make gptneox and gptj work with extended context too

AAbushady · Jul 2, 2023 · 3d2907d · 3d2907d
1 parent d6b47e6
commit 3d2907d
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 4 deletions.
diff --git a/otherarch/gptj_v3.cpp b/otherarch/gptj_v3.cpp
@@ -68,6 +68,8 @@ ModelLoadResult gptj_model_load(const std::string & fname, gptj_model & model, g
         printf("%s: ftype   = %d\n", __func__, hparams.ftype);
         printf("%s: qntvr   = %d\n", __func__, qntvr);
 
+        hparams.n_ctx = std::max(origmaxctx,hparams.n_ctx);
+
         hparams.ftype %= GGML_QNT_VERSION_FACTOR;
     }
 
@@ -474,8 +476,8 @@ bool gptj_eval(
 
         // self-attention
         {
-            struct ggml_tensor * Qcur = ggml_rope_inplace(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model.layers[il].c_attn_q_proj_w, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, 0);
-            struct ggml_tensor * Kcur = ggml_rope_inplace(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model.layers[il].c_attn_k_proj_w, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, 0);
+            struct ggml_tensor * Qcur = ggml_rope_inplace(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model.layers[il].c_attn_q_proj_w, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, n_ctx);
+            struct ggml_tensor * Kcur = ggml_rope_inplace(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model.layers[il].c_attn_k_proj_w, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, n_ctx);
 
             // store key and value to memory
             {

diff --git a/otherarch/neox_v3.cpp b/otherarch/neox_v3.cpp
@@ -68,6 +68,8 @@ ModelLoadResult gpt_neox_model_load(const std::string & fname, gpt_neox_model &
         printf("%s: ftype   = %d\n", __func__, hparams.ftype);
         printf("%s: qntvr   = %d\n", __func__, qntvr);
 
+        hparams.n_ctx = std::max(origmaxctx,hparams.n_ctx);
+
         hparams.ftype %= GGML_QNT_VERSION_FACTOR;
     }
 
@@ -502,8 +504,8 @@ bool gpt_neox_eval(
             struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_3d(ctx0, cur, n_embd/n_head, n_head, N, cur->nb[1]/n_head, cur->nb[1], 2*sizeof(float)*n_embd/n_head));
 
             // using mode = 2 for GPT-NeoX mode
-            Qcur = ggml_rope_inplace(ctx0, Qcur, n_past, n_rot, 2, 0);
-            Kcur = ggml_rope_inplace(ctx0, Kcur, n_past, n_rot, 2, 0);
+            Qcur = ggml_rope_inplace(ctx0, Qcur, n_past, n_rot, 2, n_ctx);
+            Kcur = ggml_rope_inplace(ctx0, Kcur, n_past, n_rot, 2, n_ctx);
 
             // store key and value to memory
             {