examples : add debug utility/example

common/arg.cpp

@@ -1445,7 +1445,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params, bool value) {
             params.warmup = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_COMPLETION, LLAMA_EXAMPLE_CLI, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_MTMD, LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_PERPLEXITY}));
+    ).set_examples({LLAMA_EXAMPLE_COMPLETION, LLAMA_EXAMPLE_CLI, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_MTMD, LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_PERPLEXITY, LLAMA_EXAMPLE_DEBUG}));
     add_opt(common_arg(
         {"--spm-infill"},
         string_format(
@@ -1761,7 +1761,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             else if (value == "rank") { params.pooling_type = LLAMA_POOLING_TYPE_RANK; }
             else { throw std::invalid_argument("invalid value"); }
         }
-    ).set_examples({LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_POOLING"));
+    ).set_examples({LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_DEBUG}).set_env("LLAMA_ARG_POOLING"));
     add_opt(common_arg(
         {"--attention"}, "{causal,non-causal}",
         "attention type for embeddings, use model default if unspecified",
@@ -2609,7 +2609,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params, int value) {
             params.embd_normalize = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_EMBEDDING}));
+    ).set_examples({LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_DEBUG}));
     add_opt(common_arg(
         {"--embd-output-format"}, "FORMAT",
         "empty = default, \"array\" = [[],[]...], \"json\" = openai style, \"json+\" = same \"json\" + cosine similarity matrix, \"raw\" = plain whitespace-delimited output (one embedding per line)",
@@ -2687,7 +2687,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params) {
             params.embedding = true;
         }
-    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_EMBEDDINGS"));
+    ).set_examples({LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_DEBUG}).set_env("LLAMA_ARG_EMBEDDINGS"));
     add_opt(common_arg(
         {"--rerank", "--reranking"},
         string_format("enable reranking endpoint on server (default: %s)", "disabled"),
@@ -3378,6 +3378,27 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             }
         }
     ).set_examples({ LLAMA_EXAMPLE_FINETUNE }));
+    add_opt(common_arg(
+        {"--save-logits"},
+        string_format("save final logits to files for verification (default: %s)", params.save_logits ? "true" : "false"),
+        [](common_params & params) {
+            params.save_logits = true;
+        }
+    ).set_examples({LLAMA_EXAMPLE_DEBUG}));
+    add_opt(common_arg(
+        {"--logits-output-dir"}, "PATH",
+        string_format("directory for saving logits output files (default: %s)", params.logits_output_dir.c_str()),
+        [](common_params & params, const std::string & value) {
+            params.logits_output_dir = value;
+        }
+    ).set_examples({LLAMA_EXAMPLE_DEBUG}));
+    add_opt(common_arg(
+        {"--tensor-filter"}, "REGEX",
+        "filter tensor names for debug output (regex pattern, can be specified multiple times)",
+        [](common_params & params, const std::string & value) {
+            params.tensor_filter.push_back(value);
+        }
+    ).set_examples({LLAMA_EXAMPLE_DEBUG}));
 
     // presets
     add_opt(common_arg(

common/common.h

@@ -80,6 +80,7 @@ int32_t cpu_get_num_math();
 //
 
 enum llama_example {
+    LLAMA_EXAMPLE_DEBUG,
     LLAMA_EXAMPLE_COMMON,
     LLAMA_EXAMPLE_SPECULATIVE,
     LLAMA_EXAMPLE_COMPLETION,
@@ -372,6 +373,11 @@ struct common_params {
     std::string lookup_cache_dynamic = ""; // path of dynamic ngram cache file for lookup decoding          // NOLINT
     std::string logits_file          = ""; // file for saving *all* logits                                  // NOLINT
 
+    // llama-debug specific options
+    std::string logits_output_dir = "data"; // directory for saving logits output files                     // NOLINT
+    bool        save_logits       = false;  // whether to save logits to files                              // NOLINT
+    std::vector<std::string> tensor_filter; // filter tensor names for debug output (regex)                 // NOLINT
+
     std::vector<std::string> in_files;   // all input files
     std::vector<std::string> antiprompt; // strings upon which more user input is prompted (a.k.a. reverse prompts)
     std::vector<llama_model_kv_override> kv_overrides;

examples/CMakeLists.txt

@@ -15,6 +15,7 @@ llama_add_compile_flags()
 if (EMSCRIPTEN)
 else()
     add_subdirectory(batched)
+    add_subdirectory(debug)
     add_subdirectory(embedding)
     add_subdirectory(eval-callback)
 
@@ -34,7 +35,6 @@ else()
     add_subdirectory(gen-docs)
     add_subdirectory(training)
     add_subdirectory(diffusion)
-    add_subdirectory(model-conversion)
     if (NOT GGML_BACKEND_DL)
         add_subdirectory(convert-llama2c-to-ggml)
         # these examples use the backends directly and cannot be built with dynamic loading

examples/model-conversion/CMakeLists.txt → examples/debug/CMakeLists.txt

@@ -1,5 +1,5 @@
-set(TARGET llama-logits)
-add_executable(${TARGET} logits.cpp)
+set(TARGET llama-debug)
+add_executable(${TARGET} debug.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/debug/README.md

@@ -0,0 +1,54 @@
+# llama.cpp/examples/debug
+
+This is a utility intended to help debug a model by registering a callback that
+logs GGML operations and tensor data. It can also store the generated logits or
+embeddings as well as the prompt and token ids for comparision with the original
+model.
+
+### Usage
+
+```shell
+llama-debug \
+  --hf-repo ggml-org/models \
+  --hf-file phi-2/ggml-model-q4_0.gguf \
+  --model phi-2-q4_0.gguf \
+  --prompt hello \
+  --save-logits \
+  --verbose
+```
+The tensor data is logged as debug and required the --verbose flag. The reason
+for this is that while useful for a model with many layers there can be a lot of
+output. You can filter the tensor names using the `--tensor-filter` option.
+
+A recommended approach is to first run without `--verbose` and see if the
+generated logits/embeddings are close to the original model. If they are not,
+then it might be required to inspect tensor by tensor and in that case it is
+useful to enable the `--verbose` flag along with `--tensor-filter` to focus on
+specific tensors.
+
+### Options
+This example supports all standard `llama.cpp` options and also accepts the
+following options:
+```console
+$ llama-debug --help
+...
+
+----- example-specific params -----
+
+--save-logits                           save final logits to files for verification (default: false)
+--logits-output-dir PATH                directory for saving logits output files (default: data)
+--tensor-filter REGEX                   filter tensor names for debug output (regex pattern, can be specified multiple times)
+```
+
+### Output Files
+
+When `--save-logits` is enabled, the following files are created in the output
+directory:
+
+* `llamacpp-<model>[-embeddings].bin`        - Binary output (logits or embeddings)
+* `llamacpp-<model>[-embeddings].txt`        - Text output (logits or embeddings, one per line)
+* `llamacpp-<model>[-embeddings]-prompt.txt` - Prompt text and token IDs
+* `llamacpp-<model>[-embeddings]-tokens.bin` - Binary token IDs for programmatic comparison
+
+These files can be compared against the original model's output to verify the
+converted model.