Upload folder using huggingface_hub

.gitattributes

@@ -65,3 +65,4 @@ gemma4-random-model/gguf-q4_0/gemma4-random-model-Q4_0.gguf filter=lfs diff=lfs
 qwen3-random-model/gguf-q4_0/qwen3-random-model-Q4_0.gguf filter=lfs diff=lfs merge=lfs -text
 smollm3-random-model/gguf-q4_0/smollm3-random-model-Q4_0.gguf filter=lfs diff=lfs merge=lfs -text
 gemma4-random-model/gguf-q4_k/gemma4-random-model-Q4_K.gguf filter=lfs diff=lfs merge=lfs -text
+gemma3-random-model/gguf-q4_k/gemma3-random-model-Q4_K.gguf filter=lfs diff=lfs merge=lfs -text

ARCHITECTURE_RANDOM_MODELS_REPORT.json

@@ -51,6 +51,24 @@
         "Q4_0": 29
       },
       "validation": "passed"
+    },
+    {
+      "architecture": "gemma3",
+      "case": "gemma3-random-model",
+      "eog_token_ids": [
+        106
+      ],
+      "gguf_quantization": "Q4_K",
+      "gguf_sha256": "68018d9e6377606ccd65a72de22ccdf4e891f640a8719c52045a7655cf601bd5",
+      "hf_dtype": "bfloat16",
+      "hf_sha256": "46055d484b8f97d2e81d0903bae0d9310eb3abab2a4735943b9cfab6daa2badf",
+      "parameter_count": 5938176,
+      "tensor_type_histogram": {
+        "F32": 37,
+        "Q4_K": 38,
+        "Q6_K": 5
+      },
+      "validation": "passed"
     }
   ],
   "llama_cpp_commit": "40b740ad05c531b9d57aca6698c3ed553a9e784c",

README.md

@@ -6,6 +6,7 @@ size_categories:
 tags:
 - llama
 - gemma4
+- gemma3
 - qwen3
 - smollm3
 - transformers
@@ -18,7 +19,7 @@ tags:
 
 # Deterministic Random Models
 
-This dataset contains ten small, deterministic language-model fixtures for
+This dataset contains eleven small, deterministic language-model fixtures for
 model-format, loader, inference, compatibility, and conformance testing. They
 are not trained models and must not be used for language-model quality
 evaluation.
@@ -40,9 +41,10 @@ checkpoint weights are included.
 | `gemma4-random-model` | Gemma 4 | 6,036,608 | BF16 | Q4_K | five-local/one-global attention schedule |
 | `qwen3-random-model` | Qwen 3 | 508,800 | BF16 | Q4_0 | wide Q projection and Q/K head norms |
 | `smollm3-random-model` | SmolLM3 | 4,917,504 | BF16 | Q4_0 | three-RoPE/one-no-RoPE layer schedule |
+| `gemma3-random-model` | Gemma 3 | 5,938,176 | BF16 | Q4_K | five-local/one-global attention schedule, EOS 106 |
 
 The seven Llama cases are derived from real Hugging Face configuration files by
-a preservation-first shrinker. Gemma 4, Qwen 3, and SmolLM3 retain
+a preservation-first shrinker. Gemma 4, Gemma 3, Qwen 3, and SmolLM3 retain
 architecture-specific reduced geometries that preserve important
 ratios, tensor inventories, and layer schedules observed in locally downloaded
 upstream GGUF models. Published case names use `random-model` rather than
@@ -87,7 +89,7 @@ The architecture-specific cases use:
 `-- metadata.json
 ```
 
-`manifest.json` is the machine-readable index of all ten model packages and
+`manifest.json` is the machine-readable index of all eleven model packages and
 their SHA-256 hashes and sizes.
 
 ## Synthetic weights and tokenizers
@@ -118,7 +120,7 @@ F32 or BF16 reference exactly.
 
 ## Reproducibility and scope
 
-The Hugging Face weights, configs, and GGUF outputs for Gemma 4, Qwen 3, and SmolLM3 were
+The Hugging Face weights, configs, and GGUF outputs for Gemma 4, Gemma 3, Qwen 3, and SmolLM3 were
 independently regenerated and found byte-identical. The Llama cases retain their
 source revisions, source-config hashes, shrink decisions, and generation
 provenance in each case directory.

SHA256SUMS

@@ -1,7 +1,7 @@
-d17b37c112047f3ac47f54634c1a59da55e179b4fb176ad596ecc94240750006  ARCHITECTURE_RANDOM_MODELS_REPORT.json
+23c1d84934c48197ac11e4c10cc8daebebf0207b177e24506a5ac603739d97aa  ARCHITECTURE_RANDOM_MODELS_REPORT.json
 18ec0cdad98853d7f10381790713f0b3190bd4f24962594bf576e11fdb07f76f  GGUF_Q4_K_M_ADDED_FOUR_REPORT.json
 695036ee0e223d3d79170fb3a0f795eb88f6337fb2e29c18f0b549979c057208  GGUF_Q4_K_M_REPORT.json
-fb3008e24a0fecc2bb0cb72b27770db78f72bc481dc086b8080c89f09e419816  README.md
+48dfe7e04721f8e95938d3151f26e852e3bbd482a1f2e2ba10b932c181673b99  README.md
 95d67a407dc96839250fe23e85e0a5bab9d355ee463b0c87bf9e33164c952b52  REPORT.md
 02130c5ef4fe4e00bf26cc6e0eb288b90916acf827031bf572ffc02a48d5f3f0  deepseek-coder/case.json
 616662945b2382e3c4ef36c05b40ebacea17e2e16fa3a723b7fd49eb5d0e9fdd  deepseek-coder/config-diff.json
@@ -22,6 +22,20 @@ b80f4edf3204563fd78e3c95498aa4ceb9af1770f52869c638d33be9124a43a5  deepseek-coder
 603a68ffd4df8d0b01492da00fcd7b7386f7a8bbe90e92b81b82516a9231fbfb  deepseek-coder/tokenizer/tokenizer_config.json
 7b1258e1304c5f09f3bc64cd4b569e7ddfa55a062c0581d0e5cb1a3d7a17fcc8  deepseek-coder/tokenizer/vocabulary.json
 aa20ff35795ac323bbdbc81c292f0fa74bb81f9026b88cd01c40c4d81e428d8c  deepseek-coder/validation.json
+0c39eb97dd1a63cb6b948799c00de65712451f629cf4a19d04eb790cfdfb88fc  gemma3-random-model/CONFIG_DECISION.md
+8f3d38791f1a3477d3585fb572eca5952b4c2f30588f23c69d7f24aa2397ccf5  gemma3-random-model/gguf-q4_k/convert.log
+68018d9e6377606ccd65a72de22ccdf4e891f640a8719c52045a7655cf601bd5  gemma3-random-model/gguf-q4_k/gemma3-random-model-Q4_K.gguf
+83d287b61ff13673e657d91301e047f4296ed19b6914574328a913cf2e126183  gemma3-random-model/gguf-q4_k/quantize.log
+7f5d3204aea64ffbbf686ea85afd77ba4c7d86d9af11b67e65150f5bf8b3c774  gemma3-random-model/hf-bf16/config.json
+3e4b35955db771320f4596c561ba812867070e6666cf2f01dcf439df56847e1a  gemma3-random-model/hf-bf16/generation_config.json
+46055d484b8f97d2e81d0903bae0d9310eb3abab2a4735943b9cfab6daa2badf  gemma3-random-model/hf-bf16/model.safetensors
+fb6670d673ef1a0d347ac6066d14aecf0992741f93df2e6f30cc2feed2d462fd  gemma3-random-model/hf-bf16/tokenizer.json
+47ce326acdb871124ef8fa106238d0916ffcf4ea1bf3e10f03a512c729dfcbcc  gemma3-random-model/hf-bf16/tokenizer.model
+d0fab7832fecaac8554eaebfad305d55655cc79b3404eb9a7b0a4fc9dce08a01  gemma3-random-model/hf-bf16/tokenizer_config.json
+3b8ff49c5af519e27c83c81048862e7cf05b5dc6da558ab16cb1db8f98cb9995  gemma3-random-model/metadata.json
+d380b7b0d97842cccb6dfe1e7c94baae4a31d362ea1996653e83c839f450b075  gemma3-random-model/reference/gguf-native.json
+b9a605c3702c2d3da0cd2ee00e5d1defd9b2e06be80aaa9b9809baac411a24e6  gemma3-random-model/reference/hf-outputs.safetensors
+719a2f5ea966e92e42ebcc43dd73f0714b5d740c9f426d22c942dbb1e7efb6ab  gemma3-random-model/reference/inputs.json
 4d463e831b4df46925f7ed82f898a9212d004882dacefc7012171254a70e52ed  gemma4-random-model/CONFIG_DECISION.md
 5c330d499055016556a88ee42b398fac5fa6fa7706f68cad2c4aa8e6d7286f1c  gemma4-random-model/gguf-q4_k/convert.log
 3ce6309aa4221289441b198ab63cd6c0d03075ba183ea1990001313c99d6e845  gemma4-random-model/gguf-q4_k/gemma4-random-model-Q4_K.gguf
@@ -73,7 +87,7 @@ f003d32e36ddc9d7c7a50c186da5c544b2828a0d04ca62aeac741d765727a635  livekit-turn-d
 564a5f0d3380d83d0ffcfa509a5e5369c0610f586cb6b3d0d45ccaee76610327  livekit-turn-detector/tokenizer/tokenizer_config.json
 7b1258e1304c5f09f3bc64cd4b569e7ddfa55a062c0581d0e5cb1a3d7a17fcc8  livekit-turn-detector/tokenizer/vocabulary.json
 391312ee129392e2b35b6a238593ca5ddff3dd51c6c324135f1bb0114be1e068  livekit-turn-detector/validation.json
-d3c192b83c78f0d46d46835f0e4e201dd80755dde5fd39016a051b5133b36410  manifest.json
+c7f352d7309177812e8db6def6caa5f336b4b4130b9950a81753a7f7ee2ad2e9  manifest.json
 047c4545232dc055c2c7e4a30a70c22e0e75e9e777e539732128f65c32b0c1cb  minicpm5/case.json
 282a55aae55b59573109fe37222197f298600ad691573bcd84ea5d4dc1afcc0a  minicpm5/config-diff.json
 d088266a01cd2f9a6107877932d6af9f65e6d00202292226eab76af8ea0e889f  minicpm5/gguf/metadata.json

gemma3-random-model/CONFIG_DECISION.md

@@ -0,0 +1,26 @@
+# Gemma 3 random-model configuration decision
+
+The source is `gemma-3-4b-it-Q4_K_M.gguf`. It identifies itself as
+`general.architecture = gemma3` and uses 34 layers, hidden width 2560, FFN width
+10240, eight query heads, four KV heads, head dimension 256, sliding window
+1024, context 131072, linear RoPE scaling factor 8, and tied embeddings.
+
+The selected reduced geometry is six layers, hidden width 256, FFN width 1024,
+four query heads, two KV heads, head dimension 64, sliding window 64, context
+128, and vocabulary 128. It preserves the FFN ratio of four, GQA ratio of two,
+one complete five-sliding/one-full attention period, linear RoPE scaling, RMS
+epsilon, four norm roles per layer, Q/K head norms, and tied embeddings.
+
+The source EOS token ID is 106 and remains 106. The synthetic HF tokenizer and
+converter SentencePiece tokenizer both cover IDs `0..127` and represent
+PAD/BOS/UNK/EOS as `0/2/3/106`. They are test tokenizers, not copies of the
+source linguistic vocabulary.
+
+All matrix quantization axes used by the reduced model are divisible by 256.
+The retained GGUF uses llama.cpp `Q4_K`, an alias for the mixed Q4_K_M profile,
+without `--pure`. Actual tensor types are recorded after generation.
+
+The official Transformers 5.15.0 `Gemma3ForCausalLM` construction contains
+5,938,176 trainable parameters. Norm parameters are initialized to zero because
+Gemma3 RMSNorm applies `1 + weight`; other one-dimensional parameters use their
+model defaults and matrices use the deterministic tlfloat LCG64 equation.

gemma3-random-model/gguf-q4_k/convert.log

@@ -0,0 +1,117 @@
+INFO:hf-to-gguf:Loading model: hf-bf16
+INFO:hf-to-gguf:Model architecture: Gemma3ForCausalLM
+INFO:hf-to-gguf:gguf: indexing model part 'model.safetensors'
+INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
+INFO:hf-to-gguf:Exporting model...
+INFO:hf-to-gguf:token_embd.weight,                torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.0.attn_norm.weight,           torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.0.ffn_down.weight,            torch.bfloat16 --> F32, shape = {1024, 256}
+INFO:hf-to-gguf:blk.0.ffn_gate.weight,            torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.0.ffn_up.weight,              torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.0.post_attention_norm.weight, torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.0.post_ffw_norm.weight,       torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.0.ffn_norm.weight,            torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.0.attn_k_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.0.attn_k.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.0.attn_output.weight,         torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.0.attn_q_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.0.attn_q.weight,              torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.0.attn_v.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.1.attn_norm.weight,           torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.1.ffn_down.weight,            torch.bfloat16 --> F32, shape = {1024, 256}
+INFO:hf-to-gguf:blk.1.ffn_gate.weight,            torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.1.ffn_up.weight,              torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.1.post_attention_norm.weight, torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.1.post_ffw_norm.weight,       torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.1.ffn_norm.weight,            torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.1.attn_k_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.1.attn_k.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.1.attn_output.weight,         torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.1.attn_q_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.1.attn_q.weight,              torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.1.attn_v.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.2.attn_norm.weight,           torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.2.ffn_down.weight,            torch.bfloat16 --> F32, shape = {1024, 256}
+INFO:hf-to-gguf:blk.2.ffn_gate.weight,            torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.2.ffn_up.weight,              torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.2.post_attention_norm.weight, torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.2.post_ffw_norm.weight,       torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.2.ffn_norm.weight,            torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.2.attn_k_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.2.attn_k.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.2.attn_output.weight,         torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.2.attn_q_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.2.attn_q.weight,              torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.2.attn_v.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.3.attn_norm.weight,           torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.3.ffn_down.weight,            torch.bfloat16 --> F32, shape = {1024, 256}
+INFO:hf-to-gguf:blk.3.ffn_gate.weight,            torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.3.ffn_up.weight,              torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.3.post_attention_norm.weight, torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.3.post_ffw_norm.weight,       torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.3.ffn_norm.weight,            torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.3.attn_k_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.3.attn_k.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.3.attn_output.weight,         torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.3.attn_q_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.3.attn_q.weight,              torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.3.attn_v.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.4.attn_norm.weight,           torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.4.ffn_down.weight,            torch.bfloat16 --> F32, shape = {1024, 256}
+INFO:hf-to-gguf:blk.4.ffn_gate.weight,            torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.4.ffn_up.weight,              torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.4.post_attention_norm.weight, torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.4.post_ffw_norm.weight,       torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.4.ffn_norm.weight,            torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.4.attn_k_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.4.attn_k.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.4.attn_output.weight,         torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.4.attn_q_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.4.attn_q.weight,              torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.4.attn_v.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.5.attn_norm.weight,           torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.5.ffn_down.weight,            torch.bfloat16 --> F32, shape = {1024, 256}
+INFO:hf-to-gguf:blk.5.ffn_gate.weight,            torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.5.ffn_up.weight,              torch.bfloat16 --> F32, shape = {256, 1024}
+INFO:hf-to-gguf:blk.5.post_attention_norm.weight, torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.5.post_ffw_norm.weight,       torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.5.ffn_norm.weight,            torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:blk.5.attn_k_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.5.attn_k.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:blk.5.attn_output.weight,         torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.5.attn_q_norm.weight,         torch.bfloat16 --> F32, shape = {64}
+INFO:hf-to-gguf:blk.5.attn_q.weight,              torch.bfloat16 --> F32, shape = {256, 256}
+INFO:hf-to-gguf:blk.5.attn_v.weight,              torch.bfloat16 --> F32, shape = {256, 128}
+INFO:hf-to-gguf:output_norm.weight,               torch.bfloat16 --> F32, shape = {256}
+INFO:hf-to-gguf:Set meta model
+INFO:hf-to-gguf:Set model parameters
+INFO:hf-to-gguf:gguf: context length = 128
+INFO:hf-to-gguf:gguf: embedding length = 256
+INFO:hf-to-gguf:gguf: feed forward length = 1024
+INFO:hf-to-gguf:gguf: head count = 4
+INFO:hf-to-gguf:gguf: key-value head count = 2
+INFO:hf-to-gguf:gguf: rope scaling type = LINEAR
+INFO:hf-to-gguf:gguf: rope theta = 1000000.0
+INFO:hf-to-gguf:gguf: rope theta swa = 10000.0
+INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
+INFO:hf-to-gguf:gguf: file type = 0
+WARNING:gguf.gguf_writer:Duplicated key name 'gemma3.context_length', overwriting it with new value 128 of type UINT32
+WARNING:gguf.gguf_writer:Duplicated key name 'gemma3.attention.head_count', overwriting it with new value 4 of type UINT32
+WARNING:gguf.gguf_writer:Duplicated key name 'gemma3.attention.layer_norm_rms_epsilon', overwriting it with new value 1e-06 of type FLOAT32
+WARNING:gguf.gguf_writer:Duplicated key name 'gemma3.attention.key_length', overwriting it with new value 64 of type UINT32
+WARNING:gguf.gguf_writer:Duplicated key name 'gemma3.attention.value_length', overwriting it with new value 64 of type UINT32
+WARNING:gguf.gguf_writer:Duplicated key name 'gemma3.rope.freq_base', overwriting it with new value 1000000.0 of type FLOAT32
+WARNING:gguf.gguf_writer:Duplicated key name 'gemma3.attention.head_count_kv', overwriting it with new value 2 of type UINT32
+INFO:hf-to-gguf:Set model quantization version
+INFO:hf-to-gguf:Set model tokenizer
+INFO:gguf.vocab:Setting special token type bos to 2
+INFO:gguf.vocab:Setting special token type eos to 106
+INFO:gguf.vocab:Setting special token type unk to 3
+INFO:gguf.vocab:Setting special token type pad to 0
+INFO:gguf.vocab:Setting special token type mask to 4
+INFO:gguf.gguf_writer:Writing the following files:
+INFO:gguf.gguf_writer:/home/codex/tmp/gemma3-q4k-work-20260813/gemma3-random-model-F32.gguf: n_tensors = 80, total_size = 23.8M
+
+Writing:   0%|          | 0.00/23.8M [00:00<?, ?byte/s]
+Writing: 100%|██████████| 23.8M/23.8M [00:00<00:00, 1.54Gbyte/s]
+INFO:hf-to-gguf:Model successfully exported to /home/codex/tmp/gemma3-q4k-work-20260813/gemma3-random-model-F32.gguf

gemma3-random-model/gguf-q4_k/gemma3-random-model-Q4_K.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68018d9e6377606ccd65a72de22ccdf4e891f640a8719c52045a7655cf601bd5
+size 3533984

gemma3-random-model/gguf-q4_k/quantize.log

@@ -0,0 +1,122 @@
+llama_print_build_info: build = 0 (unknown)
+llama_print_build_info: built with Clang 21.1.8 for Linux x86_64
+llama_quantize: quantizing '/home/codex/tmp/gemma3-q4k-work-20260813/gemma3-random-model-F32.gguf' to '/home/codex/conf_track/artifacts/gemma3-v0/gemma3-random-model/gguf-q4_k/gemma3-random-model-Q4_K.gguf' as Q4_K
+llama_model_loader: loaded meta data with 31 key-value pairs and 80 tensors from /home/codex/tmp/gemma3-q4k-work-20260813/gemma3-random-model-F32.gguf (version GGUF V3 (latest))
+llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
+llama_model_loader: - kv   0:                       general.architecture str              = gemma3
+llama_model_loader: - kv   1:                               general.type str              = model
+llama_model_loader: - kv   2:                               general.name str              = Gemma 3 Random Model
+llama_model_loader: - kv   3:                         general.size_label str              = 5.9M
+llama_model_loader: - kv   4:                         gemma3.block_count u32              = 6
+llama_model_loader: - kv   5:                      gemma3.context_length u32              = 128
+llama_model_loader: - kv   6:                    gemma3.embedding_length u32              = 256
+llama_model_loader: - kv   7:                 gemma3.feed_forward_length u32              = 1024
+llama_model_loader: - kv   8:                gemma3.attention.head_count u32              = 4
+llama_model_loader: - kv   9:             gemma3.attention.head_count_kv u32              = 2
+llama_model_loader: - kv  10:                   gemma3.rope.scaling.type str              = linear
+llama_model_loader: - kv  11:                 gemma3.rope.scaling.factor f32              = 8.000000
+llama_model_loader: - kv  12:                      gemma3.rope.freq_base f32              = 1000000.000000
+llama_model_loader: - kv  13:                  gemma3.rope.freq_base_swa f32              = 10000.000000
+llama_model_loader: - kv  14:    gemma3.attention.layer_norm_rms_epsilon f32              = 0.000001
+llama_model_loader: - kv  15:                gemma3.attention.key_length u32              = 64
+llama_model_loader: - kv  16:              gemma3.attention.value_length u32              = 64
+llama_model_loader: - kv  17:                          general.file_type u32              = 0
+llama_model_loader: - kv  18:            gemma3.attention.sliding_window u32              = 64
+llama_model_loader: - kv  19:               general.quantization_version u32              = 2
+llama_model_loader: - kv  20:                       tokenizer.ggml.model str              = llama
+llama_model_loader: - kv  21:                         tokenizer.ggml.pre str              = default
+llama_model_loader: - kv  22:                      tokenizer.ggml.tokens arr[str,128]     = ["<pad>", "de", "<bos>", "<unk>", "om...
+llama_model_loader: - kv  23:                      tokenizer.ggml.scores arr[f32,128]     = [0.000000, -0.000000, 0.000000, 0.000...
+llama_model_loader: - kv  24:                  tokenizer.ggml.token_type arr[i32,128]     = [3, 1, 3, 2, 1, 1, 1, 1, 1, 1, 1, 1, ...
+llama_model_loader: - kv  25:                tokenizer.ggml.bos_token_id u32              = 2
+llama_model_loader: - kv  26:                tokenizer.ggml.eos_token_id u32              = 106
+llama_model_loader: - kv  27:            tokenizer.ggml.unknown_token_id u32              = 3
+llama_model_loader: - kv  28:            tokenizer.ggml.padding_token_id u32              = 0
+llama_model_loader: - kv  29:               tokenizer.ggml.mask_token_id u32              = 4
+llama_model_loader: - kv  30:            tokenizer.ggml.add_space_prefix bool             = false
+llama_model_loader: - type  f32:   80 tensors
+[   1/  80] output_norm.weight                   - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[   2/  80] token_embd.weight                    - [   256,    128,      1,      1], type =    f32, converting to q6_K .. size =     0.12 MiB ->     0.03 MiB
+[   3/  80] blk.0.attn_k.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[   4/  80] blk.0.attn_k_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[   5/  80] blk.0.attn_norm.weight               - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[   6/  80] blk.0.attn_output.weight             - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[   7/  80] blk.0.attn_q.weight                  - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[   8/  80] blk.0.attn_q_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[   9/  80] blk.0.attn_v.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  10/  80] blk.0.ffn_down.weight                - [  1024,    256,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  11/  80] blk.0.ffn_gate.weight                - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  12/  80] blk.0.ffn_norm.weight                - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  13/  80] blk.0.ffn_up.weight                  - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  14/  80] blk.0.post_attention_norm.weight     - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  15/  80] blk.0.post_ffw_norm.weight           - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  16/  80] blk.1.attn_k.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  17/  80] blk.1.attn_k_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  18/  80] blk.1.attn_norm.weight               - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  19/  80] blk.1.attn_output.weight             - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  20/  80] blk.1.attn_q.weight                  - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  21/  80] blk.1.attn_q_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  22/  80] blk.1.attn_v.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  23/  80] blk.1.ffn_down.weight                - [  1024,    256,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  24/  80] blk.1.ffn_gate.weight                - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  25/  80] blk.1.ffn_norm.weight                - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  26/  80] blk.1.ffn_up.weight                  - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  27/  80] blk.1.post_attention_norm.weight     - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  28/  80] blk.1.post_ffw_norm.weight           - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  29/  80] blk.2.attn_k.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  30/  80] blk.2.attn_k_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  31/  80] blk.2.attn_norm.weight               - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  32/  80] blk.2.attn_output.weight             - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  33/  80] blk.2.attn_q.weight                  - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  34/  80] blk.2.attn_q_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  35/  80] blk.2.attn_v.weight                  - [   256,    128,      1,      1], type =    f32, converting to q6_K .. size =     0.12 MiB ->     0.03 MiB
+[  36/  80] blk.2.ffn_down.weight                - [  1024,    256,      1,      1], type =    f32, converting to q6_K .. size =     1.00 MiB ->     0.21 MiB
+[  37/  80] blk.2.ffn_gate.weight                - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  38/  80] blk.2.ffn_norm.weight                - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  39/  80] blk.2.ffn_up.weight                  - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  40/  80] blk.2.post_attention_norm.weight     - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  41/  80] blk.2.post_ffw_norm.weight           - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  42/  80] blk.3.attn_k.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  43/  80] blk.3.attn_k_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  44/  80] blk.3.attn_norm.weight               - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  45/  80] blk.3.attn_output.weight             - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  46/  80] blk.3.attn_q.weight                  - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  47/  80] blk.3.attn_q_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  48/  80] blk.3.attn_v.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  49/  80] blk.3.ffn_down.weight                - [  1024,    256,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  50/  80] blk.3.ffn_gate.weight                - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  51/  80] blk.3.ffn_norm.weight                - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  52/  80] blk.3.ffn_up.weight                  - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  53/  80] blk.3.post_attention_norm.weight     - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  54/  80] blk.3.post_ffw_norm.weight           - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  55/  80] blk.4.attn_k.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  56/  80] blk.4.attn_k_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  57/  80] blk.4.attn_norm.weight               - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  58/  80] blk.4.attn_output.weight             - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  59/  80] blk.4.attn_q.weight                  - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  60/  80] blk.4.attn_q_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  61/  80] blk.4.attn_v.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  62/  80] blk.4.ffn_down.weight                - [  1024,    256,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  63/  80] blk.4.ffn_gate.weight                - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  64/  80] blk.4.ffn_norm.weight                - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  65/  80] blk.4.ffn_up.weight                  - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  66/  80] blk.4.post_attention_norm.weight     - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  67/  80] blk.4.post_ffw_norm.weight           - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  68/  80] blk.5.attn_k.weight                  - [   256,    128,      1,      1], type =    f32, converting to q4_K .. size =     0.12 MiB ->     0.02 MiB
+[  69/  80] blk.5.attn_k_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  70/  80] blk.5.attn_norm.weight               - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  71/  80] blk.5.attn_output.weight             - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  72/  80] blk.5.attn_q.weight                  - [   256,    256,      1,      1], type =    f32, converting to q4_K .. size =     0.25 MiB ->     0.04 MiB
+[  73/  80] blk.5.attn_q_norm.weight             - [    64,      1,      1,      1], type =    f32, size =    0.000 MiB
+[  74/  80] blk.5.attn_v.weight                  - [   256,    128,      1,      1], type =    f32, converting to q6_K .. size =     0.12 MiB ->     0.03 MiB
+[  75/  80] blk.5.ffn_down.weight                - [  1024,    256,      1,      1], type =    f32, converting to q6_K .. size =     1.00 MiB ->     0.21 MiB
+[  76/  80] blk.5.ffn_gate.weight                - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  77/  80] blk.5.ffn_norm.weight                - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  78/  80] blk.5.ffn_up.weight                  - [   256,   1024,      1,      1], type =    f32, converting to q4_K .. size =     1.00 MiB ->     0.14 MiB
+[  79/  80] blk.5.post_attention_norm.weight     - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+[  80/  80] blk.5.post_ffw_norm.weight           - [   256,      1,      1,      1], type =    f32, size =    0.001 MiB
+llama_model_quantize_impl: model size  =    22.65 MiB (32.00 BPW)
+llama_model_quantize_impl: quant size  =     3.36 MiB (4.75 BPW)
+
+llama_quantize: quantize time =    61.30 ms
+llama_quantize:    total time =    61.30 ms

gemma3-random-model/hf-bf16/config.json

@@ -0,0 +1,51 @@
+{
+  "_sliding_window_pattern": 6,
+  "architectures": [
+    "Gemma3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "attn_logit_softcapping": null,
+  "bos_token_id": 2,
+  "dtype": "bfloat16",
+  "eos_token_id": 106,
+  "final_logit_softcapping": null,
+  "head_dim": 64,
+  "hidden_activation": "gelu_pytorch_tanh",
+  "hidden_size": 256,
+  "initializer_range": 0.02,
+  "intermediate_size": 1024,
+  "layer_types": [
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 128,
+  "model_type": "gemma3_text",
+  "num_attention_heads": 4,
+  "num_hidden_layers": 6,
+  "num_key_value_heads": 2,
+  "pad_token_id": 0,
+  "query_pre_attn_scalar": 256,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "full_attention": {
+      "factor": 8.0,
+      "rope_theta": 1000000.0,
+      "rope_type": "linear"
+    },
+    "sliding_attention": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    }
+  },
+  "sliding_window": 64,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.15.0",
+  "use_bidirectional_attention": false,
+  "use_cache": true,
+  "vocab_size": 128
+}

gemma3-random-model/hf-bf16/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 2,
+  "eos_token_id": 106,
+  "pad_token_id": 0,
+  "transformers_version": "5.15.0"
+}

gemma3-random-model/hf-bf16/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46055d484b8f97d2e81d0903bae0d9310eb3abab2a4735943b9cfab6daa2badf
+size 11885040

gemma3-random-model/hf-bf16/tokenizer.json

@@ -0,0 +1,245 @@
+{
+  "version": "1.0",
+  "truncation": null,
+  "padding": null,
+  "added_tokens": [
+    {
+      "id": 0,
+      "content": "<pad>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 2,
+      "content": "<bos>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 3,
+      "content": "<unk>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 4,
+      "content": "<mask>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 106,
+      "content": "<eos>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    }
+  ],
+  "normalizer": null,
+  "pre_tokenizer": {
+    "type": "ByteLevel",
+    "add_prefix_space": false,
+    "trim_offsets": true,
+    "use_regex": true
+  },
+  "post_processor": {
+    "type": "TemplateProcessing",
+    "single": [
+      {
+        "Sequence": {
+          "id": "A",
+          "type_id": 0
+        }
+      }
+    ],
+    "pair": [
+      {
+        "Sequence": {
+          "id": "A",
+          "type_id": 0
+        }
+      },
+      {
+        "Sequence": {
+          "id": "B",
+          "type_id": 1
+        }
+      }
+    ],
+    "special_tokens": {}
+  },
+  "decoder": {
+    "type": "Sequence",
+    "decoders": [
+      {
+        "type": "ByteFallback"
+      },
+      {
+        "type": "ByteLevel",
+        "add_prefix_space": true,
+        "trim_offsets": true,
+        "use_regex": true
+      }
+    ]
+  },
+  "model": {
+    "type": "BPE",
+    "dropout": null,
+    "unk_token": "<unk>",
+    "continuing_subword_prefix": null,
+    "end_of_word_suffix": null,
+    "fuse_unk": false,
+    "byte_fallback": true,
+    "ignore_merges": false,
+    "vocab": {
+      "<pad>": 0,
+      "<t001>": 1,
+      "<bos>": 2,
+      "<unk>": 3,
+      "<mask>": 4,
+      "a": 5,
+      "b": 6,
+      "ab": 7,
+      "<t008>": 8,
+      "<t009>": 9,
+      "<t010>": 10,
+      "<t011>": 11,
+      "<t012>": 12,
+      "<t013>": 13,
+      "<t014>": 14,
+      "<t015>": 15,
+      "<t016>": 16,
+      "<t017>": 17,
+      "<t018>": 18,
+      "<t019>": 19,
+      "<t020>": 20,
+      "<t021>": 21,
+      "<t022>": 22,
+      "<t023>": 23,
+      "<t024>": 24,
+      "<t025>": 25,
+      "<t026>": 26,
+      "<t027>": 27,
+      "<t028>": 28,
+      "<t029>": 29,
+      "<t030>": 30,
+      "<t031>": 31,
+      "<t032>": 32,
+      "<t033>": 33,
+      "<t034>": 34,
+      "<t035>": 35,
+      "<t036>": 36,
+      "<t037>": 37,
+      "<t038>": 38,
+      "<t039>": 39,
+      "<t040>": 40,
+      "<t041>": 41,
+      "<t042>": 42,
+      "<t043>": 43,
+      "<t044>": 44,
+      "<t045>": 45,
+      "<t046>": 46,
+      "<t047>": 47,
+      "<t048>": 48,
+      "<t049>": 49,
+      "<t050>": 50,
+      "<t051>": 51,
+      "<t052>": 52,
+      "<t053>": 53,
+      "<t054>": 54,
+      "<t055>": 55,
+      "<t056>": 56,
+      "<t057>": 57,
+      "<t058>": 58,
+      "<t059>": 59,
+      "<t060>": 60,
+      "<t061>": 61,
+      "<t062>": 62,
+      "<t063>": 63,
+      "<t064>": 64,
+      "<t065>": 65,
+      "<t066>": 66,
+      "<t067>": 67,
+      "<t068>": 68,
+      "<t069>": 69,
+      "<t070>": 70,
+      "<t071>": 71,
+      "<t072>": 72,
+      "<t073>": 73,
+      "<t074>": 74,
+      "<t075>": 75,
+      "<t076>": 76,
+      "<t077>": 77,
+      "<t078>": 78,
+      "<t079>": 79,
+      "<t080>": 80,
+      "<t081>": 81,
+      "<t082>": 82,
+      "<t083>": 83,
+      "<t084>": 84,
+      "<t085>": 85,
+      "<t086>": 86,
+      "<t087>": 87,
+      "<t088>": 88,
+      "<t089>": 89,
+      "<t090>": 90,
+      "<t091>": 91,
+      "<t092>": 92,
+      "<t093>": 93,
+      "<t094>": 94,
+      "<t095>": 95,
+      "<t096>": 96,
+      "<t097>": 97,
+      "<t098>": 98,
+      "<t099>": 99,
+      "<t100>": 100,
+      "<t101>": 101,
+      "<t102>": 102,
+      "<t103>": 103,
+      "<t104>": 104,
+      "<t105>": 105,
+      "<eos>": 106,
+      "<t107>": 107,
+      "<t108>": 108,
+      "<t109>": 109,
+      "<t110>": 110,
+      "<t111>": 111,
+      "<t112>": 112,
+      "<t113>": 113,
+      "<t114>": 114,
+      "<t115>": 115,
+      "<t116>": 116,
+      "<t117>": 117,
+      "<t118>": 118,
+      "<t119>": 119,
+      "<t120>": 120,
+      "<t121>": 121,
+      "<t122>": 122,
+      "<t123>": 123,
+      "<t124>": 124,
+      "<t125>": 125,
+      "<t126>": 126,
+      "<t127>": 127
+    },
+    "merges": [
+      [
+        "a",
+        "b"
+      ]
+    ]
+  }
+}

gemma3-random-model/hf-bf16/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47ce326acdb871124ef8fa106238d0916ffcf4ea1bf3e10f03a512c729dfcbcc
+size 241760

gemma3-random-model/hf-bf16/tokenizer_config.json

@@ -0,0 +1,10 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<bos>",
+  "eos_token": "<eos>",
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "<unk>"
+}

gemma3-random-model/metadata.json

@@ -0,0 +1,137 @@
+{
+  "architecture": "gemma3",
+  "gguf": {
+    "converter_command": [
+      "PYTHON",
+      "CONVERTER",
+      "HF_PACKAGE",
+      "--outfile",
+      "F32_OUTPUT",
+      "--outtype",
+      "f32",
+      "--model-name",
+      "Gemma 3 Random Model"
+    ],
+    "f32_intermediate_retained_in_package": false,
+    "f32_intermediate_sha256": "2b9c7ab9fb143deb87239f025566d7dbcb9771836ee4f9092d610af89cd8e199",
+    "llama_cpp_commit": "0b1bad14ff204627636aeb1de22ddcd5acb859d4",
+    "llama_cpp_profile": "Q4_K_M alias",
+    "model_sha256": "68018d9e6377606ccd65a72de22ccdf4e891f640a8719c52045a7655cf601bd5",
+    "model_size_bytes": 3533984,
+    "pure": false,
+    "quantizer_command": [
+      "LLAMA_QUANTIZE",
+      "IDENTITY_OVERRIDES",
+      "F32_INPUT",
+      "OUTPUT",
+      "Q4_K"
+    ],
+    "requested_quantization": "Q4_K",
+    "tensor_type_histogram": {
+      "F32": 37,
+      "Q4_K": 38,
+      "Q6_K": 5
+    },
+    "eog_token_ids": [
+      106
+    ],
+    "direct_token_validation": {
+      "load": true,
+      "prefill_token_ids": [
+        5,
+        7,
+        11,
+        13
+      ],
+      "decode_token_id": 17,
+      "eog_token_ids": [
+        106
+      ],
+      "all_logits_finite": true,
+      "reproducible": true,
+      "reference": "reference/gguf-native.json"
+    }
+  },
+  "hf": {
+    "config_sha256": "7f5d3204aea64ffbbf686ea85afd77ba4c7d86d9af11b67e65150f5bf8b3c774",
+    "construct_prefill_decode_reload": "passed",
+    "dtype": "bfloat16",
+    "model_sha256": "46055d484b8f97d2e81d0903bae0d9310eb3abab2a4735943b9cfab6daa2badf",
+    "reload_tolerance": {
+      "atol": 0.02,
+      "rtol": 0.02
+    },
+    "tokenizer_json_sha256": "fb6670d673ef1a0d347ac6066d14aecf0992741f93df2e6f30cc2feed2d462fd",
+    "tokenizer_model_sha256": "47ce326acdb871124ef8fa106238d0916ffcf4ea1bf3e10f03a512c729dfcbcc",
+    "transformers_version": "5.15.0"
+  },
+  "parameter_count": 5938176,
+  "rng": {
+    "increment": 1442695040888963407,
+    "multiplier": 6364136223846793005,
+    "name": "tlfloat LCG64 equation",
+    "warmup_steps": 10
+  },
+  "schema_version": 1,
+  "seed": 5135595944287600641,
+  "source_gguf": {
+    "architecture": "gemma3",
+    "path": "/home/codex/llm_models/gguf/gemma-3-4b-it-Q4_K_M.gguf",
+    "preserved": [
+      "GQA ratio",
+      "FFN ratio",
+      "six-layer SWA period",
+      "linear RoPE factor",
+      "RMS epsilon",
+      "tensor roles",
+      "special-token IDs",
+      "mixed K-quant profile"
+    ],
+    "tensor_count": 444,
+    "tensor_type_histogram": {
+      "F32": 205,
+      "Q4_K": 204,
+      "Q6_K": 35
+    }
+  },
+  "hashes": {
+    "hf_model": "46055d484b8f97d2e81d0903bae0d9310eb3abab2a4735943b9cfab6daa2badf",
+    "gguf_model": "68018d9e6377606ccd65a72de22ccdf4e891f640a8719c52045a7655cf601bd5",
+    "tokenizer_json": "fb6670d673ef1a0d347ac6066d14aecf0992741f93df2e6f30cc2feed2d462fd",
+    "tokenizer_model": "47ce326acdb871124ef8fa106238d0916ffcf4ea1bf3e10f03a512c729dfcbcc",
+    "hf_config": "7f5d3204aea64ffbbf686ea85afd77ba4c7d86d9af11b67e65150f5bf8b3c774"
+  },
+  "validation": {
+    "transformers": {
+      "construct": true,
+      "prefill": true,
+      "decode": true,
+      "save_reload": true,
+      "all_outputs_finite": true
+    },
+    "native": {
+      "load": true,
+      "direct_token_ids": true,
+      "prefill_token_ids": [
+        5,
+        7,
+        11,
+        13
+      ],
+      "decode_token_id": 17,
+      "eog_token_ids": [
+        106
+      ],
+      "finite": true,
+      "reproducible": true
+    },
+    "independent_regeneration": {
+      "hf_config_byte_identical": true,
+      "hf_weights_byte_identical": true,
+      "tokenizer_json_byte_identical": true,
+      "tokenizer_model_byte_identical": true,
+      "gguf_byte_identical": true,
+      "native_output_byte_identical": true
+    }
+  }
+}

gemma3-random-model/reference/gguf-native.json

@@ -0,0 +1,10 @@
+{
+  "schema_version": 1,
+  "vocab_size": 128,
+  "prefill_token_ids": [5,7,11,13],
+  "decode_token_id": 17,
+  "eog_token_ids": [106],
+  "all_logits_finite": true,
+  "prefill_last_logits": [0.241982147,0.345950752,0.222731277,0.134747565,0.132149994,0.0755203068,0.137293577,-0.332170904,0.358458996,-0.0566437244,0.307696223,-0.0512342751,-0.382270575,-0.0209459215,-0.109214738,0.0234475732,-0.148247242,-0.0502204895,-0.10334295,-0.227172077,0.144968301,0.127509624,-0.00209470093,0.562848449,-0.171330884,-0.045158267,0.14932999,0.105372697,0.267143011,0.0595521182,0.0873293728,0.21401751,-0.0756196976,-0.00524904951,0.0400382243,0.226262823,0.302781403,-0.205096096,-0.0376740247,0.245545238,0.102873266,-0.128519028,-0.167023435,0.209300846,-0.277735054,-0.0224522054,-0.163877085,-0.0211704373,0.314279139,0.313953578,-0.251288205,-0.065070644,0.0891045332,-0.0531355739,-0.0559249371,0.00801149011,0.0521210134,-0.20656988,0.354952842,0.129297733,-0.0402679443,-0.0030708909,0.196776018,-0.269928515,0.217599094,0.223678336,0.266249359,-0.0693843067,-0.207773432,0.140736535,-0.00987607241,0.201280445,0.0823513269,0.309711516,0.219206065,0.167958021,-0.556166589,-0.213413537,0.280611306,0.303692222,0.0274336338,0.205414116,-0.131083414,-0.0456702411,0.0235467926,-0.0390607715,0.0467742532,0.122853056,0.1980474,-0.234881043,0.206468701,-0.00772322714,-0.0493326783,-0.0658778399,0.0507170558,0.00719442964,0.371389389,-0.267657995,0.385891408,-0.125065327,0.257185131,0.130543932,-0.0144416392,0.0348021239,-0.120452195,-0.24567306,-0.26220715,0.162885666,-0.329199046,-0.192865208,0.23327595,-0.00328345597,0.000163927674,-0.159336865,-0.440276623,-0.13661325,-0.147915632,-0.0473099053,0.0488828793,0.0636885092,0.21004349,0.0601646155,-0.326441437,-0.0388595462,0.0450754762,0.0426438525,0.268783748,-0.198671922],
+  "decode_logits": [0.187236145,0.324398279,0.337632418,0.154553533,-0.0521036908,-0.050712049,0.138159484,-0.29411608,0.255713791,-0.115917265,0.174143761,-0.0549199656,-0.380445242,-0.169972315,-0.106870383,-0.0955797136,-0.0890652984,0.0913614482,0.0561397821,-0.0911246538,0.0752750635,0.136879474,-0.0155673176,0.157880276,-0.116104253,-0.0527709126,0.0143521428,-0.0221927762,0.112970255,0.107081681,0.0420698076,0.151492178,-0.34386301,-0.0209120512,0.0281583071,0.136092484,0.163728327,-0.253701508,-0.354121029,0.255385846,-0.0424471572,-0.366866469,0.0936541557,0.212397516,-0.25067842,-0.140878499,-0.0642191917,0.147980988,0.250819027,0.361205786,-0.113724217,0.0600615889,0.12207219,-0.199790552,-0.168662697,-0.226769507,0.256488144,-0.203081727,0.373617172,-0.0410287678,0.00627158582,-0.0474562347,0.030558791,-0.233366668,0.192625165,0.171050414,0.233616337,0.0262652934,-0.365188539,0.192161053,-0.131965041,0.141148627,-0.0869083405,-0.0832275152,0.316192001,-0.0310173482,-0.501562119,0.0149842203,0.0485771261,0.529980063,0.0530796051,0.23991479,-0.146612108,0.0109852552,0.0698975623,0.0779643357,0.0754460692,0.0628481954,0.169652537,-0.300728559,0.0718710646,-0.031375587,-0.177907974,0.000976443291,-0.0744490623,0.0906283557,0.191013828,-0.207301363,-0.00185766816,-0.205605268,0.200302869,0.279964298,-0.0932744443,-0.140320241,0.0418725386,-0.14444536,-0.0329549909,0.227150112,-0.373468518,-0.234621674,0.299688965,0.0295725614,0.100308448,-0.0359274596,-0.237680241,-0.181339249,-0.0468749702,-0.00228378177,-0.0329471305,-0.0340194926,0.0993724167,-0.241596162,-0.286910087,-0.0799263716,-0.0318076015,-0.192057967,0.397560537,-0.0597724468]
+}

gemma3-random-model/reference/hf-outputs.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9a605c3702c2d3da0cd2ee00e5d1defd9b2e06be80aaa9b9809baac411a24e6
+size 2720

gemma3-random-model/reference/inputs.json

@@ -0,0 +1,9 @@
+{
+  "decode_token_id": 17,
+  "prefill_token_ids": [
+    5,
+    7,
+    11,
+    13
+  ]
+}

manifest.json

@@ -192,6 +192,26 @@
           "size_bytes": 9839096
         }
       }
+    },
+    {
+      "architecture": "gemma3",
+      "auxiliary_tokenizer": {
+        "path": "gemma3-random-model/hf-bf16/tokenizer.json",
+        "sha256": "fb6670d673ef1a0d347ac6066d14aecf0992741f93df2e6f30cc2feed2d462fd"
+      },
+      "case": "gemma3-random-model",
+      "packages": {
+        "gguf_q4_k": {
+          "path": "gemma3-random-model/gguf-q4_k/gemma3-random-model-Q4_K.gguf",
+          "sha256": "68018d9e6377606ccd65a72de22ccdf4e891f640a8719c52045a7655cf601bd5",
+          "size_bytes": 3533984
+        },
+        "safetensors_bf16": {
+          "path": "gemma3-random-model/hf-bf16/model.safetensors",
+          "sha256": "46055d484b8f97d2e81d0903bae0d9310eb3abab2a4735943b9cfab6daa2badf",
+          "size_bytes": 11885040
+        }
+      }
     }
   ],
   "schema_version": 1