vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

adapt to recent changes

Browse source
This commit is contained in:
manikbhandari 2023-12-26 07:28:22 -05:00
parent c4d1c26710
commit 6a2a13b51c
2 changed files with 2 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

1
README.md
View file

@ -103,6 +103,7 @@ as the main playground for developing new features for the [ggml](https://github
- [x] [Qwen models](https://huggingface.co/models?search=Qwen/Qwen)
- [x] [Mixtral MoE](https://huggingface.co/models?search=mistral-ai/Mixtral)
- [x] [PLaMo-13B](https://github.com/ggerganov/llama.cpp/pull/3557)
- [x] [GPT-2](https://huggingface.co/gpt2)
**Multimodal models:**

20
llama.cpp
View file

@ -3753,14 +3753,6 @@ static bool llm_load_tensors(
model.output_norm = ml.create_tensor(ctx, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, backend_norm);
model.output_norm_b = ml.create_tensor(ctx, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, backend_norm);
model.output = ml.create_tensor(ctx, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, backend_output);
if (backend_norm == GGML_BACKEND_GPU) {
vram_weights += ggml_nbytes(model.output_norm);
vram_weights += ggml_nbytes(model.output_norm_b);
}
if (backend_output == GGML_BACKEND_GPU_SPLIT) {
vram_weights += ggml_nbytes(model.output);
}
}
const uint32_t n_ff = hparams.n_ff;
@ -3792,16 +3784,6 @@ static bool llm_load_tensors(
layer.ffn_up = ml.create_tensor(ctx, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, backend_split);
layer.ffn_up_b = ml.create_tensor(ctx, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}, backend);
if (backend == GGML_BACKEND_GPU) {
vram_weights +=
ggml_nbytes(layer.attn_norm) + ggml_nbytes(layer.attn_norm_b) +
ggml_nbytes(layer.wqkv) + ggml_nbytes(layer.bqkv) +
ggml_nbytes(layer.wo) + ggml_nbytes(layer.bo) +
ggml_nbytes(layer.ffn_norm) + ggml_nbytes(layer.ffn_norm_b) +
ggml_nbytes(layer.ffn_down) + ggml_nbytes(layer.ffn_down_b) +
ggml_nbytes(layer.ffn_up) + ggml_nbytes(layer.ffn_up_b);
}
}
} break;
default:
@ -5890,7 +5872,7 @@ struct llm_build_context {
cur = llm_build_kqv(ctx0, model, hparams, kv_self,
model.layers[il].wo, model.layers[il].bo,
Qcur, KQ_scale, KQ_mask, n_ctx, n_tokens, n_kv, -1.0f, 1.0f/sqrtf(float(n_embd_head)), cb, il);
Qcur, KQ_mask, n_ctx, n_tokens, n_kv, -1.0f, 1.0f/sqrtf(float(n_embd_head)), cb, il);
cb(cur, "kqv_out", il);
}