Movatterモバイル変換


[0]ホーム

URL:


Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Commitfbd42b6

Browse files
authored
fix: fix embeddings with quantized models (#601)
1 parent19d876e commitfbd42b6

File tree

1 file changed

+1
-1
lines changed

1 file changed

+1
-1
lines changed

‎clip.hpp‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -546,7 +546,7 @@ class CLIPEmbeddings : public GGMLBlock {
546546
int64_t num_positions;
547547

548548
voidinit_params(structggml_context* ctx, std::map<std::string,enum ggml_type>& tensor_types,const std::string prefix ="") {
549-
enum ggml_type token_wtype = (tensor_types.find(prefix +"token_embedding.weight") != tensor_types.end()) ? tensor_types[prefix +"token_embedding.weight"] : GGML_TYPE_F32;
549+
enum ggml_type token_wtype =GGML_TYPE_F32;//(tensor_types.find(prefix + "token_embedding.weight") != tensor_types.end()) ? tensor_types[prefix + "token_embedding.weight"] : GGML_TYPE_F32;
550550
enum ggml_type position_wtype = GGML_TYPE_F32;//(tensor_types.find(prefix + "position_embedding.weight") != tensor_types.end()) ? tensor_types[prefix + "position_embedding.weight"] : GGML_TYPE_F32;
551551

552552
params["token_embedding.weight"] =ggml_new_tensor_2d(ctx, token_wtype, embed_dim, vocab_size);

0 commit comments

Comments
 (0)

[8]ページ先頭

©2009-2025 Movatter.jp