Pascal
ba3c05f71d
Copilot Setup Steps / copilot-setup-steps (push) Has been cancelled
Check Pre-Tokenizer Hashes / pre-tokenizer-hashes (push) Has been cancelled
Python check requirements.txt / check-requirements (push) Has been cancelled
Python Type-Check / python type-check (push) Has been cancelled
Update Operations Documentation / update-ops-docs (push) Has been cancelled
* model: add Muse Spark Open (text, vision and audio) Muse Spark Open is a latent space MoE: the residual stream stays at 8192 while the 256 experts live in 4096, with a projection down and up around them, top-8 sigmoid routing with a correction bias. Attention interleaves three sliding layers at a 2048 window with rope for one global layer without rope, with a sigmoid output gate and a weightless QK-norm. Residuals are norm preserving, their gate is folded at conversion time. Vision reuses the Muse Glimmer tower. Audio is a causal convolutional pyramid with no attention and no spectrogram, four stages folding 16, 6, 5 and 4 samples into the channel dim, one embedding per 80 ms of 24 kHz. The quantizer keeps the latent projections and the attention gate at Q8_0, a little over 1% of the weights, since the whole residual stream passes through them and the low-bit types turned the output into a repeating loop. The model saver writes the post norm epsilon, and the arch is registered in test-llama-archs. Co-authored-by: Xuan Son Nguyen <son@huggingface.co> Co-authored-by: Pedro Cuenca <pedro@huggingface.co> * Muse Spark: drop the redundant forward expand * Muse Spark: drop the redundant forward expand and tighten the comments * Update src/llama-quant.cpp Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * Update src/llama-graph.cpp Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * Update conversion/muse_spark.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * Update conversion/muse_spark.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * Update conversion/muse_spark.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * Update conversion/muse_spark.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co> Co-authored-by: Pedro Cuenca <pedro@huggingface.co> Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
87 lines
4.4 KiB
C++
87 lines
4.4 KiB
C++
#pragma once
|
|
|
|
#include "chat.h"
|
|
#include "chat-auto-parser.h"
|
|
#include "chat-auto-parser-helpers.h"
|
|
#include "chat-peg-parser.h"
|
|
#include "common.h"
|
|
#include "ggml.h"
|
|
#include "json-schema-to-grammar.h"
|
|
#include "json.h"
|
|
|
|
#include <functional>
|
|
#include <optional>
|
|
#include <set>
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
using json = common_json;
|
|
|
|
// iterate over the function tools of an OpenAI-style tools array, passing each tool with its index in the array
|
|
void foreach_function(const json & tools, const std::function<void(size_t, const json &)> & fn);
|
|
|
|
// iterate over the parameters of a function tool, passing each parameter with its index and the document that owns it
|
|
void foreach_parameter(const json & function, const std::function<void(size_t, const common_chat_schema_property &, const common_chat_schema_document_ptr &)> & fn);
|
|
|
|
// render a template; the override arguments let a parser feed in messages, tools or context it has rewritten
|
|
common_chat_prompt common_chat_template_direct_apply_impl(
|
|
const common_chat_template & tmpl,
|
|
const autoparser::generation_params & inputs,
|
|
const std::optional<json> & messages_override = std::nullopt,
|
|
const std::optional<json> & tools_override = std::nullopt,
|
|
const std::optional<json> & additional_context = std::nullopt);
|
|
|
|
// the suffix a template appends when add_generation_prompt is set
|
|
std::string common_chat_template_generation_prompt_impl(
|
|
const common_chat_template & tmpl,
|
|
const autoparser::generation_params & inputs,
|
|
const std::optional<json> & messages_override = std::nullopt,
|
|
const std::optional<json> & tools_override = std::nullopt,
|
|
const std::optional<json> & additional_context = std::nullopt);
|
|
|
|
bool is_lfm2_template(const std::string & src);
|
|
|
|
namespace workaround {
|
|
|
|
void convert_tool_responses_gemma4(json & messages);
|
|
|
|
}
|
|
|
|
common_chat_params common_chat_params_init_cohere2moe(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_deepseek_v3_2(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_functionary_v3_2(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_gemma4(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_gigachat_v3(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_gpt_oss(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_k2_horizon(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_kimi_k2(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_kimi_k3(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_ling3(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
// tool_list_tokens preserves the LFM2 system tool-list markers; LFM2.5 renders without them
|
|
common_chat_params common_chat_params_init_lfm2(const common_chat_template & tmpl, const autoparser::generation_params & inputs, bool tool_list_tokens);
|
|
|
|
common_chat_params common_chat_params_init_llm_jp_harmony(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_minicpm5(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_minimax_m3(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_ministral_3(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_muse_glimmer(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
common_chat_params common_chat_params_init_muse_spark(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_qwen3_coder(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|
|
|
|
common_chat_params common_chat_params_init_translate_gemma(const common_chat_template & tmpl, const autoparser::generation_params & inputs);
|