Upload folder using huggingface_hub (#2)

- 31c97493cc1cb4325f363536e90a21abc3a030eb8da0a32fa2b0a12128b7e1c9 (11f39651fc747a15da2ca5d986db31c264a1e265)
- c3536383a0e138b1fbd6d10a547cc59b9f165adc34f356bff6ce706f81aa73b3 (df82927d3fe64a3eb0bce692118ba605b127e467)

config.json

@@ -1,6 +1,6 @@
 {
   "_attn_implementation_autoset": true,
-  "_name_or_path": "Open-Orca/Mistral-7B-OpenOrca",
+  "_name_or_path": "/tmp/models/tmp80dy7532/tmpuu76zfi8",
   "architectures": [
     "MistralForCausalLM"
   ],
@@ -17,6 +17,26 @@
   "num_attention_heads": 32,
   "num_hidden_layers": 32,
   "num_key_value_heads": 8,
+  "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-05,
   "rope_theta": 10000.0,
   "sliding_window": 4096,

qmodel.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6216910b5a070f317cd7badd706e8a84f2b7a0ffdbc4a9dd18319439cef8695
-size 4450966069
+oid sha256:9481ddc460753f7b5febeffe55c06aaddd5246860f18a7480a7126a670e9606d
+size 5149857497

smash_config.json

@@ -11,7 +11,7 @@
     "quant_hqq_weight_bits": 4,
     "max_batch_size": 1,
     "device": "cuda",
-    "cache_dir": "/tmp/models/tmpsn4v26kb",
+    "cache_dir": "/tmp/models/tmp80dy7532",
     "task": "",
     "save_load_fn": "hqq",
     "save_load_fn_args": {},