Skip to content

Implement DeepSeek3B-MoE-A570M (LM component) #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Nov 17, 2025
Merged

Implement DeepSeek3B-MoE-A570M (LM component) #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
eab28ed
mtmd: add DeepSeek-OCR LM support with standard attention
bluebread Nov 15, 2025
7630587
mtmd: successfully runs DeepSeek-OCR LM in llama-cli
bluebread Nov 16, 2025
2de3436
mtmd: Fix RoPE type for DeepSeek-OCR LM.
bluebread Nov 17, 2025
e8b2610
Merge branch 'sf/deepseek-ocr' of github.com:sfallah/llama.cpp into s…
bluebread Nov 17, 2025
13dc6fb
Merge branch 'sf/deepseek-ocr' into sf/deepseek-ocr
sfallah Nov 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion convert_hf_to_gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7100,7 +7100,7 @@ def set_gguf_parameters(self):
else:
# note: deepseek2 using MLA converts into MQA (ie: GQA with 1 group)
self.hparams["num_key_value_heads"] = 1

super().set_gguf_parameters()
hparams = self.hparams
kv_lora_rank = hparams["q_lora_rank"] if hparams["q_lora_rank"] is not None else 512
Expand Down
18 changes: 9 additions & 9 deletions examples/eval-callback/eval-callback.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,19 +74,19 @@ static void ggml_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne
}
}
for (int64_t i3 = 0; i3 < ne[3]; i3++) {
LOG(" [\n");
LOG(" [\n");
for (int64_t i2 = 0; i2 < ne[2]; i2++) {
if (i2 == n && ne[2] > 2*n) {
LOG(" ..., \n");
LOG(" ..., \n");
i2 = ne[2] - n;
}
LOG(" [\n");
LOG(" [\n");
for (int64_t i1 = 0; i1 < ne[1]; i1++) {
if (i1 == n && ne[1] > 2*n) {
LOG(" ..., \n");
LOG(" ..., \n");
i1 = ne[1] - n;
}
LOG(" [");
LOG(" [");
for (int64_t i0 = 0; i0 < ne[0]; i0++) {
if (i0 == n && ne[0] > 2*n) {
LOG("..., ");
Expand All @@ -98,10 +98,10 @@ static void ggml_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne
}
LOG("],\n");
}
LOG(" ],\n");
LOG(" ],\n");
}
LOG(" ]\n");
LOG(" sum = %f\n", sum);
LOG(" ]\n");
LOG(" sum = %f\n", sum);
}

// TODO: make this abort configurable/optional?
Expand Down Expand Up @@ -136,7 +136,7 @@ static bool ggml_debug(struct ggml_tensor * t, bool ask, void * user_data) {
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__,
LOG("%s: %16s = (%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 : "",
Expand Down
5 changes: 3 additions & 2 deletions src/models/deepseek2.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,6 +47,7 @@ llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_gr
// self_attention
if (is_ocr) {
const int n_embed_head = hparams.n_embd / hparams.n_head();
const int ocr_rope_type = GGML_ROPE_TYPE_NEOX;
GGML_ASSERT(n_embed_head == n_embd_head_k && n_embed_head == n_embd_head_v);

ggml_tensor * Qcur = NULL;
Expand All @@ -65,8 +66,8 @@ llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_gr
Vcur = ggml_reshape_3d(ctx0, Vcur, n_embed_head, n_head, n_tokens);

GGML_ASSERT(fabs(freq_base - 10000.0) < 1e-4);
Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_embed_head, rope_type, 0, freq_base, 1, 0, 1, 0, 0);
Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_embed_head, rope_type, 0, freq_base, 1, 0, 1, 0, 0);
Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_embed_head, ocr_rope_type, 0, freq_base, 1, 0, 1, 0, 0);
Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_embed_head, ocr_rope_type, 0, freq_base, 1, 0, 1, 0, 0);
cb(Qcur, "q_pe", il);
cb(Kcur, "k_pe", il);

Expand Down