Skip to content

Possible solution to allow K-quants on models with n_vocab!=32000 #2148

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 3 commits into from
Jul 11, 2023
+14 −4
Merged

Possible solution to allow K-quants on models with n_vocab!=32000 #2148

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
1854168
This allows LLAMA models that were previously incompatible with K qua…
LostRuins Jul 8, 2023
048dca9
Fix indentation
LostRuins Jul 10, 2023
fd9a2fd
As an alternative, to avoid failing on Metal due to lack of Q8_0 supp…
LostRuins Jul 10, 2023
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
18 changes: 14 additions & 4 deletions llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2428,15 +2428,14 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
} else {
new_type = quantized_type;
#ifdef GGML_USE_K_QUANTS
bool convert_incompatible_tensor = false;
if (quantized_type == GGML_TYPE_Q2_K || quantized_type == GGML_TYPE_Q3_K || quantized_type == GGML_TYPE_Q4_K ||
quantized_type == GGML_TYPE_Q5_K || quantized_type == GGML_TYPE_Q6_K) {
int nx = tensor.ne.at(0);
int ny = tensor.ne.at(1);
if (nx % QK_K != 0 || ny % QK_K != 0) {
fprintf(stderr, "\n\n========================= Tensor sizes %d x %d are not divisible by %d\n",nx,ny,QK_K);
fprintf(stderr, "This is required to be able to use k-quants for now!\n");
fprintf(stderr, "========================================================================================\n\n");
throw std::runtime_error("Unsupported tensor size encountered\n");
fprintf(stderr, "\n\nTensor sizes %d x %d are not divisible by %d, required for k-quants.\n",nx,ny,QK_K);
convert_incompatible_tensor = true;
}
}
if (tensor.name == "output.weight") {
Expand Down Expand Up @@ -2464,6 +2463,17 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q4_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q5_K;
}
if (convert_incompatible_tensor) {
if (tensor.name == "output.weight") {
new_type = GGML_TYPE_F16; //fall back to F16 instead of just failing.
fprintf(stderr, "F16 will be used for this tensor instead.\n");
} else if (tensor.name == "tok_embeddings.weight") {
new_type = GGML_TYPE_Q4_0; //fall back to Q4_0 instead of just failing.
fprintf(stderr, "Q4_0 will be used for this tensor instead.\n");
} else {
throw std::runtime_error("Unsupported tensor size encountered\n");
}
}
#endif

float * f32_data;
Expand Down