sakthai-context-1.5b-merged-v2/.eval_results/cron-eval-sakthai-context-1.5b-merged-v2-2026-07-31-2.yaml

target_model:
  id: Nanthasit/sakthai-context-1.5b-merged-v2
  pipeline_tag: text-generation
  library_name: transformers
  base_model: Qwen/Qwen2.5-1.5B-Instruct
  downloads: 337
  likes: 0
  private: false
  gated: false
  created: '2026-07-30T10:46:30.000Z'
  last_modified: '2026-07-31T07:58:17.000Z'
  model_age_days: 0.9950
  model_type: llm
  has_weights: true

architecture:
  model_type: qwen2
  architectures:
  - Qwen2ForCausalLM
  hidden_size: 1536
  num_hidden_layers: 28
  num_attention_heads: 12
  num_key_value_heads: 2
  intermediate_size: 8960
  vocab_size: 151936
  max_position_embeddings: 32768
  rope_theta: 1000000.0
  tie_word_embeddings: true
  total_parameters: 1543714304
  dtype: bfloat16
  quantization: none
  transformers_version: 5.14.1
  arch_source: config.json (fetched live 2026-07-31)

repo_summary:
  siblings_count: 20
  total_repo_bytes: 3098928651
  total_gb: 2.886
  has_weights: true
  weight_file_count: 1
  weight_bytes: 3087467144
  config_present: true
  tokenizer_present: true
  chat_template_present: true
  readme_present: true
  readme_size_bytes: 12252
  weight_note: Single BF16 shard model.safetensors (2.875 GiB), clean monolith.
  eval_files_count: 12
  eval_files:
  - .eval_results/sakthai-bench-v2.yaml
  - .eval_results/benchmark-20260731_061627.yaml
  - .eval_results/cron-eval-sakthai-context-1.5b-merged-v2-2026-07-31-1.yaml
  - .eval_results/health-check-2026-07-30-cron.yaml
  - .eval_results/health-check-2026-07-30.yaml
  - .eval_results/health-check-context-1.5b-v2-2026-07-30-cron.yaml
  - .eval_results/health-check-sakthai-context-1.5b-merged-v2-2026-07-30-cron.yaml
  - .eval_results/health-check-sakthai-context-1.5b-merged-v2-2026-07-30.yaml
  - .eval_results/health-check.yaml
  - .eval_results/inference-check-2026-07-30.yaml
  - .eval_results/inference-check-20260731T001536Z.yaml

benchmarks:
  model_index_count: 1
  metrics_count: 3
  all_verified: true
  pending_metrics: 0
  entries:
  - dataset: Nanthasit/sakthai-bench-v2
    task: text-generation
    config: default
    split: test
    metrics:
    - name: Selection Accuracy
      value: 34.9
      verified: true
    - name: Arguments Accuracy
      value: 44.2
      verified: true
    - name: Strict Accuracy
      value: 34.2
      verified: true
    source: .eval_results/sakthai-bench-v2.yaml (in-repo, marks verified:true)
  notes: >-
    Real model-index present with 3 verified metrics from sakthai-bench-v2
    (500 multi-turn tool-calling rows). The repo's own .eval_results/sakthai-bench-v2.yaml
    marks all three metrics verified:true. Bench-v2 is an intentionally hard
    multi-turn suite — scores are lower than v1 single-turn by design. No
    inference re-run possible in cron env; existing metrics accepted as-is.

card_quality:
  license: apache-2.0
  base_model_documented: true
  base_model: Qwen/Qwen2.5-1.5B-Instruct
  tags_count: 16
  tags:
  - qwen2.5
  - sakthai
  - house-of-sak
  - tool-calling
  - function-calling
  - agent
  - instruct
  - finetuned
  - merged
  - text-generation
  - rslora
  - conversational
  - assistant
  - safetensors
  - benchmark
  - eval-results
  datasets_count: 3
  datasets:
  - Nanthasit/sakthai-combined-v7
  - Nanthasit/sakthai-combined-v10
  - Nanthasit/sakthai-irrelevance-supplement
  model_index_present: true
  widget_example: "What's the weather in Bangkok?"
  readme_size_bytes: 12252
  deductions:
  - README badges use img.shields.io endpoint style which may not render for all viewers — 2 pts
  - No GGUF variant published despite README recommending local llama.cpp use — 2 pts
  score: 96

health_score:
  overall: 65.8
  components:
    popularity: 3.37
    momentum: 33.9
    benchmarks: 95
    card_quality: 96
    repo_hygiene: 97
  weights:
    popularity: 0.20
    momentum: 0.20
    benchmarks: 0.25
    card_quality: 0.20
    repo_hygiene: 0.15
  formula_note: >-
    0.20*popularity + 0.20*momentum + 0.25*benchmarks + 0.20*card_quality
    + 0.15*repo_hygiene. Popularity = min(100, downloads/100) = 3.37.
    Momentum = min(100, download_velocity*10) = 33.9 (338.7 dl/day * 0.1).
    Benchmarks 95 (verified:true entries in repo's own eval YAML).
    repo_hygiene 97 (all files present, 12 eval files — small deduction for
    no adapter_config.json or training artifacts since this is a merged model).

sibling_comparison:
  rank_by_downloads: 8
  total_author_models: 22
  max_sibling_downloads: 1855
  models_with_positive_downloads: 18
  velocity_rank: 1
  max_sibling_velocity: 338.69
  our_velocity: 338.69
  velocity_note: >-
    #1 download velocity among all Nanthasit models (338.7 dl/day) —
    this is the highest-velocity model in the family, likely driven by the
    v2 improvements and social reach from earlier cron cycles.
    Rank: 8/22 by absolute downloads — solid mid-upper position for a
    model published only ~1 day ago.

eval_type: metadata_cron
eval_note: >-
    SECOND cron eval for sakthai-context-1.5b-merged-v2. Since the first
    eval (run 17, ~18h ago), downloads have gone from 0 to 337 — the
    highest velocity (338.7 dl/day) of any Nanthasit model, surpassing
    even the 7B merged (1,024 dl over 30+ days). The model carries 12
    .eval_results/ files including bench-v2 metrics (all verified:true).
    The repository is well-maintained with a full README, 3 cited
    datasets (combined-v7, combined-v10, irrelevance-supplement), and
    active commits within hours. Recommendation: publish a Q4_K_M GGUF
    to extend reach to CPU/local llama.cpp users, and consider adding
    the sakthai-bench-v2 verifyToken to close the API-vs-YAML
    verification gap.

eval_metadata:
  model: Nanthasit/sakthai-context-1.5b-merged-v2
  eval_date: '2026-07-31'
  eval_time: '12:00:00Z'
  schema: llm_cron_v1
  age_days: 0.9950
  days_since_last_update: 0.0
  download_velocity: 338.69
  cron_run: 25