model: mistral le chonk (#31376)

* llama: extend DSA and sdpark to new model

* fix ci
This commit is contained in:
Xuan-Son Nguyen authored and GitHub committed 2026-10-29 15:43:33 +01:00
1 parent 0ac1a7fd35
commit 78feec062d
23 files changed
+1290 -35

No files matched your search

+11
View File
@@ -155,6 +155,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
{ LLM_ARCH_EAGLE3, "eagle3" },
{ LLM_ARCH_DFLASH, "dflash" },
{ LLM_ARCH_MISTRAL4, "mistral4" },
{ LLM_ARCH_MISTRAL_LARGE4, "mistral-large4" },
{ LLM_ARCH_PADDLEOCR, "paddleocr" },
{ LLM_ARCH_MIMO2, "mimo2" },
{ LLM_ARCH_STEP35, "step35" },
@@ -330,6 +331,7 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_ATTENTION_COMPRESS_ROPE_FREQ_BASE, "%s.attention.compress_rope_freq_base" },
{ LLM_KV_ATTENTION_COMPRESS_RATIOS, "%s.attention.compress_ratios" },
{ LLM_KV_ATTENTION_SHARED_KV_LAYERS, "%s.attention.shared_kv_layers" },
{ LLM_KV_ATTENTION_KV_SOURCE_MAP, "%s.attention.kv_source_map" },
{ LLM_KV_ATTENTION_RECURRENT_LAYERS, "%s.attention.recurrent_layers" },
{ LLM_KV_HYPER_CONNECTION_COUNT, "%s.hyper_connection.count" },
@@ -779,6 +781,7 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
{ LLM_TENSOR_POS_NET_ATTN_V, "posnet.%d.attn_v" },
{ LLM_TENSOR_POS_NET_ATTN_OUT, "posnet.%d.attn_output" },
{ LLM_TENSOR_ATTN_SINKS, "blk.%d.attn_sinks" },
{ LLM_TENSOR_ATTN_SSMAX, "blk.%d.attn_ssmax" },
{ LLM_TENSOR_SHORTCONV_CONV, "blk.%d.shortconv.conv" },
{ LLM_TENSOR_SHORTCONV_INPROJ, "blk.%d.shortconv.in_proj" },
{ LLM_TENSOR_SHORTCONV_OUTPROJ, "blk.%d.shortconv.out_proj" },
@@ -797,6 +800,7 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
{ LLM_TENSOR_INDEXER_Q_PROJ, "blk.%d.indexer.q_proj" },
{ LLM_TENSOR_INDEXER_K_PROJ, "blk.%d.indexer.k_proj" },
{ LLM_TENSOR_INDEXER_Q_NORM, "blk.%d.indexer.q_norm" },
{ LLM_TENSOR_INDEXER_SSMAX, "blk.%d.indexer.ssmax" },
{ LLM_TENSOR_INDEXER_COMPRESSOR_WKV, "blk.%d.indexer_compressor_kv" },
{ LLM_TENSOR_INDEXER_COMPRESSOR_WGATE, "blk.%d.indexer_compressor_gate" },
{ LLM_TENSOR_INDEXER_COMPRESSOR_APE, "blk.%d.indexer_compressor_ape" },
@@ -811,6 +815,8 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
{ LLM_TENSOR_DSPARK_MARKOV_W1, "markov_w1" },
{ LLM_TENSOR_DSPARK_MARKOV_W2, "markov_w2" },
{ LLM_TENSOR_DSPARK_CONF_PROJ, "conf_proj" },
{ LLM_TENSOR_DSPARK_JOINT_PROJ, "joint_proj" },
{ LLM_TENSOR_DSPARK_MASK_EMBD, "mask_embd" },
{ LLM_TENSOR_ENGRAM_KV, "blk.%d.engram_kv" },
{ LLM_TENSOR_ENGRAM_GATE, "blk.%d.engram_gate" },
{ LLM_TENSOR_DFLASH_ATTN_CONV_BASE, "blk.%d.attn_conv_base" },
@@ -943,6 +949,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_ATTN_K_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_ATTN_V_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_ATTN_SINKS, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_SCALE}},
{LLM_TENSOR_ATTN_SSMAX, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_DEC_ATTN_Q, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DEC_ATTN_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DEC_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
@@ -1124,6 +1131,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_INDEXER_Q_PROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_INDEXER_K_PROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_INDEXER_Q_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_INDEXER_SSMAX, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_INDEXER_COMPRESSOR_WKV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_INDEXER_COMPRESSOR_WGATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_INDEXER_COMPRESSOR_APE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_GET_ROWS}},
@@ -1162,6 +1170,8 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_DSPARK_MARKOV_W1, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_GET_ROWS}},
{LLM_TENSOR_DSPARK_MARKOV_W2, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DSPARK_CONF_PROJ, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DSPARK_JOINT_PROJ, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DSPARK_MASK_EMBD, {LLM_TENSOR_LAYER_INPUT, GGML_OP_GET_ROWS}},
{LLM_TENSOR_ENGRAM_KV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_ENGRAM_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_DFLASH_ATTN_CONV_BASE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
@@ -1370,6 +1380,7 @@ bool llm_arch_supports_sm_tensor(const llm_arch & arch) {
case LLM_ARCH_MINIMAX_M2:
case LLM_ARCH_MINIMAX_M3:
case LLM_ARCH_MISTRAL4:
case LLM_ARCH_MISTRAL_LARGE4:
case LLM_ARCH_MELLUM: // [TAG_MOE_AMAX_SCALE]
case LLM_ARCH_FLINT: // [TAG_MOE_AMAX_SCALE]
case LLM_ARCH_LAGUNA: // [TAG_MOE_AMAX_SCALE]