Skip to content

Support conversion of Qwen3-Embedding models #15023

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 2, 2025
Merged

Support conversion of Qwen3-Embedding models #15023

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions convert_hf_to_gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -849,6 +849,9 @@ def get_vocab_base_pre(self, tokenizer) -> str:
if chkhsh == "2085e1638f6c377a0aa4ead21b27bb4cb941bf800df86ed391011769c1758dfb":
# ref: https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B
res = "exaone4"
if chkhsh == "d4540891389ea895b53b399da6ac824becc30f2fba0e9ddbb98f92e55ca0e97c":
# ref: https://huggingface.co/Qwen/Qwen3-Embedding-8B
res = "qwen2"

if res is None:
logger.warning("\n")
Expand Down
16 changes: 14 additions & 2 deletions gguf-py/gguf/tensor_mapping.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,7 @@ class TensorNameMap:
"language_model.model.embed_tokens", # llama4
"encoder", # neobert
"model.transformer.wte", # llada
"embed_tokens", # qwen3-embedding
),

# Token type embeddings
Expand Down Expand Up @@ -143,6 +144,7 @@ class TensorNameMap:
"transformer_encoder.{bid}.attention_norm", # neobert
"model.layers.{bid}.operator_norm", # lfm2
"model.transformer.blocks.{bid}.attn_norm", # llada
"layers.{bid}.input_layernorm", # qwen3-embedding
),

# Attention norm 2
Expand Down Expand Up @@ -188,6 +190,7 @@ class TensorNameMap:
"transformer.h.{bid}.attn.attention.q_proj", # exaone
"model.layers.{bid}.self_attn.q_proj", # llama4
"model.transformer.blocks.{bid}.q_proj", # llada
"layers.{bid}.self_attn.q_proj", # qwen3-embedding
),

# Attention key
Expand All @@ -205,6 +208,7 @@ class TensorNameMap:
"transformer.h.{bid}.attn.attention.k_proj", # exaone
"model.layers.{bid}.self_attn.k_proj", # llama4
"model.transformer.blocks.{bid}.k_proj", # llada
"layers.{bid}.self_attn.k_proj", # qwen3-embedding
),

# Attention value
Expand All @@ -221,6 +225,7 @@ class TensorNameMap:
"transformer.h.{bid}.attn.attention.v_proj", # exaone
"model.layers.{bid}.self_attn.v_proj", # llama4
"model.transformer.blocks.{bid}.v_proj", # llada
"layers.{bid}.self_attn.v_proj", # qwen3-embedding
),

# Attention output
Expand Down Expand Up @@ -254,6 +259,7 @@ class TensorNameMap:
"model.layers.{bid}.self_attn.o_proj", # llama4
"transformer_encoder.{bid}.wo", # neobert
"model.transformer.blocks.{bid}.attn_out", # llada
"layers.{bid}.self_attn.o_proj", # qwen3-embedding
),

# Attention output norm
Expand Down Expand Up @@ -300,6 +306,7 @@ class TensorNameMap:
"transformer_encoder.{bid}.ffn_norm", # neobert
"model.layers.layers.{bid}.pre_mlp_norm", # plamo2
"model.transformer.blocks.{bid}.ff_norm", # llada
"layers.{bid}.post_attention_layernorm", # qwen3-embedding
),

# Post feed-forward norm
Expand Down Expand Up @@ -373,7 +380,8 @@ class TensorNameMap:
"model.layers.{bid}.feed_forward.up_proj", # llama4 jamba granite-hybrid
"transformer_encoder.{bid}.ffn.w12", # neobert
"model.layers.{bid}.block_sparse_moe.up", # smallthinker
"model.transformer.blocks.{bid}.up_proj", # llada
"model.transformer.blocks.{bid}.up_proj", # llada
"layers.{bid}.mlp.up_proj", # qwen3-embedding
),

MODEL_TENSOR.FFN_UP_EXP: (
Expand Down Expand Up @@ -416,6 +424,7 @@ class TensorNameMap:
"model.layers.{bid}.feed_forward.gate_proj", # llama4 jamba granite-hybrid
"model.layers.{bid}.block_sparse_moe.gate", # smallthinker
"model.transformer.blocks.{bid}.ff_proj", # llada
"layers.{bid}.mlp.gate_proj", # qwen3-embedding
),

MODEL_TENSOR.FFN_GATE_EXP: (
Expand Down Expand Up @@ -465,7 +474,8 @@ class TensorNameMap:
"model.layers.{bid}.feed_forward.down_proj", # llama4 jamba granite-hybrid
"transformer_encoder.{bid}.ffn.w3", # neobert
"model.layers.{bid}.block_sparse_moe.down", # smallthinker
"model.transformer.blocks.{bid}.ff_out", # llada
"model.transformer.blocks.{bid}.ff_out", # llada
"layers.{bid}.mlp.down_proj", # qwen3-embedding
),

MODEL_TENSOR.FFN_DOWN_EXP: (
Expand Down Expand Up @@ -497,6 +507,7 @@ class TensorNameMap:
"encoder.layer.{bid}.attention.self.layer_norm_q", # jina-bert-v2
"transformer.layers.{bid}.attn.q_norm", # openelm
"model.layers.layers.{bid}.mixer.q", # plamo2
"layers.{bid}.self_attn.q_norm", # qwen3-embedding
),

MODEL_TENSOR.ATTN_K_NORM: (
Expand All @@ -508,6 +519,7 @@ class TensorNameMap:
"encoder.layer.{bid}.attention.self.layer_norm_k", # jina-bert-v2
"transformer.layers.{bid}.attn.k_norm", # openelm
"model.layers.layers.{bid}.mixer.k", # plamo2
"layers.{bid}.self_attn.k_norm", # qwen3-embedding
),

MODEL_TENSOR.ROPE_FREQS: (
Expand Down
1 change: 1 addition & 0 deletions src/llama-model.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -899,6 +899,7 @@ void llama_model::load_hparams(llama_model_loader & ml) {
} break;
case LLM_ARCH_QWEN3:
{
ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type, false);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 28: type = hparams.n_embd == 1024 ? LLM_TYPE_0_6B : LLM_TYPE_1_7B; break;
Expand Down
Loading