Upload folder using huggingface_hub (#5)

- 76a9f3229c0d1aa7d577eb3a45e8a2df9ad9fc14c060b645c91913c43be12490 (7854c0c16a77587f06d94514342041ed47a93c7c)
- f2accc381628497ef754aa39324aac57145c0574448f042e051887b91707563d (23988bca5ad1f87ff0e70e471afb3c9cdd23ebc5)

config.json

@@ -1,6 +1,6 @@
 {
   "_attn_implementation_autoset": true,
-  "_name_or_path": "HuggingFaceTB/SmolLM2-1.7B-Instruct",
+  "_name_or_path": "/tmp/models/tmpuszyy52l/tmpu5eb0hd8",
   "architectures": [
     "LlamaForCausalLM"
   ],
@@ -21,6 +21,26 @@
   "num_key_value_heads": 32,
   "pad_token_id": 2,
   "pretraining_tp": 1,
+  "quantization_config": {
+    "quant_config": {
+      "offload_meta": false,
+      "scale_quant_params": null,
+      "weight_quant_params": {
+        "axis": 1,
+        "channel_wise": true,
+        "group_size": 64,
+        "nbits": 8,
+        "optimize": true,
+        "round_zero": false,
+        "view_as_float": false
+      },
+      "zero_quant_params": null
+    },
+    "quant_method": "hqq",
+    "skip_modules": [
+      "lm_head"
+    ]
+  },
   "rms_norm_eps": 1e-05,
   "rope_scaling": null,
   "rope_theta": 130000,

qmodel.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:81cd9322efd0bf8fd087d644232db97b9a282b56ee59f97c7ddbd81951cee622
-size 1912983472
+oid sha256:d54f27b8eec5c7e905958321aee5898dae3a5223ee7775ca2fba42ffd20cd1a0
+size 2215166128

smash_config.json

@@ -11,7 +11,7 @@
     "quant_hqq_weight_bits": 8,
     "max_batch_size": 1,
     "device": "cuda",
-    "cache_dir": "/tmp/models/tmpcdqz_pr1",
+    "cache_dir": "/tmp/models/tmpuszyy52l",
     "task": "",
     "save_load_fn": "hqq",
     "save_load_fn_args": {},