vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

modify clip

Browse source
This commit is contained in:
caitianchi 2024-08-10 19:21:27 +08:00
parent bffbe1cf44
commit 28d6a0f43d
1 changed files with 1 additions and 66 deletions
Download patch file Download diff file Expand all files Collapse all files

67
examples/llava/clip.cpp
View file

@ -1006,72 +1006,6 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
GGML_ASSERT(false);
}
}
// minicpmv projector
else if (ctx->has_minicpmv_projector) {
if (ctx->proj_type == PROJECTOR_TYPE_RESAMPLER) {
struct ggml_tensor * q = model.mm_model_query;
{ // layernorm
q = ggml_norm(ctx0, q, eps);
q = ggml_add(ctx0, ggml_mul(ctx0, q, model.mm_model_ln_q_w), model.mm_model_ln_q_b);
}
struct ggml_tensor *k, *v = ggml_mul_mat(ctx0, model.mm_model_kv_proj, embeddings);
{ // layernorm
v = ggml_norm(ctx0, v, eps);
v = ggml_add(ctx0, ggml_mul(ctx0, v, model.mm_model_ln_kv_w), model.mm_model_ln_kv_b);
}
{ // position
// q = ggml_add(ctx0, q, model.mm_model_pos_embed);
k = ggml_add(ctx0, v, pos_embed);
}
{ // attention
int hidden_size = 4096;
const int d_head = 128;
int n_head = hidden_size/d_head;
int num_query = 96;
if (ctx->minicpmv_version == 2) {
hidden_size = 4096;
n_head = hidden_size/d_head;
num_query = 96;
}
else if (ctx->minicpmv_version == 3) {
hidden_size = 3584;
n_head = hidden_size/d_head;
num_query = 64;
}
struct ggml_tensor * Q = ggml_add(ctx0, ggml_mul_mat(ctx0, model.mm_model_attn_q_w, q), model.mm_model_attn_q_b);
Q = ggml_scale_inplace(ctx0, Q, 1.0f / sqrt((float)d_head));
struct ggml_tensor * K = ggml_add(ctx0, ggml_mul_mat(ctx0, model.mm_model_attn_k_w, k), model.mm_model_attn_k_b);
struct ggml_tensor * V = ggml_add(ctx0, ggml_mul_mat(ctx0, model.mm_model_attn_v_w, v), model.mm_model_attn_v_b);
// permute
Q = ggml_reshape_4d(ctx0, Q, d_head, n_head, num_query, batch_size);
Q = ggml_cont(ctx0, ggml_permute(ctx0, Q, 0, 2, 1, 3));
Q = ggml_reshape_3d(ctx0, Q, d_head, num_query, n_head * batch_size);
K = ggml_reshape_4d(ctx0, K, d_head, n_head, num_positions, batch_size);
K = ggml_cont(ctx0, ggml_permute(ctx0, K, 0, 2, 1, 3));
K = ggml_reshape_3d(ctx0, K, d_head, num_positions, n_head * batch_size);
V = ggml_reshape_4d(ctx0, V, d_head, n_head, num_positions, batch_size);
V = ggml_cont(ctx0, ggml_permute(ctx0, V, 1, 2, 0, 3));
V = ggml_reshape_3d(ctx0, V, num_positions, d_head, n_head * batch_size);
struct ggml_tensor * KQ = ggml_mul_mat(ctx0, K, Q);
KQ = ggml_soft_max_inplace(ctx0, KQ);
struct ggml_tensor * KQV = ggml_mul_mat(ctx0, V, KQ);
KQV = ggml_reshape_4d(ctx0, KQV, d_head, num_query, n_head, batch_size);
KQV = ggml_permute(ctx0, KQV, 0, 2, 1, 3);
KQV = ggml_cont_3d(ctx0, KQV, hidden_size, num_query, batch_size);
embeddings = ggml_add(ctx0, ggml_mul_mat(ctx0, model.mm_model_attn_o_w, KQV), model.mm_model_attn_o_b);
}
{ // layernorm
embeddings = ggml_norm(ctx0, embeddings, eps);
embeddings = ggml_add(ctx0, ggml_mul(ctx0, embeddings, model.mm_model_ln_post_w), model.mm_model_ln_post_b);
}
embeddings = ggml_mul_mat(ctx0, model.mm_model_proj, embeddings);
}
else {
GGML_ASSERT(false);
}
}
// build the graph
ggml_build_forward_expand(gf, embeddings);
@ -1553,6 +1487,7 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
clip_image_f32_batch batch;
batch.size = 1;
ggml_cgraph * gf = clip_image_build_graph(new_clip, &batch, nullptr, false);
LOG_TEE("%s: flag\n", __func__);
ggml_gallocr_reserve(new_clip->compute_alloc, gf);
size_t compute_memory_buffer_size = ggml_gallocr_get_buffer_size(new_clip->compute_alloc, 0);
LOG_TEE("%s: compute allocated memory: %.2f MB\n", __func__, compute_memory_buffer_size /1024.0/1024.0);