feat(bid-script/gpu): support model.vram for pricing calculation and …

…AMD support in addition to NVIDIA models (#249)

* feat(bid-script/gpu): support model.vram for pricing calculation

fixes akash-network/support#148

* feat(bid-script/gpu): support AMD in addition to NVIDIA models

charts/akash-provider/scripts/price_script_generic.sh

@@ -2,7 +2,7 @@
 # WARNING: the runtime of this script should NOT exceed 5 seconds! (Perhaps can be amended via AKASH_BID_PRICE_SCRIPT_PROCESS_TIMEOUT env variable)
 # Requirements:
 # curl jq bc mawk ca-certificates
-# Version: November-16-2023
+# Version: February-27-2024
 set -o pipefail
 
 # Example:
@@ -160,9 +160,18 @@ fi
 gpu_price_total=0
 while IFS= read -r resource; do
   count=$(echo "$resource" | jq -r '.count')
-  model=$(echo "$resource" | jq -r '.gpu.attributes.vendor.nvidia.model // 0')
+  model=$(echo "$resource" | jq -r '.gpu.attributes.vendor | (.nvidia // .amd // empty).model // 0')
+  vram=$(echo "$resource" | jq -r --arg v_model "$model" '.gpu.attributes.vendor | (
+      .nvidia | select(.model == $v_model) //
+      .amd | select(.model == $v_model) //
+      empty
+  ).ram // 0')
   gpu_units=$(echo "$resource" | jq -r '.gpu.units // 0')
   # default to 100 USD/GPU per unit a month when PRICE_TARGET_GPU_MAPPINGS is not set
+  # price_target_gpu_mappings can specify <model.vram> or <model>. E.g. a100.40Gi=900,a100.80Gi=1000 or a100=950
+  if [[ "$vram" != "0" ]]; then
+    model="${model}.${vram}"
+  fi
   price="${gpu_mappings[''$model'']:-$gpu_unit_max_price}"
   ((gpu_price_total += count * gpu_units * price))