File size: 13,632 Bytes
2c7042c
 
3da4f0d
 
2c7042c
3da4f0d
 
2c7042c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3da4f0d
2c7042c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3da4f0d
2c7042c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3da4f0d
 
a7d8dfa
3da4f0d
2c7042c
3da4f0d
73f54ee
 
2c7042c
 
73f54ee
3da4f0d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2c7042c
3da4f0d
 
 
2c7042c
3da4f0d
 
 
 
 
 
 
2c7042c
3da4f0d
a7d8dfa
73f54ee
 
 
 
 
 
 
 
 
 
 
3bc3aa3
 
 
 
3da4f0d
3bc3aa3
6001782
3da4f0d
2c7042c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3da4f0d
2c7042c
 
 
 
3da4f0d
 
2c7042c
 
 
 
3da4f0d
 
 
2c7042c
 
 
3da4f0d
73f54ee
2c7042c
3da4f0d
 
2c7042c
3da4f0d
 
ca8cbba
2c7042c
3da4f0d
73f54ee
2c7042c
 
73f54ee
3da4f0d
 
73f54ee
 
2c7042c
 
 
 
 
 
 
3da4f0d
2c7042c
 
 
 
 
 
 
 
 
 
 
 
3da4f0d
 
73f54ee
 
2c7042c
 
 
 
 
 
 
 
 
 
 
3da4f0d
2c7042c
3da4f0d
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
# app.py — DeepSeek-OCR + DeepSeek-R1 Medical Mini (remoto HF o local GGUF) — Gradio 5
import os, tempfile, traceback
import gradio as gr
import torch
from PIL import Image
from transformers import AutoModel, AutoTokenizer
import spaces
from huggingface_hub import hf_hub_download, InferenceClient
from llama_cpp import Llama

# ===============================================================
# Configuración LLM (CHAT) — DeepSeek-R1 Medical Mini
#  - Remoto (HF Inference): R1_REMOTE=1  y  (opcional) R1_MODEL_ID, HF_TOKEN
#  - Local GGUF (CPU/Zero): R1_REMOTE=0  y  GGUF_REPO / GGUF_FILE
# ===============================================================
R1_REMOTE = os.getenv("R1_REMOTE", "0") == "1"
R1_MODEL_ID = os.getenv("R1_MODEL_ID", "Mouhib007/DeepSeek-r1-Medical-Mini")
HF_TOKEN = os.getenv("HF_TOKEN")  # público -> puede ser None

# ---- Local GGUF (fallback / modo offline) ----
GGUF_CANDIDATES = []
ENV_REPO = os.getenv("GGUF_REPO", "").strip()
ENV_FILE = os.getenv("GGUF_FILE", "").strip()
if ENV_REPO and ENV_FILE:
    GGUF_CANDIDATES.append((ENV_REPO, ENV_FILE))
# Candidato por defecto (ajústalo si usas otro)
GGUF_CANDIDATES.append((
    "mradermacher/DeepSeek-r1-Medical-Mini-GGUF",
    "DeepSeek-r1-Medical-Mini.f16.gguf"
))

N_CTX = int(os.getenv("N_CTX", "2048"))
N_THREADS = int(os.getenv("N_THREADS", str(os.cpu_count() or 4)))
N_GPU_LAYERS = int(os.getenv("N_GPU_LAYERS", "0"))
N_BATCH = int(os.getenv("N_BATCH", "96"))

# ---- Cliente remoto (HF Inference) ----
_remote_client = None
def get_remote_client():
    global _remote_client
    if _remote_client is None:
        _remote_client = InferenceClient(model=R1_MODEL_ID, token=HF_TOKEN, timeout=60)
    return _remote_client

# ---- Formato ChatML (compatible con DeepSeek/Qwen) ----
def _format_chatml(messages):
    parts = []
    for m in messages:
        role = m.get("role", "user")
        content = m.get("content", "")
        parts.append(f"<|im_start|>{role}\n{content}<|im_end|>\n")
    parts.append("<|im_start|>assistant\n")
    return "".join(parts)

def r1_chat(messages, temperature=0.2, max_tokens=384):
    """Remoto (HF) o local (llama-cpp) para DeepSeek-R1 Medical Mini."""
    if R1_REMOTE:
        client = get_remote_client()
        try:
            # Algunos endpoints soportan chat_completion
            resp = client.chat_completion(messages=messages, temperature=temperature, max_tokens=max_tokens)
            return resp.choices[0].message["content"]
        except Exception:
            # Fallback universal a text_generation con ChatML
            try:
                prompt = _format_chatml(messages)
                return client.text_generation(
                    prompt,
                    max_new_tokens=max_tokens,
                    temperature=temperature,
                    stop_sequences=["<|im_end|>"],
                    stream=False,
                )
            except Exception:
                # Si remoto falla (401/429/etc), caemos a local si hay GGUF
                pass
    # Local GGUF
    llm = get_llm()
    out = llm.create_chat_completion(messages=messages, temperature=temperature, max_tokens=max_tokens)
    return out["choices"][0]["message"]["content"]

# ---- Loader local (GGUF) ----
_llm = None
def _download_gguf():
    last_err = None
    for repo, fname in GGUF_CANDIDATES:
        try:
            return hf_hub_download(repo_id=repo, filename=fname), repo, fname
        except Exception as e:
            last_err = e
    raise RuntimeError(f"No se pudo descargar ningún GGUF. Último error: {last_err}")

def get_llm():
    global _llm
    if _llm is not None:
        return _llm
    gguf_path, _, _ = _download_gguf()
    _llm = Llama(
        model_path=gguf_path,
        # No forzamos chat_format; usamos el del GGUF del R1
        n_ctx=N_CTX,
        n_threads=N_THREADS,
        n_gpu_layers=N_GPU_LAYERS,
        n_batch=N_BATCH,
        verbose=False,
    )
    return _llm

# Warmup opcional (para no esperar en el primer mensaje si usas local)
if os.getenv("WARMUP", "0") == "1" and not R1_REMOTE:
    try:
        get_llm()
    except Exception:
        pass

# ===============================================================
# DeepSeek-OCR (INTACTO — con fallback si no hay FlashAttention2)
# ===============================================================
def _best_dtype():
    if torch.cuda.is_available():
        return torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float16
    return torch.float32

def _load_ocr_model():
    model_name = "deepseek-ai/DeepSeek-OCR"
    ocr_tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
    attn_impl = os.getenv("OCR_ATTN_IMPL", "flash_attention_2")  # por defecto igual que antes
    try:
        ocr_model = AutoModel.from_pretrained(
            model_name,
            _attn_implementation=attn_impl,
            trust_remote_code=True,
            use_safetensors=True,
        ).eval()
        return ocr_tokenizer, ocr_model
    except Exception as e:
        # Si falla por FlashAttention2, reintenta en modo "eager" (CPU/compat)
        msg = str(e)
        if "flash_attn" in msg or "FlashAttention2" in msg or "flash_attention_2" in msg:
            ocr_model = AutoModel.from_pretrained(
                model_name,
                _attn_implementation="eager",
                trust_remote_code=True,
                use_safetensors=True,
            ).eval()
            return ocr_tokenizer, ocr_model
        raise

tokenizer, model = _load_ocr_model()

@spaces.GPU
def process_image(image, model_size, task_type, is_eval_mode):
    """
    Devuelve: imagen anotada, markdown y texto (o markdown si no hay texto).
    """
    if image is None:
        return None, "Please upload an image first.", "Please upload an image first."
    dtype = _best_dtype()
    model_device = model.cuda().to(dtype) if torch.cuda.is_available() else model.to(dtype)

    with tempfile.TemporaryDirectory() as output_path:
        if task_type == "Free OCR":
            prompt = "<image>\nFree OCR. "
        elif task_type == "Convert to Markdown":
            prompt = "<image>\n<|grounding|>Convert the document to markdown. "
        else:
            prompt = "<image>\nFree OCR. "

        temp_image_path = os.path.join(output_path, "temp_image.jpg")
        image.save(temp_image_path)

        size_configs = {
            "Tiny": {"base_size": 512, "image_size": 512, "crop_mode": False},
            "Small": {"base_size": 640, "image_size": 640, "crop_mode": False},
            "Base": {"base_size": 1024, "image_size": 1024, "crop_mode": False},
            "Large": {"base_size": 1280, "image_size": 1280, "crop_mode": False},
            "Gundam (Recommended)": {"base_size": 1024, "image_size": 640, "crop_mode": True},
        }
        config = size_configs.get(model_size, size_configs["Gundam (Recommended)"])

        plain_text_result = model_device.infer(
            tokenizer,
            prompt=prompt,
            image_file=temp_image_path,
            output_path=output_path,
            base_size=config["base_size"],
            image_size=config["image_size"],
            crop_mode=config["crop_mode"],
            save_results=True,
            test_compress=True,
            eval_mode=is_eval_mode,
        )

        image_result_path = os.path.join(output_path, "result_with_boxes.jpg")
        markdown_result_path = os.path.join(output_path, "result.mmd")

        if os.path.exists(markdown_result_path):
            with open(markdown_result_path, "r", encoding="utf-8") as f:
                markdown_content = f.read()
        else:
            markdown_content = "Markdown result was not generated. This is expected for 'Free OCR' task."

        result_image = None
        if os.path.exists(image_result_path):
            result_image = Image.open(image_result_path)
            result_image.load()

        text_result = plain_text_result if plain_text_result else markdown_content
        return result_image, markdown_content, text_result

# ===============================================================
# Chat (inyecta OCR en el primer system) — usando R1
# ===============================================================
def _truncate(text, max_chars=3000):
    return (text or "")[:max_chars]

def _system_prompt():
    return (
        "Eres un asistente clínico educativo. No sustituyes el juicio médico. "
        "Usa CONTEXTO_OCR si existe; si falta, pídelo. Evita diagnósticos definitivos."
    )

def _ocr_context(ocr_md, ocr_txt):
    return _truncate(ocr_md) or _truncate(ocr_txt) or ""

def to_chat_messages(chat_msgs, ocr_md, ocr_txt):
    sys = _system_prompt()
    ctx = _ocr_context(ocr_md, ocr_txt)
    if ctx:
        sys += (
            "\n\n---\n"
            "CONTEXTO_OCR (fuente principal; si falta un dato, dilo explícitamente):\n"
            f"{ctx}\n---"
        )
    msgs = [{"role": "system", "content": sys}]
    for m in (chat_msgs or []):
        if m.get("role") in ("user", "assistant"):
            msgs.append({"role": m["role"], "content": m.get("content", "")})
    return msgs

def r1_reply(user_msg, chat_msgs, ocr_md, ocr_txt):
    if not user_msg:
        user_msg = "Analiza el CONTEXTO_OCR anterior y responde a partir de ese contenido."
    try:
        msgs = to_chat_messages(chat_msgs, ocr_md, ocr_txt) + [{"role": "user", "content": user_msg}]
        answer = r1_chat(msgs, temperature=0.2, max_tokens=512)
        updated = (chat_msgs or []) + [
            {"role": "user", "content": user_msg},
            {"role": "assistant", "content": answer},
        ]
        return updated, "", gr.update(value="")
    except Exception as e:
        err = f"{e.__class__.__name__}: {str(e) or repr(e)}"
        tb = traceback.format_exc(limit=2)
        updated = (chat_msgs or []) + [
            {"role": "user", "content": user_msg or ""},
            {"role": "assistant", "content": f"⚠️ Error LLM: {err}"},
        ]
        return updated, "", gr.update(value=f"{err}\n{tb}")

def clear_chat():
    return [], "", gr.update(value="")

# ===============================================================
# UI (Gradio 5)
# ===============================================================
with gr.Blocks(title="DeepSeek-OCR + DeepSeek-R1 Medical Mini", theme=gr.themes.Soft()) as demo:
    gr.Markdown(
        """
        # DeepSeek-OCR → Chat Médico con **DeepSeek-R1 Medical Mini** (remoto HF o local GGUF)
        1) **Sube una imagen** y corre **OCR** (imagen anotada, Markdown y texto).  
        2) **Chatea** con **DeepSeek-R1 Medical Mini** usando automáticamente el **OCR** como contexto.  
        *Uso educativo; no reemplaza consejo médico.*
        """
    )

    ocr_md_state = gr.State("")
    ocr_txt_state = gr.State("")

    with gr.Row():
        with gr.Column(scale=1):
            image_input = gr.Image(type="pil", label="Upload Image", sources=["upload", "clipboard", "webcam"])
            model_size = gr.Dropdown(
                choices=["Tiny", "Small", "Base", "Large", "Gundam (Recommended)"],
                value="Gundam (Recommended)", label="Model Size",
            )
            task_type = gr.Dropdown(
                choices=["Free OCR", "Convert to Markdown"],
                value="Convert to Markdown", label="Task Type",
            )
            eval_mode_checkbox = gr.Checkbox(
                value=False, label="Enable Evaluation Mode",
                info="Solo texto (más rápido). Desmárcalo para ver imagen anotada y markdown.",
            )
            submit_btn = gr.Button("Process Image", variant="primary")

        with gr.Column(scale=2):
            with gr.Tabs():
                with gr.TabItem("Annotated Image"): output_image = gr.Image(interactive=False)
                with gr.TabItem("Markdown Preview"): output_markdown = gr.Markdown()
                with gr.TabItem("Markdown Source (or Eval Output)"):
                    output_text = gr.Textbox(lines=18, show_copy_button=True, interactive=False)
            with gr.Row():
                md_preview = gr.Textbox(label="Snapshot Markdown OCR", lines=10, interactive=False)
                txt_preview = gr.Textbox(label="Snapshot Texto OCR", lines=10, interactive=False)

    gr.Markdown("## Chat Clínico (DeepSeek-R1 Medical Mini)")
    with gr.Row():
        with gr.Column(scale=2):
            chatbot = gr.Chatbot(label="Asistente OCR (R1 Medical Mini)", type="messages", height=420)
            user_in = gr.Textbox(label="Mensaje", placeholder="Escribe tu consulta… (vacío = analiza solo el OCR)", lines=2)
            with gr.Row():
                send_btn = gr.Button("Enviar", variant="primary")
                clear_btn = gr.Button("Limpiar")
        with gr.Column(scale=1):
            error_box = gr.Textbox(label="Debug (si hay error)", lines=8, interactive=False)

    # OCR → outputs y estados
    submit_btn.click(
        fn=process_image,
        inputs=[image_input, model_size, task_type, eval_mode_checkbox],
        outputs=[output_image, output_markdown, output_text],
    ).then(
        fn=lambda md, tx: (md, tx, md, tx),
        inputs=[output_markdown, output_text],
        outputs=[ocr_md_state, ocr_txt_state, md_preview, txt_preview],
    )

    # Chat
    send_btn.click(
        fn=r1_reply,
        inputs=[user_in, chatbot, ocr_md_state, ocr_txt_state],
        outputs=[chatbot, user_in, error_box],
    )
    clear_btn.click(fn=clear_chat, outputs=[chatbot, user_in, error_box])

if __name__ == "__main__":
    demo.queue(max_size=20)
    demo.launch()