Upload folder using huggingface_hub (#3)

- dc400d8fd524722417bd05af7bdce15d61b24d164a6de0cd944057aa0ffa5aa1 (4898d22f2d967a4d4bd8c661b223d6a1bd77085e)
- 38ad0cf6f4c491e46b36139deb5e104cfd67b7349f9a8a6ccaafe8dd40612b27 (ed731c0ae3dc92f566f13cfd5adb19a20aa0da45)

config.json

@@ -1,6 +1,6 @@
 {
   "_attn_implementation_autoset": true,
-  "_name_or_path": "PowerInfer/SmallThinker-3B-Preview",
+  "_name_or_path": "/tmp/models/tmpvj9nwc0p/tmprnq9n9q5",
   "architectures": [
     "Qwen2ForCausalLM"
   ],
@@ -17,6 +17,26 @@
   "num_attention_heads": 16,
   "num_hidden_layers": 36,
   "num_key_value_heads": 2,
+  "quantization_config": {
+    "quant_config": {
+      "offload_meta": false,
+      "scale_quant_params": null,
+      "weight_quant_params": {
+        "axis": 1,
+        "channel_wise": true,
+        "group_size": 64,
+        "nbits": 4,
+        "optimize": true,
+        "round_zero": true,
+        "view_as_float": false
+      },
+      "zero_quant_params": null
+    },
+    "quant_method": "hqq",
+    "skip_modules": [
+      "lm_head"
+    ]
+  },
   "rms_norm_eps": 1e-06,
   "rope_scaling": null,
   "rope_theta": 1000000.0,

qmodel.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:352a84ab9ca7251d5e65363d875b902b6514b641aace0ad937a8aec6263a23f6
-size 2183789680
+oid sha256:123d4312b2da135c1b96183ca6f882ea2ab124574bc1433432c78668e0d9353d
+size 2979815984

smash_config.json

@@ -11,7 +11,7 @@
     "quant_hqq_weight_bits": 4,
     "max_batch_size": 1,
     "device": "cuda",
-    "cache_dir": "/tmp/models/tmpjnwwvoec",
+    "cache_dir": "/tmp/models/tmpvj9nwc0p",
     "task": "",
     "save_load_fn": "hqq",
     "save_load_fn_args": {},