Upload folder using huggingface_hub (#5)

- a0bd9681fd6a02791b4197e85b5b221a7bed57dfa66d261808082b9009005f6a (8b31ce9061e559e2af8049ead6ce72998cd98d45)
- 72cc8f19f6f8d2751017cecb7d53e9ce7bbd8d1016ab28b079b1e577eec7c728 (964ebe306923f61e39e9bd9416f2a24e0f0ac3a2)

config.json

@@ -1,6 +1,6 @@
 {
   "_attn_implementation_autoset": true,
-  "_name_or_path": "google/gemma-2-2b-it",
+  "_name_or_path": "/tmp/models/tmplvpqq264/tmp7lo1_uef",
   "architectures": [
     "Gemma2ForCausalLM"
   ],
@@ -26,6 +26,26 @@
   "num_hidden_layers": 26,
   "num_key_value_heads": 4,
   "pad_token_id": 0,
+  "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"
+    ]
+  },
   "query_pre_attn_scalar": 256,
   "rms_norm_eps": 1e-06,
   "rope_theta": 10000.0,

qmodel.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e340e16e22c752bcadb364e206f53ae536c5a8e2c1f5fa935079d9b3f85cd3ca
-size 2318999562
+oid sha256:e0687ab70e1c10731f4dd5f4fe76677f9571cf72fd7d317b8201fe624c47de99
+size 3625639370

smash_config.json

@@ -11,7 +11,7 @@
     "quant_hqq_weight_bits": 4,
     "max_batch_size": 1,
     "device": "cuda",
-    "cache_dir": "/tmp/models/tmpv_g8jhwn",
+    "cache_dir": "/tmp/models/tmplvpqq264",
     "task": "",
     "save_load_fn": "hqq",
     "save_load_fn_args": {},