llama, mtmd: follow the reference precision of ggml_prec_set_acc calls (#31851)
* llama, mtmd: follow the reference precision of ggml_prec_set_acc calls Requests that the reference implementation does not keep in F32 move to GGML_PREC_BF16: the GLM4 family down and output projections, the MiMoVL vision matmuls and the Qwen3-TTS speaker encoder conv1d. The DeepSeek V4/V4.1 routers and the MiniMax-M3 indexer stay in F32 with a link to the reference line that computes them in F32. * ggml: handle GGML_PREC_BF16 the same way as GGML_PREC_F32 in backends Backends that reject GGML_PREC_F32 on the src1 of MUL_MAT_ID now also reject GGML_PREC_BF16 (CUDA, Hexagon, OpenCL, WebGPU, Vulkan), and backends that accept GGML_PREC_F32 for flash attention now also accept GGML_PREC_BF16 with the same F32 accumulation (CPU, SpaceMIT, ET). Vulkan keeps the F32 transfer between devices for both. The CUDA cuBLAS path already handles GGML_PREC_BF16 and is unchanged. * tests: cover GGML_PREC_BF16 in the flash attention tests The flash attention cases run with GGML_PREC_BF16 next to F32 and DEFAULT, and var_to_str prints bf16 so these cases keep their own name. * llama: request GGML_PREC_BF16 for the KQ multiplication * cont : add TODO in clip.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
1 parent
a2b301831d
commit
d6be065572
17 files changed
+32
-27
No files matched your search
@@ -875,7 +875,7 @@ llama_model_deepseek41::graph::graph(const llm_graph_params & params) :
|
||||
cur = build_norm(cur, layer.ffn_norm, nullptr, LLM_NORM_RMS, il);
|
||||
cb(cur, "ffn_norm", il);
|
||||
|
||||
// F32 router accumulation: expert selection is near-tie sensitive
|
||||
// the reference computes the router in F32: https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017/blob/1bae0e1088993b9b82de08f59de5be5a34e42170/inference/model.py#L813
|
||||
ggml_tensor * logits = build_lora_mm(layer.ffn_gate_inp, cur);
|
||||
ggml_prec_set_acc(logits, GGML_PREC_F32);
|
||||
cb(logits, "ffn_moe_logits", il);
|
||||
|
||||
Reference in new issue
Block a user