Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

convert : refactor Hy4-preview conversion - move HC tensor mapping to the global map (#28451)

Browse Source 
Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com>
This commit is contained in:
fairydreaming
2026-09-07 15:20:58 +02:00
committed by GitHub
co-authored by Stanisław Szymczyk
parent 73ab7599b5
commit d23c47f2a9
2 changed files with 60 additions and 89 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gguf-py/gguf/tensor_mapping.py
+38
View File
@@ -385,6 +385,7 @@ class TensorNameMap:
MODEL_TENSOR.ATTN_SINKS: (
"model.layers.{bid}.self_attn.sinks", # openai-moe
"model.layers.{bid}.self_attn.attention_sink_bias", # mimov2
"model.layers.{bid}.self_attn.learnable_sink_param", # hy-v4
),
MODEL_TENSOR.ATTN_GATE: (
@@ -392,6 +393,7 @@ class TensorNameMap:
"model.layers.{bid}.linear_attn.in_proj_z", # qwen3.5
"model.layers.{bid}.self_attn.g_proj", # step3.5 head-wise attention gate
"model.layers.{bid}.self_attn.output_gate", # minimax-01
"model.layers.{bid}.self_attn.linear_gate", # hy-v4
),
# Feed-forward norm
@@ -1329,6 +1331,42 @@ class TensorNameMap:
"model.layers.{bid}.self_attn.index_q_norm", # MSA
),
MODEL_TENSOR.HC_ATTN_FN: (
"model.layers.{bid}.hc_attn_layer.hc_pre.hc_fn", # hy-v4
),
MODEL_TENSOR.HC_ATTN_BASE: (
"model.layers.{bid}.hc_attn_layer.hc_pre.hc_base", # hy-v4
),
MODEL_TENSOR.HC_ATTN_SCALE: (
"model.layers.{bid}.hc_attn_layer.hc_pre.hc_scale", # hy-v4
),
MODEL_TENSOR.HC_FFN_FN: (
"model.layers.{bid}.hc_mlp_layer.hc_pre.hc_fn", # hy-v4
),
MODEL_TENSOR.HC_FFN_BASE: (
"model.layers.{bid}.hc_mlp_layer.hc_pre.hc_base", # hy-v4
),
MODEL_TENSOR.HC_FFN_SCALE: (
"model.layers.{bid}.hc_mlp_layer.hc_pre.hc_scale", # hy-v4
),
MODEL_TENSOR.HC_HEAD_FN: (
"model.hc_head.hc_head_fn", # hy-v4
),
MODEL_TENSOR.HC_HEAD_BASE: (
"model.hc_head.hc_head_base", # hy-v4
),
MODEL_TENSOR.HC_HEAD_SCALE: (
"model.hc_head.hc_head_scale", # hy-v4
),
############################################################################
# TODO: these do not belong to block_mappings_cfg - move them to mappings_cfg
MODEL_TENSOR.ENC_OUTPUT_NORM: (