CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

Support diffusion models: Add Dream 7B (#14644)

Browse Source 
* Support diffusion models: Add Dream 7B

* Move diffusion to examples

* Move stuff to examples. Add patch to not use kv-cache

* Address review comments

* Make sampling fast

* llama: remove diffusion functions

* Add basic timings + cleanup

* More cleanup

* Review comments: better formating, use LOG instead std::cerr, re-use batch, use ubatch instead of max_length

* fixup!

* Review: move everything to diffusion-cli for now
This commit is contained in:
Aman Gupta
2025-07-16 20:03:51 +08:00
committed by GitHub
parent 64978340b0
commit ab14019821
13 changed files with 804 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
common/arg.cpp
View File

@@ -3423,5 +3423,34 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
}
).set_examples({LLAMA_EXAMPLE_SERVER}));
// diffusion parameters
add_opt(common_arg(
{ "--diffusion-steps" }, "N",
string_format("number of diffusion steps (default: %d)", params.diffusion.steps),
[](common_params & params, int value) { params.diffusion.steps = value; }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-eps" }, "F",
string_format("epsilon for timesteps (default: %.6f)", (double) params.diffusion.eps),
[](common_params & params, const std::string & value) { params.diffusion.eps = std::stof(value); }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-algorithm" }, "N",
string_format("diffusion algorithm: 0=ORIGIN, 1=MASKGIT_PLUS, 2=TOPK_MARGIN, 3=ENTROPY (default: %d)",
params.diffusion.algorithm),
[](common_params & params, int value) { params.diffusion.algorithm = value; }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-alg-temp" }, "F",
string_format("algorithm temperature (default: %.3f)", (double) params.diffusion.alg_temp),
[](common_params & params, const std::string & value) { params.diffusion.alg_temp = std::stof(value); }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-visual" },
string_format("enable visual diffusion mode (show progressive generation) (default: %s)",
params.diffusion.visual_mode ? "true" : "false"),
[](common_params & params) { params.diffusion.visual_mode = true; }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
return ctx_arg;
}

10
common/common.h
View File

@@ -81,6 +81,7 @@ enum llama_example {
LLAMA_EXAMPLE_LOOKUP,
LLAMA_EXAMPLE_PARALLEL,
LLAMA_EXAMPLE_TTS,
LLAMA_EXAMPLE_DIFFUSION,
LLAMA_EXAMPLE_COUNT,
};
@@ -218,6 +219,14 @@ struct common_params_vocoder {
bool use_guide_tokens = false; // enable guide tokens to improve TTS accuracy // NOLINT
};
struct common_params_diffusion {
int32_t steps = 64; // number of diffusion steps
float eps = 1e-3f; // epsilon for timesteps
int32_t algorithm = 0; // diffusion algorithm (0=ORIGIN, 1=MASKGIT_PLUS, 2=TOPK_MARGIN, 3=ENTROPY)
float alg_temp = 0.0f; // algorithm temperature
bool visual_mode = false; // show progressive diffusion on screen
};
enum common_reasoning_format {
COMMON_REASONING_FORMAT_NONE,
COMMON_REASONING_FORMAT_DEEPSEEK_LEGACY, // Extract thinking tag contents and return as `message.reasoning_content`, or leave inline in <think> tags in stream mode
@@ -269,6 +278,7 @@ struct common_params {
struct common_params_sampling sampling;
struct common_params_speculative speculative;
struct common_params_vocoder vocoder;
struct common_params_diffusion diffusion;
struct common_params_model model;