Add v75 bundle (LFM2-2.5-230M, Hexagon v75, W8)

v75/lfm2-5-230m-2048.json

@@ -0,0 +1,68 @@
+{
+  "schema_version": 1,
+  "_comment": "LFM2.5-230M @ MAXCTX=2048, W8 weight-only (int8 weights, fp16 activations) decode + batched prefill (PN=512) + NPU lm-head. Sliding-window KV ring continues past 2048 (remembers the last 2048 tokens). 14-layer hybrid (6 attn + 8 conv); per-layer schedule via attn_idx/conv_idx.",
+  "model": {
+    "name": "lfm2-5-230m-2048",
+    "family": "llm",
+    "dsp_arch": "v75"
+  },
+  "params": {
+    "hidden": 1024,
+    "vocab": 65536,
+    "n_layers": 14,
+    "max_ctx": 2048,
+    "kv_dim": 512,
+    "head_dim": 64,
+    "rope_theta": 1000000.0,
+    "eos_token_id": 7
+  },
+  "artifacts": {
+    "contexts": {
+      "decode": {
+        "bin": "lfm230_dec_2048_w8.bin"
+      },
+      "prefill": {
+        "bin": "lfm230_pf_2048_w8.bin"
+      },
+      "lmhead": {
+        "bin": "lfm230_lmh_w8.bin"
+      }
+    },
+    "embed": "lfm_embed_f16.bin",
+    "tokenizer": "tokenizer.json"
+  },
+  "plan": {
+    "steps": [
+      {
+        "host": "lfm_generate",
+        "params": {
+          "prefill": "lfm230_pf_2048_w8",
+          "decode": "lfm230_dec_2048_w8",
+          "lmhead": "lfm230_lmh_w8",
+          "attn_idx": "2,4,6,8,10,12",
+          "conv_idx": "0,1,3,5,7,9,11,13"
+        }
+      },
+      {
+        "emit": "all"
+      }
+    ]
+  },
+  "chat": {
+    "bos": [
+      1
+    ],
+    "system": {
+      "prefix": "<|im_start|>system\n",
+      "suffix": "<|im_end|>\n"
+    },
+    "user": {
+      "prefix": "<|im_start|>user\n",
+      "suffix": "<|im_end|>\n"
+    },
+    "assistant": {
+      "prefix": "<|im_start|>assistant\n",
+      "suffix": "<|im_end|>\n"
+    }
+  }
+}

v75/lfm2-5-230m.json

@@ -0,0 +1,68 @@
+{
+  "schema_version": 1,
+  "_comment": "LFM2.5-230M @ MAXCTX=512, W8 weight-only (int8 weights, fp16 activations) + GQA-native decode + batched prefill (PN=512) + NPU lm-head. 14-layer hybrid = 8 short-conv + 6 GQA-attention; per-layer schedule via attn_idx/conv_idx (the op defaults to the 16-layer 350M when absent).",
+  "model": {
+    "name": "lfm2-5-230m",
+    "family": "llm",
+    "dsp_arch": "v75"
+  },
+  "params": {
+    "hidden": 1024,
+    "vocab": 65536,
+    "n_layers": 14,
+    "max_ctx": 512,
+    "kv_dim": 512,
+    "head_dim": 64,
+    "rope_theta": 1000000.0,
+    "eos_token_id": 7
+  },
+  "artifacts": {
+    "contexts": {
+      "decode": {
+        "bin": "lfm230_dec_512_w8.bin"
+      },
+      "prefill": {
+        "bin": "lfm230_pf_512_w8.bin"
+      },
+      "lmhead": {
+        "bin": "lfm230_lmh_w8.bin"
+      }
+    },
+    "embed": "lfm_embed_f16.bin",
+    "tokenizer": "tokenizer.json"
+  },
+  "plan": {
+    "steps": [
+      {
+        "host": "lfm_generate",
+        "params": {
+          "prefill": "lfm230_pf_512_w8",
+          "decode": "lfm230_dec_512_w8",
+          "lmhead": "lfm230_lmh_w8",
+          "attn_idx": "2,4,6,8,10,12",
+          "conv_idx": "0,1,3,5,7,9,11,13"
+        }
+      },
+      {
+        "emit": "all"
+      }
+    ]
+  },
+  "chat": {
+    "bos": [
+      1
+    ],
+    "system": {
+      "prefix": "<|im_start|>system\n",
+      "suffix": "<|im_end|>\n"
+    },
+    "user": {
+      "prefix": "<|im_start|>user\n",
+      "suffix": "<|im_end|>\n"
+    },
+    "assistant": {
+      "prefix": "<|im_start|>assistant\n",
+      "suffix": "<|im_end|>\n"
+    }
+  }
+}

v75/lfm230_dec_2048_w8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c49c012b270e0bfa1fc54b9f388d6914252c554f9e562a5bb120990256d9d1c4
+size 164552704

v75/lfm230_dec_512_w8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d7266526c0924d490df333ec60d996d14f78d767458b4f86703f79be7d156c2
+size 164204544

v75/lfm230_lmh_w8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05a48852021bf3e038ecd71d2b6bbddd6f206f26c01833211e56dd71e9712364
+size 68198400

v75/lfm230_pf_2048_w8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6dc77e553369c6a0b90fa3d7b88d8914f930c416dd2e0a47f4392239a4e3f204
+size 183644160

v75/lfm230_pf_512_w8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9059963f00bc8fb9f09e3f620a344fe676fc9cb9f6ea36f6489401295d2b2869
+size 167231488

v75/lfm_embed_f16.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19f7edcbb7399d6ac2f75d532d08b7fe4f2784a2f4da750c021561af05be0b1b
+size 134217728