Upload folder using huggingface_hub

config.json

@@ -23,23 +23,6 @@
   "quantization_config": {
     "config_groups": {
       "group_0": {
-        "input_activations": null,
-        "output_activations": null,
-        "targets": [
-          "Embedding"
-        ],
-        "weights": {
-          "block_structure": null,
-          "group_size": null,
-          "num_bits": 8,
-          "observer": "minmax",
-          "observer_kwargs": {},
-          "strategy": "tensor",
-          "symmetric": true,
-          "type": "int"
-        }
-      },
-      "group_1": {
         "input_activations": {
           "block_structure": null,
           "group_size": null,
@@ -47,7 +30,7 @@
           "observer": "minmax",
           "observer_kwargs": {},
           "strategy": "tensor",
-          "symmetric": false,
+          "symmetric": true,
           "type": "int"
         },
         "output_activations": null,
@@ -67,7 +50,7 @@
       }
     },
     "format": "fakequant",
-    "global_compression_ratio": 1.4416825559554713,
+    "global_compression_ratio": 1.4375595368095078,
     "ignore": [
       "model.layers.0.mlp.down_proj",
       "lm_head"

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3f94cc1d9b92c63daae79a819a45fbe669bbf1afaba13435ab0f0de7aa47130f
-size 4400287663
+oid sha256:5aae4cc5021049424ebba130bff524467cf1575b37914eb22011e67fd9a5e531
+size 4400287458

recipe.yaml

@@ -5,14 +5,9 @@ test_stage:
       config_groups:
         group_0:
           weights: {num_bits: 8, type: int, symmetric: true, strategy: tensor}
-          input_activations: {num_bits: 8, type: int, symmetric: false, strategy: tensor}
+          input_activations: {num_bits: 8, type: int, symmetric: true, strategy: tensor}
           output_activations: null
           targets: [Linear]
-        group_1:
-          weights: {num_bits: 8, type: int, symmetric: true, strategy: tensor}
-          input_activations: null
-          output_activations: null
-          targets: [Embedding]
     SparseGPTModifier:
       sparsity: 0.0
       block_size: 128