fix(config): fall back to DefaultContextSize for unparseable GGUFs; pin NVFP4 gallery context_size (#10563)

The GGUF metadata parser (gpustack/gguf-parser-go) cannot read NVFP4-quantized
GGUFs at all: it errors with "read tensor info 0: This quantized type is
currently unsupported" because NVFP4 is a ggml tensor type it does not know.
When ParseGGUFFile errors, the llama-cpp defaults hook skips guessGGUFFromFile
entirely and the deferred fallback sets the context window to the conservative
GGUFFallbackContextSize (1024). The result: a model that trains to 262144
tokens runs with n_ctx=1024, and every prompt over ~1k tokens fails with
"request (N tokens) exceeds the available context size (1024 tokens)".

Two changes:

- Drop GGUFFallbackContextSize (1024) and fall back to DefaultContextSize
  (4096) in both the GGUF run-estimate path (gguf.go) and the deferred hook
  fallback (hooks_llamacpp.go). 1024 is a sensible floor for a tiny CPU GGUF
  but a footgun for a large, long-context model whose header simply cannot be
  parsed. Strengthen the existing "GGUF unreadable" test to assert the value.

- Set context_size explicitly on the four NVFP4 gallery entries
  (qwen3.6-35b-a3b-nvfp4-mtp, qwopus3.6-27b-v2-mtp-nvfp4,
  qwopus3.6-27b-coder-mtp-nvfp4, qwen3.6-27b-nvfp4-mtp) so the parser failure
  is irrelevant for them. 32768 matches sibling Qwen entries and is safe on
  memory; operators can raise it toward the 262144 train length.


Assisted-by: Claude:claude-opus-4-8 [Claude Code]

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>

core/config/defaults.go

@@ -12,14 +12,12 @@ package config
 // these; config never imports backend.
 const (
 	// DefaultContextSize is the fallback context window when none is configured
-	// or estimable from the model.
+	// or estimable from the model. It is also the fallback for a GGUF whose
+	// metadata yields no usable estimate or that the parser cannot read at all
+	// (e.g. a quant type it does not know, such as NVFP4): a model-agnostic
+	// safe default beats a tiny, surprising window that truncates real prompts.
 	DefaultContextSize = 4096
 
-	// GGUFFallbackContextSize is the context window for a GGUF model whose
-	// metadata yields no usable estimate (see guessGGUFFromFile). Deliberately
-	// smaller than DefaultContextSize to stay conservative on memory there.
-	GGUFFallbackContextSize = 1024
-
 	// DefaultNGPULayers means "offload all layers"; the backend (fit_params)
 	// clamps to what actually fits in device memory.
 	DefaultNGPULayers = 99999999

core/config/gguf.go

@@ -33,7 +33,7 @@ func guessGGUFFromFile(cfg *ModelConfig, f *gguf.GGUFFile, defaultCtx int) {
 			cSize := int(ctxSize)
 			cfg.ContextSize = &cSize
 		} else {
-			defaultCtx = GGUFFallbackContextSize
+			defaultCtx = DefaultContextSize
 			cfg.ContextSize = &defaultCtx
 		}
 	}

core/config/hooks_llamacpp.go

@@ -34,7 +34,7 @@ func llamaCppDefaults(cfg *ModelConfig, modelPath string) {
 	// Default context size if not set, regardless of whether GGUF parsing succeeds
 	defer func() {
 		if cfg.ContextSize == nil {
-			ctx := GGUFFallbackContextSize
+			ctx := DefaultContextSize
 			cfg.ContextSize = &ctx
 		}
 	}()

core/config/hooks_test.go

@@ -248,7 +248,11 @@ var _ = Describe("Backend hooks and parser defaults", func() {
 			}
 			cfg.SetDefaults(ModelPath(dir))
 
+			// An unreadable/unparseable GGUF (e.g. a quant type the parser does
+			// not know, such as NVFP4) yields no estimate, so the hook must fall
+			// back to DefaultContextSize rather than a tiny, surprising value.
 			Expect(cfg.ContextSize).NotTo(BeNil())
+			Expect(*cfg.ContextSize).To(Equal(DefaultContextSize))
 		})
 	})
 

gallery/index.yaml

@@ -579,6 +579,10 @@
   icon: https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3.6/Figures/qwen3.6_35b_a3b_score.png
   overrides:
     backend: llama-cpp
+    # NVFP4 GGUFs use a quant type the GGUF metadata parser cannot read, so
+    # context size cannot be auto-derived; set it explicitly (the model trains
+    # to 262144, 32768 is a safe default operators can raise).
+    context_size: 32768
     function:
       automatic_tool_parsing_fallback: true
       grammar:
@@ -611,6 +615,9 @@
     - gguf
   overrides:
     backend: llama-cpp
+    # NVFP4 GGUFs use a quant type the GGUF metadata parser cannot read, so
+    # context size cannot be auto-derived; set it explicitly.
+    context_size: 32768
     function:
       automatic_tool_parsing_fallback: true
       grammar:
@@ -638,6 +645,9 @@
   icon: https://cdn-uploads.huggingface.co/production/uploads/66309bd090589b7c65950665/sGQKmrMc6L6guMoaB5_Y2.png
   overrides:
     backend: llama-cpp
+    # NVFP4 GGUFs use a quant type the GGUF metadata parser cannot read, so
+    # context size cannot be auto-derived; set it explicitly.
+    context_size: 32768
     function:
       automatic_tool_parsing_fallback: true
       grammar:
@@ -688,6 +698,10 @@
   icon: https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3.6/Figures/qwen3.6_27b_score.png
   overrides:
     backend: llama-cpp
+    # NVFP4 GGUFs use a quant type the GGUF metadata parser cannot read, so
+    # context size cannot be auto-derived; set it explicitly (the model trains
+    # to 262144, 32768 is a safe default operators can raise).
+    context_size: 32768
     function:
       automatic_tool_parsing_fallback: true
       grammar: