Upload folder using huggingface_hub

config.json

@@ -1,6 +1,6 @@
 {
   "_attn_implementation_autoset": true,
-  "_name_or_path": "deepseek-ai/deepseek-coder-6.7b-instruct",
+  "_name_or_path": "/tmp/models/tmpkeb95yvm/tmpdikbvq1u",
   "architectures": [
     "LlamaForCausalLM"
   ],
@@ -20,6 +20,26 @@
   "num_hidden_layers": 32,
   "num_key_value_heads": 32,
   "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-06,
   "rope_scaling": {
     "factor": 4.0,

qmodel.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:59f908d3159c3d0473c0a0b6cb7e4b8a527ab7f391f72bfde65f1669170fde2d
-size 7410020113
+oid sha256:6be1e95f80fe53a8f2de4d3c07a98c35c66709d8057649bea216acd9b8c19310
+size 8079534477

smash_config.json

@@ -11,7 +11,7 @@
     "quant_hqq_weight_bits": 8,
     "max_batch_size": 1,
     "device": "cuda",
-    "cache_dir": "/tmp/models/tmpj7ftrlmy",
+    "cache_dir": "/tmp/models/tmpkeb95yvm",
     "task": "",
     "save_load_fn": "hqq",
     "save_load_fn_args": {},