public_git/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/whisper.cpp.git synced 2026-10-21 01:31:53 +02:00
Code Issues Packages Projects Releases Wiki Activity

model : support LiquidAI LFM2 hybrid family (llama/14620)

Browse Source 
**Important**
LFM2 was [merged ](https://github.com/huggingface/transformers/pull/39340)into transformers, but has not yet been released.
To convert into gguf, install transformers from source
```shell
pip install "transformers @ git+https://github.com/huggingface/transformers.git@main"
```
This commit is contained in:
Tarek Dakhran
2025-07-12 19:23:56 +03:00
committed by Georgi Gerganov
parent a4ff4ec9cb
commit 74f6d47904
1 changed files with 10 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cuda/ssm-conv.cu
+10 -2
View File
@@ -107,8 +107,11 @@ static void ssm_conv_f32_cuda(const float * src0, const float * src1, const int
if (nc == 4) { if (nc == 4) {
ssm_conv_f32<threads, 4><<<blocks, threads, 0, stream>>>(src0, src1, src0_nb0, src0_nb1, src0_nb2, src1_nb1, ssm_conv_f32<threads, 4><<<blocks, threads, 0, stream>>>(src0, src1, src0_nb0, src0_nb1, src0_nb2, src1_nb1,
dst, dst_nb0, dst_nb1, dst_nb2, n_t); dst, dst_nb0, dst_nb1, dst_nb2, n_t);
} else if (nc == 3) {
ssm_conv_f32<threads, 3><<<blocks, threads, 0, stream>>>(src0, src1, src0_nb0, src0_nb1, src0_nb2, src1_nb1,
dst, dst_nb0, dst_nb1, dst_nb2, n_t);
} else { } else {
GGML_ABORT("Only support kernel size = 4 now."); GGML_ABORT("Only support kernel size = 3 or size = 4 right now.");
} }
} else { } else {
if (nc == 4) { if (nc == 4) {
@@ -116,8 +119,13 @@ static void ssm_conv_f32_cuda(const float * src0, const float * src1, const int
dim3 blocks(n_s, (nr + threads - 1) / threads, (n_t + split_n_t - 1) / split_n_t); dim3 blocks(n_s, (nr + threads - 1) / threads, (n_t + split_n_t - 1) / split_n_t);
ssm_conv_long_token_f32<threads, 4, split_n_t><<<blocks, threads, 0, stream>>>( ssm_conv_long_token_f32<threads, 4, split_n_t><<<blocks, threads, 0, stream>>>(
src0, src1, src0_nb0, src0_nb1, src0_nb2, src1_nb1, dst, dst_nb0, dst_nb1, dst_nb2, n_t); src0, src1, src0_nb0, src0_nb1, src0_nb2, src1_nb1, dst, dst_nb0, dst_nb1, dst_nb2, n_t);
} else if (nc == 3) {
const int64_t split_n_t = 32;
dim3 blocks(n_s, (nr + threads - 1) / threads, (n_t + split_n_t - 1) / split_n_t);
ssm_conv_long_token_f32<threads, 3, split_n_t><<<blocks, threads, 0, stream>>>(
src0, src1, src0_nb0, src0_nb1, src0_nb2, src1_nb1, dst, dst_nb0, dst_nb1, dst_nb2, n_t);
} else { } else {
GGML_ABORT("Only support kernel size = 4 right now."); GGML_ABORT("Only support kernel size = 3 or size = 4 right now.");
} }
} }
} }