Guinius Giro 002.14.26 — 11,950 examples, fresh Qwen3-0.6B, mega engine v3

mlc-chat-config.json

@@ -14,7 +14,7 @@
     "rope_theta": 1000000,
     "vocab_size": 151936,
     "tie_word_embeddings": true,
-    "context_window_size": 2048,
+    "context_window_size": 40960,
     "prefill_chunk_size": 2048,
     "tensor_parallel_shards": 1,
     "head_dim": 128,
@@ -23,7 +23,7 @@
     "weight_block_size": null
   },
   "vocab_size": 151936,
-  "context_window_size": 2048,
+  "context_window_size": 40960,
   "sliding_window_size": -1,
   "prefill_chunk_size": 2048,
   "attention_sink_size": -1,
@@ -48,9 +48,9 @@
     "strip_space_in_decode": false
   },
   "conv_template": {
-    "name": "chatml",
+    "name": "qwen2",
     "system_template": "<|im_start|>system\n{system_message}<|im_end|>\n",
-    "system_message": "A conversation between a user and an LLM-based AI assistant. The assistant gives helpful and honest answers.",
+    "system_message": "You are a helpful assistant.",
     "system_prefix_token_ids": null,
     "add_role_after_system_message": true,
     "roles": {
@@ -69,10 +69,12 @@
     "role_content_sep": "\n",
     "role_empty_sep": "\n",
     "stop_str": [
+      "<|endoftext|>",
       "<|im_end|>"
     ],
     "stop_token_ids": [
-      2
+      151643,
+      151645
     ],
     "function_string": "",
     "use_function_calling": false

params_shard_1.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:02f9c200445b294d1a3240a016878b3d1c350531bbdaa54234ae59a28a423f7c
+oid sha256:d22ca55dfddcb1c72d72448e898be381acbdf7a98f91e83494deb9ea2f16532d
 size 32740608

params_shard_2.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:63c1bb2bd3e8a1c6438dd805053b8b88ead73242172bb9e58907571445e0d3ae
+oid sha256:220d64af94f86f942f406ce100ae8aafc23ae19d5a24b52b58956e7616bef8f1
 size 31866624

params_shard_3.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d8438278c7a6dce7501d24782435c464c2bfe76d630eea096e653346c6e3d47d
+oid sha256:365cfc4b4e943bc48e8221e0fa51109d33c1c5eb139bc5f65dd89ec58a8f9d3f
 size 33505024

params_shard_4.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a707ffcd1d7a4d79bc93f33575c0e84e0b1e30312f4a5c889c78721a6b7cec18
+oid sha256:81d0b7beee8280e60595b4088c1c954857b43f081d408aad98b2725d87c4f986
 size 32000000

params_shard_5.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8b7027bc8f618ed4092cc4451e7e2ee6117021fc5a29617a5e94ebd473b900ca
+oid sha256:326d77ac87c7cb2f1c04be82c4f9f31fffecb8acd826a64f00b93287476c8f27
 size 31866624

params_shard_6.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fb9fa45fa933d6534ffeecf7b7d8e628443731370a1b47feadeb61920ae3e705
+oid sha256:7fd482e2cea91a8d8b3ba0734afe1e8adb95488fe02921fa59bc6f11b87998e7
 size 33505024

params_shard_7.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:af54a139bca844b549219f7d22672b0ee7f7b6ead37cd628bc68ef6fd0081fc1
+oid sha256:5008178aa4d503208cbdedcdf9a2fc4ad7486b9c64ea88b33cf0164b988c85a5
 size 32000000

params_shard_8.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8ab0c526c003d12d3eea6f65383c486baaeef145c8c921da69004277de85e119
+oid sha256:773c7a6edcb21dd22f8b13842c9ff99e688000f80d8c2ae5b531764821213d97
 size 30097152

tensor-cache.json

@@ -371,7 +371,7 @@
                     "byteOffset": 32740352
                 }
             ],
-            "md5sum": "2a5c35e098e4b349642351a85c1159c4"
+            "md5sum": "37a319196de8632e0350c71e79af9830"
         },
         {
             "dataPath": "params_shard_2.bin",
@@ -849,7 +849,7 @@
                     "byteOffset": 31670016
                 }
             ],
-            "md5sum": "0196d572e69c4eeacc5c278701ac9055"
+            "md5sum": "5f8af7ca75856ddb27973c72dd0c6beb"
         },
         {
             "dataPath": "params_shard_3.bin",
@@ -1316,7 +1316,7 @@
                     "byteOffset": 32456448
                 }
             ],
-            "md5sum": "2718bb150cfe7427bd7e1ebc0d7f271f"
+            "md5sum": "d25b760cebf3448975dd0e4180068fed"
         },
         {
             "dataPath": "params_shard_4.bin",
@@ -1803,7 +1803,7 @@
                     "byteOffset": 31999744
                 }
             ],
-            "md5sum": "bca8507cfb9c4237cb29eeb0e8312751"
+            "md5sum": "e9822aee8fd1a54f134a2f3020ce82cb"
         },
         {
             "dataPath": "params_shard_5.bin",
@@ -2281,7 +2281,7 @@
                     "byteOffset": 31670016
                 }
             ],
-            "md5sum": "20b3f87dd71e1fc43784a711e7abc581"
+            "md5sum": "64e80d7e07771a6ab88eed4dcb3cd7dd"
         },
         {
             "dataPath": "params_shard_6.bin",
@@ -2748,7 +2748,7 @@
                     "byteOffset": 32456448
                 }
             ],
-            "md5sum": "cdc024bc8641803155333f54fbaa57ec"
+            "md5sum": "abe158fd2db8670885a749a747f41d1b"
         },
         {
             "dataPath": "params_shard_7.bin",
@@ -3235,7 +3235,7 @@
                     "byteOffset": 31999744
                 }
             ],
-            "md5sum": "6e4cb1503a258515bdacce4e348848ba"
+            "md5sum": "4b6290fd58a55ade50174f3d5d1e2ca6"
         },
         {
             "dataPath": "params_shard_8.bin",
@@ -3691,7 +3691,7 @@
                     "byteOffset": 30095104
                 }
             ],
-            "md5sum": "8aaeaeef9928133cc53701d3af235f33"
+            "md5sum": "53cd04b508f80194e315ad451673d67a"
         }
     ]
 }

tokenizer_config.json

@@ -231,8 +231,9 @@
   "eos_token": "<|im_end|>",
   "errors": "replace",
   "extra_special_tokens": {},
-  "model_max_length": 131072,
-  "pad_token": "<|endoftext|>",
+  "model_max_length": 40960,
+  "pad_token": "<|vision_pad|>",
+  "padding_side": "left",
   "split_special_tokens": false,
   "tokenizer_class": "Qwen2Tokenizer",
   "unk_token": null