# Model Blob SHA-256 Hashes — Day 1 Sampling + Model-A/B Lab
# This lab uses BOTH vendored Qwen sizes so the A/B phase can compare them.
# Hashes recorded from the same registry manifests as the other Day-1..4 labs.
#
# These are content-addressed Ollama model blob hashes (ADR-6).
# If the SHA-256 matches after pulling, you have exactly the same model weights.
#
# Verify after pulling:
#   docker compose exec ollama ollama show qwen2.5:1.5b
#   docker compose exec ollama ollama show qwen2.5:3b
#   (manifests live at ~/.ollama/models/manifests/registry.ollama.ai/library/qwen2.5/...)
#
# Air-gapped verification: check that blob files exist in the volume:
#   docker compose exec ollama ls /root/.ollama/models/blobs/sha256-<first-64-chars>

# ── qwen2.5:1.5b (Model A — smaller/faster) ──────────────────────────────────
# Architecture: qwen2 | Parameters: 1.5B | Quantization: Q4_K_M
# Context length: 32768 | Embedding length: 1536 | License: Apache 2.0
#
# Main GGUF blob (986 MB, Q4_K_M quantized):
QWEN2_5_1_5B_BLOB_SHA256=183715c435899236895da3869489cc30ac241476b4971a20285b1a462818a5b4
#
# Config/manifest SHA-256:
QWEN2_5_1_5B_CONFIG_SHA256=377ac4d7aeefd5b870c9fccff9a6d4df36901d99fe3277c2f755bc401601ba1c

# ── qwen2.5:3b (Model B — larger/slower) ─────────────────────────────────────
# Architecture: qwen2 | Parameters: 3.1B | Quantization: Q4_K_M
# Context length: 32768 | Embedding length: 2048 | License: Apache 2.0
#
# Main GGUF blob (~1.93 GB, Q4_K_M quantized):
QWEN2_5_3B_BLOB_SHA256=5ee4f07cdb9beadbbb293e85803c569b01bd37ed059d2715faa7bb405f31caa6
#
# Config/manifest SHA-256 (from Ollama registry OCI manifest):
QWEN2_5_3B_CONFIG_SHA256=161ddde4c9cd07c9f1ccb4e0167c434bce72caeb3fc1844262fa66bc877b0426

# ── Total download size ───────────────────────────────────────────────────────
# qwen2.5:1.5b:  986 MB
# qwen2.5:3b:  1,930 MB
# Total (approximate):  2,916 MB (~2.9 GB)
#
# Ollama model registry: registry.ollama.ai (not Docker Hub)
# These hashes are for the Ollama model format (GGUF + metadata layers),
# not the Ollama Docker image digest.
