTag;

#define GET_F32(slot, name, ...)                                                                          \
    do {                                                                                                  \
        ggml_tensor * _t =                                                                                \
            transcribe::weights::find_tensor(ctx_meta, (name), { GGML_TYPE_F32 }, { __VA_ARGS__ }, kTag); \
        if (_t == nullptr)                                                                                \
            return TRANSCRIBE_ERR_GGUF;                                                                   \
        (slot) = _t;                                                                                      \
    } while (0)

#define GET_CONV(slot, name, ...)                                                                              \
    do {                                                                                                       \
        ggml_tensor * _t = transcribe::weights::find_tensor(ctx_meta, (name), { TRANSCRIBE_QUANT_CONV_TYPES }, \
                                                            { __VA_ARGS__ }, kTag);                            \
        if (_t == nullptr)                                                                                     \
            return TRANSCRIBE_ERR_GGUF;                                                                        \
        (slot) = _t;                                                                                           \
    } while (0)

#define GET_LIN(slot, name, ...)                                                                                 \
    do {                                                                                                         \
        ggml_tensor * _t = transcribe::weights::find_tensor(ctx_meta, (name), { TRANSCRIBE_QUANT_LINEAR_TYPES }, \
                                                            { __VA_ARGS__ }, kTag);                              \
        if (_t == nullptr)                                                                                       \
            return TRANSCRIBE_ERR_GGUF;                                                                          \
        (slot) = _t;                                                                                             \
    } while (0)

}  // namespace

transcribe_status build_cohere_weights(ggml_context * ctx_meta, const CohereHParams & hp, CohereWeights & weights) {
    if (ctx_meta == nullptr) {
        return TRANSCRIBE_ERR_INVALID_ARG;
    }

    const int64_t channels = hp.enc_subsampling_channels;
    const int64_t d_model  = hp.enc_d_model;
    const int64_t d_ff     = hp.enc_d_ff;
    const int64_t n_heads  = hp.enc_n_heads;
    const int64_t head_dim = hp.enc_head_dim();
    const int64_t k        = hp.enc_conv_kernel;
    const int64_t dec_h    = hp.dec_hidden;
    const int64_t dec_nh   = hp.dec_n_heads;
    const int64_t dec_hd   = hp.dec_head_dim();
    const int64_t dec_in   = hp.dec_inner;

    (void) dec_nh;
    (void) dec_hd;

    const int64_t pre_encode_freq = channels * (hp.fe_num_mels / hp.enc_subsampling_factor);
    const int64_t pre_encode_in   = pre_encode_freq;

    // ----- pre_encode -----
    GET_CONV(weights.pre_encode.conv0_w, "enc.pre_encode.conv.0.weight", 3, 3, 1, channels);
    GET_F32(weights.pre_encode.conv0_b, "enc.pre_encode.conv.0.bias", channels);
    GET_CONV(weights.pre_encode.conv2_w, "enc.pre_encode.conv.2.weight", 3, 3, 1, channels);
    GET_F32(weights.pre_encode.conv2_b, "enc.pre_encode.conv.2.bias", channels);
    GET_CONV(weights.pre_encode.conv3_w, "enc.pre_encode.conv.3.weight", 1, 1, channels, channels);
    GET_F32(weights.pre_encode.conv3_b, "enc.pre_encode.conv.3.bias", channels);
    GET_CONV(weights.pre_encode.conv5_w, "enc.pre_encode.conv.5.weight", 3, 3, 1, channels);
    GET_F32(weights.pre_encode.conv5_b, "enc.pre_encode.conv.5.bias", channels);
    GET_CONV(weights.pre_encode.conv6_w, "enc.pre_encode.conv.6.weight", 1, 1, channels, channels);
    GET_F32(weights.pre_encode.conv6_b, "enc.pre_encode.conv.6.bias", channels);
    GET_LIN(weights.pre_encode.out_w, "enc.pre_encode.out.weight", pre_encode_in, d_model);
    GET_F32(weights.pre_encode.out_b, "enc.pre_encode.out.bias", d_model);

    // ----- encoder blocks -----
    weights.blocks.assign(hp.enc_n_layers, CohereBlock{});
    for (int i = 0; i < hp.enc_n_layers; ++i) {
        auto & b = weights.blocks[i];

        // Macaron FF1 (with bias).
        GET_F32(b.norm_ff1_w, lname("enc.blocks.%d.norm_ff1.weight", i), d_model);
        GET_F32(b.norm_ff1_b, lname("enc.blocks.%d.norm_ff1.bias", i), d_model);
        GET_LIN(b.ff1_lin1_w, lname("enc.blocks.%d.ff1.linear1.weight", i), d_model, d_ff);
        GET_F32(b.ff1_lin1_b, lname("enc.blocks.%d.ff1.linear1.bias", i), d_ff);
        GET_LIN(b.ff1_lin2_w, lname("enc.blocks.%d.ff1.linear2.weight", i), d_ff, d_model);
        GET_F32(b.ff1_lin2_b, lname("enc.blocks.%d.ff1.linear2.bias", i), d_model);

        // Self-attention with relative position (with bias on Q, K, V, out).
        GET_F32(b.norm_attn_w, lname("enc.blocks.%d.norm_attn.weight", i), d_model);
        GET_F32(b.norm_attn_b, lname("enc.blocks.%d.norm_attn.bias", i), d_model);
        GET_LIN(b.attn_q_w, lname("enc.blocks.%d.attn.linear_q.weight", i), d_model, d_model);
        GET_F32(b.attn_q_b, lname("enc.blocks.%d.attn.linear_q.bias", i), d_model);
        GET_LIN(b.attn_k_w, lname("enc.blocks.%d.attn.linear_k.weight", i), d_model, d_model);
        GET_F32(b.attn_k_b, lname("enc.blocks.%d.attn.linear_k.bias", i), d_model);
        GET_LIN(b.attn_v_w, lname("enc.blocks.%d.attn.linear_v.weight", i), d_model, d_model);
        GET_F32(b.attn_v_b, lname("enc.blocks.%d.attn.linear_v.bias", i), d_model);
        GET_LIN(b.attn_out_w, lname("enc.blocks.%d.attn.linear_out.weight", i), d_model, d_model);
        GET_F32(b.attn_out_b, lname("enc.blocks.%d.attn.linear_out.bias", i), d_model);
        GET_LIN(b.attn_pos_w, lname("enc.blocks.%d.attn.linear_pos.weight", i), d_model, d_model);
        GET_F32(b.attn_pos_u, lname("enc.blocks.%d.attn.pos_bias_u", i), head_dim, n_heads);
        GET_F32(b.attn_pos_v, lname("enc.blocks.%d.attn.pos_bias_v", i), head_dim, n_heads);

        // Conv module (with bias on pointwise and depthwise).
        GET_F32(b.norm_conv_w, lname("enc.blocks.%d.norm_conv.weight", i), d_model);
        GET_F32(b.norm_conv_b, lname("enc.blocks.%d.norm_conv.bias", i), d_model);
        GET_CONV(b.conv_pw1_w, lname("enc.blocks.%d.conv.pointwise1.weight", i), 1, d_model, 2 * d_model);
        GET_F32(b.conv_pw1_b, lname("enc.blocks.%d.conv.pointwise1.bias", i), 2 * d_model);
        GET_CONV(b.conv_dw_w, lname("enc.blocks.%d.conv.depthwise.weight", i), k, 1, d_model);
        GET_F32(b.conv_dw_b, lname("enc.blocks.%d.conv.depthwise.bias", i), d_model);
        GET_CONV(b.conv_pw2_w, lname("enc.blocks.%d.conv.pointwise2.weight", i), 1, d_model, d_model);
        GET_F32(b.conv_pw2_b, lname("enc.blocks.%d.conv.pointwise2.bias", i), d_model);
        GET_F32(b.conv_bn_w, lname("enc.blocks.%d.conv.bn.weight", i), d_model);
        GET_F32(b.conv_bn_b, lname("enc.blocks.%d.conv.bn.bias", i), d_model);
        GET_F32(b.conv_bn_rm, lname("enc.blocks.%d.conv.bn.running_mean", i), d_model);
        GET_F32(b.conv_bn_rv, lname("enc.blocks.%d.conv.bn.running_var", i), d_model);

        // Macaron FF2 (with bias).
        GET_F32(b.norm_ff2_w, lname("enc.blocks.%d.norm_ff2.weight", i), d_model);
        GET_F32(b.norm_ff2_b, lname("enc.blocks.%d.norm_ff2.bias", i), d_model);
        GET_LIN(b.ff2_lin1_w, lname("enc.blocks.%d.ff2.linear1.weight", i), d_model, d_ff);
        GET_F32(b.ff2_lin1_b, lname("enc.blocks.%d.ff2.linear1.bias", i), d_ff);
        GET_LIN(b.ff2_lin2_w, lname("enc.blocks.%d.ff2.linear2.weight", i), d_ff, d_model);
        GET_F32(b.ff2_lin2_b, lname("enc.blocks.%d.ff2.linear2.bias", i), d_model);

        // Final per-block layer norm.
        GET_F32(b.norm_out_w, lname("enc.blocks.%d.norm_out.weight", i), d_model);
        GET_F32(b.norm_out_b, lname("enc.blocks.%d.norm_out.bias", i), d_model);
    }

    // ----- encoder-decoder projection -----
    GET_LIN(weights.enc_dec_proj.weight, "enc_dec_proj.weight", d_model, dec_h);
    GET_F32(weights.enc_dec_proj.bias, "enc_dec_proj.bias", dec_h);

    // ----- decoder embedding -----
    // Token embedding ne=[dec_hidden, vocab_size]; vocab_size is read from
    // the tensor shape (no separate KV).
    {
        ggml_tensor * tw = ggml_get_tensor(ctx_meta, "dec.embed.token.weight");
        if (tw == nullptr) {
            log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: missing tensor \"dec.embed.token.weight\"");
            return TRANSCRIBE_ERR_GGUF;
        }
        if (tw->ne[0] != dec_h) {
            log_msg(TRANSCRIBE_LOG_LEVEL_ERROR, "cohere: dec.embed.token.weight ne[0]=%lld, expected %lld",
                    static_cast<long long>(tw->ne[0]), static_cast<long long>(dec_h));
            return TRANSCRIBE_ERR_GGUF;
        }
        weights.dec_embed.token_w = tw;
    }

    const int64_t vocab_size = weights.dec_embed.token_w->ne[1];

    GET_F32(weights.dec_embed.pos_enc, "dec.embed.pos_enc", dec_h, hp.dec_max_seq);
    GET_F32(weights.dec_embed.norm_w, "dec.embed.norm.weight", dec_h);
    GET_F32(weights.dec_embed.norm_b, "dec.embed.norm.bias", dec_h);

    // ----- decoder blocks -----
    weights.dec_blocks.assign(hp.dec_n_layers, CohereDecBlock{});
    for (int i = 0; i < hp.dec_n_layers; ++i) {
        auto & db = weights.dec_blocks[i];

        // Self-attention.
        GET_F32(db.norm_self_w, lname("dec.blocks.%d.norm_self.weight", i), dec_h);
        GET_F32(db.norm_self_b, lname("dec.blocks.%d.norm_self.bias", i), dec_h);
        GET_LIN(db.self_q_w, lname("dec.blocks.%d.self_attn.q.weight", i), dec_h, dec_h);
        GET_F32(db.self_q_b, lname("dec.blocks.%d.self_attn.q.bias", i), dec_h);
        GET_LIN(db.self_k_w, lname("dec.blocks.%d.self_attn.k.weight", i), dec_h, dec_h);
        GET_F32(db.self_k_b, lname("dec.blocks.%d.self_attn.k.bias", i), dec_h);
        GET_LIN(db.self_v_w, lname("dec.blocks.%d.self_attn.v.weight", i), dec_h, dec_h);
        GET_F32(db.self_v_b, lname("dec.blocks.%d.self_attn.v.bias", i), dec_h);
        GET_LIN(db.self_out_w, lname("dec.blocks.%d.self_attn.out.weight", i), dec_h, dec_h);
        GET_F32(db.self_out_b, lname("dec.blocks.%d.self_attn.out.bias", i), dec_h);

        // Cross-attention.
        GET_F32(db.norm_cross_w, lname("dec.blocks.%d.norm_cross.weight", i), dec_h);
        GET_F32(db.norm_cross_b, lname("dec.blocks.%d.norm_cross.bias", i), dec_h);
        GET_LIN(db.cross_q_w, lname("dec.blocks.%d.cross_attn.q.weight", i), dec_h, dec_h);
        GET_F32(db.cross_q_b, lname("dec.blocks.%d.cross_attn.q.bias", i), dec_h);
        GET_LIN(db.cross_k_w, lname("dec.blocks.%d.cross_attn.k.weight", i), dec_h, dec_h);
        GET_F32(db.cross_k_b, lname("dec.blocks.%d.cross_attn.k.bias", i), dec_h);
        GET_LIN(db.cross_v_w, lname("dec.blocks.%d.cross_attn.v.weight", i), dec_h, dec_h);
        GET_F32(db.cross_v_b, lname("dec.blocks.%d.cross_attn.v.bias", i), dec_h);
        GET_LIN(db.cross_out_w, lname("dec.blocks.%d.cross_attn.out.weight", i), dec_h, dec_h);
        GET_F32(db.cross_out_b, lname("dec.blocks.%d.cross_attn.out.bias", i), dec_h);

        // FFN.
        GET_F32(db.norm_ff_w, lname("dec.blocks.%d.norm_ff.weight", i), dec_h);
        GET_F32(db.norm_ff_b, lname("dec.blocks.%d.norm_ff.bias", i), dec_h);
        GET_LIN(db.ff_in_w, lname("dec.blocks.%d.ff.dense_in.weight", i), dec_h, dec_in);
        GET_F32(db.ff_in_b, lname("dec.blocks.%d.ff.dense_in.bias", i), dec_in);
        GET_LIN(db.ff_out_w, lname("dec.blocks.%d.ff.dense_out.weight", i), dec_in, dec_h);
        GET_F32(db.ff_out_b, lname("dec.blocks.%d.ff.dense_out.bias", i), dec_h);
    }

    // ----- decoder final norm -----
    GET_F32(weights.dec_final.norm_w, "dec.final_norm.weight", dec_h);
    GET_F32(weights.dec_final.norm_b, "dec.final_norm.bias", dec_h);

    // ----- head -----
    GET_F32(weights.head.bias, "head.bias", vocab_size);

    return TRANSCRIBE_OK;
}

#undef GET_F32
#undef GET_CONV
#undef GET_LIN

}  // namespace transcribe::cohere