Text Generation
Transformers
TensorBoard
Safetensors
gemma3_text
Generated from Trainer
sft
trl
conversational
text-generation-inference
Instructions to use satendrakumar/MyGemmaNPC with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use satendrakumar/MyGemmaNPC with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="satendrakumar/MyGemmaNPC") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("satendrakumar/MyGemmaNPC") model = AutoModelForCausalLM.from_pretrained("satendrakumar/MyGemmaNPC", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use satendrakumar/MyGemmaNPC with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "satendrakumar/MyGemmaNPC" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "satendrakumar/MyGemmaNPC", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/satendrakumar/MyGemmaNPC
- SGLang
How to use satendrakumar/MyGemmaNPC with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "satendrakumar/MyGemmaNPC" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "satendrakumar/MyGemmaNPC", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "satendrakumar/MyGemmaNPC" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "satendrakumar/MyGemmaNPC", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use satendrakumar/MyGemmaNPC with Docker Model Runner:
docker model run hf.co/satendrakumar/MyGemmaNPC
File size: 7,984 Bytes
e71db58 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 | #!/bin/bash
set -e
# Parse command line arguments
TARGET="$1" # Optional target parameter
# Validate target if provided
if [[ -n "$TARGET" ]] && [[ ! "$TARGET" =~ ^(stable|latest|[0-9]+\.[0-9]+\.[0-9]+(-[^[:space:]]+)?)$ ]]; then
echo "Usage: $0 [stable|latest|VERSION]" >&2
exit 1
fi
# Refuse to run under sudo from a regular user's shell. This installer puts
# everything under $HOME, which under sudo typically resolves to root's home:
# the binary lands in /root/.local/bin (or is left root-owned in the user's
# home, depending on the distro's sudo configuration), and the 'claude'
# command is then not found in the user's own shell. Plain root with no sudo
# (containers, CI, root-only systems) is unaffected by this check.
if [ "$(id -u)" -eq 0 ] && [ -n "${SUDO_USER:-}" ] && [ "$SUDO_USER" != "root" ] && [ -z "${CLAUDE_INSTALL_ALLOW_SUDO:-}" ]; then
echo "Error: do not run this installer with sudo." >&2
echo "" >&2
echo "Claude Code installs into your home directory and does not need root access." >&2
echo "With sudo, the installation would go into root's home directory instead of" >&2
echo "yours, and the 'claude' command would not work from your own shell." >&2
echo "" >&2
echo "Please re-run the same command without sudo, e.g.:" >&2
# pinned-dep-allow: display-only guidance text in an error message, not an executed install; install.sh is Anthropic's own installer
echo " curl -fsSL https://claude.ai/install.sh | bash" >&2
echo "" >&2
echo "To intentionally install Claude Code for the root user, re-run with" >&2
echo "CLAUDE_INSTALL_ALLOW_SUDO=1 set in the installer's environment, e.g.:" >&2
# pinned-dep-allow: display-only guidance text in an error message, not an executed install; install.sh is Anthropic's own installer
echo " curl -fsSL https://claude.ai/install.sh | sudo CLAUDE_INSTALL_ALLOW_SUDO=1 bash" >&2
exit 1
fi
DOWNLOAD_BASE_URL="https://downloads.claude.ai/claude-code-releases"
DOWNLOAD_DIR="$HOME/.claude/downloads"
# Check for required dependencies
DOWNLOADER=""
if command -v curl >/dev/null 2>&1; then
DOWNLOADER="curl"
elif command -v wget >/dev/null 2>&1; then
DOWNLOADER="wget"
else
echo "Either curl or wget is required but neither is installed" >&2
exit 1
fi
# Check if jq is available (optional)
HAS_JQ=false
if command -v jq >/dev/null 2>&1; then
HAS_JQ=true
fi
# Download function that works with both curl and wget
download_file() {
local url="$1"
local output="$2"
if [ "$DOWNLOADER" = "curl" ]; then
if [ -n "$output" ]; then
curl -fsSL -o "$output" "$url"
else
curl -fsSL "$url"
fi
elif [ "$DOWNLOADER" = "wget" ]; then
if [ -n "$output" ]; then
wget -q -O "$output" "$url"
else
wget -q -O - "$url"
fi
else
return 1
fi
}
# Simple JSON parser for extracting checksum when jq is not available
get_checksum_from_manifest() {
local json="$1"
local platform="$2"
# Normalize JSON to single line and extract checksum
json=$(echo "$json" | tr -d '\n\r\t' | sed 's/ \+/ /g')
# Extract checksum for platform using bash regex
if [[ $json =~ \"$platform\"[^}]*\"checksum\"[[:space:]]*:[[:space:]]*\"([a-f0-9]{64})\" ]]; then
echo "${BASH_REMATCH[1]}"
return 0
fi
return 1
}
# Detect platform
case "$(uname -s)" in
Darwin) os="darwin" ;;
Linux) os="linux" ;;
MINGW*|MSYS*|CYGWIN*) echo "Windows is not supported by this script. See https://code.claude.com/docs for installation options." >&2; exit 1 ;;
*) echo "Unsupported operating system: $(uname -s). See https://code.claude.com/docs for supported platforms." >&2; exit 1 ;;
esac
case "$(uname -m)" in
x86_64|amd64) arch="x64" ;;
arm64|aarch64) arch="arm64" ;;
*) echo "Unsupported architecture: $(uname -m)" >&2; exit 1 ;;
esac
# Detect Rosetta 2 on macOS: if the shell is running as x64 under Rosetta on an ARM Mac,
# download the native arm64 binary instead of the x64 one
if [ "$os" = "darwin" ] && [ "$arch" = "x64" ]; then
if [ "$(sysctl -n sysctl.proc_translated 2>/dev/null)" = "1" ]; then
arch="arm64"
fi
fi
# Check for musl on Linux and adjust platform accordingly
if [ "$os" = "linux" ]; then
if [ -f /lib/libc.musl-x86_64.so.1 ] || [ -f /lib/libc.musl-aarch64.so.1 ] || ldd /bin/ls 2>&1 | grep -q musl; then
platform="linux-${arch}-musl"
else
platform="linux-${arch}"
fi
else
platform="${os}-${arch}"
fi
mkdir -p "$DOWNLOAD_DIR"
# Always download latest version (which has the most up-to-date installer)
version=$(download_file "$DOWNLOAD_BASE_URL/latest")
# Reject non-version content (e.g. an HTML error page) before it reaches the manifest URL
if [[ ! "$version" =~ ^[0-9]+\.[0-9]+\.[0-9]+ ]]; then
echo "Failed to get a valid version from downloads.claude.ai (got unexpected content)." >&2
echo "This can happen if the download service is unreachable or not available in your region - see https://www.anthropic.com/supported-countries" >&2
exit 1
fi
# Download manifest and extract checksum
manifest_json=$(download_file "$DOWNLOAD_BASE_URL/$version/manifest.json")
# Use jq if available, otherwise fall back to pure bash parsing
if [ "$HAS_JQ" = true ]; then
checksum=$(echo "$manifest_json" | jq -r ".platforms[\"$platform\"].checksum // empty")
else
checksum=$(get_checksum_from_manifest "$manifest_json" "$platform")
fi
# Validate checksum format (SHA256 = 64 hex characters)
if [ -z "$checksum" ] || [[ ! "$checksum" =~ ^[a-f0-9]{64}$ ]]; then
echo "Platform $platform not found in manifest" >&2
exit 1
fi
# Download and verify
binary_path="$DOWNLOAD_DIR/claude-$version-$platform"
if ! download_file "$DOWNLOAD_BASE_URL/$version/$platform/claude" "$binary_path"; then
echo "Download failed" >&2
rm -f "$binary_path"
exit 1
fi
# Pick the right checksum tool
if [ "$os" = "darwin" ]; then
actual=$(shasum -a 256 "$binary_path" | cut -d' ' -f1)
else
actual=$(sha256sum "$binary_path" | cut -d' ' -f1)
fi
if [ "$actual" != "$checksum" ]; then
echo "Checksum verification failed" >&2
rm -f "$binary_path"
exit 1
fi
chmod +x "$binary_path"
# Run claude install to set up launcher and shell integration
echo "Setting up Claude Code..."
install_code=0
"$binary_path" install ${TARGET:+"$TARGET"} || install_code=$?
# Clean up downloaded file
rm -f "$binary_path"
if [ "$install_code" -ne 0 ]; then
# A signal death mid-install kills the binary's TUI with no chance to
# restore the terminal, leaving the user's shell in raw mode (typed
# characters stop echoing). Restore it before printing anything.
if [ "$install_code" -ge 128 ] && [ -t 0 ]; then
stty sane 2>/dev/null || true
fi
# Red when stderr is a terminal, so the explanation stands out from the
# surrounding install output; plain when piped or captured
red="" reset=""
if [ -t 2 ]; then
red=$'\033[31m'
reset=$'\033[0m'
fi
# Signal deaths (exit code 128+N) print nothing of their own — bash shows
# only e.g. "Killed". 137 = SIGKILL, which on Linux is almost always the
# kernel OOM killer on small hosts; macOS has no equivalent OOM kill, so
# the out-of-memory explanation is Linux-only.
if [ "$install_code" -eq 137 ] && [ "$os" = "linux" ]; then
echo "${red}Installation was killed before it could finish (exit code 137). This usually means the system ran out of memory.${reset}" >&2
echo "${red}Claude Code needs roughly 512MB of free memory to install. Free up memory, then run this script again.${reset}" >&2
elif [ "$install_code" -ge 128 ]; then
echo "${red}Installation was killed before it could finish (exit code $install_code).${reset}" >&2
fi
exit "$install_code"
fi
echo ""
echo "✅ Installation complete!"
echo ""
|