llama + spec: MTP Support (#22673)

* spec: support MTP

* fix batch size

* rename files

* cont : simplify (#7)

* MTP: clean-up (#9)

* MTP: clean-up

* review: use llama_context_type instead of llama_graph_type

* review: remove llama_model_has_mtp

* review: fix convert issues

* convert: fix pycheck

* review: formatting

* use `mtp-` for identifying mtp models

* convert: fix mtp conversion

* mtp -> draft-mtp

* remove unused llama_arch

* add need_embd in speculative

* llama: allow partial seq_rm for GDN models for speculative decoding

Currently speculative checkpoint needs to restart from a checkpoint
after some draft tokens are not accepted, this leads to some wastage in
running the target again. This PR adds the ability to rollback upto
`draft_max` by storing the GDN intermediates.

* fix pending state

* vulkan: add GDN partial rollback

* meta: extend check to axis 1

* metal: add GDN partial rollback

Extend the gated delta net kernel to store intermediate states for
partial rollback support on the Metal backend.

- Add K (snapshot slot count) as a function constant
- Read input state from slot 0 of the 3D state tensor
- Write intermediate states to different slots during token loop
- For K=1, maintain backward-compatible single-slot behavior

Ref: https://github.com/ggml-org/llama.cpp/commit/8c05923630110223669f069af2000e9cf10c02bc

Assisted-by: llama.cpp:local pi

* delta_net_base: use ggml_pad instead of new_tensor

* review: add need_rs_seq

* review: rename part_bounded to n_rs

* review: deslop comments

* review: rename, add asserts

* server : adjust checkpoint logic (#11)

* server : adjust checkpoint logic

* cont : rm asserts

* server-context: fix early exit

* spec : fix compatibility with n-gram and add TODOs (#13)

* metal : cleanup

* llama : fix faulty bitwise check in recurrent memory

* server : disable RS-based MTP in combination with other spec types

* spec : add TODOs

* cont : fix comment

* cont : update comment

* common : fix logic for ngram + mtp compat

* llama-memory: enable checkpointing with partial rollback

* cont: add test-case for loading into a dirty ctx

* llama-memory-recurrent: clear rs_idx in clear

* download: fix mtp path

* llama-arch: fix enorm op

* docs: update docs

* conversion: fix type annotations

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

ggml/src/ggml-metal/ggml-metal-device.cpp

@@ -590,6 +590,8 @@ ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_gated_delta_net(
     const int ne20 = op->src[2]->ne[0]; // S_v
     const int ne21 = op->src[2]->ne[1]; // H
     const int ne30 = op->src[3]->ne[0]; // G
+    // state is src[5], 3D (S_v*S_v*H, K, n_seqs); K is the snapshot slot count.
+    const int K = op->src[5]->ne[1];
 
     const int nsg = op->src[2]->ne[0]/32;
 
@@ -598,7 +600,7 @@ ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_gated_delta_net(
     GGML_ASSERT(ne20 % 32 == 0);
 
     snprintf(base, 256, "kernel_gated_delta_net_%s_%d", ggml_type_name(op->src[0]->type), nsg);
-    snprintf(name, 256, "%s_ne20=%d_ne30=%d", base, ne20, ne30);
+    snprintf(name, 256, "%s_ne20=%d_ne30=%d_K=%d", base, ne20, ne30, K);
 
     ggml_metal_pipeline_with_params res = ggml_metal_library_get_pipeline(lib, name);
     if (!res.pipeline) {
@@ -606,6 +608,7 @@ ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_gated_delta_net(
 
         ggml_metal_cv_set_int16(cv, ne20, FC_GATED_DELTA_NET + 0);
         ggml_metal_cv_set_int16(cv, ne30, FC_GATED_DELTA_NET + 1);
+        ggml_metal_cv_set_int16(cv, K,    FC_GATED_DELTA_NET + 2);
 
         res = ggml_metal_library_compile_pipeline(lib, base, name, cv);