models : add Nemotron Super 3.5 VL and Nemotron Bolt 74B A9B support (#30520)

* add support

* remove unneeded bos fix

* add video flag

* fix flake error

* remove video-specific budgeting

* convert : fix lilicorr-dflash bid clash

* spec : enforce lilicorr n_tokens condition

* spec : fix lilicorr walk

[no release]
This commit is contained in:
Kartik Gulia authored and GitHub committed 2026-10-15 16:34:56 +03:00
1 parent a11b6530b5
commit cd0a5b6a18
15 files changed
+265 -22

No files matched your search

+2 -2
View File
@@ -930,7 +930,7 @@ llama_model_dflash::graph<false>::graph(const llama_model & model, const llm_gra
res->t_inp_tokens = inp_tokens;
ggml_tensor * inp_anchor = nullptr;
if (model.lilicorr_pass_hidden_proj && ubatch.embd_state) {
if (model.lilicorr_pass_hidden_proj && ubatch.embd_state && n_tokens == (int64_t) hparams.dflash_block_size * ubatch.n_seqs_unq) {
// LiliCorr sets inp_anchor to pass the context features to the lattice scorer
auto inp = std::make_unique<llm_graph_input_embd_h>(n_embd, n_embd);
inp->tokens = inp_tokens;
@@ -1100,7 +1100,7 @@ llama_model_dflash::graph<false>::graph(const llama_model & model, const llm_gra
if (model.dflash_selector_hidden) {
build_dflash2_selector(*this, model, inp_tokens);
} else if (model.lilicorr_pass_hidden_proj && ubatch.embd_state) {
} else if (model.lilicorr_pass_hidden_proj && ubatch.embd_state && n_tokens == (int64_t) hparams.dflash_block_size * ubatch.n_seqs_unq) {
build_lilicorr_scores(*this, model, tok_embd, inp_anchor);
}
}