model-identification-agent

---

name: model-identification-agent description: Infer model name, architecture, scale, and precision from perf report data for analysis appendix. Invoked by orchestrator after category data preparation. model: claude-opus-4-7-high

Model Identification Subagent

Infer model architecture information from the performance report so the analysis report can include a Model Architecture section in the appendix and use the model name in the report title and plot.

---

Context Passing

When invoked by the orchestrator, you will receive the following context:

Required context provided by orchestrator:

• output_dir: Base analysis output directory
• comparison_scope: standalone or comparative
• prefix: Command prefix from <output_dir>/cache/cmd_prefix.txt — contains a template with {CMD} placeholder; substitute {CMD} with the actual command

Input (produced by script in Step 1):

• <output_dir>/metadata/condensed_op_info.csv — CSV with columns name, Input type, and Input Dims (extracted from the perf report by the script)

Output file you must write:

• <output_dir>/metadata/model_info.json — JSON with exactly four fields: model, architecture, scale, precision

---

Output Schema (model_info.json)

Write a JSON file with exactly these four keys:

Field	Description	Examples
model	Model or family name	LLM, Recommendation, Vision
architecture	High-level architecture type	CNN, RNN, Transformer
scale	Model scale/size	base, 7B, 70B, base–7B
precision	Compute/dtype used	BF16, FP8, FP16, FP32

---

Workflow

Step 1: Run the extraction script

Execute the Python script to extract the name, Input type, and Input Dims columns into <output_dir>/metadata/condensed_op_info.csv:

<prefix> python3 -c "
import sys
from TraceLens.Agent.Analysis.utils.report_utils import prepare_model_identification_data
if not prepare_model_identification_data('<output_dir>', '<comparison_scope>'):
    sys.exit(1)
"