mtmd/clip+converter: address #16574; fold CLI into mtmd-cli, fix converter keys, rope_ext + bicubic

mtmd-cli: move the standalone Jina CLI into mtmd-cli (projector-only path); drop the extra binary.

Author: liyang

convert_hf_to_gguf.py

@@ -3057,6 +3057,7 @@ class VisionProjectorType:
     QWEN25VL = "qwen2.5vl_merger"
     ULTRAVOX = "ultravox"
     INTERNVL = "internvl"
+    JINACLIP2 = "jinaclip2"
     QWEN2A = "qwen2a" # audio
     QWEN25O = "qwen2.5o" # omni
     VOXTRAL = "voxtral"

gguf-py/gguf/constants.py

@@ -39,6 +39,8 @@
 #define KEY_FEATURE_LAYER       "clip.vision.feature_layer"
 #define KEY_PROJ_SCALE_FACTOR   "clip.vision.projector.scale_factor"
 #define KEY_SPATIAL_MERGE_SIZE  "clip.vision.spatial_merge_size"
+// RoPE base (theta) for vision encoder
+#define KEY_VISION_ROPE_THETA   "clip.vision.rope_theta"
 
 #define KEY_MM_PATCH_MERGE_TYPE   "clip.vision.mm_patch_merge_type"
 #define KEY_IMAGE_GRID_PINPOINTS  "clip.vision.image_grid_pinpoints"
@@ -67,14 +69,15 @@
 #define TN_ATTN_Q          "%s.blk.%d.attn_q.%s"
 #define TN_ATTN_V          "%s.blk.%d.attn_v.%s"
 #define TN_ATTN_OUTPUT     "%s.blk.%d.attn_out.%s"
+#define TN_ATTN_LN         "%s.blk.%d.attn_ln.%s"  // inner attention LayerNorm
 #define TN_ATTN_K_NORM     "%s.blk.%d.attn_k_norm.%s"
 #define TN_ATTN_Q_NORM     "%s.blk.%d.attn_q_norm.%s"
 #define TN_FFN_DOWN        "%s.blk.%d.ffn_down.%s"
 #define TN_FFN_GATE        "%s.blk.%d.ffn_gate.%s"
 #define TN_FFN_UP          "%s.blk.%d.ffn_up.%s"
-#define TN_FFN_GATE        "%s.blk.%d.ffn_gate.%s"
-#define TN_LN_1            "%s.blk.%d.ln1.%s" // layer norm
-#define TN_LN_2            "%s.blk.%d.ln2.%s" // layer norm
+#define TN_FFN_NORM        "%s.blk.%d.ffn_norm.%s"
+#define TN_LN_1            "%s.blk.%d.ln_1.%s"  // layer norm
+#define TN_LN_2            "%s.blk.%d.ln_2.%s"  // layer norm
 #define TN_LS_1            "%s.blk.%d.ls1.%s" // layer scale
 #define TN_LS_2            "%s.blk.%d.ls2.%s" // layer scale
 #define TN_LN_PRE          "%s.pre_ln.%s"
@@ -137,30 +140,32 @@ enum projector_type {
     PROJECTOR_TYPE_QWEN2A,
     PROJECTOR_TYPE_QWEN25O, // will be replaced by QWEN2A or QWEN25VL depending on clip_ctx
     PROJECTOR_TYPE_VOXTRAL,
+    PROJECTOR_TYPE_JINACLIP2, // JinaCLIP v2
     PROJECTOR_TYPE_LFM2,
     PROJECTOR_TYPE_KIMIVL,
     PROJECTOR_TYPE_UNKNOWN,
 };
 
 static std::map<projector_type, std::string> PROJECTOR_TYPE_NAMES = {
-    { PROJECTOR_TYPE_MLP,       "mlp" },
-    { PROJECTOR_TYPE_LDP,       "ldp" },
-    { PROJECTOR_TYPE_LDPV2,     "ldpv2"},
-    { PROJECTOR_TYPE_MINICPMV,  "resampler"},
-    { PROJECTOR_TYPE_GLM_EDGE,  "adapter"},
-    { PROJECTOR_TYPE_QWEN2VL,   "qwen2vl_merger"},
-    { PROJECTOR_TYPE_QWEN25VL,  "qwen2.5vl_merger"},
-    { PROJECTOR_TYPE_GEMMA3,    "gemma3"},
-    { PROJECTOR_TYPE_IDEFICS3,  "idefics3"},
-    { PROJECTOR_TYPE_PIXTRAL,   "pixtral"},
-    { PROJECTOR_TYPE_ULTRAVOX,  "ultravox"},
-    { PROJECTOR_TYPE_INTERNVL,  "internvl"},
-    { PROJECTOR_TYPE_LLAMA4,    "llama4"},
-    { PROJECTOR_TYPE_QWEN2A,    "qwen2a"},
-    { PROJECTOR_TYPE_QWEN25O,   "qwen2.5o"},
-    { PROJECTOR_TYPE_VOXTRAL,   "voxtral"},
-    { PROJECTOR_TYPE_LFM2,      "lfm2"},
-    { PROJECTOR_TYPE_KIMIVL,    "kimivl"},
+    { PROJECTOR_TYPE_MLP,      "mlp"              },
+    { PROJECTOR_TYPE_LDP,      "ldp"              },
+    { PROJECTOR_TYPE_LDPV2,    "ldpv2"            },
+    { PROJECTOR_TYPE_MINICPMV, "resampler"        },
+    { PROJECTOR_TYPE_GLM_EDGE, "adapter"          },
+    { PROJECTOR_TYPE_QWEN2VL,  "qwen2vl_merger"   },
+    { PROJECTOR_TYPE_QWEN25VL, "qwen2.5vl_merger" },
+    { PROJECTOR_TYPE_GEMMA3,   "gemma3"           },
+    { PROJECTOR_TYPE_IDEFICS3, "idefics3"         },
+    { PROJECTOR_TYPE_PIXTRAL,  "pixtral"          },
+    { PROJECTOR_TYPE_ULTRAVOX, "ultravox"         },
+    { PROJECTOR_TYPE_INTERNVL, "internvl"         },
+    { PROJECTOR_TYPE_LLAMA4,   "llama4"           },
+    { PROJECTOR_TYPE_QWEN2A,   "qwen2a"           },
+    { PROJECTOR_TYPE_QWEN25O,  "qwen2.5o"         },
+    { PROJECTOR_TYPE_VOXTRAL,  "voxtral"          },
+    { PROJECTOR_TYPE_JINACLIP2, "jinaclip2"        },
+    { PROJECTOR_TYPE_LFM2,     "lfm2"             },
+    { PROJECTOR_TYPE_KIMIVL,   "kimivl"           },
 };
 
 static projector_type clip_projector_type_from_string(const std::string & str) {