diff options
| author | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-06-26 17:38:18 +0300 |
|---|---|---|
| committer | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-06-26 17:38:18 +0300 |
| commit | 0a3a2c4cd47943dc4c3c43be75728402584a3732 (patch) | |
| tree | 9e3b58646f002f023de8991dc8b30f2d0251796a /examples/imatrix/imatrix.cpp | |
| parent | 71725a918f9edee559a978397779486dce7c703a (diff) | |
imatrix: be able to specify the name of the output tensor
For some models the same tensor is used for token embeddings and
output. This tensor tends to be named token_embedding.weight rather
than output.weight, which prevernts us from collecting imatrix data
for this tensor. With this commit we can tell the name of the
output tensor to the imatrix tool.
Diffstat (limited to 'examples/imatrix/imatrix.cpp')
| -rw-r--r-- | examples/imatrix/imatrix.cpp | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/examples/imatrix/imatrix.cpp b/examples/imatrix/imatrix.cpp index 574f5ed9..c7d73cdb 100644 --- a/examples/imatrix/imatrix.cpp +++ b/examples/imatrix/imatrix.cpp @@ -83,7 +83,8 @@ bool IMatrixCollector::collect_imatrix(struct ggml_tensor * t, bool ask, void * if (t->op != GGML_OP_MUL_MAT) return false; // why are small batches ignored (<16 tokens)? if (src1->ne[1] < 16 || src1->type != GGML_TYPE_F32) return false; - if (!(wname.substr(0, 4) == "blk." || (m_params.process_output && wname == "output.weight"))) return false; + //printf("wname = %s\n", wname.c_str()); + if (!(wname.substr(0, 4) == "blk." || (m_params.process_output && wname == m_params.output_tensor_name))) return false; return true; } |