Upload folder using huggingface_hub

generation_config.json

@@ -0,0 +1,14 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.05,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.45.1"
+}

recipe.yaml

@@ -0,0 +1,12 @@
+quant_stage:
+  quant_modifiers:
+    GPTQModifier:
+      sequential_update: true
+      dampening_frac: 0.1
+      ignore: [lm_head]
+      config_groups:
+        group_0:
+          weights: {num_bits: 4, type: int, symmetric: true, strategy: group, group_size: 128,
+            actorder: group}
+          targets: [Linear]
+          observer: mse