seamless_communication/ggml/examples/unity/fairseq2.cpp

#include "ggml.h"
#include "fairseq2.h"

/// allocate the fairseq2 model and hyperparameters
extern "C" fairseq2_model* fairseq2_model_alloc() {
    // pre-allocate some memory to write hyperparameters and tensors pointers
    auto* model = new fairseq2_model;
    model->hparams = new std::uint8_t[8 * 1024];
    model->arch = new std::uint64_t[16 * 1024];  // max tensors allowed
    model->tensors_ctx = nullptr;
    return model;
};

extern "C" void fairseq2_model_free(fairseq2_model* model) {
    if (model->tensors_ctx) ggml_free(model->tensors_ctx);
    delete (std::uint64_t*)(model->arch);
    delete (std::uint8_t*)model->hparams;
    delete model;
};

extern "C" void fairseq2_model_set_inference_ctx(fairseq2_model* model, ggml_context* ctx) {
    model->ctx = ctx;
}

extern "C" std::string* std_string_alloc(char* c_str) {
    return new std::string(c_str);
}

extern "C" void std_string_free(std::string* str) {
    delete str;
}



// Linear

std::size_t Linear_size(int32_t input_dim, int32_t output_dim)
{
    return (input_dim * output_dim * ggml_type_size(GGML_TYPE_F32)) // weight
        + (output_dim * ggml_type_size(GGML_TYPE_F32)); // bias
};

void Linear_init(
    Linear& self,
    fairseq2_model& model,
    const std::string &prefix,
    int input_dim,
    int output_dim,
    bool bias
) {
    self.weight = ggml_new_tensor_2d(model.ctx, GGML_TYPE_F32, output_dim, input_dim);
    model.tensors[prefix + ".weight"] = self.weight;
    if (bias) {
        self.bias = ggml_new_tensor_1d(model.ctx, GGML_TYPE_F32, output_dim);
        model.tensors[prefix + ".inner_proj.bias"] = self.bias;
    }
}

extern "C" ggml_tensor*
Linear_forward(
    fairseq2_model& model,
    const std::string &prefix,
    ggml_tensor* input  // (d_in)
) {
    // Note: for now we assumed un-batched input
    ggml_tensor* weight = model.tensors[prefix + ".weight"];  // (d_in, d_out)
    ggml_tensor* bias = model.tensors[prefix + ".bias"];  // (d_out)

    return ggml_add(
        model.ctx,
        ggml_mul_mat(model.ctx, weight, input),  // (d_out)
        bias
    );
}

// LayerNorm

std::size_t LayerNorm_size(int32_t dim)
{
    return 2 * dim * ggml_type_size(GGML_TYPE_F32); // weight and bias
};

void LayerNorm_init(
    LayerNorm& self,
    fairseq2_model& model,
    const std::string &prefix,
    int dim
) {
    self.weight = ggml_new_tensor_1d(model.ctx, GGML_TYPE_F32, dim);
    model.tensors[prefix + ".weight"] = self.weight;
    self.bias = ggml_new_tensor_1d(model.ctx, GGML_TYPE_F32, dim);
    model.tensors[prefix + ".bias"] = self.bias;
}

extern "C" ggml_tensor* LayerNorm_forward(
    fairseq2_model& model,
    const std::string &prefix,
    ggml_tensor* input) {
    ggml_tensor* weight = model.tensors[prefix + ".weight"];
    ggml_tensor* bias = model.tensors[prefix + ".bias"];

    auto ctx = model.ctx;
    // TODO: should `eps` be part of unity hparams ?
    input = ggml_norm(ctx, input, /*eps*/1e-5);
    return ggml_add(
        ctx,
        ggml_mul(ctx, ggml_repeat(ctx, weight, input), input),
        ggml_repeat(ctx, bias, input)
    );
}


std::size_t StandardFeedForwardNetwork_size(int32_t dim, int32_t inner_dim)
{
    return LayerNorm_size(dim) + Linear_size(dim, inner_dim) + Linear_size(inner_dim, dim);
};

void StandardFeedForwardNetwork_init(
    StandardFeedForwardNetwork& self,
    fairseq2_model& model,
    const std::string &prefix,
    int model_dim,
    int inner_dim
) {
    Linear_init(self.inner_proj, model, prefix + ".inner_proj", model_dim, inner_dim, true);
    LayerNorm_init(self.inner_layer_norm, model, prefix + ".inner_layer_norm", inner_dim);
    Linear_init(self.output_proj, model, prefix + ".output_proj", inner_dim, model_dim, true);
}

extern "C" ggml_tensor* StandardFeedForwardNetwork_forward(
    fairseq2_model& model,
    const std::string& prefix,
    ggml_tensor* seqs
) {
    seqs = Linear_forward(model, prefix + ".inner_proj", seqs);
    // inner_activation = ReLu // TODO: allow other activation
    seqs = ggml_relu(model.ctx, seqs);

    if (model.tensors.find(prefix + ".inner_layer_norm.weight") != model.tensors.end()) {
        seqs = LayerNorm_forward(model, prefix + ".inner_layer_norm", seqs);
    }

    // TODO: inference dropout
    // if self.inner_dropout is not None:
    //     seqs = self.inner_dropout(seqs)
    seqs = Linear_forward(model, prefix + ".output_proj", seqs);
    return seqs;
}

void MultiheadAttention_init(
    MultiheadAttention& self,
    fairseq2_model& model,
    const std::string &prefix,
    int model_dim,
    int num_heads
) {
    int bias = true;
    int num_key_value_heads = num_heads;
    int head_dim = model_dim / num_heads;

    Linear_init(self.q_proj, model, prefix + ".q_proj", model_dim, model_dim, bias);
    Linear_init(self.k_proj, model, prefix + ".k_proj", model_dim, head_dim * num_key_value_heads, bias);
    Linear_init(self.v_proj, model, prefix + ".v_proj", model_dim, model_dim, bias);

    // (H, 1, K_h)
    self.bias_k = ggml_new_tensor_3d(model.ctx, GGML_TYPE_F32, num_heads, 1, head_dim * num_key_value_heads/ num_heads);
    // (H, 1, V_h)
    self.bias_v = ggml_new_tensor_3d(model.ctx, GGML_TYPE_F32, num_heads, 1, model_dim / num_heads);
}

ggml_tensor* reshape_num_head(ggml_context* ctx, ggml_tensor* x, int num_heads) {
    int slen = x->ne[1];
    int model_dim = x->ne[0];
    // (S, dim) -> (S, H, H_dim)
    x = ggml_reshape_3d(ctx, x, model_dim / num_heads, num_heads, slen);
    // (S, H, H_dim) -> (H, S, H_dim)
    x = ggml_permute(ctx, x, 0, 2, 1, 3);
    return x;
}



extern "C" ggml_tensor* // (slen, d_in)
MultiheadAttention_forward(
    fairseq2_model& model,
    const std::string &prefix,
    ggml_tensor* queries,  // (slen, d_in)
    ggml_tensor* keys,  // (klen, d_in)
    ggml_tensor* values,  // (klen, d_out)
    ggml_tensor* _ // (klen, slen)  TODO: do we need to pass mask here ?
) {
    int slen = queries->ne[1];
    int num_heads = 16;
    int head_dim = queries->ne[0] / num_heads;
    ggml_context* ctx = model.ctx;
    ggml_tensor* q = Linear_forward(model, prefix + ".q_proj", queries);
    q = reshape_num_head(ctx, q, num_heads);  // (H, S, H_dim)
    ggml_tensor* k = Linear_forward(model, prefix + ".k_proj", keys);
    k = reshape_num_head(ctx, k, num_heads);  // (H, S, H_dim)
    ggml_tensor* v = Linear_forward(model, prefix + ".v_proj", values);
    v = ggml_reshape_3d(ctx, v, head_dim, num_heads, slen); // (S, H, H_dim)
    // v = ggml_permute(ctx, v, 1, 2, 0, 3);  // (H, H_dim, S)
    v = ggml_permute(ctx, v, 1, 0, 2, 3);  // (S, H_dim, H)
    v = ggml_cont(ctx, v);

    // ggml_tensor* attn = ggml_flash_attn(ctx, q, k, v, /*masked*/false);  // (H, S, H_dim)
    attn = ggml_permute(ctx, attn, 0, 2, 1, 3);  // (S, H, H_dim)
    attn = ggml_cont(ctx, attn);
    attn = ggml_reshape_2d(ctx, attn, num_heads * head_dim, slen);   // (S, H * V_h)
    attn = Linear_forward(model, prefix + ".output_proj", attn);              // (S, d_out)

    return attn;
}

// ggml_tensor* attn_weights = ggml_mul_mat(ctx, q, k);  // (H, S, S)
//     attn_weights = ggm_mul * (q.size(-1) ** -0.5)

//     if mask is not None:
//         attn_weights = attn_weights + mask

//     # For numerical stability run in single precision.
//     attn_weights = softmax(attn_weights, dim=-1, dtype=torch.float32)

//     attn_weights = attn_weights.type_as(q)

//     if training and dropout_p > 0.0:
//         attn_weights = dropout(attn_weights, dropout_p)

//     # (*, S, S_kv) @ (*, S_kv, V) = (*, S, V)
//     attn = torch.matmul(attn_weights, values)

//     return attn, attn_weights if needs_weights else None

// extern "C" ggml_tensor* // (d_out, seq_len)
// SDPA_forward(
//     fairseq2_model& model,
//     const std::string &prefix,
//     ggml_tensor* queries,  // (d_in, len_q)
//     ggml_tensor* keys,  // (d_in, len_k)
//     ggml_tensor* values,  // (d_out, len_k)
//     ggml_tensor* mask // (seq_len, len_q)
// ) {
//     return queries;
// }