初始化项目，由ModelHub XC社区提供模型

Model: Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758
Source: Original Platform

.gitattributes

@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+phi-4.Q8_0.gguf filter=lfs diff=lfs merge=lfs -text

Modelfile

@@ -0,0 +1,8 @@
+
+FROM phi-4.Q8_0.gguf
+TEMPLATE """{{ if .System }}<|im_start|><|system|><|im_sep|>{{ .System }}<|im_end|>{{ end }}{{ if .Prompt }}<|im_start|><|user|><|im_sep|>{{ .Prompt }}<|im_end|>{{ end }}<|im_start|><|assistant|><|im_sep|>{{ .Response }}<|im_end|>"""
+PARAMETER stop "<|im_end|>"
+PARAMETER stop "<|im_start|>"
+PARAMETER stop "<|im_sep|>"
+PARAMETER temperature 1.5
+PARAMETER min_p 0.1

README.md

@@ -0,0 +1,54 @@
+---
+pipeline_tag: text-generation
+tags:
+  - gguf
+  - llama.cpp
+  - unsloth
+  - conversational
+base_model:
+  - unsloth/Phi-4-unsloth-bnb-4bit
+datasets:
+  - Mathieu-Thomas-JOSSET/michael_abab_conversations_infini_instruct.jsonl
+---
+# joke-finetome-model-gguf-phi4-20260112-081758 : GGUF
+
+This model was finetuned and converted to GGUF format using [Unsloth](https://github.com/unslothai/unsloth).
+
+**Example usage**:
+- For text only LLMs:    `./llama.cpp/llama-cli -hf Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758 --jinja`
+- For multimodal models: `./llama.cpp/llama-mtmd-cli -hf Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758 --jinja`
+
+## Available Model files:
+- `phi-4.Q8_0.gguf`
+
+## Ollama
+An Ollama Modelfile is included for easy deployment.
+This was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth)
+[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
+
+## Training artifacts
+
+- Plot (interactive): [`reports/training_loss_step.html`](reports/training_loss_step.html)
+- Run manifest: [`reports/run_manifest.json`](reports/run_manifest.json)
+- Inference sample: [`reports/inference_sample.json`](reports/inference_sample.json)
+- Config snapshot: [`reports/config_snapshot.json`](reports/config_snapshot.json)
+
+## Inference
+This repository contains a **GGUF** model intended to be used with **llama.cpp** and/or deployed on **Hugging Face Inference Endpoints (llama.cpp container)**.
+
+Recommended Inference Endpoints knobs:
+- Max tokens / request: **1024**
+- Max concurrent requests: **2**
+
+### Local llama.cpp (Phi-4 template)
+
+```bash
+llama-cli -hf Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758:q8_0 -cnv --chat-template phi4
+```
+
+### Hugging Face Inference Endpoint (llama.cpp)
+
+When creating an endpoint, select this repo and the GGUF file **<your_model>.gguf** (quant: **q8_0**).
+Recommended settings are stored in: `inference/endpoint_recipe.json`.
+
+Python client example: `inference/hf_endpoint_client.py`

config.json

@@ -0,0 +1,33 @@
+{
+    "architectures": [
+        "LlamaForCausalLM"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 100257,
+    "torch_dtype": "bfloat16",
+    "eos_token_id": 100265,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 5120,
+    "initializer_range": 0.02,
+    "intermediate_size": 17920,
+    "max_position_embeddings": 16384,
+    "mlp_bias": false,
+    "model_type": "llama",
+    "num_attention_heads": 40,
+    "num_hidden_layers": 40,
+    "num_key_value_heads": 10,
+    "original_max_position_embeddings": 16384,
+    "pad_token_id": 100351,
+    "pretraining_tp": 1,
+    "rms_norm_eps": 1e-05,
+    "rope_scaling": null,
+    "rope_theta": 250000,
+    "tie_word_embeddings": false,
+    "transformers_version": "4.56.2",
+    "unsloth_fixed": true,
+    "unsloth_version": "2026.1.2",
+    "use_cache": true,
+    "vocab_size": 100352
+}

inference/endpoint_recipe.json

@@ -0,0 +1,18 @@
+{
+  "engine": "llama.cpp",
+  "recommended_endpoint_settings": {
+    "max_tokens_per_request": 1024,
+    "max_concurrent_requests": 2,
+    "notes": "Memory scales roughly with (max_concurrent_requests * max_tokens_per_request)."
+  },
+  "recommended_generation_defaults": {
+    "temperature": 1.2,
+    "top_p": 0.95,
+    "min_p": 0.05,
+    "repeat_penalty": 1.08,
+    "max_tokens": 2560
+  },
+  "chat_template": "phi4",
+  "gguf_file": "",
+  "gguf_quant": "q8_0"
+}

inference/hf_endpoint_client.py

@@ -0,0 +1,23 @@
+import os
+from huggingface_hub import InferenceClient
+
+# Required env vars:
+#   export HF_TOKEN="..."
+#   export HF_ENDPOINT_BASE_URL="https://xxxx.endpoints.huggingface.cloud"
+
+client = InferenceClient(
+    base_url=os.environ["HF_ENDPOINT_BASE_URL"],
+    api_key=os.environ["HF_TOKEN"],
+)
+
+resp = client.chat.completions.create(
+    messages=[
+        {"role": "system", "content": "You are a helpful assistant."},
+        {"role": "user", "content": "Write a short joke in the style of The Office."},
+    ],
+    max_tokens=2560,
+    temperature=1.2,
+    top_p=0.95,
+)
+
+print(resp.choices[0].message.content)

inference/llama_cli_examples.md

@@ -0,0 +1,12 @@
+### Local inference (llama.cpp)
+
+```bash
+llama-cli -hf {REPO_ID}:q8_0 -cnv --chat-template phi4
+```
+
+### Server (OpenAI-compatible)
+
+```bash
+llama-server -hf {REPO_ID}:q8_0
+# /v1/chat/completions will be available (OpenAI-compatible)
+```

phi-4.Q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70be651211a3ca863ad8bc6c9b2596bf131f434686a5d06808c79bad7e5743b8
+size 15580507200

reports/config_snapshot.json

@@ -0,0 +1,28 @@
+{
+  "MODEL_NAME": "unsloth/Phi-4-unsloth-bnb-4bit",
+  "CHAT_TEMPLATE": "phi-4",
+  "MAX_SEQ_LENGTH": 2048,
+  "LOAD_IN_4BIT": true,
+  "DATASET_NAME": "Mathieu-Thomas-JOSSET/michael_abab_conversations_infini_instruct.jsonl",
+  "DATASET_SPLIT": "train",
+  "PER_DEVICE_TRAIN_BATCH_SIZE": 2,
+  "GRADIENT_ACCUMULATION_STEPS": 4,
+  "WARMUP_STEPS": 10,
+  "MAX_STEPS": 2000,
+  "LEARNING_RATE": 9.95267419777795e-06,
+  "LR_AUTO_ENABLED": true,
+  "LR_AUTO_USE_N": "train",
+  "LR_AUTO_N_REF": 1436,
+  "LR_AUTO_BASE": 1e-05,
+  "LR_AUTO_MULT": 0.5,
+  "LR_AUTO_FINAL": 5e-06,
+  "WEIGHT_DECAY": 0.009206070410847844,
+  "LR_SCHEDULER_TYPE": "linear",
+  "SEED": 3407,
+  "PLOTLY_DARK_MODE": true,
+  "PLOTLY_BASE_COLOR": "#00CC96",
+  "PLOTLY_EMA_SPAN": 25,
+  "HF_REPO_ID_MERGED_RESOLVED": "Mathieu-Thomas-JOSSET/joke-20260112-081758",
+  "HF_REPO_ID_GGUF_RESOLVED": "Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758",
+  "HF_ARTIFACTS_DIR_IN_REPO": "reports"
+}

reports/inference_sample.json

@@ -0,0 +1,10 @@
+{
+  "source": "dataset",
+  "index": 232,
+  "messages": [
+    {
+      "role": "user",
+      "content": "Dwight: \"Oh, man.\"\nMichael: \"How did we do it?\"\nDwight: \"I don’t … have no idea.\""
+    }
+  ]
+}

reports/run_manifest.json

@@ -0,0 +1,57 @@
+{
+  "started_at": "2026-01-12T08:52:02+01:00",
+  "repos": {
+    "merged": "Mathieu-Thomas-JOSSET/joke-20260112-081758",
+    "gguf": "Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758"
+  },
+  "model_name": "unsloth/Phi-4-unsloth-bnb-4bit",
+  "dataset": {
+    "name": "Mathieu-Thomas-JOSSET/michael_abab_conversations_infini_instruct.jsonl",
+    "split": "train"
+  },
+  "training": {
+    "max_steps": 2000,
+    "learning_rate": 9.95267419777795e-06,
+    "per_device_train_batch_size": 2,
+    "gradient_accumulation_steps": 4,
+    "max_seq_length": 2048,
+    "seed": 3407,
+    "optimizer": "adamw_8bit",
+    "lr_scheduler_type": "linear"
+  },
+  "auto_lr": {
+    "enabled": true,
+    "use_n": "train",
+    "n_ref": 1436,
+    "base": 1e-05,
+    "mult": 0.5,
+    "final": 5e-06
+  },
+  "metrics": {
+    "train_runtime": 2220.807,
+    "train_samples_per_second": 5.944,
+    "train_steps_per_second": 0.743,
+    "total_flos": 4.35104765343744e+16,
+    "train_loss": 1.654274252106746,
+    "epoch": 3.3342618384401113
+  },
+  "best": {
+    "checkpoint": "/content/outputs/continue_r1_from_350_20260112_073729/checkpoint-100",
+    "metric": 2.2380564212799072,
+    "metric_name": "eval_loss"
+  },
+  "plotly": {
+    "html": "reports/training_loss_step.html",
+    "png": null
+  },
+  "inference_sample": {
+    "source": "dataset",
+    "index": 232,
+    "messages": [
+      {
+        "role": "user",
+        "content": "Dwight: \"Oh, man.\"\nMichael: \"How did we do it?\"\nDwight: \"I don’t … have no idea.\""
+      }
+    ]
+  }
+}