Create llama-hparams-ms.cpp

Signed-off-by: Brad Hutchings <[email protected]>

Author: BradHutchings

src/llama-hparams-ms.cpp

@@ -0,0 +1,157 @@
+#include "llama-hparams.h"
+
+#include "ggml.h"
+
+void llama_hparams::set_swa_pattern(uint32_t n_pattern) {
+    for (uint32_t il = 0; il < n_layer; ++il) {
+        swa_layers[il] = n_pattern == 0 || (il % n_pattern < (n_pattern - 1));
+    }
+}
+
+bool llama_hparams::is_swa_any() const {
+    for (uint32_t il = 0; il < n_layer; ++il) {
+        if (swa_layers[il]) {
+            return true;
+        }
+    }
+
+    return false;
+}
+
+uint32_t llama_hparams::n_head(uint32_t il) const {
+    if (il < n_layer) {
+        return n_head_arr[il];
+    }
+
+    GGML_ABORT("fatal error");
+}
+
+uint32_t llama_hparams::n_head_kv(uint32_t il) const {
+    if (il < n_layer) {
+        return n_head_kv_arr[il];
+    }
+
+    GGML_ABORT("fatal error");
+}
+
+uint32_t llama_hparams::n_ff(uint32_t il) const {
+    if (il < n_layer) {
+        return n_ff_arr[il];
+    }
+
+    GGML_ABORT("fatal error");
+}
+
+uint32_t llama_hparams::n_gqa(uint32_t il) const {
+    const uint32_t n_head    = this->n_head(il);
+    const uint32_t n_head_kv = this->n_head_kv(il);
+
+    if (n_head_kv == 0) {
+        return 0;
+    }
+
+    return n_head/n_head_kv;
+}
+
+uint32_t llama_hparams::n_embd_k_gqa(uint32_t il) const {
+    const uint32_t n_head_kv = this->n_head_kv(il);
+
+    return n_embd_head_k * n_head_kv;
+}
+
+uint32_t llama_hparams::n_embd_v_gqa(uint32_t il) const {
+    const uint32_t n_head_kv = this->n_head_kv(il);
+
+    return n_embd_head_v * n_head_kv;
+}
+
+bool llama_hparams::is_n_embd_k_gqa_variable() const {
+    const uint32_t val = n_embd_k_gqa();
+    for (uint32_t il = 0; il < n_layer; ++il) {
+        if (val != n_embd_k_gqa(il)) {
+            return true;
+        }
+    }
+
+    return false;
+}
+
+bool llama_hparams::is_n_embd_v_gqa_variable() const {
+    const uint32_t val = n_embd_v_gqa();
+    for (uint32_t il = 0; il < n_layer; ++il) {
+        if (val != n_embd_v_gqa(il)) {
+            return true;
+        }
+    }
+
+    return false;
+}
+
+uint32_t llama_hparams::n_embd_k_gqa_max() const {
+    uint32_t val = n_embd_k_gqa();
+    for (uint32_t il = 0; il < n_layer; ++il) {
+        #ifndef COSMOCC
+        val = std::max(val, n_embd_k_gqa(il));
+        #else
+        val = (val > n_embd_k_gqa(il)) ? val : n_embd_k_gqa(il);
+        #endif
+    }
+
+    return val;
+}
+
+uint32_t llama_hparams::n_embd_v_gqa_max() const {
+    uint32_t val = n_embd_v_gqa();
+    for (uint32_t il = 0; il < n_layer; ++il) {
+        #ifndef COSMOCC
+        val = std::max(val, n_embd_v_gqa(il));
+        #else
+        val = (val > n_embd_v_gqa(il)) ? val : n_embd_v_gqa(il);
+        #endif
+    }
+
+    return val;
+}
+
+uint32_t llama_hparams::n_embd_r() const {
+    if (wkv_head_size != 0) {
+        // for RWKV models
+        return token_shift_count * n_embd;
+    }
+
+    if (n_shortconv_l_cache != 0) {
+        // for LFM2 models
+        return n_embd * (n_shortconv_l_cache - 1);
+    }
+
+    // TODO: maybe support other convolution strides than 1
+    // NOTE: since the first column of the conv_state is shifted out each time, it's not actually needed
+    // Corresponds to Mamba's conv_states size
+    return (ssm_d_conv > 0 ? ssm_d_conv - 1 : 0) * (ssm_d_inner + 2*ssm_n_group*ssm_d_state);
+}
+
+uint32_t llama_hparams::n_embd_s() const {
+    if (wkv_head_size != 0) {
+        // corresponds to RWKV's wkv_states size
+        return n_embd * wkv_head_size;
+    }
+
+    // corresponds to Mamba's ssm_states size
+    return ssm_d_state * ssm_d_inner;
+}
+
+bool llama_hparams::is_recurrent(uint32_t il) const {
+    return recurrent_layer_arr[il];
+}
+
+uint32_t llama_hparams::n_pos_per_embd() const {
+    return rope_type == LLAMA_ROPE_TYPE_MROPE ? 4 : 1;
+}
+
+bool llama_hparams::is_swa(uint32_t il) const {
+    if (il < n_layer) {
+        return swa_layers[il];
+    }
+
+    GGML_ABORT("fatal error");
+}