diff --git a/.devops/full-cuda.Dockerfile b/.devops/full-cuda.Dockerfile
index 360602d6567b8..77a9ddc145d0b 100644
--- a/.devops/full-cuda.Dockerfile
+++ b/.devops/full-cuda.Dockerfile
@@ -14,7 +14,8 @@ ARG CUDA_DOCKER_ARCH=all
 RUN apt-get update && \
     apt-get install -y build-essential python3 python3-pip git
 
-COPY requirements.txt requirements.txt
+COPY requirements.txt   requirements.txt
+COPY requirements       requirements
 
 RUN pip install --upgrade pip setuptools wheel \
     && pip install -r requirements.txt
diff --git a/.devops/full-rocm.Dockerfile b/.devops/full-rocm.Dockerfile
index 6c521e9b4101f..8b9633dc4ebf5 100644
--- a/.devops/full-rocm.Dockerfile
+++ b/.devops/full-rocm.Dockerfile
@@ -23,7 +23,8 @@ ARG ROCM_DOCKER_ARCH=\
     gfx1101 \
     gfx1102
 
-COPY requirements.txt requirements.txt
+COPY requirements.txt   requirements.txt
+COPY requirements       requirements
 
 RUN pip install --upgrade pip setuptools wheel \
     && pip install -r requirements.txt
diff --git a/.devops/full.Dockerfile b/.devops/full.Dockerfile
index 687628b35e996..cef1297d3e156 100644
--- a/.devops/full.Dockerfile
+++ b/.devops/full.Dockerfile
@@ -5,7 +5,8 @@ FROM ubuntu:$UBUNTU_VERSION as build
 RUN apt-get update && \
     apt-get install -y build-essential python3 python3-pip git
 
-COPY requirements.txt requirements.txt
+COPY requirements.txt   requirements.txt
+COPY requirements       requirements
 
 RUN pip install --upgrade pip setuptools wheel \
     && pip install -r requirements.txt
diff --git a/.devops/main-rocm.Dockerfile b/.devops/main-rocm.Dockerfile
index 789deff6dc8c1..0a706dc73227d 100644
--- a/.devops/main-rocm.Dockerfile
+++ b/.devops/main-rocm.Dockerfile
@@ -23,7 +23,8 @@ ARG ROCM_DOCKER_ARCH=\
     gfx1101 \
     gfx1102
 
-COPY requirements.txt requirements.txt
+COPY requirements.txt   requirements.txt
+COPY requirements       requirements
 
 RUN pip install --upgrade pip setuptools wheel \
     && pip install -r requirements.txt
diff --git a/.github/workflows/python-check-requirements.yml b/.github/workflows/python-check-requirements.yml
new file mode 100644
index 0000000000000..92e1108b3af88
--- /dev/null
+++ b/.github/workflows/python-check-requirements.yml
@@ -0,0 +1,29 @@
+name: Python check requirements.txt
+
+on:
+  push:
+    paths:
+      - 'scripts/check-requirements.sh'
+      - 'convert*.py'
+      - 'requirements.txt'
+      - 'requirements/*.txt'
+  pull_request:
+    paths:
+      - 'scripts/check-requirements.sh'
+      - 'convert*.py'
+      - 'requirements.txt'
+      - 'requirements/*.txt'
+
+jobs:
+  python-check-requirements:
+    runs-on: ubuntu-latest
+    name: check-requirements
+    steps:
+      - name: Check out source repository
+        uses: actions/checkout@v3
+      - name: Set up Python environment
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.11"
+      - name: Run check-requirements.sh script
+        run:  bash scripts/check-requirements.sh nocleanup
diff --git a/convert-hf-to-gguf.py b/convert-hf-to-gguf.py
index 3557a825eb357..51724c0dfca56 100755
--- a/convert-hf-to-gguf.py
+++ b/convert-hf-to-gguf.py
@@ -242,7 +242,7 @@ def _set_vocab_gpt2(self):
         tokens: list[bytearray] = []
         toktypes: list[int] = []
 
-        from transformers import AutoTokenizer  # type: ignore[attr-defined]
+        from transformers import AutoTokenizer
         tokenizer = AutoTokenizer.from_pretrained(dir_model)
         vocab_size = hparams.get("vocab_size", len(tokenizer.vocab))
         assert max(tokenizer.vocab.values()) < vocab_size
@@ -856,7 +856,7 @@ def set_gguf_parameters(self):
         hparams = self.hparams
         block_count = hparams["num_hidden_layers"]
 
-        self.gguf_writer.add_name(dir_model.name)
+        self.gguf_writer.add_name(self.dir_model.name)
         self.gguf_writer.add_context_length(hparams["max_position_embeddings"])
         self.gguf_writer.add_embedding_length(hparams["hidden_size"])
         self.gguf_writer.add_block_count(block_count)
@@ -902,7 +902,7 @@ def set_vocab(self):
         tokens: list[bytearray] = []
         toktypes: list[int] = []
 
-        from transformers import AutoTokenizer  # type: ignore[attr-defined]
+        from transformers import AutoTokenizer
         tokenizer = AutoTokenizer.from_pretrained(dir_model, trust_remote_code=True)
         vocab_size = hparams["vocab_size"]
         assert max(tokenizer.get_vocab().values()) < vocab_size
@@ -1185,57 +1185,62 @@ def parse_args() -> argparse.Namespace:
     return parser.parse_args()
 
 
-args = parse_args()
+def main() -> None:
+    args = parse_args()
 
-dir_model = args.model
+    dir_model = args.model
 
-if args.awq_path:
-    sys.path.insert(1, str(Path(__file__).parent / 'awq-py'))
-    from awq.apply_awq import add_scale_weights
-    tmp_model_path = args.model / "weighted_model"
-    dir_model = tmp_model_path
-    if tmp_model_path.is_dir():
-        print(f"{tmp_model_path} exists as a weighted model.")
+    if args.awq_path:
+        sys.path.insert(1, str(Path(__file__).parent / 'awq-py'))
+        from awq.apply_awq import add_scale_weights
+        tmp_model_path = args.model / "weighted_model"
+        dir_model = tmp_model_path
+        if tmp_model_path.is_dir():
+            print(f"{tmp_model_path} exists as a weighted model.")
+        else:
+            tmp_model_path.mkdir(parents=True, exist_ok=True)
+            print("Saving new weighted model ...")
+            add_scale_weights(str(args.model), str(args.awq_path), str(tmp_model_path))
+            print(f"Saved weighted model at {tmp_model_path}.")
+
+    if not dir_model.is_dir():
+        print(f'Error: {args.model} is not a directory', file=sys.stderr)
+        sys.exit(1)
+
+    ftype_map = {
+        "f32": gguf.GGMLQuantizationType.F32,
+        "f16": gguf.GGMLQuantizationType.F16,
+    }
+
+    if args.outfile is not None:
+        fname_out = args.outfile
     else:
-        tmp_model_path.mkdir(parents=True, exist_ok=True)
-        print("Saving new weighted model ...")
-        add_scale_weights(str(args.model), str(args.awq_path), str(tmp_model_path))
-        print(f"Saved weighted model at {tmp_model_path}.")
-
-if not dir_model.is_dir():
-    print(f'Error: {args.model} is not a directory', file=sys.stderr)
-    sys.exit(1)
+        # output in the same directory as the model by default
+        fname_out = dir_model / f'ggml-model-{args.outtype}.gguf'
 
-ftype_map = {
-    "f32": gguf.GGMLQuantizationType.F32,
-    "f16": gguf.GGMLQuantizationType.F16,
-}
+    print(f"Loading model: {dir_model.name}")
 
-if args.outfile is not None:
-    fname_out = args.outfile
-else:
-    # output in the same directory as the model by default
-    fname_out = dir_model / f'ggml-model-{args.outtype}.gguf'
+    hparams = Model.load_hparams(dir_model)
 
-print(f"Loading model: {dir_model.name}")
+    with torch.inference_mode():
+        model_class = Model.from_model_architecture(hparams["architectures"][0])
+        model_instance = model_class(dir_model, ftype_map[args.outtype], fname_out, args.bigendian)
 
-hparams = Model.load_hparams(dir_model)
+        print("Set model parameters")
+        model_instance.set_gguf_parameters()
 
-with torch.inference_mode():
-    model_class = Model.from_model_architecture(hparams["architectures"][0])
-    model_instance = model_class(dir_model, ftype_map[args.outtype], fname_out, args.bigendian)
+        print("Set model tokenizer")
+        model_instance.set_vocab()
 
-    print("Set model parameters")
-    model_instance.set_gguf_parameters()
+        if args.vocab_only:
+            print(f"Exporting model vocab to '{fname_out}'")
+            model_instance.write_vocab()
+        else:
+            print(f"Exporting model to '{fname_out}'")
+            model_instance.write()
 
-    print("Set model tokenizer")
-    model_instance.set_vocab()
+        print(f"Model successfully exported to '{fname_out}'")
 
-    if args.vocab_only:
-        print(f"Exporting model vocab to '{fname_out}'")
-        model_instance.write_vocab()
-    else:
-        print(f"Exporting model to '{fname_out}'")
-        model_instance.write()
 
-    print(f"Model successfully exported to '{fname_out}'")
+if __name__ == '__main__':
+    main()
diff --git a/convert-lora-to-ggml.py b/convert-lora-to-ggml.py
index 53bb8a3d97a05..35ce152f4248d 100755
--- a/convert-lora-to-ggml.py
+++ b/convert-lora-to-ggml.py
@@ -47,95 +47,96 @@ def write_tensor_header(fout: BinaryIO, name: str, shape: Sequence[int], data_ty
     fout.seek((fout.tell() + 31) & -32)
 
 
-if len(sys.argv) < 2:
-    print(f"Usage: python {sys.argv[0]} <path> [arch]")
-    print(
-        "Path must contain HuggingFace PEFT LoRA files 'adapter_config.json' and 'adapter_model.bin'"
-    )
-    print(f"Arch must be one of {list(gguf.MODEL_ARCH_NAMES.values())} (default: llama)")
-    sys.exit(1)
-
-input_json = os.path.join(sys.argv[1], "adapter_config.json")
-input_model = os.path.join(sys.argv[1], "adapter_model.bin")
-output_path = os.path.join(sys.argv[1], "ggml-adapter-model.bin")
-
-model = torch.load(input_model, map_location="cpu")
-arch_name = sys.argv[2] if len(sys.argv) == 3 else "llama"
-
-if arch_name not in gguf.MODEL_ARCH_NAMES.values():
-    print(f"Error: unsupported architecture {arch_name}")
-    sys.exit(1)
-
-arch = list(gguf.MODEL_ARCH_NAMES.keys())[list(gguf.MODEL_ARCH_NAMES.values()).index(arch_name)]
-name_map = gguf.TensorNameMap(arch, 200) # 200 layers ought to be enough for anyone
-
-with open(input_json, "r") as f:
-    params = json.load(f)
-
-if params["peft_type"] != "LORA":
-    print(f"Error: unsupported adapter type {params['peft_type']}, expected LORA")
-    sys.exit(1)
-
-if params["fan_in_fan_out"] is True:
-    print("Error: param fan_in_fan_out is not supported")
-    sys.exit(1)
-
-if params["bias"] is not None and params["bias"] != "none":
-    print("Error: param bias is not supported")
-    sys.exit(1)
-
-# TODO: these seem to be layers that have been trained but without lora.
-# doesn't seem widely used but eventually should be supported
-if params["modules_to_save"] is not None and len(params["modules_to_save"]) > 0:
-    print("Error: param modules_to_save is not supported")
-    sys.exit(1)
-
-with open(output_path, "wb") as fout:
-    fout.truncate()
-
-    write_file_header(fout, params)
-    for k, v in model.items():
-        orig_k = k
-        if k.endswith(".default.weight"):
-            k = k.replace(".default.weight", ".weight")
-        if k in ["llama_proj.weight", "llama_proj.bias"]:
-            continue
-        if k.endswith("lora_A.weight"):
-            if v.dtype != torch.float16 and v.dtype != torch.float32:
+if __name__ == '__main__':
+    if len(sys.argv) < 2:
+        print(f"Usage: python {sys.argv[0]} <path> [arch]")
+        print(
+            "Path must contain HuggingFace PEFT LoRA files 'adapter_config.json' and 'adapter_model.bin'"
+        )
+        print(f"Arch must be one of {list(gguf.MODEL_ARCH_NAMES.values())} (default: llama)")
+        sys.exit(1)
+
+    input_json = os.path.join(sys.argv[1], "adapter_config.json")
+    input_model = os.path.join(sys.argv[1], "adapter_model.bin")
+    output_path = os.path.join(sys.argv[1], "ggml-adapter-model.bin")
+
+    model = torch.load(input_model, map_location="cpu")
+    arch_name = sys.argv[2] if len(sys.argv) == 3 else "llama"
+
+    if arch_name not in gguf.MODEL_ARCH_NAMES.values():
+        print(f"Error: unsupported architecture {arch_name}")
+        sys.exit(1)
+
+    arch = list(gguf.MODEL_ARCH_NAMES.keys())[list(gguf.MODEL_ARCH_NAMES.values()).index(arch_name)]
+    name_map = gguf.TensorNameMap(arch, 200) # 200 layers ought to be enough for anyone
+
+    with open(input_json, "r") as f:
+        params = json.load(f)
+
+    if params["peft_type"] != "LORA":
+        print(f"Error: unsupported adapter type {params['peft_type']}, expected LORA")
+        sys.exit(1)
+
+    if params["fan_in_fan_out"] is True:
+        print("Error: param fan_in_fan_out is not supported")
+        sys.exit(1)
+
+    if params["bias"] is not None and params["bias"] != "none":
+        print("Error: param bias is not supported")
+        sys.exit(1)
+
+    # TODO: these seem to be layers that have been trained but without lora.
+    # doesn't seem widely used but eventually should be supported
+    if params["modules_to_save"] is not None and len(params["modules_to_save"]) > 0:
+        print("Error: param modules_to_save is not supported")
+        sys.exit(1)
+
+    with open(output_path, "wb") as fout:
+        fout.truncate()
+
+        write_file_header(fout, params)
+        for k, v in model.items():
+            orig_k = k
+            if k.endswith(".default.weight"):
+                k = k.replace(".default.weight", ".weight")
+            if k in ["llama_proj.weight", "llama_proj.bias"]:
+                continue
+            if k.endswith("lora_A.weight"):
+                if v.dtype != torch.float16 and v.dtype != torch.float32:
+                    v = v.float()
+                v = v.T
+            else:
                 v = v.float()
-            v = v.T
-        else:
-            v = v.float()
-
-        t = v.detach().numpy()
-
-        prefix = "base_model.model."
-        if k.startswith(prefix):
-            k = k[len(prefix) :]
-
-        lora_suffixes = (".lora_A.weight", ".lora_B.weight")
-        if k.endswith(lora_suffixes):
-            suffix = k[-len(lora_suffixes[0]):]
-            k = k[: -len(lora_suffixes[0])]
-        else:
-            print(f"Error: unrecognized tensor name {orig_k}")
-            sys.exit(1)
-
-        tname = name_map.get_name(k)
-        if tname is None:
-            print(f"Error: could not map tensor name {orig_k}")
-            print(" Note: the arch parameter must be specified if the model is not llama")
-            sys.exit(1)
-
-        if suffix == ".lora_A.weight":
-            tname += ".weight.loraA"
-        elif suffix == ".lora_B.weight":
-            tname += ".weight.loraB"
-        else:
-            assert False
-
-        print(f"{k} => {tname} {t.shape} {t.dtype} {t.nbytes/1024/1024:.2f}MB")
-        write_tensor_header(fout, tname, t.shape, t.dtype)
-        t.tofile(fout)
-
-print(f"Converted {input_json} and {input_model} to {output_path}")
+
+            t = v.detach().numpy()
+
+            prefix = "base_model.model."
+            if k.startswith(prefix):
+                k = k[len(prefix) :]
+
+            lora_suffixes = (".lora_A.weight", ".lora_B.weight")
+            if k.endswith(lora_suffixes):
+                suffix = k[-len(lora_suffixes[0]):]
+                k = k[: -len(lora_suffixes[0])]
+            else:
+                print(f"Error: unrecognized tensor name {orig_k}")
+                sys.exit(1)
+
+            tname = name_map.get_name(k)
+            if tname is None:
+                print(f"Error: could not map tensor name {orig_k}")
+                print(" Note: the arch parameter must be specified if the model is not llama")
+                sys.exit(1)
+
+            if suffix == ".lora_A.weight":
+                tname += ".weight.loraA"
+            elif suffix == ".lora_B.weight":
+                tname += ".weight.loraB"
+            else:
+                assert False
+
+            print(f"{k} => {tname} {t.shape} {t.dtype} {t.nbytes/1024/1024:.2f}MB")
+            write_tensor_header(fout, tname, t.shape, t.dtype)
+            t.tofile(fout)
+
+    print(f"Converted {input_json} and {input_model} to {output_path}")
diff --git a/convert-persimmon-to-gguf.py b/convert-persimmon-to-gguf.py
old mode 100644
new mode 100755
index 206b7d5ff9e31..1ba5864dc25ec
--- a/convert-persimmon-to-gguf.py
+++ b/convert-persimmon-to-gguf.py
@@ -1,3 +1,4 @@
+#!/usr/bin/env python3
 import torch
 import os
 from pprint import pprint
diff --git a/requirements-hf-to-gguf.txt b/requirements-hf-to-gguf.txt
deleted file mode 100644
index f4600539e27ac..0000000000000
--- a/requirements-hf-to-gguf.txt
+++ /dev/null
@@ -1,3 +0,0 @@
--r requirements.txt
-torch==2.1.1
-transformers==4.35.2
diff --git a/requirements.txt b/requirements.txt
index 1a116256671e5..d36f745201ef0 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,5 +1,12 @@
-numpy==1.24.4
-sentencepiece==0.1.98
-transformers>=4.34.0
-gguf>=0.1.0
-protobuf>=4.21.0
+# These requirements include all dependencies for all top-level python scripts
+# for llama.cpp. Avoid adding packages here directly.
+#
+# Package versions must stay compatible across all top-level python scripts.
+#
+
+-r ./requirements/requirements-convert.txt
+
+-r ./requirements/requirements-convert-hf-to-gguf.txt
+-r ./requirements/requirements-convert-llama-ggml-to-gguf.txt
+-r ./requirements/requirements-convert-lora-to-ggml.txt
+-r ./requirements/requirements-convert-persimmon-to-gguf.txt
diff --git a/requirements/requirements-convert-hf-to-gguf.txt b/requirements/requirements-convert-hf-to-gguf.txt
new file mode 100644
index 0000000000000..6ac4026107fbe
--- /dev/null
+++ b/requirements/requirements-convert-hf-to-gguf.txt
@@ -0,0 +1,2 @@
+-r ./requirements-convert.txt
+torch~=2.1.1
diff --git a/requirements/requirements-convert-llama-ggml-to-gguf.txt b/requirements/requirements-convert-llama-ggml-to-gguf.txt
new file mode 100644
index 0000000000000..a0f37cd1c71e4
--- /dev/null
+++ b/requirements/requirements-convert-llama-ggml-to-gguf.txt
@@ -0,0 +1 @@
+-r ./requirements-convert.txt
diff --git a/requirements/requirements-convert-lora-to-ggml.txt b/requirements/requirements-convert-lora-to-ggml.txt
new file mode 100644
index 0000000000000..6ac4026107fbe
--- /dev/null
+++ b/requirements/requirements-convert-lora-to-ggml.txt
@@ -0,0 +1,2 @@
+-r ./requirements-convert.txt
+torch~=2.1.1
diff --git a/requirements/requirements-convert-persimmon-to-gguf.txt b/requirements/requirements-convert-persimmon-to-gguf.txt
new file mode 100644
index 0000000000000..6ac4026107fbe
--- /dev/null
+++ b/requirements/requirements-convert-persimmon-to-gguf.txt
@@ -0,0 +1,2 @@
+-r ./requirements-convert.txt
+torch~=2.1.1
diff --git a/requirements/requirements-convert.txt b/requirements/requirements-convert.txt
new file mode 100644
index 0000000000000..a3d6ecec0ac04
--- /dev/null
+++ b/requirements/requirements-convert.txt
@@ -0,0 +1,5 @@
+numpy~=1.24.4
+sentencepiece~=0.1.98
+transformers>=4.35.2,<5.0.0
+gguf>=0.1.0
+protobuf>=4.21.0,<5.0.0
diff --git a/scripts/check-requirements.sh b/scripts/check-requirements.sh
new file mode 100755
index 0000000000000..af7bab7533789
--- /dev/null
+++ b/scripts/check-requirements.sh
@@ -0,0 +1,174 @@
+#!/bin/bash
+set -euo pipefail
+
+#
+# check-requirements.sh checks all requirements files for each top-level
+# convert*.py script.
+#
+# WARNING: This is quite IO intensive, because a fresh venv is set up for every
+# python script. As of 2023-12-22, this writes ~2.7GB of data. An adequately
+# sized tmpfs /tmp or ramdisk is recommended if running this frequently.
+#
+# usage:    check-requirements.sh [<working_dir>]
+#           check-requirements.sh nocleanup [<working_dir>]
+#
+# where:
+#           - <working_dir> is a directory that can be used as the base for
+#               setting up the venvs. Defaults to `/tmp`.
+#           - 'nocleanup' as the first argument will disable automatic cleanup
+#               of the files created by this script.
+#
+# requires:
+#           - bash >= 3.2.57
+#           - shellcheck
+#
+# For each script, it creates a fresh venv, `pip install`s the requirements, and
+# finally imports the python script to check for `ImportError`.
+#
+
+log() {
+    local level=$1 msg=$2
+    printf >&2 '%s: %s\n' "$level" "$msg"
+}
+
+debug() {
+    log DEBUG "$@"
+}
+
+info() {
+    log INFO "$@"
+}
+
+fatal() {
+    log FATAL "$@"
+    exit 1
+}
+
+cleanup() {
+    if [[ -n ${workdir+x} && -d $workdir && -w $workdir ]]; then
+        info "Removing $workdir"
+        local count=0
+        rm -rfv -- "$workdir" | while read -r; do
+            if (( count++ > 750 )); then
+                printf .
+                count=0
+            fi
+        done
+        printf '\n'
+        info "Removed $workdir"
+    fi
+}
+
+do_cleanup=1
+if [[ ${1-} == nocleanup ]]; then
+    do_cleanup=0; shift
+fi
+
+if (( do_cleanup )); then
+    trap exit INT TERM
+    trap cleanup EXIT
+fi
+
+this=$(realpath -- "$0"); readonly this
+cd "$(dirname "$this")/.." # PWD should stay in llama.cpp project directory
+
+shellcheck "$this"
+
+readonly reqs_dir=requirements
+
+if [[ ${1+x} ]]; then
+    tmp_dir=$(realpath -- "$1")
+    if [[ ! ( -d $tmp_dir && -w $tmp_dir ) ]]; then
+        fatal "$tmp_dir is not a writable directory"
+    fi
+else
+    tmp_dir=/tmp
+fi
+
+workdir=$(mktemp -d "$tmp_dir/check-requirements.XXXX"); readonly workdir
+info "Working directory: $workdir"
+
+check_requirements() {
+    local reqs=$1
+
+    info "$reqs: beginning check"
+    pip --disable-pip-version-check install -qr "$reqs"
+    info "$reqs: OK"
+}
+
+check_convert_script() {
+    local py=$1             # e.g. ./convert-hf-to-gguf.py
+    local pyname=${py##*/}  # e.g. convert-hf-to-gguf.py
+    pyname=${pyname%.py}    # e.g. convert-hf-to-gguf
+
+    info "$py: beginning check"
+
+    local reqs="$reqs_dir/requirements-$pyname.txt"
+    if [[ ! -r $reqs ]]; then
+        fatal "$py missing requirements. Expected: $reqs"
+    fi
+
+    local venv="$workdir/$pyname-venv"
+    python3 -m venv "$venv"
+
+    (
+        # shellcheck source=/dev/null
+        source "$venv/bin/activate"
+
+        check_requirements "$reqs"
+
+        python - "$py" "$pyname" <<'EOF'
+import sys
+from importlib.machinery import SourceFileLoader
+py, pyname = sys.argv[1:]
+SourceFileLoader(pyname, py).load_module()
+EOF
+    )
+
+    if (( do_cleanup )); then
+        rm -rf -- "$venv"
+    fi
+
+    info "$py: imports OK"
+}
+
+readonly ignore_eq_eq='check_requirements: ignore "=="'
+
+for req in "$reqs_dir"/*; do
+    # Check that all sub-requirements are added to top-level requirements.txt
+    if ! grep -qF "$req" requirements.txt; then
+        fatal "$req needs to be added to requirements.txt"
+    fi
+
+    # Make sure exact release versions aren't being pinned in the requirements
+    # Filters out the ignore string
+    if grep -vF "$ignore_eq_eq" "$req" | grep -q '=='; then
+        tab=$'\t'
+        cat >&2 <<EOF
+FATAL: Avoid pinning exact package versions. Use '~=' instead.
+You can suppress this error by appending the following to the line:
+$tab# $ignore_eq_eq
+EOF
+        exit 1
+    fi
+done
+
+all_venv="$workdir/all-venv"
+python3 -m venv "$all_venv"
+
+(
+    # shellcheck source=/dev/null
+    source "$all_venv/bin/activate"
+    check_requirements requirements.txt
+)
+
+if (( do_cleanup )); then
+    rm -rf -- "$all_venv"
+fi
+
+check_convert_script convert.py
+for py in convert-*.py; do
+    check_convert_script "$py"
+done
+
+info 'Done! No issues found.'