model: DSV4.1 flash (#31086)

* model : add DeepSeek-V4.1-Flash (deepseek41)

* deepseek41 : remove MXFP8, convert the Engram table to BF16

* deepseek41 : compute the compressed-KV read layers once in the cache

* deepseek41 : use get_cells() for the shared-history check, drop llama_kv_cache::seq_pos_tok_le

* deepseek41 : gather the Engram rows in-graph like qwen4exp, drop the host gather

* deepseek41 : keep the Engram hash constants in hparams.ple_*, one module per Engram layer

* deepseek41 : drop the LLAMA_DSV41_KV_MIRROR toggle

* deepseek41 : keep il last in build_hc_mixes

* deepseek41 : sort the top-k rows before the sparse gather

* tests : add deepseek41 to the Metal fusion baseline

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
This commit is contained in:
Aman GuptaandDaniel Han authored and GitHub committed 2026-10-25 09:43:51 +08:00
1 parent d327bf1f07
commit 74bca39500
39 files changed
+2798 -104

No files matched your search

+3
View File
@@ -316,6 +316,9 @@ static bool tensor_allows_quantization(const llama_model_quantize_params * param
// do not quantize the i32 token-id -> expert-id routing table (DeepSeek-V4)
quantize &= name.find("ffn_gate_tid2eid.weight") == std::string::npos;
// Engram gate is an elementwise weight (ggml_mul), not a matmul operand
quantize &= name.find("engram_gate.weight") == std::string::npos;
// these are very small (e.g. 4x4)
quantize &= name.find("altup") == std::string::npos;
quantize &= name.find("laurel") == std::string::npos;