fix #270

davidkoski · davidkoski · commit 01b974f622d0 · 2025-04-17T13:53:44.000-07:00
diff --git a/Applications/VLMEval/ContentView.swift b/Applications/VLMEval/ContentView.swift
@@ -411,13 +411,22 @@ class VLMEvaluator {
                     if !images.isEmpty || !videos.isEmpty {
                         [
                             [
-                                "role": "user",
+                                "role": "system",
                                 "content": [
                                     [
                                         "type": "text",
                                         "text": videoURL != nil
                                             ? videoSystemPrompt : imageSystemPrompt,
                                     ]
+                                ],
+                            ],
+                            [
+                                "role": "user",
+                                "content": [
+                                    [
+                                        "type": "text",
+                                        "text": prompt,
+                                    ]
                                 ]
                                     // Messages format for Qwen 2 VL, Qwen 2.5 VL. May need to be adapted for other models.
                                     + images.map { _ in
@@ -426,7 +435,7 @@ class VLMEvaluator {
                                     + videos.map { _ in
                                         ["type": "video"]
                                     },
-                            ]
+                            ],
                         ]
                     } else {
                         [
diff --git a/Tools/llm-tool/LLMTool.swift b/Tools/llm-tool/LLMTool.swift
@@ -239,7 +239,7 @@ struct EvaluateCommand: AsyncParsableCommand {
             chat: [
                 .system(generate.system),
                 .user(prompt, images: images, videos: videos),
-            ],
+            ]
         )
         if !resize.isEmpty {
             let size: CGSize

Original file line number	Diff line number	Diff line change
`@@ -411,13 +411,22 @@ class VLMEvaluator {`
`411`	`411`	`if !images.isEmpty \|\| !videos.isEmpty {`
`412`	`412`	`[`
`413`	`413`	`[`
`414`		`- "role": "user",`
	`414`	`+ "role": "system",`
`415`	`415`	`"content": [`
`416`	`416`	`[`
`417`	`417`	`"type": "text",`
`418`	`418`	`"text": videoURL != nil`
`419`	`419`	`? videoSystemPrompt : imageSystemPrompt,`
`420`	`420`	`]`
	`421`	`+ ],`
	`422`	`+ ],`
	`423`	`+ [`
	`424`	`+ "role": "user",`
	`425`	`+ "content": [`
	`426`	`+ [`
	`427`	`+ "type": "text",`
	`428`	`+ "text": prompt,`
	`429`	`+ ]`
`421`	`430`	`]`
`422`	`431`	`// Messages format for Qwen 2 VL, Qwen 2.5 VL. May need to be adapted for other models.`
`423`	`432`	`+ images.map { _ in`
`@@ -426,7 +435,7 @@ class VLMEvaluator {`
`426`	`435`	`+ videos.map { _ in`
`427`	`436`	`["type": "video"]`
`428`	`437`	`},`
`429`		`- ]`
	`438`	`+ ],`
`430`	`439`	`]`
`431`	`440`	`} else {`
`432`	`441`	`[`
Original file line number	Diff line number	Diff line change
`@@ -239,7 +239,7 @@ struct EvaluateCommand: AsyncParsableCommand {`
`239`	`239`	`chat: [`
`240`	`240`	`.system(generate.system),`
`241`	`241`	`.user(prompt, images: images, videos: videos),`
`242`		`- ],`
	`242`	`+ ]`
`243`	`243`	`)`
`244`	`244`	`if !resize.isEmpty {`
`245`	`245`	`let size: CGSize`