tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-26 19:55:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml : reduce hash table reset cost (#8698)

Browse Source 
* ggml : reduce hash table reset cost

* fix unreachable code warnings after GGML_ASSERT(false)

* GGML_ASSERT(false) -> GGML_ABORT("fatal error")

* GGML_ABORT use format string
This commit is contained in:
slaren
2024-07-27 04:41:55 +02:00
committed by GitHub
parent 01245f5b16
commit 2b1f616b20
46 changed files with 851 additions and 754 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/test-backend-ops.cpp
View File

@ -94,7 +94,7 @@ static void init_tensor_uniform(ggml_tensor * tensor, float min = -1.0f, float m
// This is going to create some weird integers though.
ggml_backend_tensor_set(tensor, data.data(), 0, ggml_nbytes(tensor));
} else {
GGML_ASSERT(false);
GGML_ABORT("fatal error");
}
}
@ -132,7 +132,7 @@ static std::vector<float> tensor_to_float(const ggml_tensor * t) {
tt.to_float(&buf[i], vq.data(), bs);
tv.insert(tv.end(), vq.begin(), vq.end());
} else {
GGML_ASSERT(false);
GGML_ABORT("fatal error");
}
}
}
@ -1435,7 +1435,7 @@ struct test_argsort : public test_case {
ggml_backend_tensor_set(t, data.data(), r * t->nb[1], t->ne[0] * sizeof(float));
}
} else {
GGML_ASSERT(false);
GGML_ABORT("fatal error");
}
}
}
@ -2462,7 +2462,7 @@ static bool test_backend(ggml_backend_t backend, test_mode mode, const char * op
return true;
}
GGML_ASSERT(false);
GGML_ABORT("fatal error");
return false;
}

10
tests/test-sampling.cpp
View File

@ -166,12 +166,12 @@ static void test_sampler_queue(
for (auto s : samplers_sequence) {
switch (s){
case 'k': llama_sample_top_k (nullptr, &candidates_p, top_k, 1); break;
case 'f': GGML_ASSERT(false && "tail_free test not implemented"); break;
case 'y': GGML_ASSERT(false && "typical test not implemented"); break;
case 'f': GGML_ABORT("tail_free test not implemented"); break;
case 'y': GGML_ABORT("typical test not implemented"); break;
case 'p': llama_sample_top_p (nullptr, &candidates_p, top_p, 1); break;
case 'm': llama_sample_min_p (nullptr, &candidates_p, min_p, 1); break;
case 't': GGML_ASSERT(false && "temperature test not implemented"); break;
default : GGML_ASSERT(false && "Unknown sampler"); break;
case 't': GGML_ABORT("temperature test not implemented"); break;
default : GGML_ABORT("Unknown sampler"); break;
}
llama_sample_softmax(nullptr, &candidates_p); // make sure tokens are sorted for tests
@ -222,7 +222,7 @@ static void test_sampler_queue(
GGML_ASSERT(candidates_p.data[0].id == max_token_id);
GGML_ASSERT(candidates_p.data[expected_size-1].id == min_token_id);
} else {
GGML_ASSERT(false);
GGML_ABORT("fatal error");
}
}