diff --git a/common/CMakeLists.txt b/common/CMakeLists.txt index f7b99159e3..45c48b047e 100644 --- a/common/CMakeLists.txt +++ b/common/CMakeLists.txt @@ -60,6 +60,8 @@ add_library(${TARGET} STATIC common.h console.cpp console.h + debug.cpp + debug.h download.cpp download.h http.h diff --git a/common/debug.cpp b/common/debug.cpp new file mode 100644 index 0000000000..19f0f5d98b --- /dev/null +++ b/common/debug.cpp @@ -0,0 +1,150 @@ +#include "debug.h" +#include "log.h" + +#include +#include + +std::string ggml_ne_string(const ggml_tensor * t) { + std::string str; + for (int i = 0; i < GGML_MAX_DIMS; ++i) { + str += std::to_string(t->ne[i]); + if (i + 1 < GGML_MAX_DIMS) { + str += ", "; + } + } + return str; +} + +float ggml_get_float_value(const uint8_t * data, ggml_type type, const size_t * nb, size_t i0, size_t i1, size_t i2, size_t i3) { + size_t i = i3 * nb[3] + i2 * nb[2] + i1 * nb[1] + i0 * nb[0]; + float v; + if (type == GGML_TYPE_F16) { + v = ggml_fp16_to_fp32(*(const ggml_fp16_t *) &data[i]); + } else if (type == GGML_TYPE_F32) { + v = *(const float *) &data[i]; + } else if (type == GGML_TYPE_I64) { + v = (float) *(const int64_t *) &data[i]; + } else if (type == GGML_TYPE_I32) { + v = (float) *(const int32_t *) &data[i]; + } else if (type == GGML_TYPE_I16) { + v = (float) *(const int16_t *) &data[i]; + } else if (type == GGML_TYPE_I8) { + v = (float) *(const int8_t *) &data[i]; + } else if (type == GGML_TYPE_BF16) { + v = ggml_bf16_to_fp32(*(const ggml_bf16_t *) &data[i]); + } else { + GGML_ABORT("fatal error"); + } + return v; +} + +template +void ggml_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne, const size_t * nb, int64_t n) { + GGML_ASSERT(n > 0); + float sum = 0; + for (int64_t i3 = 0; i3 < ne[3]; i3++) { + for (int64_t i2 = 0; i2 < ne[2]; i2++) { + for (int64_t i1 = 0; i1 < ne[1]; i1++) { + for (int64_t i0 = 0; i0 < ne[0]; i0++) { + const float v = ggml_get_float_value(data, type, nb, i0, i1, i2, i3); + sum += v; + } + } + } + } + for (int64_t i3 = 0; i3 < ne[3]; i3++) { + LOG_ERR(" [\n"); + for (int64_t i2 = 0; i2 < ne[2]; i2++) { + if (i2 == n && ne[2] > 2*n) { + LOG_ERR(" ..., \n"); + i2 = ne[2] - n; + } + LOG_ERR(" [\n"); + for (int64_t i1 = 0; i1 < ne[1]; i1++) { + if (i1 == n && ne[1] > 2*n) { + LOG_ERR(" ..., \n"); + i1 = ne[1] - n; + } + LOG_ERR(" ["); + for (int64_t i0 = 0; i0 < ne[0]; i0++) { + if (i0 == n && ne[0] > 2*n) { + LOG_ERR("..., "); + i0 = ne[0] - n; + } + const float v = ggml_get_float_value(data, type, nb, i0, i1, i2, i3); + LOG_ERR("%12.4f", v); + if (i0 < ne[0] - 1) { + LOG_ERR(", "); + } + } + LOG_ERR("],\n"); + } + LOG_ERR(" ],\n"); + } + LOG_ERR(" ]\n"); + LOG_ERR(" sum = %f\n", sum); + } + + if constexpr (abort) { + if (std::isnan(sum)) { + LOG_ERR("encountered NaN - aborting\n"); + exit(0); + } + } +} + +/** + * GGML operations callback during the graph execution. + * + * @param t current tensor + * @param ask when ask is true, the scheduler wants to know if we are interested in data from this tensor + * if we return true, a follow-up call will be made with ask=false in which we can do the actual collection. + * see ggml_backend_sched_eval_callback + * @param user_data user data to pass at each call back + * @return true to receive data or continue the graph, false otherwise + */ +template +bool ggml_debug(struct ggml_tensor * t, bool ask, void * user_data) { + auto * cb_data = (base_callback_data *) user_data; + + const struct ggml_tensor * src0 = t->src[0]; + const struct ggml_tensor * src1 = t->src[1]; + + if (ask) { + return true; // Always retrieve data + } + + char src1_str[128] = {0}; + if (src1) { + snprintf(src1_str, sizeof(src1_str), "%s{%s}", src1->name, ggml_ne_string(src1).c_str()); + } + + LOG_ERR("%s: %24s = (%s) %10s(%s{%s}, %s}) = {%s}\n", __func__, + t->name, ggml_type_name(t->type), ggml_op_desc(t), + src0->name, ggml_ne_string(src0).c_str(), + src1 ? src1_str : "", + ggml_ne_string(t).c_str()); + + + // copy the data from the GPU memory if needed + const bool is_host = ggml_backend_buffer_is_host(t->buffer); + + if (!is_host) { + auto n_bytes = ggml_nbytes(t); + cb_data->data.resize(n_bytes); + ggml_backend_tensor_get(t, cb_data->data.data(), 0, n_bytes); + } + + if (!ggml_is_quantized(t->type)) { + uint8_t * data = is_host ? (uint8_t *) t->data : cb_data->data.data(); + ggml_print_tensor(data, t->type, t->ne, t->nb, 3); + } + + return true; +} + +// Explicit template instantiations +template bool ggml_debug(ggml_tensor*, bool, void*); +template bool ggml_debug(ggml_tensor*, bool, void*); +template void ggml_print_tensor(uint8_t*, ggml_type, const int64_t*, const size_t*, int64_t); +template void ggml_print_tensor(uint8_t*, ggml_type, const int64_t*, const size_t*, int64_t); diff --git a/common/debug.h b/common/debug.h new file mode 100644 index 0000000000..1c540f4dc0 --- /dev/null +++ b/common/debug.h @@ -0,0 +1,19 @@ +#pragma once +#include "llama.h" + +float ggml_get_float_value(const uint8_t * data, enum ggml_type type, const size_t * nb, size_t i0, size_t i1, size_t i2, size_t i3); + +#ifdef __cplusplus +#include +#include + +// common debug functions and structs +struct base_callback_data { + std::vector data; +}; + +std::string ggml_ne_string(const ggml_tensor * t); +template void ggml_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne, const size_t * nb, int64_t n); +template bool ggml_debug(struct ggml_tensor * t, bool ask, void * user_data); +#endif + diff --git a/examples/eval-callback/eval-callback.cpp b/examples/eval-callback/eval-callback.cpp index 408338f1af..847eedc84a 100644 --- a/examples/eval-callback/eval-callback.cpp +++ b/examples/eval-callback/eval-callback.cpp @@ -1,165 +1,12 @@ #include "arg.h" #include "common.h" +#include "debug.h" #include "log.h" #include "llama.h" -#include "ggml.h" - -#include -#include +#include "llama-cpp.h" #include #include -/** - * This the arbitrary data which will be passed to each callback. - * Later on we can for example add operation or tensor name filter from the CLI arg, or a file descriptor to dump the tensor. - */ -struct callback_data { - std::vector data; -}; - -static std::string ggml_ne_string(const ggml_tensor * t) { - std::string str; - for (int i = 0; i < GGML_MAX_DIMS; ++i) { - str += std::to_string(t->ne[i]); - if (i + 1 < GGML_MAX_DIMS) { - str += ", "; - } - } - return str; -} - -static inline float ggml_compute_bf16_to_fp32(ggml_bf16_t h) { - union { - float f; - uint32_t i; - } u; - u.i = (uint32_t)h.bits << 16; - return u.f; -} - -static float ggml_get_float_value(const uint8_t * data, ggml_type type, const size_t * nb, size_t i0, size_t i1, size_t i2, size_t i3) { - size_t i = i3 * nb[3] + i2 * nb[2] + i1 * nb[1] + i0 * nb[0]; - float v; - if (type == GGML_TYPE_F16) { - v = ggml_fp16_to_fp32(*(const ggml_fp16_t *) &data[i]); - } else if (type == GGML_TYPE_F32) { - v = *(const float *) &data[i]; - } else if (type == GGML_TYPE_I64) { - v = (float) *(const int64_t *) &data[i]; - } else if (type == GGML_TYPE_I32) { - v = (float) *(const int32_t *) &data[i]; - } else if (type == GGML_TYPE_I16) { - v = (float) *(const int16_t *) &data[i]; - } else if (type == GGML_TYPE_I8) { - v = (float) *(const int8_t *) &data[i]; - } else if (type == GGML_TYPE_BF16) { - v = ggml_compute_bf16_to_fp32(*(const ggml_bf16_t *) &data[i]); - } else { - GGML_ABORT("fatal error"); - } - return v; -} - -static void ggml_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne, const size_t * nb, int64_t n) { - GGML_ASSERT(n > 0); - float sum = 0; - for (int64_t i3 = 0; i3 < ne[3]; i3++) { - for (int64_t i2 = 0; i2 < ne[2]; i2++) { - for (int64_t i1 = 0; i1 < ne[1]; i1++) { - for (int64_t i0 = 0; i0 < ne[0]; i0++) { - const float v = ggml_get_float_value(data, type, nb, i0, i1, i2, i3); - sum += v; - } - } - } - } - for (int64_t i3 = 0; i3 < ne[3]; i3++) { - LOG(" [\n"); - for (int64_t i2 = 0; i2 < ne[2]; i2++) { - if (i2 == n && ne[2] > 2*n) { - LOG(" ..., \n"); - i2 = ne[2] - n; - } - LOG(" [\n"); - for (int64_t i1 = 0; i1 < ne[1]; i1++) { - if (i1 == n && ne[1] > 2*n) { - LOG(" ..., \n"); - i1 = ne[1] - n; - } - LOG(" ["); - for (int64_t i0 = 0; i0 < ne[0]; i0++) { - if (i0 == n && ne[0] > 2*n) { - LOG("..., "); - i0 = ne[0] - n; - } - const float v = ggml_get_float_value(data, type, nb, i0, i1, i2, i3); - LOG("%12.4f", v); - if (i0 < ne[0] - 1) LOG(", "); - } - LOG("],\n"); - } - LOG(" ],\n"); - } - LOG(" ]\n"); - LOG(" sum = %f\n", sum); - } - - // TODO: make this abort configurable/optional? - if (std::isnan(sum)) { - LOG_ERR("encountered NaN - aborting\n"); - exit(0); - } -} - -/** - * GGML operations callback during the graph execution. - * - * @param t current tensor - * @param ask when ask is true, the scheduler wants to know if we are interested in data from this tensor - * if we return true, a follow-up call will be made with ask=false in which we can do the actual collection. - * see ggml_backend_sched_eval_callback - * @param user_data user data to pass at each call back - * @return true to receive data or continue the graph, false otherwise - */ -static bool ggml_debug(struct ggml_tensor * t, bool ask, void * user_data) { - auto * cb_data = (callback_data *) user_data; - - const struct ggml_tensor * src0 = t->src[0]; - const struct ggml_tensor * src1 = t->src[1]; - - if (ask) { - return true; // Always retrieve data - } - - char src1_str[128] = {0}; - if (src1) { - snprintf(src1_str, sizeof(src1_str), "%s{%s}", src1->name, ggml_ne_string(src1).c_str()); - } - - LOG("%s: %24s = (%s) %10s(%s{%s}, %s}) = {%s}\n", __func__, - t->name, ggml_type_name(t->type), ggml_op_desc(t), - src0->name, ggml_ne_string(src0).c_str(), - src1 ? src1_str : "", - ggml_ne_string(t).c_str()); - - - // copy the data from the GPU memory if needed - const bool is_host = ggml_backend_buffer_is_host(t->buffer); - - if (!is_host) { - auto n_bytes = ggml_nbytes(t); - cb_data->data.resize(n_bytes); - ggml_backend_tensor_get(t, cb_data->data.data(), 0, n_bytes); - } - - if (!ggml_is_quantized(t->type)) { - uint8_t * data = is_host ? (uint8_t *) t->data : cb_data->data.data(); - ggml_print_tensor(data, t->type, t->ne, t->nb, 3); - } - - return true; -} - static bool run(llama_context * ctx, const common_params & params) { const llama_model * model = llama_get_model(ctx); const llama_vocab * vocab = llama_model_get_vocab(model); @@ -182,7 +29,7 @@ static bool run(llama_context * ctx, const common_params & params) { } int main(int argc, char ** argv) { - callback_data cb_data; + base_callback_data cb_data; common_params params; @@ -197,7 +44,7 @@ int main(int argc, char ** argv) { // pass the callback to the backend scheduler // it will be executed for each node during the graph computation - params.cb_eval = ggml_debug; + params.cb_eval = ggml_debug; params.cb_eval_user_data = &cb_data; params.warmup = false; diff --git a/tools/mtmd/clip.cpp b/tools/mtmd/clip.cpp index 9c9abd8d2e..86010843a5 100644 --- a/tools/mtmd/clip.cpp +++ b/tools/mtmd/clip.cpp @@ -9,6 +9,7 @@ #include "ggml-alloc.h" #include "ggml-backend.h" #include "gguf.h" +#include "common/debug.h" #include #include @@ -159,7 +160,7 @@ struct clip_ctx { // for debugging bool debug_graph = false; - std::vector debug_print_tensors; + base_callback_data cb_data; clip_ctx(clip_context_params & ctx_params) { flash_attn_type = ctx_params.flash_attn_type; @@ -204,6 +205,10 @@ struct clip_ctx { sched.reset( ggml_backend_sched_new(backend_ptrs.data(), backend_buft.data(), backend_ptrs.size(), 8192, false, true) ); + + if (debug_graph) { + ggml_backend_sched_set_eval_callback(sched.get(), ggml_debug, &cb_data); + } } ~clip_ctx() { @@ -252,16 +257,1478 @@ clip_graph::clip_graph(clip_ctx * ctx, const clip_image_f32 & img) : gf = ggml_new_graph_custom(ctx0, ctx->max_nodes, false); } -void clip_graph::cb(ggml_tensor * cur0, const char * name, int il) const { - if (debug_graph) { - ggml_tensor * cur = ggml_cpy(ctx0, cur0, ggml_dup_tensor(ctx0, cur0)); - std::string cur_name = il >= 0 ? std::string(name) + "_" + std::to_string(il) : name; - ggml_set_name(cur, cur_name.c_str()); - ggml_set_output(cur); + ggml_cgraph * build_siglip() { + ggml_tensor * inp = build_inp(); + + ggml_tensor * learned_pos_embd = model.position_embeddings; + if (ctx->proj_type() == PROJECTOR_TYPE_LFM2) { + learned_pos_embd = resize_position_embeddings(); + } + + ggml_tensor * cur = build_vit( + inp, n_patches, + NORM_TYPE_NORMAL, + hparams.ffn_op, + learned_pos_embd, + nullptr); + + if (ctx->proj_type() == PROJECTOR_TYPE_GEMMA3) { + const int batch_size = 1; + GGML_ASSERT(n_patches_x == n_patches_y); + const int patches_per_image = n_patches_x; + const int kernel_size = hparams.n_merge; + + cur = ggml_transpose(ctx0, cur); + cur = ggml_cont_4d(ctx0, cur, patches_per_image, patches_per_image, n_embd, batch_size); + + // doing a pool2d to reduce the number of output tokens + cur = ggml_pool_2d(ctx0, cur, GGML_OP_POOL_AVG, kernel_size, kernel_size, kernel_size, kernel_size, 0, 0); + cur = ggml_reshape_3d(ctx0, cur, cur->ne[0] * cur->ne[0], n_embd, batch_size); + cur = ggml_cont(ctx0, ggml_transpose(ctx0, cur)); + + // apply norm before projection + cur = ggml_rms_norm(ctx0, cur, eps); + cur = ggml_mul(ctx0, cur, model.mm_soft_emb_norm_w); + + // apply projection + cur = ggml_mul_mat(ctx0, + ggml_cont(ctx0, ggml_transpose(ctx0, model.mm_input_proj_w)), + cur); + + } else if (ctx->proj_type() == PROJECTOR_TYPE_IDEFICS3) { + // pixel_shuffle + // https://github.com/huggingface/transformers/blob/0a950e0bbe1ed58d5401a6b547af19f15f0c195e/src/transformers/models/idefics3/modeling_idefics3.py#L578 + const int scale_factor = model.hparams.n_merge; + cur = build_patch_merge_permute(cur, scale_factor); + cur = ggml_mul_mat(ctx0, model.projection, cur); + + } else if (ctx->proj_type() == PROJECTOR_TYPE_LFM2) { + // pixel unshuffle block + const int scale_factor = model.hparams.n_merge; + cur = build_patch_merge_permute(cur, scale_factor); + + // projection + cur = ggml_norm(ctx0, cur, 1e-5); // default nn.LayerNorm + cur = ggml_mul(ctx0, cur, model.mm_input_norm_w); + cur = ggml_add(ctx0, cur, model.mm_input_norm_b); + + cur = ggml_mul_mat(ctx0, model.mm_1_w, cur); + cur = ggml_add(ctx0, cur, model.mm_1_b); + cur = ggml_gelu(ctx0, cur); + cur = ggml_mul_mat(ctx0, model.mm_2_w, cur); + cur = ggml_add(ctx0, cur, model.mm_2_b); + + } else if (ctx->proj_type() == PROJECTOR_TYPE_JANUS_PRO) { + cur = build_ffn(cur, + model.mm_0_w, model.mm_0_b, + nullptr, nullptr, + model.mm_1_w, model.mm_1_b, + hparams.ffn_op, + -1); + + } else { + GGML_ABORT("SigLIP: Unsupported projector type"); + } + + // build the graph ggml_build_forward_expand(gf, cur); - debug_print_tensors.push_back(cur); + + return gf; + } + + ggml_cgraph * build_pixtral() { + const int n_merge = hparams.n_merge; + + // 2D input positions + ggml_tensor * pos_h = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_patches); + ggml_set_name(pos_h, "pos_h"); + ggml_set_input(pos_h); + + ggml_tensor * pos_w = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_patches); + ggml_set_name(pos_w, "pos_w"); + ggml_set_input(pos_w); + + auto add_pos = [&](ggml_tensor * cur, const clip_layer &) { + return build_rope_2d(ctx0, cur, pos_h, pos_w, hparams.rope_theta, true); + }; + + ggml_tensor * inp = build_inp(); + ggml_tensor * cur = build_vit( + inp, n_patches, + NORM_TYPE_RMS, + hparams.ffn_op, + nullptr, // no learned pos embd + add_pos); + + // mistral small 3.1 patch merger + // ref: https://github.com/huggingface/transformers/blob/7a3e208892c06a5e278144eaf38c8599a42f53e7/src/transformers/models/mistral3/modeling_mistral3.py#L67 + if (model.mm_patch_merger_w) { + GGML_ASSERT(hparams.n_merge > 0); + + cur = ggml_mul(ctx0, ggml_rms_norm(ctx0, cur, eps), model.mm_input_norm_w); + + // reshape image tokens to 2D grid + cur = ggml_reshape_3d(ctx0, cur, n_embd, n_patches_x, n_patches_y); + cur = ggml_permute(ctx0, cur, 2, 0, 1, 3); // [x, y, n_embd] + cur = ggml_cont(ctx0, cur); + + // torch.nn.functional.unfold is just an im2col under the hood + // we just need a dummy kernel to make it work + ggml_tensor * kernel = ggml_view_3d(ctx0, cur, n_merge, n_merge, cur->ne[2], 0, 0, 0); + cur = ggml_im2col(ctx0, kernel, cur, n_merge, n_merge, 0, 0, 1, 1, true, inp->type); + + // project to n_embd + cur = ggml_reshape_2d(ctx0, cur, cur->ne[0], cur->ne[1] * cur->ne[2]); + cur = ggml_mul_mat(ctx0, model.mm_patch_merger_w, cur); + } + + // LlavaMultiModalProjector (always using GELU activation) + { + cur = ggml_mul_mat(ctx0, model.mm_1_w, cur); + if (model.mm_1_b) { + cur = ggml_add(ctx0, cur, model.mm_1_b); + } + + cur = ggml_gelu(ctx0, cur); + cur = ggml_mul_mat(ctx0, model.mm_2_w, cur); + if (model.mm_2_b) { + cur = ggml_add(ctx0, cur, model.mm_2_b); + } + } + + // arrangement of the [IMG_BREAK] token + if (model.token_embd_img_break) { + // not efficient, but works + // the trick is to view the embeddings as a 3D tensor with shape [n_embd, n_patches_per_row, n_rows] + // and then concatenate the [IMG_BREAK] token to the end of each row, aka n_patches_per_row dimension + // after the concatenation, we have a tensor with shape [n_embd, n_patches_per_row + 1, n_rows] + + const int p_y = n_merge > 0 ? n_patches_y / n_merge : n_patches_y; + const int p_x = n_merge > 0 ? n_patches_x / n_merge : n_patches_x; + const int p_total = p_x * p_y; + const int n_embd_text = cur->ne[0]; + const int n_tokens_output = p_total + p_y - 1; // one [IMG_BREAK] per row, except the last row + + ggml_tensor * tmp = ggml_reshape_3d(ctx0, cur, n_embd_text, p_x, p_y); + ggml_tensor * tok = ggml_new_tensor_3d(ctx0, tmp->type, n_embd_text, 1, p_y); + tok = ggml_scale(ctx0, tok, 0.0); // clear the tensor + tok = ggml_add(ctx0, tok, model.token_embd_img_break); + tmp = ggml_concat(ctx0, tmp, tok, 1); + cur = ggml_view_2d(ctx0, tmp, + n_embd_text, n_tokens_output, + ggml_row_size(tmp->type, n_embd_text), 0); + } + + // build the graph + ggml_build_forward_expand(gf, cur); + + return gf; + } + + // Qwen2VL and Qwen2.5VL use M-RoPE + ggml_cgraph * build_qwen2vl() { + GGML_ASSERT(model.patch_bias == nullptr); + GGML_ASSERT(model.class_embedding == nullptr); + + const int batch_size = 1; + const bool use_window_attn = hparams.n_wa_pattern > 0; + const int n_wa_pattern = hparams.n_wa_pattern; + const int n_pos = n_patches; + const int num_position_ids = n_pos * 4; // m-rope requires 4 dim per position + + norm_type norm_t = ctx->proj_type() == PROJECTOR_TYPE_QWEN25VL + ? NORM_TYPE_RMS // qwen 2.5 vl + : NORM_TYPE_NORMAL; // qwen 2 vl + + int mrope_sections[4] = {d_head/4, d_head/4, d_head/4, d_head/4}; + + ggml_tensor * inp_raw = build_inp_raw(); + ggml_tensor * inp = ggml_conv_2d(ctx0, model.patch_embeddings_0, inp_raw, patch_size, patch_size, 0, 0, 1, 1); + + GGML_ASSERT(img.nx % (patch_size * 2) == 0); + GGML_ASSERT(img.ny % (patch_size * 2) == 0); + + // second conv dimension + { + auto inp_1 = ggml_conv_2d(ctx0, model.patch_embeddings_1, inp_raw, patch_size, patch_size, 0, 0, 1, 1); + inp = ggml_add(ctx0, inp, inp_1); + + inp = ggml_permute(ctx0, inp, 1, 2, 0, 3); // [w, h, c, b] -> [c, w, h, b] + inp = ggml_cont_4d( + ctx0, inp, + n_embd * 2, n_patches_x / 2, n_patches_y, batch_size); + inp = ggml_reshape_4d( + ctx0, inp, + n_embd * 2, n_patches_x / 2, 2, batch_size * (n_patches_y / 2)); + inp = ggml_permute(ctx0, inp, 0, 2, 1, 3); + inp = ggml_cont_3d( + ctx0, inp, + n_embd, n_patches_x * n_patches_y, batch_size); + } + + ggml_tensor * inpL = inp; + ggml_tensor * window_mask = nullptr; + ggml_tensor * window_idx = nullptr; + ggml_tensor * inv_window_idx = nullptr; + + ggml_tensor * positions = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, num_position_ids); + ggml_set_name(positions, "positions"); + ggml_set_input(positions); + + // pre-layernorm + if (model.pre_ln_w) { + inpL = build_norm(inpL, model.pre_ln_w, model.pre_ln_b, norm_t, eps, -1); + } + + if (use_window_attn) { + // handle window attention inputs + inv_window_idx = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_pos / 4); + ggml_set_name(inv_window_idx, "inv_window_idx"); + ggml_set_input(inv_window_idx); + // mask for window attention + window_mask = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_pos, n_pos); + ggml_set_name(window_mask, "window_mask"); + ggml_set_input(window_mask); + + // if flash attn is used, we need to pad the mask and cast to f16 + if (ctx->flash_attn_type == CLIP_FLASH_ATTN_TYPE_ENABLED) { + int n_pad = GGML_PAD(window_mask->ne[1], GGML_KQ_MASK_PAD) - window_mask->ne[1]; + if (n_pad > 0) { + window_mask = ggml_pad(ctx0, window_mask, 0, n_pad, 0, 0); + } + window_mask = ggml_cast(ctx0, window_mask, GGML_TYPE_F16); + } + + // inpL shape: [n_embd, n_patches_x * n_patches_y, batch_size] + GGML_ASSERT(batch_size == 1); + inpL = ggml_reshape_2d(ctx0, inpL, n_embd * 4, n_patches_x * n_patches_y * batch_size / 4); + inpL = ggml_get_rows(ctx0, inpL, inv_window_idx); + inpL = ggml_reshape_3d(ctx0, inpL, n_embd, n_patches_x * n_patches_y, batch_size); + } + + // loop over layers + for (int il = 0; il < n_layer; il++) { + auto & layer = model.layers[il]; + const bool full_attn = use_window_attn ? (il + 1) % n_wa_pattern == 0 : true; + + ggml_tensor * cur = inpL; // inpL = residual, cur = hidden_states + + // layernorm1 + cur = build_norm(cur, layer.ln_1_w, layer.ln_1_b, norm_t, eps, il); + cb(cur, "ln1", il); + + // self-attention + { + ggml_tensor * Qcur = ggml_add(ctx0, + ggml_mul_mat(ctx0, layer.q_w, cur), layer.q_b); + ggml_tensor * Kcur = ggml_add(ctx0, + ggml_mul_mat(ctx0, layer.k_w, cur), layer.k_b); + ggml_tensor * Vcur = ggml_add(ctx0, + ggml_mul_mat(ctx0, layer.v_w, cur), layer.v_b); + + Qcur = ggml_reshape_3d(ctx0, Qcur, d_head, n_head, n_patches); + Kcur = ggml_reshape_3d(ctx0, Kcur, d_head, n_head, n_patches); + Vcur = ggml_reshape_3d(ctx0, Vcur, d_head, n_head, n_patches); + + cb(Qcur, "Qcur", il); + cb(Kcur, "Kcur", il); + cb(Vcur, "Vcur", il); + + // apply M-RoPE + Qcur = ggml_rope_multi( + ctx0, Qcur, positions, nullptr, + d_head/2, mrope_sections, GGML_ROPE_TYPE_VISION, 32768, 10000, 1, 0, 1, 32, 1); + Kcur = ggml_rope_multi( + ctx0, Kcur, positions, nullptr, + d_head/2, mrope_sections, GGML_ROPE_TYPE_VISION, 32768, 10000, 1, 0, 1, 32, 1); + + cb(Qcur, "Qcur_rope", il); + cb(Kcur, "Kcur_rope", il); + + ggml_tensor * attn_mask = full_attn ? nullptr : window_mask; + + cur = build_attn(layer.o_w, layer.o_b, + Qcur, Kcur, Vcur, attn_mask, kq_scale, il); + cb(cur, "attn_out", il); + } + + // re-add the layer input, e.g., residual + cur = ggml_add(ctx0, cur, inpL); + + inpL = cur; // inpL = residual, cur = hidden_states + + cb(cur, "ffn_inp", il); + + // layernorm2 + cur = build_norm(cur, layer.ln_2_w, layer.ln_2_b, norm_t, eps, il); + cb(cur, "ffn_inp_normed", il); + + // ffn + cur = build_ffn(cur, + layer.ff_up_w, layer.ff_up_b, + layer.ff_gate_w, layer.ff_gate_b, + layer.ff_down_w, layer.ff_down_b, + hparams.ffn_op, il); + + cb(cur, "ffn_out", il); + + // residual 2 + cur = ggml_add(ctx0, inpL, cur); + cb(cur, "layer_out", il); + + inpL = cur; + } + + // post-layernorm + if (model.post_ln_w) { + inpL = build_norm(inpL, model.post_ln_w, model.post_ln_b, norm_t, eps, n_layer); + } + + // multimodal projection + ggml_tensor * embeddings = inpL; + embeddings = ggml_reshape_3d(ctx0, embeddings, n_embd * 4, n_pos / 4, batch_size); + + embeddings = ggml_mul_mat(ctx0, model.mm_0_w, embeddings); + embeddings = ggml_add(ctx0, embeddings, model.mm_0_b); + + // GELU activation + embeddings = ggml_gelu(ctx0, embeddings); + + // Second linear layer + embeddings = ggml_mul_mat(ctx0, model.mm_1_w, embeddings); + embeddings = ggml_add(ctx0, embeddings, model.mm_1_b); + + if (use_window_attn) { + window_idx = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_pos / 4); + ggml_set_name(window_idx, "window_idx"); + ggml_set_input(window_idx); + + // embeddings shape: [n_embd, n_patches_x * n_patches_y, batch_size] + GGML_ASSERT(batch_size == 1); + embeddings = ggml_reshape_2d(ctx0, embeddings, hparams.projection_dim, n_patches_x * n_patches_y / 4); + embeddings = ggml_get_rows(ctx0, embeddings, window_idx); + embeddings = ggml_reshape_3d(ctx0, embeddings, hparams.projection_dim, n_patches_x * n_patches_y / 4, batch_size); + } + + // build the graph + ggml_build_forward_expand(gf, embeddings); + + return gf; + } + + // Qwen3VL + ggml_cgraph * build_qwen3vl() { + GGML_ASSERT(model.patch_bias != nullptr); + GGML_ASSERT(model.position_embeddings != nullptr); + GGML_ASSERT(model.class_embedding == nullptr); + + const int batch_size = 1; + const int n_pos = n_patches; + const int num_position_ids = n_pos * 4; // m-rope requires 4 dim per position + + norm_type norm_t = NORM_TYPE_NORMAL; + + int mrope_sections[4] = {d_head/4, d_head/4, d_head/4, d_head/4}; + + ggml_tensor * inp_raw = build_inp_raw(); + ggml_tensor * inp = ggml_conv_2d(ctx0, model.patch_embeddings_0, inp_raw, patch_size, patch_size, 0, 0, 1, 1); + + GGML_ASSERT(img.nx % (patch_size * 2) == 0); + GGML_ASSERT(img.ny % (patch_size * 2) == 0); + + // second conv dimension + { + auto inp_1 = ggml_conv_2d(ctx0, model.patch_embeddings_1, inp_raw, patch_size, patch_size, 0, 0, 1, 1); + inp = ggml_add(ctx0, inp, inp_1); + + inp = ggml_permute(ctx0, inp, 1, 2, 0, 3); // [w, h, c, b] -> [c, w, h, b] + inp = ggml_cont_4d( + ctx0, inp, + n_embd * 2, n_patches_x / 2, n_patches_y, batch_size); + inp = ggml_reshape_4d( + ctx0, inp, + n_embd * 2, n_patches_x / 2, 2, batch_size * (n_patches_y / 2)); + inp = ggml_permute(ctx0, inp, 0, 2, 1, 3); + inp = ggml_cont_3d( + ctx0, inp, + n_embd, n_patches_x * n_patches_y, batch_size); + } + + // add patch bias + if (model.patch_bias != nullptr) { + inp = ggml_add(ctx0, inp, model.patch_bias); + cb(inp, "patch_bias", -1); + } + + // calculate absolute position embedding and apply + ggml_tensor * learned_pos_embd = resize_position_embeddings(); + learned_pos_embd = ggml_cont_4d( + ctx0, learned_pos_embd, + n_embd * 2, n_patches_x / 2, n_patches_y, batch_size); + learned_pos_embd = ggml_reshape_4d( + ctx0, learned_pos_embd, + n_embd * 2, n_patches_x / 2, 2, batch_size * (n_patches_y / 2)); + learned_pos_embd = ggml_permute(ctx0, learned_pos_embd, 0, 2, 1, 3); + learned_pos_embd = ggml_cont_3d( + ctx0, learned_pos_embd, + n_embd, n_patches_x * n_patches_y, batch_size); + inp = ggml_add(ctx0, inp, learned_pos_embd); + cb(inp, "inp_pos_emb", -1); + + ggml_tensor * inpL = inp; + + ggml_tensor * positions = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, num_position_ids); + ggml_set_name(positions, "positions"); + ggml_set_input(positions); + + // pre-layernorm + if (model.pre_ln_w) { + inpL = build_norm(inpL, model.pre_ln_w, model.pre_ln_b, norm_t, eps, -1); + } + + // deepstack features (stack along the feature dimension), [n_embd * len(deepstack_layers), n_patches_x * n_patches_y, batch_size] + ggml_tensor * deepstack_features = nullptr; + const int merge_factor = hparams.n_merge > 0 ? hparams.n_merge * hparams.n_merge : 4; // default 2x2=4 for qwen3vl + + // loop over layers + for (int il = 0; il < n_layer; il++) { + auto & layer = model.layers[il]; + + ggml_tensor * cur = inpL; // inpL = residual, cur = hidden_states + + // layernorm1 + cur = build_norm(cur, layer.ln_1_w, layer.ln_1_b, norm_t, eps, il); + cb(cur, "ln1", il); + + // self-attention + { + cur = ggml_mul_mat(ctx0, layer.qkv_w, cur); + cur = ggml_add(ctx0, cur, layer.qkv_b); + + ggml_tensor * Qcur = ggml_view_3d(ctx0, cur, d_head, n_head, n_pos, + /* nb1 */ ggml_row_size(cur->type, d_head), + /* nb2 */ cur->nb[1], + /* offset */ 0); + + ggml_tensor * Kcur = ggml_view_3d(ctx0, cur, d_head, n_head, n_pos, + /* nb1 */ ggml_row_size(cur->type, d_head), + /* nb2 */ cur->nb[1], + /* offset */ ggml_row_size(cur->type, n_embd)); + + ggml_tensor * Vcur = ggml_view_3d(ctx0, cur, d_head, n_head, n_pos, + /* nb1 */ ggml_row_size(cur->type, d_head), + /* nb2 */ cur->nb[1], + /* offset */ ggml_row_size(cur->type, 2 * n_embd)); + + cb(Qcur, "Qcur", il); + cb(Kcur, "Kcur", il); + cb(Vcur, "Vcur", il); + + // apply M-RoPE + Qcur = ggml_rope_multi( + ctx0, Qcur, positions, nullptr, + d_head/2, mrope_sections, GGML_ROPE_TYPE_VISION, 32768, 10000, 1, 0, 1, 32, 1); + Kcur = ggml_rope_multi( + ctx0, Kcur, positions, nullptr, + d_head/2, mrope_sections, GGML_ROPE_TYPE_VISION, 32768, 10000, 1, 0, 1, 32, 1); + + cb(Qcur, "Qcur_rope", il); + cb(Kcur, "Kcur_rope", il); + + cur = build_attn(layer.o_w, layer.o_b, + Qcur, Kcur, Vcur, nullptr, kq_scale, il); + cb(cur, "attn_out", il); + } + + // re-add the layer input, e.g., residual + cur = ggml_add(ctx0, cur, inpL); + + inpL = cur; // inpL = residual, cur = hidden_states + + cb(cur, "ffn_inp", il); + + // layernorm2 + cur = build_norm(cur, layer.ln_2_w, layer.ln_2_b, norm_t, eps, il); + cb(cur, "ffn_inp_normed", il); + + // ffn + cur = build_ffn(cur, + layer.ff_up_w, layer.ff_up_b, + layer.ff_gate_w, layer.ff_gate_b, + layer.ff_down_w, layer.ff_down_b, + hparams.ffn_op, il); + + cb(cur, "ffn_out", il); + + // residual 2 + cur = ggml_add(ctx0, inpL, cur); + cb(cur, "layer_out", il); + + if (layer.has_deepstack()) { + ggml_tensor * feat = ggml_reshape_3d(ctx0, cur, n_embd * merge_factor, n_pos / merge_factor, batch_size); + feat = build_norm(feat, layer.deepstack_norm_w, layer.deepstack_norm_b, norm_t, eps, il); + feat = build_ffn(feat, + layer.deepstack_fc1_w, layer.deepstack_fc1_b, + nullptr, nullptr, + layer.deepstack_fc2_w, layer.deepstack_fc2_b, + ffn_op_type::FFN_GELU, il); + + if(!deepstack_features) { + deepstack_features = feat; + } else { + // concat along the feature dimension + deepstack_features = ggml_concat(ctx0, deepstack_features, feat, 0); + } + } + + inpL = cur; + } + + // post-layernorm + if (model.post_ln_w) { + inpL = build_norm(inpL, model.post_ln_w, model.post_ln_b, norm_t, eps, n_layer); + } + + // multimodal projection + ggml_tensor * embeddings = inpL; + embeddings = ggml_reshape_3d(ctx0, embeddings, n_embd * 4, n_pos / 4, batch_size); + + embeddings = build_ffn(embeddings, + model.mm_0_w, model.mm_0_b, + nullptr, nullptr, + model.mm_1_w, model.mm_1_b, + ffn_op_type::FFN_GELU, -1); + + embeddings = ggml_concat(ctx0, embeddings, deepstack_features, 0); // concat along the feature dimension + + // build the graph + ggml_build_forward_expand(gf, embeddings); + + return gf; + } + + ggml_cgraph * build_minicpmv() { + GGML_ASSERT(model.class_embedding == nullptr); + const int n_pos = n_patches; + const int n_embd_proj = clip_n_mmproj_embd(ctx); + + // position embeddings for the projector (not for ViT) + // see: https://huggingface.co/openbmb/MiniCPM-o-2_6/blob/main/resampler.py#L70 + // base frequency omega + ggml_tensor * omega = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, n_embd_proj / 4); + ggml_set_name(omega, "omega"); + ggml_set_input(omega); + + // 2D input positions (using float for sinusoidal embeddings) + ggml_tensor * pos_h = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, 1, n_pos); + ggml_set_name(pos_h, "pos_h"); + ggml_set_input(pos_h); + ggml_tensor * pos_w = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, 1, n_pos); + ggml_set_name(pos_w, "pos_w"); + ggml_set_input(pos_w); + + // for selecting learned pos embd, used by ViT + struct ggml_tensor * positions = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_pos); + ggml_set_name(positions, "positions"); + ggml_set_input(positions); + + ggml_tensor * learned_pos_embd = ggml_get_rows(ctx0, model.position_embeddings, positions); + + ggml_tensor * inp = build_inp(); + ggml_tensor * embeddings = build_vit( + inp, n_pos, + NORM_TYPE_NORMAL, + hparams.ffn_op, + learned_pos_embd, + nullptr); + + // resampler projector (it is just another transformer) + + ggml_tensor * q = model.mm_model_query; + ggml_tensor * v = ggml_mul_mat(ctx0, model.mm_model_kv_proj, embeddings); + + // norm + q = build_norm(q, model.mm_model_ln_q_w, model.mm_model_ln_q_b, NORM_TYPE_NORMAL, eps, -1); + v = build_norm(v, model.mm_model_ln_kv_w, model.mm_model_ln_kv_b, NORM_TYPE_NORMAL, eps, -1); + + // calculate sinusoidal pos embd + ggml_tensor * pos_embed = nullptr; + { + // outer product + ggml_tensor * omega_b = ggml_repeat_4d(ctx0, omega, omega->ne[0], n_pos, 1, 1); // n_pos rows + ggml_tensor * theta_x = ggml_mul(ctx0, omega_b, pos_w); + ggml_tensor * theta_y = ggml_mul(ctx0, omega_b, pos_h); + // sin and cos + ggml_tensor * pos_embd_x = ggml_concat( + ctx0, + ggml_sin(ctx0, theta_x), + ggml_cos(ctx0, theta_x), + 0 // concat on first dim + ); + ggml_tensor * pos_embd_y = ggml_concat( + ctx0, + ggml_sin(ctx0, theta_y), + ggml_cos(ctx0, theta_y), + 0 // concat on first dim + ); + pos_embed = ggml_concat(ctx0, pos_embd_x, pos_embd_y, 0); + } + + // k = v + pos_embed + ggml_tensor * k = ggml_add(ctx0, v, pos_embed); + + // attention + { + const int d_head = 128; + int n_head = n_embd_proj/d_head; + // Use actual config value if available, otherwise fall back to hardcoded values + int num_query = ctx->model.hparams.minicpmv_query_num; + ggml_tensor * Q = ggml_add(ctx0, + ggml_mul_mat(ctx0, model.mm_model_attn_q_w, q), + model.mm_model_attn_q_b); + ggml_tensor * K = ggml_add(ctx0, + ggml_mul_mat(ctx0, model.mm_model_attn_k_w, k), + model.mm_model_attn_k_b); + ggml_tensor * V = ggml_add(ctx0, + ggml_mul_mat(ctx0, model.mm_model_attn_v_w, v), + model.mm_model_attn_v_b); + + Q = ggml_reshape_3d(ctx0, Q, d_head, n_head, num_query); + K = ggml_reshape_3d(ctx0, K, d_head, n_head, n_pos); + V = ggml_reshape_3d(ctx0, V, d_head, n_head, n_pos); + + cb(Q, "resampler_Q", -1); + cb(K, "resampler_K", -1); + cb(V, "resampler_V", -1); + + float resampler_kq_scale = 1.0f/ sqrtf(float(d_head)); + embeddings = build_attn( + model.mm_model_attn_o_w, + model.mm_model_attn_o_b, + Q, K, V, nullptr, resampler_kq_scale, -1); + cb(embeddings, "resampler_attn_out", -1); + } + // layernorm + embeddings = build_norm(embeddings, model.mm_model_ln_post_w, model.mm_model_ln_post_b, NORM_TYPE_NORMAL, eps, -1); + + // projection + embeddings = ggml_mul_mat(ctx0, model.mm_model_proj, embeddings); + + // build the graph + ggml_build_forward_expand(gf, embeddings); + + return gf; + } + + ggml_cgraph * build_internvl() { + GGML_ASSERT(model.class_embedding != nullptr); + GGML_ASSERT(model.position_embeddings != nullptr); + + const int n_pos = n_patches + 1; + ggml_tensor * inp = build_inp(); + + // add CLS token + inp = ggml_concat(ctx0, inp, model.class_embedding, 1); + + // The larger models use a different ViT, which uses RMS norm instead of layer norm + // ref: https://github.com/ggml-org/llama.cpp/pull/13443#issuecomment-2869786188 + norm_type norm_t = (hparams.n_embd == 3200 && hparams.n_layer == 45) + ? NORM_TYPE_RMS // 6B ViT (Used by InternVL 2.5/3 - 26B, 38B, 78B) + : NORM_TYPE_NORMAL; // 300M ViT (Used by all smaller InternVL models) + + ggml_tensor * cur = build_vit( + inp, n_pos, + norm_t, + hparams.ffn_op, + model.position_embeddings, + nullptr); + + // remove CLS token + cur = ggml_view_2d(ctx0, cur, + n_embd, n_patches, + ggml_row_size(cur->type, n_embd), 0); + + // pixel shuffle + { + const int scale_factor = model.hparams.n_merge; + const int bsz = 1; // batch size, always 1 for now since we don't support batching + const int height = n_patches_y; + const int width = n_patches_x; + GGML_ASSERT(scale_factor > 0); + cur = ggml_reshape_4d(ctx0, cur, n_embd * scale_factor, height / scale_factor, width, bsz); + cur = ggml_permute(ctx0, cur, 0, 2, 1, 3); + cur = ggml_cont_4d(ctx0, cur, + n_embd * scale_factor * scale_factor, + height / scale_factor, + width / scale_factor, + bsz); + cur = ggml_permute(ctx0, cur, 0, 2, 1, 3); + // flatten to 2D + cur = ggml_cont_2d(ctx0, cur, + n_embd * scale_factor * scale_factor, + cur->ne[1] * cur->ne[2]); + } + + // projector (always using GELU activation) + { + // projector LayerNorm uses pytorch's default eps = 1e-5 + // ref: https://huggingface.co/OpenGVLab/InternVL3-8B-Instruct/blob/a34d3e4e129a5856abfd6aa6de79776484caa14e/modeling_internvl_chat.py#L79 + cur = build_norm(cur, model.mm_0_w, model.mm_0_b, NORM_TYPE_NORMAL, 1e-5, -1); + cur = ggml_mul_mat(ctx0, model.mm_1_w, cur); + cur = ggml_add(ctx0, cur, model.mm_1_b); + cur = ggml_gelu(ctx0, cur); + cur = ggml_mul_mat(ctx0, model.mm_3_w, cur); + cur = ggml_add(ctx0, cur, model.mm_3_b); + } + + // build the graph + ggml_build_forward_expand(gf, cur); + + return gf; + } + + ggml_cgraph * build_llama4() { + GGML_ASSERT(model.class_embedding != nullptr); + GGML_ASSERT(model.position_embeddings != nullptr); + + const int n_pos = n_patches + 1; // +1 for [CLS] + + // 2D input positions + ggml_tensor * pos_h = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_pos); + ggml_set_name(pos_h, "pos_h"); + ggml_set_input(pos_h); + + ggml_tensor * pos_w = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_pos); + ggml_set_name(pos_w, "pos_w"); + ggml_set_input(pos_w); + + ggml_tensor * inp = build_inp_raw(); + + // Llama4UnfoldConvolution + { + ggml_tensor * kernel = ggml_reshape_4d(ctx0, model.patch_embeddings_0, + patch_size, patch_size, 3, n_embd); + inp = ggml_im2col(ctx0, kernel, inp, patch_size, patch_size, 0, 0, 1, 1, true, inp->type); + inp = ggml_mul_mat(ctx0, model.patch_embeddings_0, inp); + inp = ggml_reshape_2d(ctx0, inp, n_embd, n_patches); + cb(inp, "patch_conv", -1); + } + + // add CLS token + inp = ggml_concat(ctx0, inp, model.class_embedding, 1); + + // build ViT with 2D position embeddings + auto add_pos = [&](ggml_tensor * cur, const clip_layer &) { + // first half is X axis and second half is Y axis + // ref: https://github.com/huggingface/transformers/blob/40a493c7ed4f19f08eadb0639cf26d49bfa5e180/src/transformers/models/llama4/modeling_llama4.py#L1312 + // ref: https://github.com/Blaizzy/mlx-vlm/blob/a57156aa87b33cca6e5ee6cfc14dd4ef8f611be6/mlx_vlm/models/llama4/vision.py#L441 + return build_rope_2d(ctx0, cur, pos_w, pos_h, hparams.rope_theta, false); + }; + ggml_tensor * cur = build_vit( + inp, n_pos, + NORM_TYPE_NORMAL, + hparams.ffn_op, + model.position_embeddings, + add_pos); + + // remove CLS token + cur = ggml_view_2d(ctx0, cur, + n_embd, n_patches, + ggml_row_size(cur->type, n_embd), 0); + + // pixel shuffle + // based on Llama4VisionPixelShuffleMLP + // https://github.com/huggingface/transformers/blob/2932f318a20d9e54cc7aea052e040164d85de7d6/src/transformers/models/llama4/modeling_llama4.py#L1151 + { + const int scale_factor = model.hparams.n_merge; + const int bsz = 1; // batch size, always 1 for now since we don't support batching + GGML_ASSERT(scale_factor > 0); + GGML_ASSERT(n_patches_x == n_patches_y); // llama4 only supports square images + cur = ggml_reshape_4d(ctx0, cur, + n_embd * scale_factor, + n_patches_x / scale_factor, + n_patches_y, + bsz); + cur = ggml_permute(ctx0, cur, 0, 2, 1, 3); + cur = ggml_cont_4d(ctx0, cur, + n_embd * scale_factor * scale_factor, + n_patches_x / scale_factor, + n_patches_y / scale_factor, + bsz); + //cur = ggml_permute(ctx0, cur, 0, 2, 1, 3); + // flatten to 2D + cur = ggml_cont_2d(ctx0, cur, + n_embd * scale_factor * scale_factor, + n_patches / scale_factor / scale_factor); + cb(cur, "pixel_shuffle", -1); + } + + // based on Llama4VisionMLP2 (always uses GELU activation, no bias) + { + cur = ggml_mul_mat(ctx0, model.mm_model_mlp_1_w, cur); + cur = ggml_gelu(ctx0, cur); + cur = ggml_mul_mat(ctx0, model.mm_model_mlp_2_w, cur); + cur = ggml_gelu(ctx0, cur); + cb(cur, "adapter_mlp", -1); + } + + // Llama4MultiModalProjector + cur = ggml_mul_mat(ctx0, model.mm_model_proj, cur); + cb(cur, "projected", -1); + + // build the graph + ggml_build_forward_expand(gf, cur); + + return gf; + } + + ggml_cgraph * build_kimivl() { + // 2D input positions + ggml_tensor * pos_h = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_patches); + ggml_set_name(pos_h, "pos_h"); + ggml_set_input(pos_h); + + ggml_tensor * pos_w = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_patches); + ggml_set_name(pos_w, "pos_w"); + ggml_set_input(pos_w); + + ggml_tensor * learned_pos_embd = resize_position_embeddings(); + + // build ViT with 2D position embeddings + auto add_pos = [&](ggml_tensor * cur, const clip_layer &) { + // first half is X axis and second half is Y axis + return build_rope_2d(ctx0, cur, pos_w, pos_h, hparams.rope_theta, false); + }; + + ggml_tensor * inp = build_inp(); + ggml_tensor * cur = build_vit( + inp, n_patches, + NORM_TYPE_NORMAL, + hparams.ffn_op, + learned_pos_embd, + add_pos); + + cb(cur, "vit_out", -1); + + { + // patch_merger + const int scale_factor = model.hparams.n_merge; + cur = build_patch_merge_permute(cur, scale_factor); + + // projection norm + int proj_inp_dim = cur->ne[0]; + cur = ggml_view_2d(ctx0, cur, + n_embd, cur->ne[1] * scale_factor * scale_factor, + ggml_row_size(cur->type, n_embd), 0); + cur = ggml_norm(ctx0, cur, 1e-5); // default nn.LayerNorm + cur = ggml_mul(ctx0, cur, model.mm_input_norm_w); + cur = ggml_add(ctx0, cur, model.mm_input_norm_b); + cur = ggml_view_2d(ctx0, cur, + proj_inp_dim, cur->ne[1] / scale_factor / scale_factor, + ggml_row_size(cur->type, proj_inp_dim), 0); + cb(cur, "proj_inp_normed", -1); + + // projection mlp + cur = ggml_mul_mat(ctx0, model.mm_1_w, cur); + cur = ggml_add(ctx0, cur, model.mm_1_b); + cur = ggml_gelu(ctx0, cur); + cur = ggml_mul_mat(ctx0, model.mm_2_w, cur); + cur = ggml_add(ctx0, cur, model.mm_2_b); + cb(cur, "proj_out", -1); + } + + // build the graph + ggml_build_forward_expand(gf, cur); + + return gf; + } + + // this graph is used by llava, granite and glm + // due to having embedding_stack (used by granite), we cannot reuse build_vit + ggml_cgraph * build_llava() { + const int batch_size = 1; + const int n_pos = n_patches + (model.class_embedding ? 1 : 0); + + GGML_ASSERT(n_patches_x == n_patches_y && "only square images supported"); + + // Calculate the deepest feature layer based on hparams and projector type + int max_feature_layer = n_layer; + { + // Get the index of the second to last layer; this is the default for models that have a llava projector + int il_last = hparams.n_layer - 1; + int deepest_feature_layer = -1; + + if (ctx->proj_type() == PROJECTOR_TYPE_MINICPMV || ctx->proj_type() == PROJECTOR_TYPE_GLM_EDGE) { + il_last += 1; + } + + // If we set explicit vision feature layers, only go up to the deepest one + // NOTE: only used by granite-vision models for now + for (const auto & feature_layer : hparams.vision_feature_layer) { + if (feature_layer > deepest_feature_layer) { + deepest_feature_layer = feature_layer; + } + } + max_feature_layer = deepest_feature_layer < 0 ? il_last : deepest_feature_layer; + } + + ggml_tensor * inp = build_inp(); + + // concat class_embeddings and patch_embeddings + if (model.class_embedding) { + inp = ggml_concat(ctx0, inp, model.class_embedding, 1); + } + + ggml_tensor * positions = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_pos); + ggml_set_name(positions, "positions"); + ggml_set_input(positions); + + inp = ggml_add(ctx0, inp, ggml_get_rows(ctx0, model.position_embeddings, positions)); + + ggml_tensor * inpL = inp; + + // pre-layernorm + if (model.pre_ln_w) { + inpL = build_norm(inpL, model.pre_ln_w, model.pre_ln_b, NORM_TYPE_NORMAL, eps, -1); + cb(inpL, "pre_ln", -1); + } + + std::vector embedding_stack; + const auto & vision_feature_layer = hparams.vision_feature_layer; + + // loop over layers + for (int il = 0; il < max_feature_layer; il++) { + auto & layer = model.layers[il]; + ggml_tensor * cur = inpL; // inpL = residual, cur = hidden_states + + // If this is an embedding feature layer, save the output. + // NOTE: 0 index here refers to the input to the encoder. + if (vision_feature_layer.find(il) != vision_feature_layer.end()) { + embedding_stack.push_back(cur); + } + + // layernorm1 + cur = build_norm(cur, layer.ln_1_w, layer.ln_1_b, NORM_TYPE_NORMAL, eps, il); + cb(cur, "layer_inp_normed", il); + + // self-attention + { + ggml_tensor * Qcur = ggml_mul_mat(ctx0, layer.q_w, cur); + if (layer.q_b) { + Qcur = ggml_add(ctx0, Qcur, layer.q_b); + } + + ggml_tensor * Kcur = ggml_mul_mat(ctx0, layer.k_w, cur); + if (layer.k_b) { + Kcur = ggml_add(ctx0, Kcur, layer.k_b); + } + + ggml_tensor * Vcur = ggml_mul_mat(ctx0, layer.v_w, cur); + if (layer.v_b) { + Vcur = ggml_add(ctx0, Vcur, layer.v_b); + } + + Qcur = ggml_reshape_3d(ctx0, Qcur, d_head, n_head, n_pos); + Kcur = ggml_reshape_3d(ctx0, Kcur, d_head, n_head, n_pos); + Vcur = ggml_reshape_3d(ctx0, Vcur, d_head, n_head, n_pos); + + cb(Qcur, "Qcur", il); + cb(Kcur, "Kcur", il); + cb(Vcur, "Vcur", il); + + cur = build_attn(layer.o_w, layer.o_b, + Qcur, Kcur, Vcur, nullptr, kq_scale, il); + cb(cur, "attn_out", il); + } + + // re-add the layer input, e.g., residual + cur = ggml_add(ctx0, cur, inpL); + + inpL = cur; // inpL = residual, cur = hidden_states + + cb(cur, "ffn_inp", il); + + // layernorm2 + cur = build_norm(cur, layer.ln_2_w, layer.ln_2_b, NORM_TYPE_NORMAL, eps, il); + cb(cur, "ffn_inp_normed", il); + + // ffn + cur = build_ffn(cur, + layer.ff_up_w, layer.ff_up_b, + layer.ff_gate_w, layer.ff_gate_b, + layer.ff_down_w, layer.ff_down_b, + hparams.ffn_op, il); + + cb(cur, "ffn_out", il); + + // residual 2 + cur = ggml_add(ctx0, inpL, cur); + cb(cur, "layer_out", il); + + inpL = cur; + } + + // post-layernorm + if (model.post_ln_w) { + inpL = build_norm(inpL, model.post_ln_w, model.post_ln_b, NORM_TYPE_NORMAL, eps, -1); + } + + ggml_tensor * embeddings = inpL; + + // process vision feature layers (used by granite) + { + // final layer is a vision feature layer + if (vision_feature_layer.find(max_feature_layer) != vision_feature_layer.end()) { + embedding_stack.push_back(inpL); + } + + // If feature layers are explicitly set, stack them (if we have multiple) + if (!embedding_stack.empty()) { + embeddings = embedding_stack[0]; + for (size_t i = 1; i < embedding_stack.size(); i++) { + embeddings = ggml_concat(ctx0, embeddings, embedding_stack[i], 0); + } + } + } + + // llava projector (also used by granite) + if (ctx->model.hparams.has_llava_projector) { + embeddings = ggml_reshape_2d(ctx0, embeddings, embeddings->ne[0], embeddings->ne[1]); + + ggml_tensor * patches = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_patches); + ggml_set_name(patches, "patches"); + ggml_set_input(patches); + + // shape [1, 576, 1024] + // ne is whcn, ne = [1024, 576, 1, 1] + embeddings = ggml_get_rows(ctx0, embeddings, patches); + + // print_tensor_info(embeddings, "embeddings"); + + // llava projector + if (ctx->proj_type() == PROJECTOR_TYPE_MLP) { + embeddings = ggml_mul_mat(ctx0, model.mm_0_w, embeddings); + embeddings = ggml_add(ctx0, embeddings, model.mm_0_b); + + embeddings = ggml_gelu(ctx0, embeddings); + if (model.mm_2_w) { + embeddings = ggml_mul_mat(ctx0, model.mm_2_w, embeddings); + embeddings = ggml_add(ctx0, embeddings, model.mm_2_b); + } + } + else if (ctx->proj_type() == PROJECTOR_TYPE_MLP_NORM) { + embeddings = ggml_mul_mat(ctx0, model.mm_0_w, embeddings); + embeddings = ggml_add(ctx0, embeddings, model.mm_0_b); + // ggml_tensor_printf(embeddings, "mm_0_w",0,true,false); + // First LayerNorm + embeddings = ggml_norm(ctx0, embeddings, eps); + embeddings = ggml_add(ctx0, ggml_mul(ctx0, embeddings, model.mm_1_w), + model.mm_1_b); + + // GELU activation + embeddings = ggml_gelu(ctx0, embeddings); + + // Second linear layer + embeddings = ggml_mul_mat(ctx0, model.mm_3_w, embeddings); + embeddings = ggml_add(ctx0, embeddings, model.mm_3_b); + + // Second LayerNorm + embeddings = ggml_norm(ctx0, embeddings, eps); + embeddings = ggml_add(ctx0, ggml_mul(ctx0, embeddings, model.mm_4_w), + model.mm_4_b); + } + else if (ctx->proj_type() == PROJECTOR_TYPE_LDP) { + // MobileVLM projector + int n_patch = 24; + ggml_tensor * mlp_1 = ggml_mul_mat(ctx0, model.mm_model_mlp_1_w, embeddings); + mlp_1 = ggml_add(ctx0, mlp_1, model.mm_model_mlp_1_b); + mlp_1 = ggml_gelu(ctx0, mlp_1); + ggml_tensor * mlp_3 = ggml_mul_mat(ctx0, model.mm_model_mlp_3_w, mlp_1); + mlp_3 = ggml_add(ctx0, mlp_3, model.mm_model_mlp_3_b); + // mlp_3 shape = [1, 576, 2048], ne = [2048, 576, 1, 1] + + // block 1 + ggml_tensor * block_1 = nullptr; + { + // transpose from [1, 576, 2048] --> [1, 2048, 576] --> [1, 2048, 24, 24] + mlp_3 = ggml_permute(ctx0, mlp_3, 1, 0, 2, 3); + mlp_3 = ggml_cont_4d(ctx0, mlp_3, n_patch, n_patch, mlp_3->ne[1], mlp_3->ne[2]); + // stride = 1, padding = 1, bias is nullptr + block_1 = ggml_conv_2d_dw(ctx0, model.mm_model_block_1_block_0_0_w, mlp_3, 1, 1, 1, 1, 1, 1); + + // layer norm + // // block_1 shape = [1, 2048, 24, 24], ne = [24, 24, 2048, 1] + block_1 = ggml_cont(ctx0, ggml_permute(ctx0, block_1, 1, 2, 0, 3)); + // block_1 shape = [1, 24, 24, 2048], ne = [2048, 24, 24, 1] + block_1 = ggml_norm(ctx0, block_1, eps); + block_1 = ggml_add(ctx0, ggml_mul(ctx0, block_1, model.mm_model_block_1_block_0_1_w), model.mm_model_block_1_block_0_1_b); + block_1 = ggml_cont(ctx0, ggml_permute(ctx0, block_1, 2, 0, 1, 3)); + + // block_1 shape = [1, 2048, 24, 24], ne = [24, 24, 2048, 1] + // hardswish + ggml_tensor * block_1_hw = ggml_hardswish(ctx0, block_1); + + block_1 = ggml_pool_2d(ctx0, block_1_hw, GGML_OP_POOL_AVG, block_1_hw->ne[0], block_1_hw->ne[1], block_1_hw->ne[0], block_1_hw->ne[1], 0, 0); + // block_1 shape = [1, 2048, 1, 1], ne = [1, 1, 2048, 1] + // pointwise conv + block_1 = ggml_reshape_2d(ctx0, block_1, block_1->ne[0]*block_1->ne[1]*block_1->ne[2], block_1->ne[3]); + block_1 = ggml_mul_mat(ctx0, model.mm_model_block_1_block_1_fc1_w, block_1); + block_1 = ggml_add(ctx0, block_1, model.mm_model_block_1_block_1_fc1_b); + block_1 = ggml_relu(ctx0, block_1); + block_1 = ggml_mul_mat(ctx0, model.mm_model_block_1_block_1_fc2_w, block_1); + block_1 = ggml_add(ctx0, block_1, model.mm_model_block_1_block_1_fc2_b); + block_1 = ggml_hardsigmoid(ctx0, block_1); + // block_1_hw shape = [1, 2048, 24, 24], ne = [24, 24, 2048, 1], block_1 shape = [1, 2048], ne = [2048, 1, 1, 1] + block_1 = ggml_reshape_4d(ctx0, block_1, 1, 1, block_1->ne[0], block_1->ne[1]); + block_1 = ggml_mul(ctx0, block_1_hw, block_1); + + int w = block_1->ne[0], h = block_1->ne[1]; + block_1 = ggml_reshape_3d(ctx0, block_1, w*h, block_1->ne[2], block_1->ne[3]); + block_1 = ggml_cont(ctx0, ggml_permute(ctx0, block_1, 1, 0, 2, 3)); + + // block_1 shape = [1, 24*24, 2048], ne = [24*24, 2048, 1] + block_1 = ggml_mul_mat(ctx0, model.mm_model_block_1_block_2_0_w, block_1); + block_1 = ggml_reshape_4d(ctx0, block_1, block_1->ne[0], w, h, block_1->ne[3]); + + // block_1 shape = [1, 24, 24, 2048], ne = [2048, 24, 24, 1] + block_1 = ggml_norm(ctx0, block_1, eps); + block_1 = ggml_add(ctx0, ggml_mul(ctx0, block_1, model.mm_model_block_1_block_2_1_w), model.mm_model_block_1_block_2_1_b); + block_1 = ggml_cont(ctx0, ggml_permute(ctx0, block_1, 2, 0, 1, 3)); + // block1 shape = [1, 2048, 24, 24], ne = [24, 24, 2048, 1] + // residual + block_1 = ggml_add(ctx0, mlp_3, block_1); + } + + // block_2 + { + // stride = 2 + block_1 = ggml_conv_2d_dw(ctx0, model.mm_model_block_2_block_0_0_w, block_1, 2, 2, 1, 1, 1, 1); + + // block_1 shape = [1, 2048, 12, 12], ne = [12, 12, 2048, 1] + // layer norm + block_1 = ggml_cont(ctx0, ggml_permute(ctx0, block_1, 1, 2, 0, 3)); + // block_1 shape = [1, 12, 12, 2048], ne = [2048, 12, 12, 1] + block_1 = ggml_norm(ctx0, block_1, eps); + block_1 = ggml_add(ctx0, ggml_mul(ctx0, block_1, model.mm_model_block_2_block_0_1_w), model.mm_model_block_2_block_0_1_b); + block_1 = ggml_cont(ctx0, ggml_permute(ctx0, block_1, 2, 0, 1, 3)); + // block_1 shape = [1, 2048, 12, 12], ne = [12, 12, 2048, 1] + // hardswish + ggml_tensor * block_1_hw = ggml_hardswish(ctx0, block_1); + + // not sure the parameters is right for globalAvgPooling + block_1 = ggml_pool_2d(ctx0, block_1_hw, GGML_OP_POOL_AVG, block_1_hw->ne[0], block_1_hw->ne[1], block_1_hw->ne[0], block_1_hw->ne[1], 0, 0); + // block_1 shape = [1, 2048, 1, 1], ne = [1, 1, 2048, 1] + // pointwise conv + block_1 = ggml_reshape_2d(ctx0, block_1, block_1->ne[0]*block_1->ne[1]*block_1->ne[2], block_1->ne[3]); + block_1 = ggml_mul_mat(ctx0, model.mm_model_block_2_block_1_fc1_w, block_1); + block_1 = ggml_add(ctx0, block_1, model.mm_model_block_2_block_1_fc1_b); + block_1 = ggml_relu(ctx0, block_1); + block_1 = ggml_mul_mat(ctx0, model.mm_model_block_2_block_1_fc2_w, block_1); + block_1 = ggml_add(ctx0, block_1, model.mm_model_block_2_block_1_fc2_b); + block_1 = ggml_hardsigmoid(ctx0, block_1); + + // block_1_hw shape = [1, 2048, 12, 12], ne = [12, 12, 2048, 1], block_1 shape = [1, 2048, 1, 1], ne = [1, 1, 2048, 1] + block_1 = ggml_reshape_4d(ctx0, block_1, 1, 1, block_1->ne[0], block_1->ne[1]); + block_1 = ggml_mul(ctx0, block_1_hw, block_1); + + int w = block_1->ne[0], h = block_1->ne[1]; + block_1 = ggml_reshape_3d(ctx0, block_1, w*h, block_1->ne[2], block_1->ne[3]); + block_1 = ggml_cont(ctx0, ggml_permute(ctx0, block_1, 1, 0, 2, 3)); + // block_1 shape = [1, 24*24, 2048], ne = [24*24, 2048, 1] + block_1 = ggml_mul_mat(ctx0, model.mm_model_block_2_block_2_0_w, block_1); + block_1 = ggml_reshape_4d(ctx0, block_1, block_1->ne[0], w, h, block_1->ne[3]); + + + // block_1 shape = [1, 12, 12, 2048], ne = [2048, 12, 12, 1] + block_1 = ggml_norm(ctx0, block_1, eps); + block_1 = ggml_add(ctx0, ggml_mul(ctx0, block_1, model.mm_model_block_2_block_2_1_w), model.mm_model_block_2_block_2_1_b); + block_1 = ggml_reshape_3d(ctx0, block_1, block_1->ne[0], block_1->ne[1] * block_1->ne[2], block_1->ne[3]); + // block_1 shape = [1, 144, 2048], ne = [2048, 144, 1] + } + embeddings = block_1; + } + else if (ctx->proj_type() == PROJECTOR_TYPE_LDPV2) + { + int n_patch = 24; + ggml_tensor * mlp_0 = ggml_mul_mat(ctx0, model.mm_model_mlp_0_w, embeddings); + mlp_0 = ggml_add(ctx0, mlp_0, model.mm_model_mlp_0_b); + mlp_0 = ggml_gelu(ctx0, mlp_0); + ggml_tensor * mlp_2 = ggml_mul_mat(ctx0, model.mm_model_mlp_2_w, mlp_0); + mlp_2 = ggml_add(ctx0, mlp_2, model.mm_model_mlp_2_b); + // mlp_2 ne = [2048, 576, 1, 1] + // // AVG Pool Layer 2*2, strides = 2 + mlp_2 = ggml_permute(ctx0, mlp_2, 1, 0, 2, 3); + // mlp_2 ne = [576, 2048, 1, 1] + mlp_2 = ggml_cont_4d(ctx0, mlp_2, n_patch, n_patch, mlp_2->ne[1], mlp_2->ne[2]); + // mlp_2 ne [24, 24, 2048, 1] + mlp_2 = ggml_pool_2d(ctx0, mlp_2, GGML_OP_POOL_AVG, 2, 2, 2, 2, 0, 0); + // weight ne = [3, 3, 2048, 1] + ggml_tensor * peg_0 = ggml_conv_2d_dw(ctx0, model.mm_model_peg_0_w, mlp_2, 1, 1, 1, 1, 1, 1); + peg_0 = ggml_cont(ctx0, ggml_permute(ctx0, peg_0, 1, 2, 0, 3)); + peg_0 = ggml_add(ctx0, peg_0, model.mm_model_peg_0_b); + mlp_2 = ggml_cont(ctx0, ggml_permute(ctx0, mlp_2, 1, 2, 0, 3)); + peg_0 = ggml_add(ctx0, peg_0, mlp_2); + peg_0 = ggml_reshape_3d(ctx0, peg_0, peg_0->ne[0], peg_0->ne[1] * peg_0->ne[2], peg_0->ne[3]); + embeddings = peg_0; + } + else { + GGML_ABORT("fatal error"); + } + } + + // glm projector + else if (ctx->proj_type() == PROJECTOR_TYPE_GLM_EDGE) { + size_t gridsz = (size_t)sqrt(embeddings->ne[1]); + embeddings = ggml_permute(ctx0,embeddings,1,0,2,3); + embeddings = ggml_cont_3d(ctx0, embeddings, gridsz, gridsz, embeddings->ne[1]); + embeddings = ggml_conv_2d(ctx0, model.mm_model_adapter_conv_w, embeddings, 2, 2, 0, 0, 1, 1); + embeddings = ggml_reshape_3d(ctx0, embeddings,embeddings->ne[0]*embeddings->ne[1] , embeddings->ne[2], batch_size); + embeddings = ggml_cont(ctx0, ggml_permute(ctx0,embeddings, 1, 0, 2, 3)); + embeddings = ggml_add(ctx0, embeddings, model.mm_model_adapter_conv_b); + // GLU + { + embeddings = ggml_mul_mat(ctx0, model.mm_model_mlp_0_w, embeddings); + embeddings = ggml_norm(ctx0, embeddings, eps); + embeddings = ggml_add(ctx0, ggml_mul(ctx0, embeddings, model.mm_model_ln_q_w), model.mm_model_ln_q_b); + embeddings = ggml_gelu_inplace(ctx0, embeddings); + ggml_tensor * x = embeddings; + embeddings = ggml_mul_mat(ctx0, model.mm_model_mlp_2_w, embeddings); + x = ggml_mul_mat(ctx0, model.mm_model_mlp_1_w,x); + embeddings = ggml_swiglu_split(ctx0, embeddings, x); + embeddings = ggml_mul_mat(ctx0, model.mm_model_mlp_3_w, embeddings); + } + // arrangement of BOI/EOI token embeddings + // note: these embeddings are not present in text model, hence we cannot process them as text tokens + // see: https://huggingface.co/THUDM/glm-edge-v-2b/blob/main/siglip.py#L53 + { + embeddings = ggml_concat(ctx0, model.mm_boi, embeddings, 1); // BOI + embeddings = ggml_concat(ctx0, embeddings, model.mm_eoi, 1); // EOI + } + } + + else { + GGML_ABORT("llava: unknown projector type"); + } + + // build the graph + ggml_build_forward_expand(gf, embeddings); + + return gf; + } + // whisper encoder with custom projector + ggml_cgraph * build_whisper_enc() { + const int n_frames = img.nx; + const int n_pos = n_frames / 2; + GGML_ASSERT(model.position_embeddings->ne[1] >= n_pos); + + ggml_tensor * inp = build_inp_raw(1); + + // conv1d block + { + // convolution + gelu + ggml_tensor * cur = ggml_conv_1d_ph(ctx0, model.conv1d_1_w, inp, 1, 1); + cur = ggml_add(ctx0, cur, model.conv1d_1_b); + + cur = ggml_gelu_erf(ctx0, cur); + + cur = ggml_conv_1d_ph(ctx0, model.conv1d_2_w, cur, 2, 1); + cur = ggml_add(ctx0, cur, model.conv1d_2_b); + + cur = ggml_gelu_erf(ctx0, cur); + // transpose + inp = ggml_cont(ctx0, ggml_transpose(ctx0, cur)); + cb(inp, "after_conv1d", -1); + } + + // sanity check (only check one layer, but it should be the same for all) + GGML_ASSERT(model.layers[0].ln_1_w && model.layers[0].ln_1_b); + GGML_ASSERT(model.layers[0].ln_2_w && model.layers[0].ln_2_b); + GGML_ASSERT(model.layers[0].q_b); + GGML_ASSERT(model.layers[0].v_b); + GGML_ASSERT(!model.layers[0].k_b); // no bias for k + GGML_ASSERT(model.post_ln_w && model.post_ln_b); + + ggml_tensor * pos_embd_selected = ggml_view_2d( + ctx0, model.position_embeddings, + model.position_embeddings->ne[0], n_pos, + model.position_embeddings->nb[1], 0 + ); + ggml_tensor * cur = build_vit( + inp, n_pos, + NORM_TYPE_NORMAL, + hparams.ffn_op, + pos_embd_selected, + nullptr); + + cb(cur, "after_transformer", -1); + + if (model.audio_has_stack_frames()) { + // StackAudioFrames + // https://huggingface.co/fixie-ai/ultravox-v0_5-llama-3_2-1b/blob/main/ultravox_model.py + int64_t stride = n_embd * hparams.proj_stack_factor; + int64_t padded_len = GGML_PAD(ggml_nelements(cur), stride); + int64_t pad = padded_len - ggml_nelements(cur); + if (pad > 0) { + cur = ggml_view_1d(ctx0, cur, ggml_nelements(cur), 0); + cur = ggml_pad(ctx0, cur, pad, 0, 0, 0); + } + cur = ggml_view_2d(ctx0, cur, stride, padded_len / stride, + ggml_row_size(cur->type, stride), 0); + cb(cur, "after_stacked", -1); + } + + if (ctx->proj_type() == PROJECTOR_TYPE_ULTRAVOX) { + // UltravoxProjector + // pre-norm + cur = ggml_rms_norm(ctx0, cur, 1e-6); + cur = ggml_mul(ctx0, cur, model.mm_norm_pre_w); + + // ffn in + cur = ggml_mul_mat(ctx0, model.mm_1_w, cur); + + // swiglu + // see SwiGLU in ultravox_model.py, the second half passed through is silu, not the first half + cur = ggml_swiglu_swapped(ctx0, cur); + + // mid-norm + cur = ggml_rms_norm(ctx0, cur, 1e-6); + cur = ggml_mul(ctx0, cur, model.mm_norm_mid_w); + + // ffn out + cur = ggml_mul_mat(ctx0, model.mm_2_w, cur); + + } else if (ctx->proj_type() == PROJECTOR_TYPE_QWEN2A) { + // projector + cur = ggml_mul_mat(ctx0, model.mm_fc_w, cur); + cur = ggml_add(ctx0, cur, model.mm_fc_b); + + } else if (ctx->proj_type() == PROJECTOR_TYPE_VOXTRAL) { + // projector + cur = ggml_mul_mat(ctx0, model.mm_1_w, cur); + cur = ggml_gelu_erf(ctx0, cur); + cur = ggml_mul_mat(ctx0, model.mm_2_w, cur); + + } else { + GGML_ABORT("%s: unknown projector type", __func__); + } + + cb(cur, "projected", -1); + + ggml_build_forward_expand(gf, cur); + + return gf; + } + + // cogvlm vision encoder + ggml_cgraph * build_cogvlm() { + GGML_ASSERT(model.class_embedding != nullptr); + GGML_ASSERT(model.position_embeddings != nullptr); + + const int n_pos = n_patches + 1; // +1 for [CLS] + + // build input and concatenate class embedding + ggml_tensor * inp = build_inp(); + inp = ggml_concat(ctx0, inp, model.class_embedding, 1); + + inp = ggml_add(ctx0, inp, model.position_embeddings); + cb(inp, "inp_pos", -1); + + ggml_tensor * inpL = inp; + + for (int il = 0; il < n_layer; il++) { + auto & layer = model.layers[il]; + ggml_tensor * cur = inpL; + + cur = ggml_mul_mat(ctx0, layer.qkv_w, cur); + + cur = ggml_add(ctx0, cur, layer.qkv_b); + + ggml_tensor * Qcur = ggml_view_3d(ctx0, cur, d_head, n_head, n_pos, d_head*sizeof(float), + cur->nb[1], 0); + ggml_tensor * Kcur = ggml_view_3d(ctx0, cur, d_head, n_head, n_pos, d_head*sizeof(float), + cur->nb[1], n_embd * sizeof(float)); + ggml_tensor * Vcur = ggml_view_3d(ctx0, cur, d_head, n_head, n_pos, d_head*sizeof(float), + cur->nb[1], 2 * n_embd * sizeof(float)); + + cb(Qcur, "Qcur", il); + cb(Kcur, "Kcur", il); + cb(Vcur, "Vcur", il); + + cur = build_attn(layer.o_w, layer.o_b, + Qcur, Kcur, Vcur, nullptr, kq_scale, il); + cb(cur, "attn_out", il); + + cur = build_norm(cur, layer.ln_1_w, layer.ln_1_b, NORM_TYPE_NORMAL, eps, il); + cb(cur, "attn_post_norm", il); + + cur = ggml_add(ctx0, cur, inpL); + inpL = cur; + + cur = build_ffn(cur, + layer.ff_up_w, layer.ff_up_b, + layer.ff_gate_w, layer.ff_gate_b, + layer.ff_down_w, layer.ff_down_b, + hparams.ffn_op, il); + + cb(cur, "ffn_out", il); + + cur = build_norm(cur, layer.ln_2_w, layer.ln_2_b, NORM_TYPE_NORMAL, eps, il); + cb(cur, "ffn_post_norm", il); + + cur = ggml_add(ctx0, cur, inpL); + cb(cur, "layer_out", il); + inpL = cur; + + } + + // remove CLS token (like build_llama4 does) + ggml_tensor * cur = ggml_view_2d(ctx0, inpL, + n_embd, n_patches, + ggml_row_size(inpL->type, n_embd), 0); + + // Multiply with mm_model_proj + cur = ggml_mul_mat(ctx0, model.mm_model_proj, cur); + + // Apply layernorm, weight, bias + cur = build_norm(cur, model.mm_post_fc_norm_w, model.mm_post_fc_norm_b, NORM_TYPE_NORMAL, 1e-5, -1); + + // Apply GELU + cur = ggml_gelu_inplace(ctx0, cur); + + // Branch 1: multiply with mm_h_to_4h_w + ggml_tensor * h_to_4h = ggml_mul_mat(ctx0, model.mm_h_to_4h_w, cur); + + // Branch 2: multiply with mm_gate_w + ggml_tensor * gate = ggml_mul_mat(ctx0, model.mm_gate_w, cur); + + // Apply silu + gate = ggml_swiglu_split(ctx0, gate, h_to_4h); + + // Apply mm_4h_to_h_w + cur = ggml_mul_mat(ctx0, model.mm_4h_to_h_w, gate); + + // Concatenate with boi and eoi + cur = ggml_concat(ctx0, model.mm_boi, cur, 1); + cur = ggml_concat(ctx0, cur, model.mm_eoi, 1); + + // build the graph + ggml_build_forward_expand(gf, cur); + + return gf; + } + +private: + // + // utility functions + // + + void cb(ggml_tensor * cur, const char * name, int il) const { + if (il >= 0) { + ggml_format_name(cur, "%s-%d", name, il); + } else { + ggml_set_name(cur, name); + } } -} // siglip2 naflex ggml_tensor * clip_graph::resize_position_embeddings(uint32_t interpolation_mode) { @@ -3208,7 +4675,6 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima } // build the inference graph - ctx->debug_print_tensors.clear(); ggml_backend_sched_reset(ctx->sched.get()); ggml_cgraph * gf = clip_image_build_graph(ctx, imgs); ggml_backend_sched_alloc_graph(ctx->sched.get(), gf); @@ -3577,18 +5043,6 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima return false; } - // print debug nodes - if (ctx->debug_graph) { - LOG_INF("\n\n---\n\n"); - LOG_INF("\n\nDebug graph:\n\n"); - for (ggml_tensor * t : ctx->debug_print_tensors) { - std::vector data(ggml_nbytes(t)); - ggml_backend_tensor_get(t, data.data(), 0, ggml_nbytes(t)); - print_tensor_shape(t); - print_tensor_data(t, data.data(), 3); - } - } - // the last node is the embedding tensor ggml_tensor * embeddings = ggml_graph_node(gf, -1); diff --git a/tools/server/public/index.html.gz b/tools/server/public/index.html.gz index a3fcf8dcdb..d674e4236c 100644 Binary files a/tools/server/public/index.html.gz and b/tools/server/public/index.html.gz differ