root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-30 13:24:35 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Update gptneox-main.cpp

Browse Source
This commit is contained in:
klosax 2023-08-11 23:09:49 +02:00 committed by GitHub
parent 2a5ac7af44
commit e76c59d524
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
gptneox-main.cpp
View File

@ -565,8 +565,8 @@ bool gpt_neox_model_load(const std::string & fname, gpt_neox_model & model, gpt2
std::string blocknamestart = "transformer.blocks." + std::to_string(i) + "."; std::string blocknamestart = "transformer.blocks." + std::to_string(i) + ".";
layer.ln_1_g = get_tensor_ex(ctx, blocknamestart + "attn_norm_1.weight" ); layer.ln_1_g = get_tensor_ex(ctx, blocknamestart + "attn_norm.weight" );
layer.ln_1_b = get_tensor_ex(ctx, blocknamestart + "attn_norm_1.bias" ); layer.ln_1_b = get_tensor_ex(ctx, blocknamestart + "attn_norm.bias" );
layer.c_attn_attn_w = get_tensor_ex(ctx, blocknamestart + "attn_qkv.weight" ); layer.c_attn_attn_w = get_tensor_ex(ctx, blocknamestart + "attn_qkv.weight" );
layer.c_attn_attn_b = get_tensor_ex(ctx ,blocknamestart + "attn_qkv.bias" ); layer.c_attn_attn_b = get_tensor_ex(ctx ,blocknamestart + "attn_qkv.bias" );
@ -584,8 +584,8 @@ bool gpt_neox_model_load(const std::string & fname, gpt_neox_model & model, gpt2
layer.c_mlp_proj_b = get_tensor_ex(ctx, blocknamestart + "ffn_down.bias" ); layer.c_mlp_proj_b = get_tensor_ex(ctx, blocknamestart + "ffn_down.bias" );
// map by name // map by name
model.tensors[blocknamestart + "attn_norm_1.weight"] = layer.ln_1_g; model.tensors[blocknamestart + "attn_norm.weight"] = layer.ln_1_g;
model.tensors[blocknamestart + "attn_norm_1.bias"] = layer.ln_1_b; model.tensors[blocknamestart + "attn_norm.bias"] = layer.ln_1_b;
model.tensors[blocknamestart + "attn_qkv.weight"] = layer.c_attn_attn_w; model.tensors[blocknamestart + "attn_qkv.weight"] = layer.c_attn_attn_w;
model.tensors[blocknamestart + "attn_qkv.bias"] = layer.c_attn_attn_b; model.tensors[blocknamestart + "attn_qkv.bias"] = layer.c_attn_attn_b;