Instructions to use josephmayo/Holo-3.1-4B-Coder with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use josephmayo/Holo-3.1-4B-Coder with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="josephmayo/Holo-3.1-4B-Coder") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("josephmayo/Holo-3.1-4B-Coder") model = AutoModelForCausalLM.from_pretrained("josephmayo/Holo-3.1-4B-Coder", 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 josephmayo/Holo-3.1-4B-Coder with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "josephmayo/Holo-3.1-4B-Coder" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "josephmayo/Holo-3.1-4B-Coder", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/josephmayo/Holo-3.1-4B-Coder
- SGLang
How to use josephmayo/Holo-3.1-4B-Coder 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 "josephmayo/Holo-3.1-4B-Coder" \ --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": "josephmayo/Holo-3.1-4B-Coder", "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 "josephmayo/Holo-3.1-4B-Coder" \ --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": "josephmayo/Holo-3.1-4B-Coder", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use josephmayo/Holo-3.1-4B-Coder with Docker Model Runner:
docker model run hf.co/josephmayo/Holo-3.1-4B-Coder
Add local evidence artifacts and model-card evidence links
Browse files
README.md
CHANGED
|
@@ -56,3 +56,16 @@ Use this checkpoint for coding assistance experiments, Python function generatio
|
|
| 56 |
## Reproducibility And Provenance
|
| 57 |
|
| 58 |
The model was produced by merging a PEFT LoRA coding adapter into `Hcompany/Holo-3.1-4B` and saving the result as sharded `safetensors`. Companion evaluation and training provenance artifacts are available in the LoRA repository.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
## Reproducibility And Provenance
|
| 57 |
|
| 58 |
The model was produced by merging a PEFT LoRA coding adapter into `Hcompany/Holo-3.1-4B` and saving the result as sharded `safetensors`. Companion evaluation and training provenance artifacts are available in the LoRA repository.
|
| 59 |
+
|
| 60 |
+
<!-- evidence-files:start -->
|
| 61 |
+
|
| 62 |
+
## Evidence files
|
| 63 |
+
|
| 64 |
+
Run evidence for this release is stored in the repository under `evidence/`:
|
| 65 |
+
|
| 66 |
+
- [`evidence/holo_4b_repair38_merge_cpu_out_v3_merge_summary.json`](./evidence/holo_4b_repair38_merge_cpu_out_v3_merge_summary.json)
|
| 67 |
+
- [`evidence/holo_4b_repair38_merge_cpu_out_v3_holo_4b_repair38_adapter_release_summary.json`](./evidence/holo_4b_repair38_merge_cpu_out_v3_holo_4b_repair38_adapter_release_summary.json)
|
| 68 |
+
|
| 69 |
+
These files are compact local/Kaggle run artifacts used to document training, evaluation, merge, or quantization evidence for this model family.
|
| 70 |
+
|
| 71 |
+
<!-- evidence-files:end -->
|
evidence/README.md
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Evidence files
|
| 2 |
+
|
| 3 |
+
This folder contains local training/evaluation/merge/quantization evidence artifacts copied from the original run outputs.
|
| 4 |
+
|
| 5 |
+
- `evidence/holo_4b_repair38_merge_cpu_out_v3_merge_summary.json`
|
| 6 |
+
- `evidence/holo_4b_repair38_merge_cpu_out_v3_holo_4b_repair38_adapter_release_summary.json`
|
evidence/holo_4b_repair38_merge_cpu_out_v3_holo_4b_repair38_adapter_release_summary.json
ADDED
|
@@ -0,0 +1,105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"run_slug": "holo-3-1-4b-repair38-v1",
|
| 3 |
+
"base_model_id": "Hcompany/Holo-3.1-4B",
|
| 4 |
+
"adapter_repo": "josephmayo/Holo-3.1-4B-Coding-Repair38-LoRA",
|
| 5 |
+
"method": "4-bit QLoRA SFT on curated local coding splits plus external coding instruction data; gated merge only",
|
| 6 |
+
"target": ">=38/80 pass count on held-out HumanEval+MBPP greedy probe after adapter repair",
|
| 7 |
+
"push_to_hf": true,
|
| 8 |
+
"push_even_if_gate_fail": false,
|
| 9 |
+
"seed": 42,
|
| 10 |
+
"stage": "done",
|
| 11 |
+
"errors": [],
|
| 12 |
+
"warnings": [],
|
| 13 |
+
"hf_token_secret_name": "hardcoded_fallback",
|
| 14 |
+
"hf_token_present": true,
|
| 15 |
+
"dataset_selection": {
|
| 16 |
+
"root": "/kaggle/input/datasets/josephayanda/curated-agent-coding-dataset-sft-v1",
|
| 17 |
+
"used": [
|
| 18 |
+
{
|
| 19 |
+
"file": "targeted_sft_2000.jsonl",
|
| 20 |
+
"available": 2000,
|
| 21 |
+
"selected": 2000,
|
| 22 |
+
"purpose": "targeted failure families"
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"file": "eval_maxxing_3500.jsonl",
|
| 26 |
+
"available": 3500,
|
| 27 |
+
"selected": 3200,
|
| 28 |
+
"purpose": "hidden-test style coding"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"file": "coding_sft_5000.jsonl",
|
| 32 |
+
"available": 5000,
|
| 33 |
+
"selected": 3200,
|
| 34 |
+
"purpose": "broad language coding"
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"file": "real_world_coding_3500.jsonl",
|
| 38 |
+
"available": 3500,
|
| 39 |
+
"selected": 1800,
|
| 40 |
+
"purpose": "production coding tasks"
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"file": "heavy_real_world_agentic_5000.jsonl",
|
| 44 |
+
"available": 5000,
|
| 45 |
+
"selected": 120,
|
| 46 |
+
"purpose": "long repo-agent examples"
|
| 47 |
+
}
|
| 48 |
+
],
|
| 49 |
+
"skipped_preference_pairs": [
|
| 50 |
+
"orpo_gold_1000.jsonl",
|
| 51 |
+
"reasoning_preference_1000.jsonl"
|
| 52 |
+
],
|
| 53 |
+
"external": [
|
| 54 |
+
{
|
| 55 |
+
"dataset": "ise-uiuc/Magicoder-Evol-Instruct-110K",
|
| 56 |
+
"selected": 96
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"dataset": "m-a-p/CodeFeedback-Filtered-Instruction",
|
| 60 |
+
"selected": 96
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"dataset": "HuggingFaceH4/CodeAlpaca_20K",
|
| 64 |
+
"selected": 96
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"dataset": "glaiveai/glaive-code-assistant-v3",
|
| 68 |
+
"selected": 96
|
| 69 |
+
}
|
| 70 |
+
]
|
| 71 |
+
},
|
| 72 |
+
"train_rows_loaded": 10704,
|
| 73 |
+
"cuda_available": true,
|
| 74 |
+
"cuda_device_count": 2,
|
| 75 |
+
"devices": [
|
| 76 |
+
"Tesla T4",
|
| 77 |
+
"Tesla T4"
|
| 78 |
+
],
|
| 79 |
+
"eval_task_count": 80,
|
| 80 |
+
"before_counts": {
|
| 81 |
+
"pass": 24,
|
| 82 |
+
"fail": 56
|
| 83 |
+
},
|
| 84 |
+
"before_eval_skipped_reused": true,
|
| 85 |
+
"lora_target_modules": [
|
| 86 |
+
"q_proj",
|
| 87 |
+
"k_proj",
|
| 88 |
+
"v_proj",
|
| 89 |
+
"o_proj"
|
| 90 |
+
],
|
| 91 |
+
"repair_source_adapter": "josephmayo/Holo-3.1-4B-Coding-30pct-LoRA",
|
| 92 |
+
"after_counts": {
|
| 93 |
+
"pass": 31,
|
| 94 |
+
"fail": 49
|
| 95 |
+
},
|
| 96 |
+
"improvement_gate": {
|
| 97 |
+
"ran_eval": true,
|
| 98 |
+
"before_pass": 24,
|
| 99 |
+
"after_pass": 31,
|
| 100 |
+
"target_pass_for_repair38": 38,
|
| 101 |
+
"relative_improvement": 0.2917,
|
| 102 |
+
"passed": false
|
| 103 |
+
},
|
| 104 |
+
"push_skipped": "repair38 gate failed; set PUSH_EVEN_IF_GATE_FAIL=1 to push adapter anyway"
|
| 105 |
+
}
|
evidence/holo_4b_repair38_merge_cpu_out_v3_merge_summary.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"stage": "error",
|
| 3 |
+
"base_model_id": "Hcompany/Holo-3.1-4B",
|
| 4 |
+
"adapter_repo": "josephmayo/Holo-3.1-4B-Coding-Repair38-LoRA",
|
| 5 |
+
"merged_repo": "josephmayo/Holo-3.1-4B-Coding-Repair38-Merged",
|
| 6 |
+
"model_name": "Holo-3.1-4B-Coding-Repair38",
|
| 7 |
+
"errors": [
|
| 8 |
+
"ImportError('Found an incompatible version of torchao. Found version 0.10.0, but only versions above 0.16.0 are supported')"
|
| 9 |
+
]
|
| 10 |
+
}
|