llama : add gpt-oss (#15091)

* oai moe

* compat with new checkpoint

* add attn sink impl

* add rope scaling yarn

* logits match with latest transformers code

* wip chat template

* rm trailing space

* use ggml_scale_bias

* rm redundant is_swa_all

* convert interleaved gate_up

* graph : fix activation function to match reference (#7)

* vocab : handle o200k_harmony special tokens

* ggml : add attention sinks support (#1)

* llama : add attn sinks

* ggml : add attn sinks

* cuda : add attn sinks

* vulkan : add support for sinks in softmax

remove unnecessary return

* ggml : add fused swiglu_oai op (#11)

* ggml : add fused swiglu_oai op

* Update ggml/src/ggml-cpu/ops.cpp

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* update CUDA impl

* cont : metal impl

* add vulkan impl

* test-backend-ops : more test cases, clean up

* llama : remove unfused impl

* remove extra lines

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

---------

Co-authored-by: slaren <slarengh@gmail.com>

* repack mxfp4 upon conversion

* clean up a bit

* enable thinking

* add quick hack to render only some special tokens

* fix bf16 conversion

* remove vocab hack

* webui ok

* support chat parsing for gpt-oss

* fix webui

* direct mapping mxfp4, FINALLY

* force using mxfp4

* properly use lazy tensor

* ggml : add mxfp4

ggml : use e8m0 conversion instead of powf

Co-authored-by: Diego Devesa <slarengh@gmail.com>

change kvalues_mxfp4 table to match e2m1 (#6)

metal : remove quantization for now (not used)

cuda : fix disabled CUDA graphs due to ffn moe bias

vulkan : add support for mxfp4

cont : add cm2 dequant

* ggml : add ggml_add_id (#13)

* ggml : add ggml_add_id

* add cuda impl

* llama : add weight support check for add_id

* perf opt

* add vulkan impl

* rename cuda files

* add metal impl

* allow in-place ggml_add_id

* llama : keep biases on CPU with --cpu-moe

* llama : fix compile error

ggml-ci

* cuda : add fallback for __nv_cvt_e8m0_to_bf16raw

ggml-ci

* cleanup

ggml-ci

* sycl : fix supports_op for MXFP4

ggml-ci

* fix Unknown reasoning format

* ggml-cpu : fix AVX build

ggml-ci

* fix hip build

ggml-ci

* cuda : add mxfp4 dequantization support for cuBLAS

ggml-ci

* ggml-cpu : fix mxfp4 fallback definitions for some architectures

ggml-ci

* cuda : fix version required for __nv_cvt_e8m0_to_bf16raw

---------

Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
Co-authored-by: slaren <slarengh@gmail.com>

ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp

@@ -20,6 +20,7 @@ layout (push_constant) uniform parameter
     float m1;
     uint n_head_log2;
     uint nrows_x;
+    uint has_sinks;
 } p;
 
 #include "types.comp"
@@ -29,7 +30,8 @@ layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
 
 layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
 layout (binding = 1) readonly buffer Y {B_TYPE data_b[];};
-layout (binding = 2) buffer D {D_TYPE data_d[];};
+layout (binding = 2) readonly buffer Z {float data_c[];};
+layout (binding = 3) buffer D {D_TYPE data_d[];};
 
 shared FLOAT_TYPE vals[BLOCK_SIZE];
 
@@ -60,13 +62,13 @@ void soft_max(uint num_iters) {
         const uint h = (rowx / p.ne01) % p.ne02; // head index
 
         const float base = h < p.n_head_log2 ? p.m0 : p.m1;
-        const uint   exp  = h < p.n_head_log2 ? h + 1 : 2*(h - p.n_head_log2) + 1;
+        const uint   exp = h < p.n_head_log2 ? h + 1 : 2*(h - p.n_head_log2) + 1;
 
         slope = pow(base, exp);
     }
 
     // Find max
-    FLOAT_TYPE max_val = uintBitsToFloat(0xFF800000);
+    FLOAT_TYPE max_val = p.has_sinks == 0 ? uintBitsToFloat(0xFF800000) : data_c[i02];
 
     // Cache values while we compute the max, so we don't need to read them
     // again when we're ready to compute exp(x-max).
@@ -148,6 +150,10 @@ void soft_max(uint num_iters) {
     }
     sum = vals[0];
 
+    if (p.has_sinks != 0) {
+        sum += FLOAT_TYPE(exp(FLOAT_TYPE(data_c[i02]) - max_val));
+    }
+
     FLOAT_TYPE rcpdivisor = 1.0/sum;
 
     [[unroll]] for (uint col0 = 0, idx = 0; idx < num_iters; col0 += BLOCK_SIZE, ++idx) {