Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Sync master with upstream release b6848 #305

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 10 commits into from
Oct 27, 2025
Merged

Sync master with upstream release b6848 #305

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
5d195f1
convert : handle mmproj filename/path properly (#16760)
Galunid Oct 25, 2025
3cfa9c3
vulkan: deduplicate Microsoft Direct3D12 devices (#16689)
giladgd Oct 26, 2025
f77c13b
CUDA: General GEMV fusion (#16715)
am17an Oct 26, 2025
8d88628
docs : add Jamba to Text-only models list (#16778)
amirai21 Oct 26, 2025
7cce4f8
model : set res->t_embd in SmallThinker models (#16782)
CISC Oct 26, 2025
f696428
graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (#16…
CISC Oct 26, 2025
73a48c9
convert : enable expert group selection for all models with it (#16691)
CISC Oct 26, 2025
bbac6a2
ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs…
leejet Oct 26, 2025
bd562fe
cuda : use fast copy when src and dst are of different type and conti…
CISC Oct 26, 2025
3470a5c
ggml-alloc : make gallocr prefer chunks that allow memory reuse (#16788)
Acly Oct 26, 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
Prev Previous commit
Next Next commit
convert : enable expert group selection for all models with it (ggml-… 
…org#16691)
  • Loading branch information
@CISC
CISC authored Oct 26, 2025
commit 73a48c9790d320476b3e5ef75bda09f2f8269e6e
8 changes: 6 additions & 2 deletions convert_hf_to_gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -742,6 +742,12 @@ def set_gguf_parameters(self):
if (n_experts_used := self.hparams.get("num_experts_per_tok")) is not None:
self.gguf_writer.add_expert_used_count(n_experts_used)
logger.info(f"gguf: experts used count = {n_experts_used}")
if (n_expert_groups := self.hparams.get("n_group")) is not None:
self.gguf_writer.add_expert_group_count(n_expert_groups)
logger.info(f"gguf: expert groups count = {n_expert_groups}")
if (n_group_used := self.hparams.get("topk_group")) is not None:
self.gguf_writer.add_expert_group_used_count(n_group_used)
logger.info(f"gguf: expert groups used count = {n_group_used}")

if (head_dim := self.hparams.get("head_dim")) is not None:
self.gguf_writer.add_key_length(head_dim)
Expand Down Expand Up @@ -8233,8 +8239,6 @@ def set_gguf_parameters(self):
self.gguf_writer.add_expert_weights_scale(hparams["routed_scaling_factor"])
self.gguf_writer.add_expert_count(hparams["num_experts"])
self.gguf_writer.add_expert_shared_count(hparams["num_shared_experts"])
self.gguf_writer.add_expert_group_count(hparams["n_group"])
self.gguf_writer.add_expert_group_used_count(hparams["topk_group"])
self.gguf_writer.add_expert_weights_norm(hparams["norm_topk_prob"])

if hparams["score_function"] == "sigmoid":
Expand Down
4 changes: 2 additions & 2 deletions src/llama-model.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6369,6 +6369,8 @@ void llama_model::print_info() const {
LLAMA_LOG_INFO("%s: n_ff = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_ff(il); }, hparams.n_layer).c_str());
LLAMA_LOG_INFO("%s: n_expert = %u\n", __func__, hparams.n_expert);
LLAMA_LOG_INFO("%s: n_expert_used = %u\n", __func__, hparams.n_expert_used);
LLAMA_LOG_INFO("%s: n_expert_groups = %d\n", __func__, hparams.n_expert_groups);
LLAMA_LOG_INFO("%s: n_group_used = %d\n", __func__, hparams.n_group_used);
LLAMA_LOG_INFO("%s: causal attn = %d\n", __func__, hparams.causal_attn);
LLAMA_LOG_INFO("%s: pooling type = %d\n", __func__, hparams.pooling_type);
LLAMA_LOG_INFO("%s: rope type = %d\n", __func__, hparams.rope_type);
Expand Down Expand Up @@ -6469,8 +6471,6 @@ void llama_model::print_info() const {
LLAMA_LOG_INFO("%s: n_ff_exp = %d\n", __func__, hparams.n_ff_exp);
LLAMA_LOG_INFO("%s: n_ff_shexp = %d\n", __func__, hparams.n_ff_shexp);
LLAMA_LOG_INFO("%s: n_expert_shared = %d\n", __func__, hparams.n_expert_shared);
LLAMA_LOG_INFO("%s: n_expert_groups = %d\n", __func__, hparams.n_expert_groups);
LLAMA_LOG_INFO("%s: n_group_used = %d\n", __func__, hparams.n_group_used);
LLAMA_LOG_INFO("%s: expert_weights_scale = %.1f\n", __func__, hparams.expert_weights_scale);
LLAMA_LOG_INFO("%s: expert_weights_norm = %d\n", __func__, hparams.expert_weights_norm);
LLAMA_LOG_INFO("%s: expert_gating_func = %s\n", __func__, llama_expert_gating_func_name((llama_expert_gating_func_type) hparams.expert_gating_func));
Expand Down