初始化项目，由ModelHub XC社区提供模型
Model: OmkarShewale/clintrial-qwen2.5-7b-sft Source: Original Platform
This commit is contained in:
ModelHub XC
36
.gitattributes
vendored
Normal file
36
.gitattributes
vendored
Normal file
@@ -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
|
||||
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
||||
135
README.md
Normal file
135
README.md
Normal file
@@ -0,0 +1,135 @@
|
||||
---
|
||||
license: apache-2.0
|
||||
base_model: Qwen/Qwen2.5-7B-Instruct
|
||||
library_name: transformers
|
||||
pipeline_tag: text-generation
|
||||
language:
|
||||
- en
|
||||
tags:
|
||||
- clinical-trials
|
||||
- healthcare
|
||||
- biomedical
|
||||
- information-extraction
|
||||
- qlora
|
||||
- peft
|
||||
- trl
|
||||
- grpo
|
||||
---
|
||||
|
||||
# ClinTrial-LM — Qwen2.5-7B fine-tuned for clinical-trial understanding
|
||||
|
||||
A **QLoRA**-fine-tuned [Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct)
|
||||
specialised for clinical-trial tasks, trained on ~26k instruction examples derived from the
|
||||
[ClinicalTrials.gov](https://clinicaltrials.gov) registry.
|
||||
|
||||
📦 **Code, full pipeline & write-up:** https://github.com/omkar-droid/clintrial-lm
|
||||
|
||||
> ⚕️ **Research/education only. Not medical advice.** Do not use for clinical decision-making.
|
||||
|
||||
## What it does
|
||||
|
||||
| Task | Description |
|
||||
|---|---|
|
||||
| **Eligibility extraction** | Free-text inclusion/exclusion criteria → structured JSON |
|
||||
| **Plain-language summary** | Trial description → patient-friendly summary |
|
||||
| **Condition Q&A** | "What conditions does this trial study?" |
|
||||
| **Phase classification** | Identify the trial phase |
|
||||
|
||||
## Results
|
||||
|
||||
Evaluated on a **held-out test set split by trial ID** (no trial appears in both train and test),
|
||||
500 sampled examples.
|
||||
|
||||
| Task | Metric | Base Qwen2.5-7B | **This model (SFT)** |
|
||||
|---|---|---|---|
|
||||
| Eligibility extraction | criterion F1 | 0.840 | **0.968** |
|
||||
| Eligibility output | **JSON validity** | 0.986 | **1.000** |
|
||||
| Phase classification | exact match | 0.000 | **0.794** |
|
||||
| Condition Q&A | token F1 | 0.298 | **0.742** |
|
||||
| Plain-language summary | ROUGE-L | 0.195 | **0.290** |
|
||||
|
||||
Eligibility extraction reaches **precision 0.962 / recall 0.978** with **100% schema-valid JSON**.
|
||||
|
||||
The base model scores 0.000 on phase classification not because it lacks the knowledge, but because
|
||||
it won't answer in the required format — fine-tuning buys **format discipline and faithfulness**.
|
||||
|
||||
## Usage
|
||||
|
||||
```python
|
||||
import torch
|
||||
from transformers import AutoModelForCausalLM, AutoTokenizer
|
||||
|
||||
MODEL_ID = "OmkarShewale/clintrial-qwen2.5-7b-sft"
|
||||
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
||||
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, dtype=torch.bfloat16, device_map="auto")
|
||||
|
||||
SYSTEM = (
|
||||
"You are a clinical research assistant. You help patients and clinicians understand "
|
||||
"clinical trials. Answer only from the information provided, be precise, and never "
|
||||
"invent eligibility criteria, conditions, or outcomes."
|
||||
)
|
||||
|
||||
criteria = """Inclusion Criteria:
|
||||
1. Adults aged 18 years or older with confirmed type 2 diabetes
|
||||
2. HbA1c between 7.0% and 10.5% at screening
|
||||
|
||||
Exclusion Criteria:
|
||||
1. History of severe hypoglycaemia within the last 6 months
|
||||
2. Pregnancy or breastfeeding
|
||||
"""
|
||||
|
||||
messages = [
|
||||
{"role": "system", "content": SYSTEM},
|
||||
{"role": "user", "content":
|
||||
'Extract the eligibility criteria from the trial text below into JSON with two lists, '
|
||||
'"inclusion" and "exclusion". Copy each criterion verbatim; do not add any.\n\n'
|
||||
f"Trial eligibility text:\n{criteria}"},
|
||||
]
|
||||
|
||||
enc = tokenizer.apply_chat_template(messages, add_generation_prompt=True,
|
||||
return_tensors="pt", return_dict=True).to(model.device)
|
||||
out = model.generate(**enc, max_new_tokens=1024, do_sample=False)
|
||||
print(tokenizer.decode(out[0, enc["input_ids"].shape[1]:], skip_special_tokens=True))
|
||||
```
|
||||
|
||||
**Note:** eligibility answers can be long (up to ~1.7k tokens). Use `max_new_tokens >= 1024` or the
|
||||
JSON will be truncated and fail to parse.
|
||||
|
||||
## Training
|
||||
|
||||
| | |
|
||||
|---|---|
|
||||
| Base model | Qwen2.5-7B-Instruct |
|
||||
| Method | QLoRA — 4-bit NF4 + LoRA (r=16, α=32, dropout 0.05) |
|
||||
| Target modules | q/k/v/o_proj, gate/up/down_proj |
|
||||
| Trainable params | ~0.5% of total |
|
||||
| Precision | bf16 compute, gradient checkpointing |
|
||||
| Effective batch | 32 (8 × 4 grad-accum) |
|
||||
| LR / schedule | 2e-4, cosine, 3% warmup |
|
||||
| Hardware | 1× NVIDIA H100 NVL (95 GB), ~19 GB used |
|
||||
| Framework | HuggingFace TRL + PEFT + Transformers |
|
||||
|
||||
**Best-checkpoint selection matters here:** eval loss bottomed around step 500 and then *rose* while
|
||||
train loss kept falling (overfitting). `load_best_model_at_end` on `eval_loss` selected the step-500
|
||||
checkpoint — this model. Training for 3 epochs was more than necessary.
|
||||
|
||||
A **GRPO / RLVR** variant (RL against a programmatic reward: JSON validity + criterion F1) was also
|
||||
trained; it matched this SFT model but did not beat it, because SFT had already saturated the reward.
|
||||
Details in the [repo](https://github.com/omkar-droid/clintrial-lm).
|
||||
|
||||
## Data
|
||||
|
||||
Built from 8,000 real ClinicalTrials.gov studies. Targets come from the registry's **own structured
|
||||
fields** (not a teacher LLM), so labels are auditable. Splits are partitioned **by trial**, not by
|
||||
example, to prevent leakage.
|
||||
|
||||
## Limitations
|
||||
|
||||
- **Not medical advice.** Outputs may be wrong; a human expert must review anything clinical.
|
||||
- Trained on English registry text only; performance on other formats/languages is unknown.
|
||||
- Summary quality (ROUGE-L 0.29) is the weakest task.
|
||||
- Only 8k of ~500k available trials were used — more data would likely improve generalisation.
|
||||
|
||||
## License
|
||||
|
||||
Apache-2.0, inherited from Qwen2.5. Source registry data is public-domain U.S. government work.
|
||||
54
chat_template.jinja
Normal file
54
chat_template.jinja
Normal file
@@ -0,0 +1,54 @@
|
||||
{%- if tools %}
|
||||
{{- '<|im_start|>system\n' }}
|
||||
{%- if messages[0]['role'] == 'system' %}
|
||||
{{- messages[0]['content'] }}
|
||||
{%- else %}
|
||||
{{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
|
||||
{%- endif %}
|
||||
{{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
||||
{%- for tool in tools %}
|
||||
{{- "\n" }}
|
||||
{{- tool | tojson }}
|
||||
{%- endfor %}
|
||||
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
||||
{%- else %}
|
||||
{%- if messages[0]['role'] == 'system' %}
|
||||
{{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
|
||||
{%- else %}
|
||||
{{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
|
||||
{%- endif %}
|
||||
{%- endif %}
|
||||
{%- for message in messages %}
|
||||
{%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
|
||||
{{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
|
||||
{%- elif message.role == "assistant" %}
|
||||
{{- '<|im_start|>' + message.role }}
|
||||
{%- if message.content %}
|
||||
{{- '\n' + message.content }}
|
||||
{%- endif %}
|
||||
{%- for tool_call in message.tool_calls %}
|
||||
{%- if tool_call.function is defined %}
|
||||
{%- set tool_call = tool_call.function %}
|
||||
{%- endif %}
|
||||
{{- '\n<tool_call>\n{"name": "' }}
|
||||
{{- tool_call.name }}
|
||||
{{- '", "arguments": ' }}
|
||||
{{- tool_call.arguments | tojson }}
|
||||
{{- '}\n</tool_call>' }}
|
||||
{%- endfor %}
|
||||
{{- '<|im_end|>\n' }}
|
||||
{%- elif message.role == "tool" %}
|
||||
{%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
|
||||
{{- '<|im_start|>user' }}
|
||||
{%- endif %}
|
||||
{{- '\n<tool_response>\n' }}
|
||||
{{- message.content }}
|
||||
{{- '\n</tool_response>' }}
|
||||
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
||||
{{- '<|im_end|>\n' }}
|
||||
{%- endif %}
|
||||
{%- endif %}
|
||||
{%- endfor %}
|
||||
{%- if add_generation_prompt %}
|
||||
{{- '<|im_start|>assistant\n' }}
|
||||
{%- endif %}
|
||||
61
config.json
Normal file
61
config.json
Normal file
@@ -0,0 +1,61 @@
|
||||
{
|
||||
"architectures": [
|
||||
"Qwen2ForCausalLM"
|
||||
],
|
||||
"attention_dropout": 0.0,
|
||||
"bos_token_id": 151643,
|
||||
"dtype": "bfloat16",
|
||||
"eos_token_id": 151645,
|
||||
"hidden_act": "silu",
|
||||
"hidden_size": 3584,
|
||||
"initializer_range": 0.02,
|
||||
"intermediate_size": 18944,
|
||||
"layer_types": [
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention"
|
||||
],
|
||||
"max_position_embeddings": 32768,
|
||||
"max_window_layers": 28,
|
||||
"model_type": "qwen2",
|
||||
"num_attention_heads": 28,
|
||||
"num_hidden_layers": 28,
|
||||
"num_key_value_heads": 4,
|
||||
"pad_token_id": null,
|
||||
"rms_norm_eps": 1e-06,
|
||||
"rope_parameters": {
|
||||
"rope_theta": 1000000.0,
|
||||
"rope_type": "default"
|
||||
},
|
||||
"sliding_window": null,
|
||||
"tie_word_embeddings": false,
|
||||
"transformers_version": "5.13.1",
|
||||
"use_cache": true,
|
||||
"use_sliding_window": false,
|
||||
"vocab_size": 152064
|
||||
}
|
||||
14
generation_config.json
Normal file
14
generation_config.json
Normal file
@@ -0,0 +1,14 @@
|
||||
{
|
||||
"bos_token_id": 151643,
|
||||
"do_sample": true,
|
||||
"eos_token_id": [
|
||||
151645,
|
||||
151643
|
||||
],
|
||||
"pad_token_id": 151643,
|
||||
"repetition_penalty": 1.05,
|
||||
"temperature": 0.7,
|
||||
"top_k": 20,
|
||||
"top_p": 0.8,
|
||||
"transformers_version": "5.13.1"
|
||||
}
|
||||
3
model.safetensors
Normal file
3
model.safetensors
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:0e1e5c534f46bdb2d27434016f1c62f07a8aba7b1a66483d06b22e0f9ffbba4b
|
||||
size 15231272152
|
||||
3
tokenizer.json
Normal file
3
tokenizer.json
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:3fd169731d2cbde95e10bf356d66d5997fd885dd8dbb6fb4684da3f23b2585d8
|
||||
size 11421892
|
||||
30
tokenizer_config.json
Normal file
30
tokenizer_config.json
Normal file
@@ -0,0 +1,30 @@
|
||||
{
|
||||
"add_prefix_space": false,
|
||||
"backend": "tokenizers",
|
||||
"bos_token": null,
|
||||
"clean_up_tokenization_spaces": false,
|
||||
"eos_token": "<|im_end|>",
|
||||
"errors": "replace",
|
||||
"extra_special_tokens": [
|
||||
"<|im_start|>",
|
||||
"<|im_end|>",
|
||||
"<|object_ref_start|>",
|
||||
"<|object_ref_end|>",
|
||||
"<|box_start|>",
|
||||
"<|box_end|>",
|
||||
"<|quad_start|>",
|
||||
"<|quad_end|>",
|
||||
"<|vision_start|>",
|
||||
"<|vision_end|>",
|
||||
"<|vision_pad|>",
|
||||
"<|image_pad|>",
|
||||
"<|video_pad|>"
|
||||
],
|
||||
"is_local": false,
|
||||
"local_files_only": false,
|
||||
"model_max_length": 131072,
|
||||
"pad_token": "<|endoftext|>",
|
||||
"split_special_tokens": false,
|
||||
"tokenizer_class": "Qwen2Tokenizer",
|
||||
"unk_token": null
|
||||
}
|
||||
Reference in New Issue
Block a user