ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-06-14 12:58:10 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

sync : ggml (ggml_scale, ggml_row_size, etc.) (#1677)

Browse Source 
* sync : ggml

* sync : llama.cpp

* talk-llama : fix obsolete param

* ggml-alloc : fix ggml_tallocr_is_own

* talk.wasm : update to new ggml

* ggml : fix type punning in ggml_scale

* ggml : cuda jetson + arm quants warnings
This commit is contained in:
Georgi Gerganov
2023-12-22 17:53:39 +02:00
committed by GitHub
parent d2ee117a0a
commit 3a5302108d
18 changed files with 3479 additions and 1537 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
examples/talk.wasm/gpt-2.cpp
View File

@ -155,33 +155,33 @@ bool gpt2_model_load(const std::string & fname, gpt2_model & model, gpt_vocab &
const int n_ctx = hparams.n_ctx;
const int n_vocab = hparams.n_vocab;
ctx_size += n_embd*ggml_type_sizef(GGML_TYPE_F32); // ln_f_g
ctx_size += n_embd*ggml_type_sizef(GGML_TYPE_F32); // ln_f_b
ctx_size += ggml_row_size(GGML_TYPE_F32, n_embd); // ln_f_g
ctx_size += ggml_row_size(GGML_TYPE_F32, n_embd); // ln_f_b
ctx_size += n_vocab*n_embd*ggml_type_sizef(wtype); // wte
ctx_size += n_ctx*n_embd*ggml_type_sizef(GGML_TYPE_F32); // wpe
ctx_size += n_vocab*n_embd*ggml_type_sizef(wtype); // lm_head
ctx_size += n_vocab*ggml_row_size(wtype, n_embd); // wte
ctx_size += n_ctx*ggml_row_size(GGML_TYPE_F32, n_embd); // wpe
ctx_size += n_vocab*ggml_row_size(wtype, n_embd); // lm_head
ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // ln_1_g
ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // ln_1_b
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, n_embd)); // ln_1_g
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, n_embd)); // ln_1_b
ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // ln_2_g
ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // ln_2_b
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, n_embd)); // ln_2_g
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, n_embd)); // ln_2_b
ctx_size += n_layer*(3*n_embd*n_embd*ggml_type_sizef(wtype)); // c_attn_attn_w
ctx_size += n_layer*( 3*n_embd*ggml_type_sizef(GGML_TYPE_F32)); // c_attn_attn_b
ctx_size += n_layer*(ggml_row_size(wtype, 3*n_embd*n_embd)); // c_attn_attn_w
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, 3*n_embd)); // c_attn_attn_b
ctx_size += n_layer*(n_embd*n_embd*ggml_type_sizef(wtype)); // c_attn_proj_w
ctx_size += n_layer*( n_embd*ggml_type_sizef(GGML_TYPE_F32)); // c_attn_proj_b
ctx_size += n_layer*(ggml_row_size(wtype, n_embd*n_embd)); // c_attn_proj_w
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, n_embd)); // c_attn_proj_b
ctx_size += n_layer*(4*n_embd*n_embd*ggml_type_sizef(wtype)); // c_mlp_fc_w
ctx_size += n_layer*( 4*n_embd*ggml_type_sizef(GGML_TYPE_F32)); // c_mlp_fc_b
ctx_size += n_layer*(ggml_row_size(wtype, 4*n_embd*n_embd)); // c_mlp_fc_w
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, 4*n_embd)); // c_mlp_fc_b
ctx_size += n_layer*(4*n_embd*n_embd*ggml_type_sizef(wtype)); // c_mlp_proj_w
ctx_size += n_layer*( n_embd*ggml_type_sizef(GGML_TYPE_F32)); // c_mlp_proj_b
ctx_size += n_layer*(ggml_row_size(wtype, 4*n_embd*n_embd)); // c_mlp_proj_w
ctx_size += n_layer*(ggml_row_size(GGML_TYPE_F32, n_embd)); // c_mlp_proj_b
ctx_size += n_ctx*n_layer*n_embd*ggml_type_sizef(GGML_TYPE_F32); // memory_k
ctx_size += n_ctx*n_layer*n_embd*ggml_type_sizef(GGML_TYPE_F32); // memory_v
ctx_size += n_ctx*n_layer*ggml_row_size(GGML_TYPE_F32, n_embd); // memory_k
ctx_size += n_ctx*n_layer*ggml_row_size(GGML_TYPE_F32, n_embd); // memory_v
ctx_size += (6 + 12*n_layer)*256; // object overhead
@ -524,8 +524,7 @@ bool gpt2_eval(
struct ggml_tensor * KQ_scaled =
ggml_scale(ctx0,
KQ,
ggml_new_f32(ctx0, 1.0f/sqrt(float(n_embd)/n_head))
);
1.0f/sqrt(float(n_embd)/n_head));
// KQ_masked = mask_past(KQ_scaled)
// [n_past + N, N, 12]